university of groningen (un)healthy in the city zijlema, wilma...example is the london smog of 1952....

University of Groningen

(Un)Healthy in the cityZijlema, Wilma

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite fromit. Please check the document version below.

Document VersionPublisher's PDF, also known as Version of record

Publication date:2016

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):Zijlema, W. (2016). (Un)Healthy in the city: Adverse health effects of traffic-related noise and air pollution.University of Groningen.

CopyrightOther than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of theauthor(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons thenumber of authors shown on this cover page is limited to 10 maximum.

Download date: 08-05-2021

https://research.rug.nl/en/publications/unhealthy-in-the-city(e21944bf-a6b1-4ce5-9594-15ad6c18f391).html

(Un)Healthy in the city

Adverse health effects of traffic-related noise and air pollution

Wilma L. Zijlema

© 2016, Wilma L. Zijlema

ParanymphsBrenda Samaniego Cameron and Maaike Meurs

Lay-out: Ridderprint BV - www.ridderprint.nlPrinting: Ridderprint BV - www.ridderprint.nlCover design: Wilma Zijlema

The research leading to these results has received funding from the European Union Seventh Framework Pro-gramme (FP7/2007-2013) under grant agreement n°261433 (Biobank Standardisation and Harmonisation for Research Excellence in the European Union - BioSHaRE-EU).

The Lifelines Biobank initiative has been made possible by funds from FES (Fonds Economische Structuurver-sterking), SNN (Samenwerkingsverband Noord Nederland) and REP (Ruimtelijk Economisch Programma).

Printing of this dissertation was supported by the University Medical Center Groningen, the University of Gron-ingen, and the Graduate School SHARE of the University Medical Center Groningen. Financial support bythe Dutch Heart Foundation for the publication of this thesis is gratefully acknowledged.

ISBN: 978-90-367-8509-9 (printed version)ISBN: 978-90-367-8508-2 (digital version)

(Un)Healthy in the city

Adverse health effects of traffic-related noise and air pollution

Proefschrift

ter verkrijging van de graad van doctor aan de Rijksuniversiteit Groningen

op gezag van de rector magnificus prof. dr. E. Sterken

en volgens besluit van het College voor Promoties.

De openbare verdediging zal plaatsvinden op

woensdag 3 februari 2016 om 12.45 uur

door

Wilma Louise Zijlema

geboren op 19 augustus 1986 te Groningen

PromotoresProf. dr. J.G.M. RosmalenProf. dr. R.P. Stolk

BeoordelingscommissieProf. dr. E. BuskensProf. dr. E. LebretProf. dr. M. Nieuwenhuijsen

TaBle of conTenTs

Chapter 1 General Introduction 7

Chapter 2 The LifeLines Cohort Study: a resource providing new opportunities for environmental epidemiology

19

Chapter 3 (Un)Healthy in the City: respiratory, cardiometabolic and mental health associated with urbanity

41

Chapter 4 The association of air pollution and depressed mood in 70,928 individuals from four European cohorts

57

Chapter 5 How to assess common somatic symptoms in large-scale studies: a systematic review of questionnaires

83

Chapter 6 Noise and somatic symptoms: a role for personality traits? 113

Chapter 7 Road traffic noise, blood pressure and heart rate: Pooled analyses of harmonized data from 91,718 participants from LifeLines, EPIC-Oxford, and HUNT3

131

Chapter 8 General Discussion 153

Summary 175Nederlandse samenvatting 181Dankwoord 187Curriculum Vitae 188List of publications 188List of recent SHARE publications 190

1General Introduction

9

General Introduction

1environmenTal ePidemiology

Many human diseases can be attributed to environmental factors. Environmental factors can be viewed as those that are not genetic, and include factors that are as broad as for example diet, exercise, and social relationships. In environmental epidemiology, and in this dissertation, we refer to environmental factors as those factors that are outside the immediate control of individuals [1]. Exposures of interest are for example air pollution, traffic noise, electromagnetic fields, and chemical agents. Environmental epidemiology comprises the study of the link between the environment and population health by combining data on environ-mental exposures and health.

A famous example of how a combination of disease statistics and related spatial data could trace a disease outbreak, is the case of John Snow. During the 1854 cholera outbreak in London, John Snow used spatial maps of disease clusters to identify the source of the outbreak as the public water pump on Broad Street. He is seen as the founding father of epidemiology and taught us that spatial data can help to investigate adverse health outcomes of environmental exposures. Another example is the London smog of 1952. During one week in December 1952, a dense smog descended on London. A combination of foggy weather, coal-burning homes and polluting factories led to pollution levels that were 5 to19 times higher than current regulatory standards. Mortality increased substantially during the smog period, and was still increased for a long period thereafter. It has been estimated that 12,000 excess deaths can be attributed to the acute and persistent effects of the 1952 London smog. It became clear that incidences of elevated exposures do not only result in acute effects on public health, but also in chronic health problems [2].

Reductions of harmful exposures have led to substantial improvements in public health. Examples are the introduction of proper sanitation [3], control of particulate air pollution [4], and more recently the introduction of smoke-free legislation [5]. Although the benefits for public health have been acknowledged, there is still a lot to gain in environmental epidemiology. For some exposures, health effects already occur at levels much lower than previously expected. In addition, studying harmful environmental exposures and their effects on public health is important because many individuals may be affected by the same pollu-tion source, and individuals often have no control over their exposures. Moreover, environmental exposures often interfere with a variety of bodily processes, po-tentially affecting a wide range of diseases [6].

Chapter 1

10

Perceived exPosUres and Perceived HealTH

Besides the actual exposure to environmental factors, also perceived exposure and concerns about the health risks associated with the exposure might influence health outcomes [1]. When studying particular exposures, for example elec-tromagnetic fields or living in the vicinity of nuclear power plants, the public’s perception of the health risks is as relevant to health as the exposure itself. More-over, sometimes perceived exposures have stronger associations with diminished health than the actual exposures [7–9]. People may attribute common health complaints to environmental factors, regardless of the actual level of exposure. Individuals that perceive exposure to possibly harmful environmental agents, sometimes attribute their health complaints to the environment. The assessment of common somatic symptoms such as headache, stomachache, and palpitations is therefore becoming increasingly relevant to environmental epidemiology. However, evidence that certain environmental sources cause these symptoms is currently lacking [7,10].

Adverse heAlth effects of trAffic-relAted noise And Air PollUTion

The research in this dissertation focusses on adverse health effects associated with traffic-related air pollution and noise. Many individuals are exposed to polluted air [11,12]. Evidence of harmful health effects of air pollution has accumulated over the recent years. Exposure to air pollution is associated with morbidity and mortality due to respiratory [13] and cardiovascular diseases [14–16]. Recent studies have also linked air pollution to other adverse health effects, including diabetes mellitus [17–19], depressed mood [20,21], and adverse birth outcomes [22–24]. Environmental noise has also been related to a variety of adverse effects [25], including annoyance, hearing loss, sleep disturbance, cognitive impairment, and cardiovascular disease [26]. It is estimated that in Western Europe, each year, at least one million healthy life years are lost due to traffic-related noise [27]. Both air pollution and traffic noise are associated with the highest burden of disease in Europe related to environmental exposures [12]. Road traffic is a com-mon source for both noise and air pollution. Individuals exposed to road traffic noise are probably also exposed to air pollution from traffic. Some adverse health effects can be attributed to noise or air pollution, or both. It is therefore important to distinguish between the two exposures. Previous studies often did not take into account the simultaneous effects of traffic-related noise and air pollution.

11


1However, the research community is becoming increasingly aware of the need for disentangling effects of noise and air pollution [28–30].

PossiBle mecHanisms

Mechanisms that may explain how the environment affects health are the breeder and drift hypotheses. These hypotheses try to explain spatial variations in health. The breeder hypothesis refers to the idea that health differences arise because of environmental factors to which people are directly exposed [31]. Examples could be exposure to air pollution [32] and traffic noise [33], or exposure to harmful chemicals, for example pesticides [34]. Spatial variation in behavior may also explain health differences. It has been shown that urban living can evoke certain (un)healthy lifestyles, such as smoking [35], physical activity [36], alcohol use [37], and diet choices [38]. The drift hypothesis states that selection processes might result in spatial health differences. Selection occurs when healthy persons stay, while ill persons move, or vice versa. Furthermore, specific places may at-tract or repel persons with certain health-related characteristics [31]. The gen-eral stress model has also been used to explain environmental effects on health. Environmental exposures can be seen as stressors, which directly or indirectly affect bodily stress systems. Responses including increased heart rate and blood pressure, bronchodilation, and release of stress hormones are triggered. Long-term activation of the stress system may lead to allostatic load, also known as the ‘wear and tear’ on our bodies. Over longer periods, allostatic load might lead to disease [39].

All these above mechanisms may (partly) explain how environmental factors affect health and knowledge regarding these mechanisms may provide targets for prevention strategies.

Using mUlTiPle BioBanks for environmenTal ePidemiology: BiosHare

One of the major challenges in environmental epidemiology is detecting the small risks associated with environmental exposures. Risks are small because environ-mental exposures often occur in low concentrations and in complex mixtures (e.g. traffic-related air pollution and noise). Many of the health outcomes of interest have other underlying risk factors which effects may be much stronger than those of the environmental exposure under study [6]. For example, a myocardial infarc-

Chapter 1

12

tion may result from a combination of genetic predisposition, lifestyle risk factors, and exposure to noise and air pollution [40]. As a consequence, risks associated with environmental exposures are often difficult to detect. Although effect sizes are generally small, the disease burden associated with environmental exposures is large [11,12]. Large populations are simultaneously affected by various expo-sures during their lives. Furthermore, even small increases in population risks can lead to a vast increase in population burden of disease, and vice versa [41].

Studies with large numbers of participants and broad exposure ranges are needed to investigate the complex interaction between the environment and the development of multifactorial chronic diseases. Single studies may not provide adequate numbers of subjects, but collaboration across large regions could make important contributions to understanding environmental causes of disease. Therefore, population studies must be harmonized and standardized, enabling the assembling of data in valid and effective ways. To this end, we used multiple European cohort studies to study harmful health effects of the environment.

Previously, consortia of collaborating studies (e.g. the HYENA (Hypertension and Exposure to Noise near Airports) study [42], ESCAPE (European Study of Co-horts for Air Pollution Effects)[43]) contributed largely to the current knowledge of the link between environmental exposures and disease. However, practical, ethical, legal, and consent-related restrictions of sharing and pooling individual-level study data form a challenge to collaboration across studies [44–46]. Conse-quently, previous consortia of collaborating studies mostly combined study data on an aggregated level, while combining individual-level data could result in more statistical power, and offers increasing flexibility in data analysis [46].

This dissertation was written within the framework of the European harmoni-zation initiative BioSHaRE (Biobank Standardisation and Harmonisation for Research Excellence in the European Union; www.bioshare.eu). Within BioSHaRE, tools were developed for data harmonization and federated data analyses. Har-monization tools enable the optimal extraction of common data from studies that have collected extensive data and samples under different design protocols (retrospective harmonization) [47,48]. In addition, harmonized exposure models for ambient air pollution and for road traffic noise were partly developed and implemented in several cohort studies participating in the BioSHaRE project (prospective harmonization). Federated data analyses enable researchers to com-prehensively analyze individual-level data from multiple studies, while keeping the original data strictly secure, overcoming challenges regarding data sharing and pooling [46].

13


1researcH qUesTions and oBjecTives

In this dissertation, we focused on adverse health effects of air pollution and road traffic noise. We investigated the (simultaneous) exposure to traffic-related noise and air pollution, and their relations with various adverse health outcomes. A part of this research was undertaken in multiple European cohort studies, from which data were harmonized and pooled. This has led to two central questions:1. Is the living environment, and in particular traffic-related noise and air pollu-

tion, affecting health?2. Can we harmonize and pool individual data from multiple cohort studies, and

does this enable us to better understand relations of environmental exposures and adverse health outcomes?

oUTline of disserTaTion

In chapter 2, we describe the exposure assessment methods that were used in the LifeLines Cohort Study. Lifelines is a prospective population-based cohort study investigating the biological, behavioral and environmental determinants of healthy ageing among 167,729 participants from the North East region of the Netherlands. Environmental exposure and health data assessed in LifeLines were used throughout the research presented in this dissertation. The methods for exposure assessment of road traffic noise and ambient air pollution that were used are presented here.

In chapter 3, the associations of degree of urbanity and lung function, metabolic syndrome, anxiety and depression are described. Many people live in urban areas these days, and urban residents may be more exposed to harmful factors than their rural counterparts. Here, we report on health differences associated with urbanity in the Northern part of the Netherlands.

In chapter 4, the associations of ambient air pollution and depressed mood are described. For this study, we analyzed data from four European cohorts. We also investigated whether exposure to noise influenced the relation between air pollu-tion and depressed mood.

In chapter 5, we report on a systematic review of symptom questionnaires. The aim was to identify suitable questionnaires for the assessment of common somatic symptoms in large-scale population based cohort studies. Identifying suitable questionnaires is useful for the prospective harmonization of somatic symptom assessment.

Chapter 1

14

In chapter 6, we describe the association of road traffic noise and common somatic symptoms in the LifeLines Cohort Study. Associations between road traffic noise exposure, as estimated with a noise model, annoyance from noise, and somatic symptoms are investigated. Secondly, we investigate whether certain personality facets were modifiers or confounders of the relationship between noise, noise annoyance and symptoms.

In chapter 7, we report on the results of a study investigating associations be-tween road traffic noise, air pollution, blood pressure, and heart rate. Federated analyses of harmonized data from three European cohorts were conducted at the individual level using a novel software approach developed within BioSHaRE.

In chapter 8, a general discussion of the main findings is presented. We also discuss the methodological strengths and limitations, the implications of our study findings, and directions for future research.

15


1references

1. Rothman KJ. Methodologic frontiers in environmental epidemiology. Environ Health Perspect. 1993;101 Suppl: 19–21.

2. Bell ML, Davis DL. Reassessment of the lethal London fog of 1952: novel indicators of acute and chronic consequences of acute exposure to air pollution. Environ Health Perspect. 2001;109 Suppl : 389–94.

3. Hamlin C, Sheard S. Revolutions in public health: 1848, and 1998? BMJ. 1998;317: 587–591. 4. Clancy L, Goodman P, Sinclair H, Dockery DW. Effect of air-pollution control on death rates

in Dublin, Ireland: an intervention study. Lancet. 2002;360: 1210–4. doi:10.1016/S0140-6736(02)11281-5

5. Been J V, Nurmatov UB, Cox B, Nawrot TS, van Schayck CP, Sheikh A. Effect of smoke-free legisla-tion on perinatal and child health: a systematic review and meta-analysis. Lancet. 2014;383: 1549–60. doi:10.1016/S0140-6736(14)60082-9

6. Briggs D. Environmental pollution and the global burden of disease. Br Med Bull. 2003;68: 1–24. doi:10.1093/bmb/ldg019

7. Baliatsas C, Bolte J, Yzermans J, Kelfkens G, Hooiveld M, Lebret E, et al. Actual and perceived exposure to electromagnetic fields and non-specific physical symptoms: An epidemiological study based on self-reported data and electronic medical records. Int J Hyg Environ Health. 2015; doi:10.1016/j.ijheh.2015.02.001

8. Fyhri A, Klæboe R. Road traffic noise, sensitivity, annoyance and self-reported health—A struc-tural equation model exercise. Environ Int. 2009;35: 91–97. doi:http://dx.doi.org/10.1016/j.envint.2008.08.006

9. Héritier H, Vienneau D, Frei P, Eze IC, Brink M, Probst-Hensch N, et al. The Association between Road Traffic Noise Exposure, Annoyance and Health-Related Quality of Life (HRQOL). Int J Environ Res Public Health. 2014;11: 12652–12667. doi:10.3390/ijerph111212652

10. Das-Munshi J, Rubin GJ, Wessely S. Multiple chemical sensitivities: A systematic review of prov-ocation studies. J Allergy Clin Immunol. 2006;118: 1257–64. doi:10.1016/j.jaci.2006.07.046

11. Lim SS, Vos T, Flaxman AD, Danaei G, Shibuya K, Adair-Rohani H, et al. A comparative risk as-sessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990-2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet. 2013;380: 2224–2260. doi:10.1016/S0140-6736(12)61766-8

12. Hanninen O, Knol AB, Jantunen M, Lim TA, Conrad A, Rappolder M, et al. Environmental Burden of Disease in Europe: Assessing Nine Risk Factors in Six Countries. Environ Health Perspect. 2014; doi:10.1289/ehp.1206154

13. Götschi T, Heinrich J, Sunyer J, Künzli N. Long-term effects of ambient air pollution on lung function: a review. Epidemiology. 2008;19: 690–701. doi:10.1097/EDE.0b013e318181650f

14. Hoek G, Brunekreef B, Goldbohm S, Fischer P, van den Brandt PA. Association between mortal-ity and indicators of traffic-related air pollution in the Netherlands: a cohort study. Lancet. 2002;360: 1203–1209. doi:10.1016/S0140-6736(02)11280-3

15. Brunekreef B, Holgate ST. Air pollution and health. Lancet. 2002;360: 1233–1242. doi:10.1016/S0140-6736(02)11274-8

16. Brook RD, Rajagopalan S, Pope CA, Brook JR, Bhatnagar A, Diez-Roux A V, et al. Particulate matter air pollution and cardiovascular disease: An update to the scientific statement from the Ameri-can Heart Association. Circulation. 2010;121: 2331–78. doi:10.1161/CIR.0b013e3181dbece1

Chapter 1

16

17. Chen H, Burnett RT, Kwong JC, Villeneuve PJ, Goldberg MS, Brook RD, et al. Risk of incident dia-betes in relation to long-term exposure to fine particulate matter in Ontario, Canada. Environ Health Perspect. 2013;121: 804–10. doi:10.1289/ehp.1205958

18. Krämer U, Herder C, Sugiri D, Strassburger K, Schikowski T, Ranft U, et al. Traffic-related air pollution and incident type 2 diabetes: results from the SALIA cohort study. Environ Health Perspect. 2010;118: 1273–9. doi:10.1289/ehp.0901689

19. Eze IC, Hemkens LG, Bucher HC, Hoffmann B, Schindler C, Künzli N, et al. Association between Ambient Air Pollution and Diabetes Mellitus in Europe and North America: Systematic Review and Meta-Analysis. Environ Health Perspect. 2015;123: 381–389. doi:10.1289/ehp.1307823

20. Lim YH, Kim H, Kim JH, Bae S, Park HY, Hong YC. Air pollution and symptoms of depression in elderly adults. Environ Health Perspect. 2012;120: 1023–1028. doi:10.1289/ehp.1104100

21. Szyszkowicz M, Rowe BH, Colman I. Air pollution and daily emergency department visits for depression. Int J Occup Med Environ Health. 2009;22: 355–362. doi:10.2478/v10001-009-0031-6

22. Green R, Sarovar V, Malig B, Basu R. Association of Stillbirth With Ambient Air Pollution in a California Cohort Study. Am J Epidemiol. 2015;181: 874–82. doi:10.1093/aje/kwu460

23. Schembari A, de Hoogh K, Pedersen M, Dadvand P, Martinez D, Hoek G, et al. Ambient Air Pol-lution and Newborn Size and Adiposity at Birth: Differences by Maternal Ethnicity (the Born in Bradford Study Cohort). Environ Health Perspect. 2015; doi:10.1289/ehp.1408675

24. Stieb DM, Chen L, Beckerman BS, Jerrett M, Crouse DL, Omariba DW, et al. Associations of Pregnancy Outcomes and PM2.5 in a National Canadian Study. Environ Health Perspect. 2015; doi:10.1289/ehp.1408995

25. Basner M, Babisch W, Davis A, Brink M, Clark C, Janssen S, et al. Auditory and non-auditory effects of noise on health. Lancet. 2013;383: 1325–32. doi:10.1016/S0140-6736(13)61613-X

26. Münzel T, Gori T, Babisch W, Basner M. Cardiovascular effects of environmental noise exposure. Eur Heart J. 2014;35: 829–36. doi:10.1093/eurheartj/ehu030

27. WHO. Burden of disease from environmental noise. Quantification of healthy life years lost in Europe. WHO European Centre for Environment Health, Bonn Office; 2011.

28. Tetreault LF, Perron S, Smargiassi A. Cardiovascular health, traffic-related air pollution and noise: are associations mutually confounded? A systematic review. Int J Public Health. 2013; doi:10.1007/s00038-013-0489-7

29. Tzivian L, Winkler A, Dlugaj M, Schikowski T, Vossoughi M, Fuks K, et al. Effect of long-term outdoor air pollution and noise on cognitive and psychological functions in adults. Int J Hyg Environ Health. 2015;218: 1–11. doi:10.1016/j.ijheh.2014.08.002

30. Guxens M, Sunyer J. A review of epidemiological studies on neuropsychological effects of air pollution. Swiss Med Wkly. 2012;141: w13322. doi:10.4414/smw.2011.13322

31. Verheij RA. Explaining urban-rural variations in health: A review of interactions between indi-vidual and environment. Soc Sci Med. 1996;42: 923–935. doi:10.1016/0277-9536(95)00190-5

32. Raaschou-Nielsen O, Andersen ZJ, Beelen R, Samoli E, Stafoggia M, Weinmayr G, et al. Air pol-lution and lung cancer incidence in 17 European cohorts: prospective analyses from the Eu-ropean Study of Cohorts for Air Pollution Effects (ESCAPE). Lancet Oncol. 2013;14: 813–822. doi:10.1016/S1470-2045(13)70279-1

33. Bluhm GL, Berglind N, Nordling E, Rosenlund M. Road traffic noise and hypertension. Occup Environ Med. 2007;64: 122–126. doi:10.1136/oem.2005.025866

17


1 34. Van der Mark M, Brouwer M, Kromhout H, Nijssen P, Huss A, Vermeulen R. Is pesticide use related to Parkinson disease? Some clues to heterogeneity in study results. Environ Health Perspect. 2012;120: 340–7. doi:10.1289/ehp.1103881

35. Pearce J, Barnett R, Moon G. Sociospatial inequalities in health-related behaviours: Pathways linking place and smoking. Prog Hum Geogr. 2011;36: 3–24. doi:10.1177/0309132511402710

36. Saelens BE, Sallis JF, Black JB, Chen D. Neighborhood-Based Differences in Physical Activity: An Environment Scale Evaluation. Am J Public Health. 2003;93: 1552–1558. doi:10.2105/AJPH.93.9.1552

37. Donath C, Grässel E, Baier D, Pfeiffer C, Karagülle D, Bleich S, et al. Alcohol consumption and binge drinking in adolescents: comparison of different migration backgrounds and rural vs. urban residence—a representative study. BMC Public Health. 2011;11: 84. doi:10.1186/1471-2458-11-84

38. Burgoine T, Forouhi NG, Griffin SJ, Wareham NJ, Monsivais P. Associations between exposure to takeaway food outlets, takeaway food consumption, and body weight in Cambridgeshire, UK: population based, cross sectional study. BMJ. 2014;348: g1464.

39. McEwen BS. Stress, Adaptation, and Disease: Allostasis and Allostatic Load. Ann N Y Acad Sci. 1998;840: 33–44. doi:10.1111/j.1749-6632.1998.tb09546.x

40. Cesaroni G, Forastiere F, Stafoggia M, Andersen ZJ, Badaloni C, Beelen R, et al. Long term expo-sure to ambient air pollution and incidence of acute coronary events: prospective cohort study and meta-analysis in 11 European cohorts from the ESCAPE Project. BMJ. 2014;348: f7412. doi:10.1136/bmj.f7412

41. Rose G. Sick individuals and sick populations. Int J Epidemiol. 1985;14: 32–8. 42. Jarup L, Babisch W, Houthuijs D, Pershagen G, Katsouyanni K, Cadum E, et al. Hypertension

and exposure to noise near airports: the HYENA study. Environ Health Perspect. 2008;116. doi:10.1289/ehp.10775

43. Beelen R, Raaschou-Nielsen O, Stafoggia M, Andersen ZJ, Weinmayr G, Hoffmann B, et al. Effects of long-term exposure to air pollution on natural-cause mortality: an analysis of 22 European cohorts within the multicentre ESCAPE project. Lancet. 2014;383: 785–795. doi:10.1016/S0140-6736(13)62158-3

44. Kaye J. From single biobanks to international networks: developing e-governance. Hum Genet. 2011;130: 377–82. doi:10.1007/s00439-011-1063-0

45. Knoppers BM, Harris JR, Tassé AM, Budin-Ljøsne I, Kaye J, Deschênes M, et al. Towards a data sharing Code of Conduct for international genomic research. Genome Med. 2011;3: 46. doi:10.1186/gm262

46. Gaye A, Marcon Y, Isaeva J, LaFlamme P, Turner A, Jones EM, et al. DataSHIELD: taking the analy-sis to the data, not the data to the analysis. Int J Epidemiol. 2014;43: 1929–44. doi:10.1093/ije/dyu188

47. Fortier I, Doiron D, Little J, Ferretti V, L’Heureux F, Stolk RP, et al. Is rigorous retrospective harmonization possible? Application of the DataSHaPER approach across 53 large studies. Int J Epidemiol. 2011;40: 1314–1328. doi:10.1093/ije/dyr106

48. Fortier I, Burton PR, Robson PJ, Ferretti V, Little J, L’Heureux F, et al. Quality, quantity and har-mony: the DataSHaPER approach to integrating data across bioclinical studies. Int J Epidemiol. 2010;39: 1383–1393. doi:10.1093/ije/dyq139

2 The LifeLines Cohort Study: a resource providing new opportunities for environmental epidemiology

Wilma L. Zijlema, Nynke Smidt, Bart Klijs, David W. Morley, John Gulliver, Kees de Hoogh, Salome Scholtens, Judith G.M. Rosmalen, Ronald P. Stolk

Submitted

Chapter 2

20

aBsTracT

Background

Lifelines is a population-based cohort study investigating healthy ageing among 167,729 participants from the Netherlands. We describe the reasons for choosing particular assessments of environmental exposures and consider the implications for future investigations.

methods

Exposure to ambient air pollution was estimated using two types of land use regression models. Road traffic noise exposure was assessed using a model based on Common Noise Assessment Methods in Europe. Data on perceived exposures and neighborhood characteristics are also available. A comprehensive medical assessment and questionnaires were completed.

results

Mean age was 45 years (standard deviation (SD) 13 years), and 59% were female. Median levels of NO2 and PM10 were 15.7 (interquartile range (IQR) 4.9) µg/m3 and 24.0 (IQR 0.6) µg/m3 respectively. Median level of daytime road traffic noise was 54.0 (IQR 4.2) dB(A).

conclusion

The combination of harmonized environmental exposures and extensive assess-ment of a variety of health outcomes offers opportunities for environmental epidemiology. LifeLines aims to be a resource for the international scientific community.

21

The LifeLines Cohort Study

2

BackgroUnd

Environmental epidemiology has contributed to public health by focusing on the reduction of harmful exposures from the environment. Smoke-free legislation [1], control of particulate air pollution [2], and the introduction of proper sanitation [3] are examples of policies that resulted in substantial benefits for public health. Although large efforts have been made, a lot more is to gain in environmental epidemiology, both in terms of public health policy and research. Prevention of harmful environmental exposures is important because many of these exposures are often involuntary and outside the immediate control of the individual. More-over, many individuals may be affected by the same pollution source.

Studying health effects of environmental exposures can be challenging. Envi-ronmental exposures often occur in low concentrations and in complex mixtures. Most health outcomes of interest have many underlying risk factors whose effects may be much stronger than those of the environmental exposure. Risks associ-ated with environmental exposures are generally small [4], making them often difficult to detect. Although effect sizes are small, impacts on health are still substantial because of the large number of exposed persons. Population based cohort studies with large sample sizes are needed to investigate the complex in-teraction between genetic, behavioral, and environmental factors in the develop-ment of multifactorial chronic diseases. Large population based studies can make important contributions to environmental epidemiology. Moreover, collaboration between cohort studies could help understand environmental causes of disease. The advantages of such an approach are widespread. Collaboration between stud-ies will lead to very large sample sizes and wide exposure ranges. Furthermore, results can be compared across regions, in which not only the exposures, but also genetic, social and cultural factors vary. This makes the generalizability of study results larger. Being able to generalize study results across large regions in Europe is important for establishing European exposure norms and guidelines. Therefore, population studies must be harmonized and standardized, enabling the assembling of data in valid and effective ways. To this end, LifeLines aims to implement standardized and harmonized data collection methods. LifeLines is one of eight studies participating in the BioSHaRE (Biobank Standardisation and Harmonisation for Research Excellence in the European Union, www.bioshare.eu) project. Within the framework of this European harmonization initiative, tools were developed for data harmonization, database integration and federated data analyses [5]. Other participating biobanks are located in Germany, Norway, the UK, Finland, Italy, Estonia, and Ireland. As a result, standardized and validated methods for measurement were used wherever possible, and a part of the dataset

Chapter 2

22

(including e.g. educational level, alcohol and tobacco use) has been retrospectively harmonized with other European cohorts [6].With its large sample size, and the combination of harmonized environmental exposures and extensive assessment of a variety of health outcomes, LifeLines aims to be a resource for environmental epidemiology.

Our aim is to inform researchers on the choice of data collection methods and methodology of data on environmental exposures in LifeLines. Descriptive infor-mation of selected exposure and cohort data will also be reported. By describing these features of LifeLines we aim to highlight methodological choices made in the context of large population based cohort studies exposed to multiple environ-mental stressors.

meTHods

study design and participants

The LifeLines Cohort Study was established to facilitate research on complex interactions between environmental, phenotypic and genetic factors in the devel-opment of chronic diseases [7]. The cohort profile of LifeLines has been described elsewhere [8]. Briefly, the recruitment of study participants took place between 2006-2013. In the Netherlands, all inhabitants are registered with a general practitioner. A large number of general practitioners from the Northern provinces of the Netherlands participated in the recruitment and invited all their patients between ages 25 and 50 years. Individuals who agreed to participate were asked to indicate whether their family members would also be willing to participate. In addition, individuals could also register themselves via the LifeLines website. The sample was recruited from the three Northern provinces of the Netherlands, but this was not a requirement for inclusion of family members [8]. Characteristics of adult LifeLines participants are broadly representative for the adult population of the north of the Netherlands [9].

Baseline data were collected from 167,729 participants, aged 6 months to 93 years. Follow-up is planned for at least 30 years, with questionnaires administered every 1.5 years, and a renewed physical examination scheduled every five years. Participants visited one of the LifeLines research sites for a physical examination, including spirometry, electrocardiogram (ECG), blood pressure measurements, anthropometry, cognition tests, and a psychiatric interview. Fasting blood and 24-hour urine samples were collected from all participants, and genome wide association data are currently available for a subsample of 15,638 participants. An extensive baseline questionnaire was completed at home, including questions

23


2

on history of illness, health related quality of life, lifestyle, socioeconomic status, psychosocial stress, work (profession, working hours), psychosocial characteris-tics, and medication use. Linkage will be established with records from general practitioners and health registries. A comprehensive and detailed overview of the available data is presented in the online LifeLines Data Catalogue (www.lifelines.net).

geocoding

All participants’ home addresses were geocoded. In addition to the most recent home address, address history is available from the Municipal Personal Record Database. This governmental registry contains personal data of all individuals who live or have lived in the Netherlands. Data on address history provide in-sights in residential mobility and length of exposure to different environments. This allows assessment of long-term exposures, which is relevant for life course epidemiology. Work addresses were also collected and will be geocoded as well, allowing for outdoor exposure estimation of air pollution and noise at both home and work location. Most previous studies estimate exposure to road traffic noise and air pollution at the home addresses of participants (e.g. [10–12]), assuming this to be a good indicator of personal exposure. Individuals spend a large amount of their time at the work address. Combining exposures from both home and work locations will result in better estimations of an individual’s exposure [13]. A detailed description of the environmental exposures assessed in LifeLines is described below, and is summarized in Table 1.

environmental exposures

Ambient air pollution

Air pollution has been related to various health outcomes, including respiratory diseases [14] and cardiovascular disease [15]. In LifeLines, exposure to ambient air pollution was estimated using land use regression (LUR) models developed for the European Study of Cohorts for Air Pollution Effects (ESCAPE) [16,17] and using European wide LUR models enhanced with satellite derived air pollution estimates [18]. It was chosen to implement these air pollution models in LifeLines because both models are advantageous regarding comparability across studies in other European regions.

Within the ESCAPE project, LUR models for various European study areas were developed using a standardized approach [16,17]. ESCAPE LUR models were developed for NO2 (nitrogen dioxide), NO2 background, PM2.5 (particulate matter with a diameter ≤2.5 µm), PM2.5 absorbance (reflectance on PM2.5 filters, i.e. a

Chapter 2

24

marker of black carbon), and PM10 (particulate matter with a diameter ≤10 µm), and were based on annual average concentrations from an intensive monitoring campaign and GIS (geographic information system) derived predictor variables (e.g. distance to the nearest major road, traffic intensity, built-up land, population density, altitude). LUR models were developed using measurements carried out in 2009-2010 and predictor variable data for the same years. Model performance was evaluated by leave-one-out cross validation [16,17]. The adjusted explained variability in measured concentrations (R2) was 0.85 for NO2, 0.83 for NO2 back-ground, 0.66 for PM10, 0.64 for PM2.5, and 0.91 for PM2.5 absorbance.

Table 1. Overview of environmental exposures in the LifeLines Cohort Study

Ambient air pollution

ESCAPE models

NO2 nitrogen dioxide

NO2 background background level of nitrogen dioxide

PM2.5 particulate matter with diameter ≤2.5 µm

PM2.5 absorbance reflectance on PM2.5 filters, i.e. marker of black carbon

PM10 particulate matter with diameter ≤10 µm

EU-wide models

NO2 nitrogen dioxide

PM10 particulate matter with diameter ≤10 µm

Road traffic noise (CNOSSOS-EU model)

Lday 12-hour day time period from 07:00 to 19:00 hour

Levening 4-hour evening time period from 19:00 to 23:00 hours

Lnight 8-hour night time period from 23:00 to 07:00 hours

Laeq16 16-hour day and evening time period from 07:00 to 23:00 hours

Lden day-evening-night time period of 24 hours

Laeq 0-23 hours Hourly noise estimates

Questionnaire data

Noise annoyance

Perceived exposures to power lines, mobile phone masts

Mobile phone use

Exposure to secondhand smoke

Perceived living environment

Database linkage

Neighborhood characteristics (Statistics Netherlands)

Neighborhood level demographic and socioeconomic figures

LISA employment register Location, type of establishment, number of employees

ESCAPE = European Study of Cohorts for Air Pollution Effects; EU-wide = European-wide; CNOSSOS-EU = Com-mon Noise Assessment Methods in Europe

25


2

In addition, exposure to ambient PM10 and NO2 is available from European(EU)-wide models. These EU-wide models incorporate GIS-derived land use, road network, and topographic data, as well as satellite-derived estimates of ground level concentrations for PM2.5 (as an indicator of PM10) and NO2 [18]. Model devel-opment follows the ESCAPE procedure to construct the multiple linear regression equations, and are applicable for years 2005, 2006 (NO2) and 2007 (NO2 and PM10). Models were evaluated against measured PM10 and NO2 concentrations at an independent subset of sites reserved for this purpose. The adjusted explained variability in measured concentrations (R2) was 0.48−0.58 for NO2, and 0.22−0.50 for PM10.

The main difference between the ESCAPE and EU-wide models is that the ESCAPE models are region specific, while EU-wide models are developed for a much larger area. ESCAPE models are developed for 20 (PM) to 36 (NO2) European regions and EU-wide models for 17 countries in Western Europe. In addition, monitoring data used in ESCAPE models originated from a monitoring campaign specifically conducted for the ESCAPE-project with monitoring sites selected for this purpose, whereas monitoring data for the EU-wide models were obtained from regulatory monitoring networks. ESCAPE models may perform better on a regional scale, while EU-wide models may make comparisons across multiple study areas easier, since the model is available for a larger region. A study including Lifelines and other cohorts falling in ESCAPE study areas (e.g. the British cohort EPIC-Oxford, also involved in BioSHaRE) should use air pollution exposures estimates from the relevant ESCAPE models. However, the EU-wide model can also provide air pollu-tion exposure estimates for areas falling outside the ESCAPE study areas, and will therefore allow research including Lifelines plus non-ESCAPE cohorts (e.g. the Norwegian cohort HUNT, also involved in BioSHaRE).

Road traffic noise

Environmental noise has been related to a variety of adverse outcomes, including hearing loss, annoyance, sleep disturbance, cognitive impairment, and cardiovas-cular disease [19]. In LifeLines, road traffic noise was estimated using an imple-mentation of the Common Noise Assessment Methods in Europe (CNOSSOS-EU) noise modelling framework [20,21], which was developed as a common method-ology for noise modelling across Europe. This noise model was preferred since it allows comparison of results from different countries. The CNOSSOS-EU noise model implemented within LifeLines is using lower resolution source data be-cause the highest resolution input data at national or large regional level is either unavailable, expensive or would be too computationally intensive to process. The performance of CNOSSSOS-EU using the lower resolution inputs has been shown

Chapter 2

26

to be reasonable for application in epidemiological studies. The model’s exposure ranking, i.e. prediction of noisier and quieter sites, was adequate (Spearman’s rank = 0.75; p <0.001), but the predicted noise levels have relatively large errors (root mean square error (RMSE) = 4.46 dB(A)) [22].

The CNOSSOS-EU framework contains empirically derived equations to deter-mine the initial noise level based on traffic flow and sound attenuation (i.e. sound reduction or damping) based on known environmental factors and physical processes. To estimate source noise on road segments in the Netherlands, traffic information was obtained including hourly flow of passenger cars, heavy goods vehicles and their average speeds. The sound propagation model was based on the CORINE (Coordination of information on the environment) land cover dataset that has a European wide coverage accurate to 100 meters for major land cover types [23]. In particular, the distinction between urban fabric and areas of vegeta-tion was made. Traffic data originated from year 2009 and landcover data from 2006. Full details of the noise model used for the LifeLines study are provided elsewhere [22].

Perceived exposures

Besides the actual exposure to environmental factors, also perceived exposure and concerns about the health risks associated with the exposure might influence health outcomes [24]. When studying particular exposures, for example electro-magnetic fields, the public’s perception of the health risks is as relevant to health as the exposure itself. Moreover, sometimes perceived exposures have stronger associations with diminished health than the actual exposures [25–28]. In addi-tion to the modelled estimates for air pollution and noise exposures, perceived ex-posures are measured in LifeLines. Noise annoyance from eight different sources was measured using a standardized self-report questionnaire. The sources of noise annoyance include for example air, road and rail traffic. This questionnaire originates from the International Organization for Standardization (ISO) guide-line, which provides specifications for socio-acoustic surveys and social surveys that include questions on noise effects [29]. Similar noise annoyance questions were used in the HYENA study on HYpertension and Exposure to Noise near Airports [30]. Perceived exposure to electromagnetic fields is measured using a questionnaire on perceived exposure to power lines and mobile phone masts. Participants were asked to what extent they think they are exposed to radiation from power lines and mobile phone masts, and whether they perceive this as bad for their health (adapted from [31]). In addition, a number of questions on the use of mobile phones (e.g. average time per week using mobile phone; on which side of the head) were included. These questions were adapted from the UKBiobank

27


2

questionnaire [32], which is one of the cohorts involved in the BioSHaRE project. Exposure to second-hand smoke is assessed with questions about the duration and place (in the household, at the workplace) of exposure, and originate from the European Community Respiratory Health Survey (ECRHS, [33]).

In addition to work addresses, information was collected on working hours and type of profession, which may be relevant for occupational exposure studies. In the next follow-up questionnaire, LifeLines will also measure how participants perceive their living environment. The questionnaire comprises of nine items investigating characteristics of the physical and social living environment as per-ceived by the respondent (e.g. neighborhood satisfaction, social interaction with neighbors), and was based on the 2010 health survey of the Dutch Community Health Services (Dutch name: GGD Gezondheidsenquête 2010) [34] and the 2006 WoON questionnaire (Dutch name: WoonOnderzoek Nederland) [35].

Neighborhood characteristics

Various neighborhood characteristics have been associated with health, ranging from cardiometabolic risk factors [36] to life expectancy [37]. Data on neighbor-hood characteristics from Statistics Netherlands are available for linkage to the LifeLines database. Statistics Netherlands publishes demographic and socio-economic figures for municipalities, districts and neighborhoods [38]. These figures cover various themes, for example housing, education, income, and land use. Such information enables to investigate the impact of neighborhood condi-tions on health [39].

Furthermore, the LISA employment register (www.lisa.nl) is linked to the LifeLines database. This register contains nationwide information on locations (geocoded at the address level) of establishments where paid work is done. The database also contains information about the type of establishment (i.e. restau-rants, hospitals, shops) and the number of employees. Using the LISA employment register, it is possible to investigate the density and distance to specific facilities in relation to various health outcomes.

resUlTs

The first data release of the baseline sample consisted of 95,432 participants, of which 58.7% were female. Mean age was 45.2 years (standard deviation (SD) 12.6 years), and men were more often current smokers than women (23.8% and 20.4%, respectively). Overall, mean body mass index (BMI) was 26.1 kg/m2 (SD 4.3 kg/m2), and mean systolic blood pressure was 125.6 mmHg (SD 15.3 mmHg).

Chapter 2

28

Forced vital capacity (FVC), a measure for lung function assessed with spirometry, was on average 4.5 (SD 1.1) liters. Most participants live in rural (i.e. <500 ad-dresses per km2) neighborhoods (41.2%) (Table 2).

Most participants live in the three Northern provinces of the Netherlands, but part of the participants (approximately 3%) live elsewhere in the country (Fig-ure 1). Median levels of NO2 were 15.7 (interquartile range (IQR) 4.9) µg/m3 (ES-CAPE) and 20.6 (IQR 7.9) µg/m3 (EU-wide), and 24.0 (IQR 0.6) µg/m3 (ESCAPE) and 23.6 (IQR 2.4) µg/m3 (EU-wide) for PM10.Distributions of estimated concentrations of NO2 and PM10 are shown in Figure 2. Correlation between ESCAPE-LUR modelled NO2 and satellite-enhanced LUR modelled NO2 was high (Spearman’s rho = 0.86), while correlation for PM10 from both models was moderate (Spearman’s rho = 0.54).

Table 2. Sample characteristics by sex of the LifeLines Cohort Study

Women Men Total

N (%) 56 053 (58.7) 39 379 (41.3) 95 432

Age (years) 44.9 (12.6) 45.7 (12.7) 45.2 (12.6)

Education (%)

No or primary 3.2 3.0 3.1

Lower or preparatory vocational 12.4 16.2 14.0

Lower general secondary 16.0 11.7 14.2

Intermediate vocational or apprenticeship 30.6 31.0 30.8

Higher general secondary or pre-university secondary 9.9 7.0 8.7

Higher vocational or university 28.1 31.0 29.3

Current smokers (%) 20.4 23.8 21.8

BMI (kg/m2) 25.9 (4.7) 26.4 (3.7) 26.1 (4.3)

SBP (mmHg) 122.1 (15.3) 130.7 (14.0) 125.6 (15.3)

DBP (mmHg) 71.8 (8.7) 76.6 (9.3) 73.8 (9.3)

FVC (L) 3.9 (0.6) 5.4 (0.9) 4.5 (1.1)

FEV1e (L) 3.0 (0.6) 4.1 (0.8) 3.5 (0.8)

Urbanity b (%)

Rural 40.8 41.9 41.2

Semi-rural 24.3 25.4 24.8

Intermediate urban-rural 17.5 16.7 17.1

Semi-urban 10.6 10.1 10.4

Urban 6.8 6.0 6.5

Means (SD) are presented for continuous variables, and percentages are presented for categorical variables.a Data are based on the first data release of n=95,432.bAverage number of addresses per km2 within a range of 1 kilometer, categorized into five levels ranging from rural (<500 addresses per km2) to urban (≥2500 addresses per km2).Abbreviations: BMI=body mass index; SBP=systolic blood pressure; DBP=diastolic blood pressure; FVC=forced vital capacity; FEV1= forced expiratory volume in 1 second

29


2

Five A-weighted indicators of road traffic noise were estimated from the CNOS-SOS-EU model. The noise indicators are Lday (12-hour day time period from 07:00 to 19:00 hour); Levening (4-hour evening time period from 19:00 to 23:00 hours); Lnight (8-hour night time period from 23:00 to 07:00 hours); Laeq16 (16-hour day and evening time period from 07:00 to 23:00 hours); and Lden (day-evening-night time period of 24 hours). Median levels of road traffic noise were 54.0 (IQR 4.2) dB(A) for Lday and 45.1 (IQR 4.2) dB(A) for Lnight. Figure 3 shows distributions of daytime and nighttime road traffic noise estimates for LifeLines.

LifeLines participants living in urban neighborhoods had highest exposure to air pollution (nitrogen dioxide; Figure 4) and 24-hours road traffic noise (Figure 5), compared to participants in neighborhoods of lower degree of urbanity.

figure 1. LifeLines study area and number of participants per square kilometer a

a Participants were aggregated and plotted at the center of each 1 kilometer grid cell.

Chapter 2

30

daTa access and linkage

LifeLines has adopted an open protocol, meaning that within the standing in-frastructure additional data and biomaterial collection, and linkage with other (environmental) data sources can be implemented by submitting a research pro-posal to LifeLines. Environmental biomonitoring of exposure and response could

0 5 10 15 20 25 30 35

NO2 (ESCAPE)

NO2 (EU-wide)

PM10 (ESCAPE)

PM10 (EU-wide)

µg/m3

figure 2. Distribution of nitrogen dioxide and particulate matter (≤10 µm) exposure in the LifeLines Cohort Study with estimates based on ESCAPE land use regression model and EU-wide land use re-gression model. Medians, 25th, and 75th percentiles are shown in the box, whiskers indicate 5th and 95th percentiles.ESCAPE = European Study of Cohorts for Air Pollution Effects; EU-wide = European-wide.

0 10 20 30 40 50 60 70

Lday

Lnight

Decibel (A)

figure 3. Distribution of road traffic noise exposure Lday and Lnight in the LifeLines Cohort Study. Medians, 25th, and 75th percentiles are shown in the box, whiskers indicate 5th and 95th percentiles.dB(A) = decibel (A); Lday = A-weighted equivalent noise level over the 12-hour day time period from 07:00 to 19:00 hours; Lnight = A-weighted equivalent noise level over the 8-hour night time period from 23:00 to 07:00 hours.

31


2

also be implemented. One example is LifeLines DEEP, an add-on study where in a subsample of participants additional biological samples (feces, exhaled air) were collected, additional blood analyses were undertaken, and additional question-naires were filled out [40]. Other exposures types, such as domestic radon, elec-tromagnetic fields, harmful chemicals (e.g. pesticides) are of interest for research

0

5

10

15

20

25

urban semi-urban intermediateurban-rural

semi-rural rural

µg/m

3

figure 4. Median exposure to ambient nitrogen dioxide (NO2; based on ESCAPE model) according to degree of urbanity within the LifeLines Cohort Study.

50

51

52

53

54

55

56

57

58

59

60

61


semi-rural rural

dB(A

)

figure 5. Median exposure to road traffic noise (Lden; average A-weighted noise level estimated over a 24 hour period) according to degree of urbanity within the LifeLines Cohort Study.

Chapter 2

32

in LifeLines, and proposals for additional environmental exposures assessments relevant for ‘Healthy Ageing’ are warmly welcomed. Furthermore, biological samples are stored for future analyses, enabling for example measurement of exposure biomarkers. Data and biomaterials are provided on a fee-for-service basis and may be used for scientific research only. Public and private researchers, from inside and outside the Netherlands are invited to submit a research proposal to the LifeLines Research Office. Quality control of the data of hard copy question-naires consist of various stages, including preparation for scanning the question-naire, scanning the questionnaires, verification of the data on missing values, inconsistencies, and extreme values and a final data quality control is conducted in the database. Quality control of the data is done by trained medical students and data managers, using Standard Operating Procedures (SOPs). Data is released within a remote system (LifeLines workspace) running on a high performance computer cluster, which ensures data quality and security. The LifeLines research website (www.lifelines.net) provides the details of the application process, the data collection, and an overview of publications with LifeLines data.

discUssion

The combination of harmonized environmental exposures, relevant mediators and modifiers, and extensive assessment of multiple health outcomes makes LifeLines a great resource for environmental epidemiology. This paper provides an overview of assessments in the LifeLines Cohort Study that are relevant to environmental epidemiology.

limitations and strengths

The results in this paper are based on the first 95,432 participants that were included in LifeLines. Geocoding and exposure estimation using the noise and air pollution models of the full cohort is currently still ongoing. Since the inclusion of participants was independent of their place of residence, we have no reason to suspect geographical differences between participants in the first data release and the complete sample. Our study area is relatively rural compared to other parts of the Netherlands [41]. As a consequence, levels of air pollution and noise exposures are lower than other parts of the Netherlands, due to for example lower population densities, and less extensive road networks. For example, exposure to NO2 in the EPIC-PROSPECT cohort located in the city of Utrecht and surround-ing areas was on average 26.7 µg/m3 [42], compared to 15.7 µg/m3 in LifeLines. Conclusions based on research undertaken with LifeLines data will therefore be

33


2

limited to exposure levels in that particular range. Nevertheless, participants live in both urban and rural areas and are spread across a large part of the Nether-lands, resulting in a great spatial variation. A major challenge in environmental epidemiology includes accurate exposure assessment [24,43,44]. Actual mea-surement of individual-level exposures in a cohort as large as LifeLines would be impossible. Therefore other approaches to estimate exposures are used, such as land use regression modelling. Use of these models introduces misclassification of exposure to a varying degree; for example, due to daily mobility and long-term residential mobility [45]. In LifeLines, misclassification due to residential mobil-ity can be tackled because data is available on address history, which is for some participants available from periods as early as year 1943. Future research should focus on characterizing exposures in earlier years, allowing for assessment of long-term exposures which is relevant for life course epidemiology. Data about the number of hours a person works and work location could provide insights into daily mobility. In addition, the collection of work addresses allows for outdoor exposure estimation of noise and air pollution at the work location. Combining exposures at both home and work location will result in better estimation of an individual’s exposure [13]. Another factor contributing to exposure misclassifica-tion is for example the type of housing (e.g. insulation quality, bedroom location). We did not collect information with regard to type of housing, but this is planned for the follow up questionnaires.

Ethical and legal challenges involved with the collection, storage and sharing of biobank data are addressed with procedures and facilities to ensure privacy protection. Data is released via a remote desktop from which pseudonymized data can be accessed. This workspace runs on a high-performance computer clus-ter, ensuring data quality and security. High quality standards apply to the entire process of data and biomaterial collection, storage and release. Privacy, security and traceability are ensured in all aspects, and in 2014, LifeLines received the ISO-9001 quality certificate [8]. For example, geocoded addresses combined with individual data are not available for data release to researchers. In addition to the risk of re-identification of participants, publications of local health hazards could expose (non-)participants to risks to their social rights and private eco-nomic interests. As population biobanks have a legal obligation to inform their participants about the risks of their participation, it has been recommended that they also acknowledge the risks posed by the publication of exposure studies to the legitimate interests of their participants and the public [46]. Effectively, this means that LifeLines researchers are not allowed to publish detailed exposure maps for the LifeLines area.

Chapter 2

34

With 167,729 participants, LifeLines is one of the largest population based co-hort studies of the world. A large amount of data is collected, biological samples are stored for future analyses (e.g. measurement of exposure biomarkers), and for a subsample of 15,638 participants genome wide genotype data are available. These numbers are large enough for studying effects of environmental exposures in vulnerable subgroups. This is important, because many factors (genetics, in-dividual disease states, psychosocial stress, and socioeconomic status) have the potential to interact with environmental exposures [47]. The major strength of this study is the use of harmonized exposure models for ambient air pollution and for road traffic noise, and the use of validated questionnaires. This, and also data harmonization facilitates comparability and combination with data from other cohorts and regions. This is a great advantage especially for environmental epide-miology, where large sample sizes and broad exposure ranges are needed. Existing and future collaborations with other biobanks and international consortia hold the promise of answering complex questions in environmental epidemiology. Furthermore, the prospective nature of LifeLines allows research into long-term effects of environmental exposures. In addition, both objective (modelled) and subjective (questionnaire based) exposures were assessed. This enables study-ing effects of the exposure itself, and of perception of the exposure. Key research questions that will be investigated in LifeLines are about effects of noise and air pollution on healthy ageing.

conclUsions

LifeLines aims to be a resource for the national and international scientific com-munity. Requests for data and biomaterials can be submitted to the LifeLines Research Office ([email protected]).

35


2

references


2. Clancy L, Goodman P, Sinclair H, Dockery DW. Effect of air-pollution control on death rates in Dublin, Ireland: an intervention study. Lancet. 2002;360: 1210–4. doi:10.1016/S0140-6736(02)11281-5

3. Hamlin C, Sheard S. Revolutions in public health: 1848, and 1998? BMJ. 1998;317: 587–591. 4. Pekkanen J, Pearce N. Environmental epidemiology: Challenges and opportunities. Environ

Health Perspect. 2001;109: 1–5. doi:10.2307/3434913 5. Doiron D, Burton P, Marcon Y, Gaye A, Wolffenbuttel BH, Perola M, et al. Data harmonization

and federated analysis of population-based studies: the BioSHaRE project. Emerg Themes Epidemiol. 2013;10: 12. doi:10.1186/1742-7622-10-12

6. Van Vliet-Ostaptchouk J V, Nuotio ML, Slagter SN, Doiron D, Fischer K, Foco L, et al. The prevalence of metabolic syndrome and metabolically healthy obesity in Europe: a collaborative analysis of ten large cohort studies. BMC Endocr Disord. 2014;14: 9. doi:10.1186/1472-6823-14-9

7. Stolk RP, Rosmalen JGM, Postma DS, de Boer RA, Navis G, Slaets JPJ, et al. Universal risk fac-tors for multifactorial diseases: LifeLines: a three-generation population-based study. Eur J Epidemiol. 2008;23: 67–74. doi:10.1007/s10654-007-9204-4

8. Scholtens S, Smidt N, Swertz MA, Bakker SJ, Dotinga A, Vonk JM, et al. Cohort Profile: LifeLines, a three-generation cohort study and biobank. Int J Epidemiol. 2014; doi: 10.1093/ije/dyu229. doi:10.1093/ije/dyu229

9. Klijs B, Scholtens S, Mandemakers JJ, Snieder H, Stolk RP, Smidt N. Representativeness of the LifeLines Cohort Study. Ali RI, editor. PLoS One. 2015;10: e0137203. doi:10.1371/journal.pone.0137203

10. Foraster M, Basagaña X, Aguilera I, Rivera M, Agis D, Bouso L, et al. Association of long-term exposure to traffic-related air pollution with blood pressure and hypertension in an adult population-based cohort in Spain (the REGICOR study). Environ Health Perspect. 2014;122: 404–11. doi:10.1289/ehp.1306497

11. Fuks KB, Weinmayr G, Foraster M, Dratva J, Hampel R, Houthuijs D, et al. Arterial blood pres-sure and long-term exposure to traffic-related air pollution: an analysis in the European Study of Cohorts for Air Pollution Effects (ESCAPE). Environ Health Perspect. 2014;122: 896–905. doi:10.1289/ehp.1307725

12. Fischer PH, Marra M, Ameling CB, Hoek G, Beelen R, de Hoogh K, et al. Air Pollution and Mortal-ity in Seven Million Adults: The Dutch Environmental Longitudinal Study (DUELS). Environ Health Perspect. 2015; doi:10.1289/ehp.1408254

13. Ragettli MS, Phuleria HC, Tsai M-Y, Schindler C, de Nazelle A, Ducret-Stich RE, et al. The rel-evance of commuter and work/school exposure in an epidemiological study on traffic-related air pollution. J Expo Sci Environ Epidemiol. 2014; doi:10.1038/jes.2014.83


15. Newby DE, Mannucci PM, Tell GS, Baccarelli AA, Brook RD, Donaldson K, et al. Expert position paper on air pollution and cardiovascular disease. Eur Heart J. 2014;36: 83–93b. doi:10.1093/eurheartj/ehu458

Chapter 2

36

16. Beelen R, Hoek G, Vienneau D, Eeftens M, Dimakopoulou K, Pedeli X, et al. Development of NO2 and NOx land use regression models for estimating air pollution exposure in 36 study areas in Europe – The ESCAPE project. Atmos Environ. 2013;72: 10–23. doi:10.1016/j.at-mosenv.2013.02.037

17. Eeftens M, Beelen R, de Hoogh K, Bellander T, Cesaroni G, Cirach M, et al. Development of Land Use Regression models for PM(2.5), PM(2.5) absorbance, PM(10) and PM(coarse) in 20 Euro-pean study areas; results of the ESCAPE project. Environ Sci Technol. 2012;46: 11195–11205. doi:10.1021/es301948k

18. Vienneau D, de Hoogh K, Bechle MJ, Beelen R, van Donkelaar A, Martin R V, et al. Western Euro-pean land use regression incorporating satellite- and ground-based measurements of NO2 and PM10. Environ Sci Technol. 2013;47: 13555–64. doi:10.1021/es403089q


20. Kephalopoulos S, Paviotti M, Anfosso-Lédée F, Van Maercke D, Shilton S, Jones N. Advances in the development of common noise assessment methods in Europe: The CNOSSOS-EU frame-work for strategic environmental noise mapping. Sci Total Environ. 2014;482-483: 400–10. doi:10.1016/j.scitotenv.2014.02.031

21. Kephalopoulos S, Paviotti M, Anfosso-Lédée F. Common Noise Assessment Methods in Europe (CNOSSOS-EU). 2012.

22. Morley DW, de Hoogh K, Fecht D, Fabbri F, Bell M, Goodman PS, et al. International scale imple-mentation of the CNOSSOS-EU road traffic noise prediction model for epidemiological studies. Environ Pollut. 2015;206: 332–341. doi:10.1016/j.envpol.2015.07.031

23. European Environment Agency. Corine Land Cover 2000 Seamless Vector Data (V16) [Inter-net]. 2015 [cited 1 Jul 2015]. Available: http://www.eea.europa.eu/data-and-maps/data/corine-land-cover-2000-clc2000-seamless-vector-database-4

24. Rothman KJ. Methodologic frontiers in environmental epidemiology. Environ Health Perspect. 1993;101 Suppl: 19–21.




28. Zijlema WL, Morley D, Stolk RP, Rosmalen JGM. Noise and somatic symptoms: A role for person-ality traits? Int J Hyg Environ Health. 2015; doi:10.1016/j.ijheh.2015.05.001

29. International Organization for Standardization (ISO). Acoustics – Assessment of Noise Annoy-ance by Means of Social and Socio-acoustic Surveys; ISO/TS 15666. Geneva, Switzerland; 2003.

30. Babisch W, Houthuijs D, Pershagen G, Cadum E, Katsouyanni K, Velonakis M, et al. Annoyance due to aircraft noise has increased over the years—results of the HYENA study. Environ Int. 2009;35: 1169–76. doi:10.1016/j.envint.2009.07.012

37


2

31. Porsius JT, Claassen L, Smid T, Woudenberg F, Timmermans DRM. Health responses to a new high-voltage power line route: design of a quasi-experimental prospective field study in the Netherlands. BMC Public Health. 2014;14: 237. doi:10.1186/1471-2458-14-237

32. Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J, et al. UK Biobank: An Open Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of Middle and Old Age. PLoS Med. 2015;12: e1001779. doi:10.1371/journal.pmed.1001779

33. Janson C, Chinn S, Jarvis D, Zock J-P, Torén K, Burney P. Effect of passive smoking on respira-tory symptoms, bronchial responsiveness, lung function, and total serum IgE in the European Community Respiratory Health Survey: a cross-sectional study. Lancet. 2001;358: 2103–2109. doi:10.1016/S0140-6736(01)07214-2

34. GGD. Gezondheidsenquete. In: 2015 [Internet]. [cited 1 Sep 2015]. Available: http://ggd.gron-ingen.nl/onderzoek/gezondheidsprofiel

35. Rijksoverheid. WoonOnderzoek Nederland (WoON). In: 2015 [Internet]. [cited 1 Sep 2015]. Available: https://www.rijksoverheid.nl/onderwerpen/onderzoeken-over-bouwen-wonen-en-leefomgeving/inhoud/lopende-onderzoeken/woononderzoek-nederland-woon

36. Leal C, Chaix B. The influence of geographic life environments on cardiometabolic risk factors: a systematic review, a methodological assessment and a research agenda. Obes Rev. 2011;12: 217–30. doi:10.1111/j.1467-789X.2010.00726.x

37. Moser A, Panczak R, Zwahlen M, Clough-Gorr KM, Spoerri A, Stuck AE, et al. What does your neighbourhood say about you? A study of life expectancy in 1.3 million Swiss neighbourhoods. J Epidemiol Community Health. 2014;68: 1125–32. doi:10.1136/jech-2014-204352

38. Statistics Netherlands. CBS - Kerncijfers wijken en buurten 2004-2013. In: Centraal Bureau voor de Statistiek, Den Haag/Heerlen [Internet]. 2013. Available: http://www.cbs.nl/nl-NL/menu/themas/dossiers/nederland-regionaal/cijfers/incidenteel/maatwerk/wijk-buurt-statistieken/kwb-recent/default.htm

39. Emaus MJ, Bakker MF, Beelen RMJ, Veldhuis WB, Peeters PHM, van Gils CH. Degree of urbaniza-tion and mammographic density in Dutch breast cancer screening participants: results from the EPIC-NL cohort. Breast Cancer Res Treat. 2014;148: 655–63. doi:10.1007/s10549-014-3205-2

40. Tigchelaar EF, Zhernakova A, Dekens JAM, Hermes G, Baranska A, Mujagic Z, et al. Cohort profile: LifeLines DEEP, a prospective, general population cohort study in the northern Neth-erlands: study design and baseline characteristics. BMJ Open. 2015;5: e006772. doi:10.1136/bmjopen-2014-006772

41. Maas J, Verheij RA, de Vries S, Spreeuwenberg P, Schellevis FG, Groenewegen PP. Morbidity is related to a green living environment. J Epidemiol Community Health. 2009;63: 967–973. doi:10.1136/jech.2008.079038


43. Brunekreef B. Exposure science, the exposome, and public health. Environ Mol Mutagen. 2013;54: 596–8. doi:10.1002/em.21767

44. Peters A, Hoek G, Katsouyanni K. Understanding the link between environmental exposures and health: does the exposome promise too much? J Epidemiol Community Health. 2011; doi:10.1136/jech-2011-200643

Chapter 2

38

45. Chang ET, Adami H-O, Bailey WH, Boffetta P, Krieger RI, Moolgavkar SH, et al. Validity of geo-graphically modeled environmental exposure estimates. Crit Rev Toxicol. 2014;44: 450–66. doi:10.3109/10408444.2014.902029

46. Bovenberg JA, Hansell A, de Hoogh K, Knoppers BM. Nature, nurture and exposure: Con-necting biobank data with geographic data could yield public and individual health benefits, but risks to human rights need to be assessed. EMBO Rep. 2015;16: 404–6. doi:10.15252/embr.201439902

47. Schwartz J, Bellinger D, Glass T. Exploring potential sources of differential vulnerability and susceptibility in risk from environmental hazards to expand the scope of risk assessment. Am J Public Health. 2011;101 Suppl: S94–101. doi:10.2105/AJPH.2011.300272

3(Un)Healthy in the City: respiratory, cardiometabolic and mental health associated with urbanity

Wilma L. Zijlema, Bart Klijs, Ronald P. Stolk, Judith G.M. Rosmalen

PLoS One 2015 Dec;10(12):e0143910

Chapter 3

42

aBsTracT

Background

Research has shown that health differences exist between urban and rural ar-eas. Most studies conducted, however, have focused on single health outcomes and have not assessed to what extent the association of urbanity with health is explained by population composition or socioeconomic status of the area. Our aim is to investigate associations of urbanity with four different health outcomes (i.e. lung function, metabolic syndrome, depression and anxiety) and to assess whether these associations are independent of residents’ characteristics and area socioeconomic status.

methods

Our study population consisted of 74,733 individuals (42% males, mean age 43.8) who were part of the baseline sample of the LifeLines Cohort Study. Health outcomes were objectively measured with spirometry, a physical examination, laboratory blood analyses, and a psychiatric interview. Using multilevel linear and logistic regression models, associations of urbanity with lung function, and prevalence of metabolic syndrome, major depressive disorder and generalized anxiety disorder were assessed. All models were sequentially adjusted for age, sex, highest education, household equivalent income, smoking, physical activity, and mean neighborhood income.

results

As compared with individuals living in rural areas, those in semi-urban or ur-ban areas had a poorer lung function (β -1.62, 95% CI -2.07;-1.16), and higher prevalence of major depressive disorder (OR 1.65, 95% CI 1.35;2.00), and gen-eralized anxiety disorder (OR 1.58, 95% CI 1.35;1.84). Prevalence of metabolic syndrome, however, was lower in urban areas (OR 0.51, 95% CI 0.44;0.59). These associations were only partly explained by differences in residents’ demographic, socioeconomic and lifestyle characteristics and socioeconomic status of the areas.

conclusions

Our results suggest a differential health impact of urbanity according to type of disease. Living in an urban environment appears to be beneficial for cardiometa-bolic health but to have a detrimental impact on respiratory function and mental health. Future research should investigate which underlying mechanisms explain the differential health impact of urbanity.

43

Urban-rural health differences

3

inTrodUcTion

There’s no place like home. But is this place or home a healthy one? In the present-day, more than half of the world’s population lives in urban areas [1], but this may not be the healthiest environment to live in. Living in urban areas exposes people to traffic fumes, noise, crime, and other stressful factors associated with city liv-ing. On the other hand, urban living may also be beneficial because of a variety of food choices, proximity of health care services, and other facilities such as sports amenities.

Indeed, research suggests that there are urban-rural health differences. Living in a less urbanized or rural environment rather than an urban environment is for example associated with increased blood pressure [2,3], obesity [4,5], and higher cholesterol levels [6]. However, living in more densely populated areas may dete-riorate respiratory [7] and mental health [8], and may be a risk factor for breast cancer [9,10]. Urban-rural differences in mortality have also been observed. For example, positive relations between population density and mortality from lung cancer, chronic obstructive pulmonary disease and ischemic heart disease have been observed [11], while a beneficial effect of high urbanity on all-cause mortal-ity has also been reported [12].

Although urban-rural health differences have been a popular research topic in the past 10 years, much is still to gain in the area of research. A recent review identified areas in which more research is needed [13]. For example, some of the previous studies used self-report outcome measures [13]. It has been shown that self-report of for example height and weight can lead to underestimation of overweight [14], indicating the importance of objectively measured outcomes. Furthermore, in addition to individual socioeconomic status, neighborhood level socioeconomic status should also be taken into account. Neighborhood socioeco-nomic status is related to the health of its residents and also to neighborhood characteristics including availability of services, infrastructure and social cohe-sion, which in turn influence health [15]. Previous studies often investigated one single health outcome. Because city living can have both negative and positive effects on health, we think it is important to study multiple health outcomes in relation to urbanity. We studied lung function, metabolic syndrome, depression and anxiety. The burden of disease is high for each of these diseases [16], and previous research has shown that they may be related to the degree of urban-ity. Our aim is to investigate associations of urbanity with four different health outcomes (i.e. lung function, metabolic syndrome, depression and anxiety) and to assess whether these associations are independent of individual characteristics and area socioeconomic status.

Chapter 3

44

meTHods

study design

LifeLines is a multi-disciplinary prospective population-based cohort study examining in a unique three-generation design the health and health-related behaviors of 167,729 persons living in the North East region of The Netherlands. It employs a broad range of investigative procedures in assessing the biomedical, socio-demographic, behavioral, physical and psychological factors which con-tribute to the health and disease of the general population, with a special focus on multimorbidity and complex genetics [17]. Recruitment of study participants took place between 2006-2013 through general practitioners and self-enroll-ment. Participants visited the research sites for a physical examination, including spirometry, anthropometry, collection of fasting blood samples, and a psychiatric interview. An extensive questionnaire was completed at home [18].

Participants

At the time of analysis, recruitment of participants was still ongoing. The pres-ent study is therefore a cross-sectional analysis based on the first data release of 95,432 participants, of which 74,733 had complete data on degree of urbanity. Participants were excluded in the corresponding analyses when they had missing data on the variables needed to define the outcome, resulting in complete data for 63,946 (lung function), 73,278 (metabolic syndrome), or 71,536 (MDD and GAD) participants. All participants provided written informed consent. The study protocol was carried out in accordance to the Declaration of Helsinki, and ap-proved by the medical ethical review committee of the University Medical Center Groningen.

degree of urbanity and neighborhood socioeconomic status

Degree of urbanity and neighborhood level socioeconomic status were obtained from Statistics Netherlands [19]. Home addresses of study participants were geocoded and subsequently the corresponding neighborhood was determined. This resulted in 2,524 unique neighborhoods for the current dataset. Degree of urbanity was expressed as the address density (average number of addresses per square kilometer within a range of 1 kilometer), and was categorized into five levels ranging from rural (<500 addresses per km2; e.g. very small villages) to urban (≥2500 addresses per km2; e.g. city center of a provincial capital) [19]. Neighborhood socioeconomic status was operationalized as neighborhood in-come (net income per year divided by number of inhabitants in the neighborhood in euros). Because of privacy reasons, data on neighborhood income were only

45


3

made available for neighborhoods with a minimum of 200 inhabitants [19]. For the current dataset, this resulted in 1,176 missing observations.

metabolic syndrome

Participants were considered to have metabolic syndrome if they satisfied at least three out of the five criteria according to the revised National Cholesterol Educa-tion Program’s Adult Treatment Panel III (NCEP-ATPIII) [20]. The criteria are as follows: (1) systolic blood pressure (SBP) ≥130 mmHg or diastolic blood pres-sure (DBP) ≥85 mmHg or use of blood pressure-lowering medication; (2) fasting blood glucose level ≥5.6 mmol/l or use of blood glucose-lowering medication or diagnosis of type 2 diabetes; (3) high-density lipoprotein(HDL)-cholesterol <1.03 mmol/l in men or <1.30 mmol/l in women or lipid-lowering medical treatment; (4) triglycerides ≥1.70 mmol/l or medication for elevated triglycerides; (5) and waist circumference ≥102 cm in men or ≥88 cm in women. If body mass index (BMI) is ≥30 kg/m2, abdominal obesity was assumed and waist circumference is not included as a criterion. When BMI is ≥30 kg/m2, at least two out of four other criteria should be met to fulfill the criteria for metabolic syndrome.

Anthropometric and blood pressure measurements were conducted by trained medical professionals using a standardized protocol. Fasting blood samples were collected and analyzed on the day of collection, and medication use was assessed by means of a questionnaire.

lung function

FEV1 (forced expiratory volume in 1 second) was measured with spirometry during a medical examination performed in a standardized setting following American Thoracic Society guidelines. Lung function varies with age, height, sex and ethnicity. Therefore, lung function was compared to predicted values, which were calculated with equations from the European Respiratory Society [21]. Subsequently, the percentage FEV1 relative to the predicted values (FEV1 % predicted) was calculated. Since the number of complete cases for the lung func-tion analyses were lower compared to the analyses for the other health outcomes, we imputed the missing values using a Markov Chain Monte Carlo algorithm in IBM SPSS Statistics (version 22). All variables that are used in the lung function analyses were used as predictors in the imputation model to improve imputation results. Analyses were undertaken with the original data and the imputed dataset. Analyses on the imputed dataset did not result in different conclusions compared to analyses on the original dataset. This indicates that the missing values probably had no effect on our results, and therefore results based on the original dataset are reported.

Chapter 3

46

Major depressive disorder and generalized anxiety disorder

For the assessment of major depressive disorder (MDD) and generalized anxiety disorder (GAD), participants were interviewed by trained medical professionals during their visit to the research facilities. MDD and GAD were assessed with the Mini-International Neuropsychiatric Interview (MINI). The MINI is a brief struc-tured interview for diagnosing psychiatric disorders as defined by the DSM-IV (Diagnostic and Statistical Manual of Mental Disorders) and ICD-10 (International Statistical Classification of Diseases and Related Health Problems).

covariates

Sex, age, highest obtained educational degree, household income, number of persons in the household, current smoking and physical activity were assessed with a questionnaire. Household equivalent income was calculated as net house-hold income per month divided by the square root of the number of persons in the household [22]. Participants were current smokers if they indicated current smoking or smoking in the last month. Physical activity was assessed by asking how many days per week participants are physically active for at least half an hour.

statistical analysis

Sample characteristics were described with means (with standard deviations) and prevalence (%) stratified for degree of urbanity. Because of non-normal distribution of neighborhood income medians and interquartile ranges (IQRs) were calculated instead. To account for the within-neighborhood correlation of the outcomes, associations were analyzed with multilevel analysis with a random intercept defined at the neighborhood level. First, associations of urbanity (refer-ence category rural) and lung function, metabolic syndrome, MDD and GAD were assessed. Models were subsequently adjusted for neighborhood income (model 2); age, sex (model 3), education, household equivalent income (model 4); smok-ing, and physical activity (model 5). Adjusting for these covariates enabled us to assess whether neighborhood income, individual characteristics, socioeconomic characteristics, and lifestyle factors could explain (part of) the association of urbanity and health. Outcomes were expressed as regression coefficients (β) for FEV1 % predicted, and as odds ratios (ORs) for metabolic syndrome, MDD and GAD. Model comparison was based on the change in log-likelihood. Multilevel models were analyzed with R (version 3.0.1) using the lme4 package (version 1.1-7) [23]. Associations were considered statistically significant if the 95% con-fidence intervals (CI) did not include zero (β) or one (OR).

47


3

resUlTs

Population characteristics

Of the 74,733 participants in the study sample, 42% were male, the majority lived in rural neighborhoods (41%) and the mean age was 43.8 (SD 11.6) years. Mean FEV1 % predicted was 102.4; 15.4% were classified as having metabolic syn-drome, MDD and GAD prevalence was respectively 2.3% and 4.5% (Table 1). The lowest average FEV1 % predicted and highest prevalence of MDD was observed in semi-urban neighborhoods. Highest prevalence of metabolic syndrome was ob-served in intermediate urban-rural neighborhoods, and highest GAD prevalence in urban neighborhoods. Participants living in urban neighborhoods were on average younger, more highly educated, more often current smokers, and slightly more often physically active.

Table 1. Population characteristics of the total study population (n=74,733) and stratified by urbanity

Total Rural Semi-ruralIntermediate urban-rural Semi-urban Urban

N 74,733 30,823 18,515 12,813 7,745 4,837

Males (%) 42.0 42.6 43.1 40.8 40.8 38.9

Age, mean (SD) 43.8 (11.6) 44.8 (11.1) 44.2 (10.7) 44.2(11.5) 42.0 (12.1) 36.1 (11.8)

Education level (%)

No or primary 2.1 2.1 2.0 2.2 3.1 1.3

Lower or preparatory vocational 13.1 15.1 12.7 12.4 13.2 3.7

Lower general secondary 13.5 14.7 13.9 13.8 12.9 5.6

Intermediate vocational or apprenticeship

31.7 34.4 32.2 32.0 29.4 15.8

Higher general secondary or pre-university secondary

9.0 8.0 9.0 9.3 9.1 15.2

Higher vocational or university 30.4 25.8 30.2 30.3 32.3 58.3

Household equivalent income/month (%)

<€1000 18.0 17.4 15.7 18.1 21.7 23.9

€1000 - €1300 22.4 23.7 21.1 22.6 23.2 17.5

€1300 - €1600 16.7 18.4 17.9 17.0 13.8 7.3

€1600 - €1900 21.4 20.2 24.8 20.7 19.7 20.5

>€1900 21.4 20.2 20.4 21.7 21.6 30.8

Current smokers (%) 22.8 21.5 21.5 23.4 27.6 27.2

Days PA, mean (SD) 4.3 (2.2) 4.2 (2.2) 4.2 (2.2) 4.3 (2.2) 4.3 (2.2) 4.5 (2.2)

FEV1 % predicted, mean (SD) 102.4 (13.0) 103.0 (13.1) 102.6 (13.0) 101.9(13.2) 101.5 (12.9) 102.1(12.4)

Metabolic syndrome (%) 15.4 15.9 16.1 16.5 15.0 8.1

GAD (%) 4.5 3.8 4.2 5.6 5.5 5.8

MDD (%) 2.3 2.0 2.3 2.9 3.2 2.4

Neighborhood income (x1000 euro), median (IQR)

19.0 (3.3) 19.1 (2.6) 19.4 (3.3) 19.1 (3.2) 18.3 (3.6) 18.6 (4.8)

Abbreviations: SD: standard deviation; IQR: interquartile range; FEV1 % predicted: forced expiratory volume in 1 second relative to the predicted value; MDD: major depressive disorder; GAD: generalized anxiety disorder; days PA: number of days physically active

Chapter 3

48

Multilevel models with random intercept for neighborhood

Lung function

Living in urban, semi-urban and intermediate urban-rural neighborhoods was associated with a decrease in lung function, when compared to living in rural neighborhoods. The largest association was observed for semi-urban neighbor-hoods, with a decrease in lung function of -1.62% (95% CI -2.07;-1.16), followed by intermediate urban-rural neighborhoods (β -1.06, 95% CI -1.45;-0.67), and urban (β -0.99, 95% CI -1.52;-0.47) (Table 2). The gradient of regression coef-ficients indicates a quadratic (concave) relationship between urbanity and lung function. The association for semi-rural (vs. rural) was not statistically signifi-cant. Adjustment for neighborhood income resulted in both slight increases and decreases in the regression coefficients for the urbanity categories; this was also true for additional adjustment for demographic and SES variables, as well as for smoking and physical activity. In the fully adjusted model, negative associations of urbanity and lung function were still observed, except for the urban category, which was no longer statistically significant. However, model comparison based on log likelihood ratio tests indicated that model 5 did not improve significantly over model 4. Participants in neighborhoods of higher urbanity were more often current smokers, and this may influence the association between urbanity and lung function.

Table 2. Associations of urbanity and lung function (FEV1 % predicted) estimated from linear multilevel models (n=63,946)

Model 1 Model 2 Model 3 Model 4 Model 5

β 95% CI β 95% CI β 95% CI β 95% CI β 95% CI

Urbanity

Rural reference reference reference reference reference

Semi-rural -0.32 -0.67;0.02 -0.48 -0.81;-0.15 -0.39 -0.71;-0.06 -0.51 -0.82;0.18 -0.59 -0.91;-0.26

Intermediate urban-rural

-1.06 -1.45;-0.67 -1.08 -1.44;-0.72 -1.01 -1.37;-0.65 -1.13 -1.48;-0.77 -1.11 -1.47;-0.75

Semi urban -1.62 -2.07;-1.16 -1.56 -1.99;-1.13 -1.26 -1.69;-0.83 -1.37 -1.80;-0.95 -1.17 -1.61;-0.74

Urban -0.99 -1.52;-0.47 -0.88 -1.38;-0.39 -0.01 -0.51;0.49 -0.66 -1.16;-0.15 -0.48 -1.01;0.04

Abbreviations: FEV1 % predicted: forced expiratory volume in 1 second relative to the predicted value; 95% CI: 95% confidence intervalmodel 1: unadjustedmodel 2: adjusted for neighborhood incomemodel 3: adjusted for neighborhood income, sex, agemodel 4: adjusted for neighborhood income, sex, age, education, household equivalent incomemodel 5: adjusted for neighborhood income, sex, age, education, household equivalent income, smoking, physi-cal activity

49


3

Metabolic syndrome

Living in an urban neighborhood was associated with lower odds for metabolic syndrome (OR 0.51, 95% CI 0.44;0.59), when compared to living in rural neigh-borhoods (Table 3). Associations for other urbanity categories (vs. rural) were not statistically significant. Associations remained similar after adjustment for neighborhood income. When additionally adjusted for demographic and SES variables, the OR for urban (vs. rural) for metabolic syndrome became 0.81 (95% CI 0.71;0.93), indicating attenuation of the association. After adjusting for smoking and physical activity, associations of urbanity and metabolic syndrome remained similar. Participants in neighborhoods of higher urbanity were more often physically active, but this did not explain the association between urbanity and metabolic syndrome.

Major depressive disorder

Living in intermediate urban-rural (OR 1.39, 95% CI 1.17;1.65) and semi-urban neighborhoods (OR 1.65, 95% CI 1.35;2.00) was associated with higher odds for MDD, compared to living in rural neighborhoods (Table 4). Associations for other urbanity categories (vs. rural) were not statistically significant. Also here, asso-ciations remained similar after adjustment for neighborhood income. However, additional adjustment for demographic and SES variables resulted in amplifica-tion of the associations of degree of urbanity and MDD. ORs increased and became statistically significant for all urbanity categories. When additionally adjusted for smoking and physical activity, results remained similar.

table 3. Associations of urbanity and metabolic syndrome estimated from logistic multilevel models (n=73,278)


OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI

Urbanity


Semi-rural 0.96 0.88;1.04 0.98 0.90;1.06 1.02 0.94;1.10 1.05 0.97;1.13 1.05 0.97;1.13


1.04 0.95;1.15 1.03 0.95;1.13 1.08 0.98;1.18 1.10 1.01;1.20 1.09 1.00;1.19

Semi urban 0.94 0.84;1.05 0.92 0.83;1.03 1.03 0.93;1.15 1.06 0.95;1.17 1.04 0.94;1.16

Urban 0.51 0.44;0.59 0.49 0.43;0.56 0.69 0.60;0.79 0.81 0.71;0.93 0.80 0.70;0.92

Abbreviations: OR: odds ratio; 95% CI: 95% confidence intervalmodel 1: unadjustedmodel 2: adjusted for neighborhood incomemodel 3: adjusted for neighborhood income, sex, agemodel 4: adjusted for neighborhood income, sex, age, education, household equivalent incomemodel 5: adjusted for neighborhood income, sex, age, education, household equivalent income, smoking, physi-cal activity

Chapter 3

50

Generalized anxiety disorder

We observed that living in urban (OR 1.58, 95% CI 1.35;1.84), semi-urban (OR 1.45, 95% CI 1.27;1.67), and intermediate urban-rural neighborhoods (OR 1.46, 95% CI 1.30;1.65) was associated with higher odds for GAD, when compared to living in rural neighborhoods (Table 5). When adjusted for neighborhood income, the association for semi-rural and GAD also became statistically significant (OR 1.13, 95% CI 1.01;1.26), while the associations of the other categories remained

table 4. Associations of urbanity and major depressive disorder estimated from logistic multilevel mod-els (n=71,536)



Urbanity


Semi-rural 1.08 0.92;1.27 1.16 1.00;1.36 1.16 1.00;1.62 1.22 1.05;1.42 1.21 1.04;1.40


1.39 1.17;1.65 1.58 1.19;1.64 1.38 1.18;1.62 1.44 1.23;1.69 1.43 1.22;1.67

Semi urban 1.65 1.35;2.00 1.40 1.31;1.90 1.56 1.29;1.87 1.60 1.33;1.93 1.54 1.29;1.85

Urban 1.26 0.99;1.59 1.16 0.93;1.46 1.12 0.89;1.40 1.47 1.17;1.86 1.42 1.12;1.78


table 5. Associations of urbanity and generalized anxiety disorder estimated from logistic multilevel models (n=71,536)



Urbanity


Semi-rural 1.09 0.97;1.21 1.13 1.01;1.26 1.12 1.01;1.25 1.15 1.03;1.28 1.14 1.03;1.27


1.46 1.30;1.65 1.46 1.31;1.64 1.44 1.29;1.61 1.46 1.31;1.63 1.44 1.29;1.61

Semi urban 1.45 1.27;1.67 1.43 1.25;1.64 1.38 1.21;1.58 1.39 1.22;1.59 1.36 1.19;1.55

Urban 1.58 1.35;1.84 1.53 1.31;1.77 1.37 1.18;1.60 1.52 1.30;1.78 1.47 1.26;1.72


51


3

similar. Adjustment for demographic and SES variables, as well as smoking and physical activity, only changed the results marginally.

discUssion

Our results suggest a differential health impact of urbanity according to type of disease. Living in an urban environment appears to be beneficial for cardio-metabolic health, but may be detrimental to respiratory function and mental health. These associations were mostly independent of neighborhood and indi-vidual socioeconomic status, and other covariates. This suggests an effect of the residential environment itself, as opposed to health differences originating from demographic differences.

A strength of this study is the large study population and the objectively measured health outcomes. The fact that we used more than one health outcome enabled us to investigate the relationship between degree of urbanity and health in a much broader sense. Furthermore, by using multilevel models, we accounted for the clustering of participants in neighborhoods. We also accounted for individual and neighborhood level indicators of socioeconomic status. This enabled us to investigate urban-rural health differences independent from individual and neighborhood socioeconomic status.

A limitation of this study is the missing data for lung function. Missing data were imputed, and analyses were undertaken with the original dataset and the imputed dataset. Since conclusions based on the original dataset and the imputed dataset did not differ, we assume that the missing data did not influence our results. Re-gardless of the missing data, still a large study sample could be used. Furthermore, although we aimed to investigate the relation of urbanity and health in a broader view, we acknowledge that next to lung function, metabolic syndrome, MDD and GAD, also other diseases are important. We chose to study these particular diseases because each of them can lead to a large burden of disease and because they could be measured either objectively or with a standardized and validated interview.

Mechanisms that may explain urban-rural health differences are the breeder and the drift hypotheses. The breeder hypothesis states that health differences may arise because of the environmental factors to which people are directly ex-posed [24]. Examples are air pollution [25], traffic noise [26] and noise annoy-ance [27], stress [28], overcrowding, and social isolation [8]. Spatial variation in behavior may also explain these differences. Urban living can evoke certain (un)

Chapter 3

52

healthy lifestyles, such as smoking, physical activity, and alcohol use, and may influence diet choices [29,30]. With the drift hypothesis, it is thought that selec-tion processes result in urban-rural health differences. People with certain health related characteristics may move to or from specific places [24], for example because they have a chronic disease, and want to live close to healthcare facili-ties. These mechanisms could help explain the associations found in the current study. We found a negative relation between urbanity and lung function, which may be due to, for example, higher exposure to pollution associated with city liv-ing. Our findings suggest a potential role for smoking or physical activity, since the association between lung function and urbanity attenuated after adjustment for these lifestyle factors. Similar findings were reported in other studies [7,31]. In contrast, living in urban neighborhoods was associated with a lower preva-lence of metabolic syndrome, when compared to living in rural neighborhoods. We also evaluated the individual components of metabolic syndrome and their distribution among degree of urbanity. In line with our expectations, mean SBP, DBP, fasting blood glucose, triglycerides, waist circumference and BMI increased as urbanity decreased, and HDL cholesterol decreased as urbanity decreased. One explanation may be that people in rural areas are less physically active compared to their urban counterparts [32]. Many facilities in rural areas are often not within walking or cycling distance, resulting in car use instead. Also the absence of broad walking and cycling lanes and street lights make rural areas less attractive for being physically active [33,34]. However, associations between urbanity and metabolic syndrome found in our study could not be explained by the number of days that people are physically active. Finally, residents of urban areas had an increased risk for developing MDD and GAD, which was in accordance with the conclusions of a meta-analysis [8]. Again, (in)direct exposure to environmental factors [35], or selection processes may potentially explain urban -rural varia-tions in psychopathology [8].

This study suggests a differential health impact of urbanity according to type of disease. Future research should investigate which underlying mechanisms explain the differential health impact of urbanity. Spatial planning may facilitate behavior change in order to promote health, for example through the provision of cycling infrastructure. In addition, when it becomes more clear which aspects of the environment are threatening to our health, policy makers can diminish harm-ful effects by introducing legislation. The introduction of smoke-free legislation has for example resulted in substantial benefits for public health [36]. Further-more, identification of susceptible populations is relevant for policy makers in order to initiate prevention strategies that are tailored to the population and their environment [37].

53


3

references

1. United Nations Population Fund. State of world population 2007: unleashing the potential of urban growth [Internet]. New York; 2007. Available: http://www.unfpa.org/publications/state-world-population-2007

2. Chaix B, Ducimetiere P, Lang T, Haas B, Montaye M, Ruidavets JB, et al. Residential environment and blood pressure in the PRIME Study: is the association mediated by body mass index and waist circumference? J Hypertens. 2008;26: 1078–1084. doi:10.1097/HJH.0b013e3282fd991f

3. Van Hulst A, Thomas F, Barnett TA, Kestens Y, Gauvin L, Pannier B, et al. A typology of neighbor-hoods and blood pressure in the RECORD Cohort Study. J Hypertens. 2012;30: 1336–1346. doi:10.1097/HJH.0b013e3283544863

4. Neovius M, Rasmussen F. Place of residence and obesity in 1,578,694 young Swedish men between 1969 and 2005. Obesity (Silver Spring). 2008;16: 671–6. doi:10.1038/oby.2007.115

5. Bruner MW, Lawson J, Pickett W, Boyce W, Janssen I. Rural Canadian adolescents are more likely to be obese compared with urban adolescents. Int J Pediatr Obes. 2008;3: 205–11. doi:10.1080/17477160802158477

6. Lindroth M, Lundqvist R, Lilja M, Eliasson M. Cardiovascular risk factors differ between rural and urban Sweden: the 2009 Northern Sweden MONICA cohort. BMC Public Health. 2014;14: 825. doi:10.1186/1471-2458-14-825

7. Broms K, Norback D, Eriksson M, Sundelin C, Svardsudd K. Effect of degree of urbanisation on age and sex-specific asthma prevalence in Swedish preschool children. BMC Public Health. 2009;9: 303. doi:10.1186/1471-2458-9-303

8. Peen J, Schoevers RA, Beekman AT, Dekker J. The current status of urban-rural differ-ences in psychiatric disorders. Acta Psychiatr Scand. 2010;121: 84–93. doi:10.1111/j.1600-0447.2009.01438.x

9. Robert SA, Strombom I, Trentham-Dietz A, Hampton JM, McElroy JA, Newcomb PA, et al. Socioeconomic risk factors for breast cancer: distinguishing individual- and community-level effects. Epidemiology. 2004;15: 442–50.

10. Emaus MJ, Bakker MF, Beelen RMJ, Veldhuis WB, Peeters PHM, van Gils CH. Degree of urbaniza-tion and mammographic density in Dutch breast cancer screening participants: results from the EPIC-NL cohort. Breast Cancer Res Treat. 2014;148: 655–63. doi:10.1007/s10549-014-3205-2

11. Chaix B, Rosvall M, Lynch J, Merlo J. Disentangling contextual effects on cause-specific mortal-ity in a longitudinal 23-year follow-up study: impact of population density or socioeconomic environment? Int J Epidemiol. 2006;35: 633–43. doi:10.1093/ije/dyl009

12. Van Hooijdonk C, Droomers M, Deerenberg IM, Mackenbach JP, Kunst AE. Higher mortality in urban neighbourhoods in The Netherlands: who is at risk? J Epidemiol Community Health. 2008;62: 499–505. doi:10.1136/jech.2007.060145

13. Leal C, Chaix B. The influence of geographic life environments on cardiometabolic risk factors: a systematic review, a methodological assessment and a research agenda. Obes Rev. 2011;12: 217–30. doi:10.1111/j.1467-789X.2010.00726.x

14. Rowland M. Self-reported weight and height. Am J Clin Nutr. 1990;52: 1125–1133. 15. Chaix B, Bean K, Leal C, Thomas F, Havard S, Evans D, et al. Individual/neighborhood social

factors and blood pressure in the RECORD Cohort Study: which risk factors explain the associa-tions? Hypertension. 2010;55: 769–775. doi:10.1161/HYPERTENSIONAHA.109.143206

Chapter 3

54

16. Lim SS, Vos T, Flaxman AD, Danaei G, Shibuya K, Adair-Rohani H, et al. A comparative risk as-sessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990-2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet. 2013;380: 2224–2260. doi:10.1016/S0140-6736(12)61766-8



19. Statistics Netherlands. CBS - Kerncijfers wijken en buurten 2004-2013. In: Centraal Bureau voor de Statistiek, Den Haag/Heerlen [Internet]. 2013. Available: http://www.cbs.nl/nl-NL/menu/themas/dossiers/nederland-regionaal/cijfers/incidenteel/maatwerk/wijk-buurt-statistieken/kwb-recent/default.htm

20. Alberti KG, Eckel RH, Grundy SM, Zimmet PZ, Cleeman JI, Donato KA, et al. Harmonizing the metabolic syndrome: a joint interim statement of the International Diabetes Federation Task Force on Epidemiology and Prevention; National Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International. Circulation. 2009;120: 1640–1645. doi:10.1161/CIRCULATIONAHA.109.192644

21. Quanjer PH, Stanojevic S, Cole TJ, Baur X, Hall GL, Culver BH, et al. Multi-ethnic reference values for spirometry for the 3-95-yr age range: the global lung function 2012 equations. Eur Respir J. 2012;40: 1324–1343. doi:10.1183/09031936.00080312

22. Buhrman B, Rainwater L, Schmaus G, Smeeding T. Sensitivity estimates across ten countries using the luxembourg income study (LIS) data base. Rev Income Wealth. 1988;2: 115–42.

23. Bates D, Maechler M, Bolker B, Walker S. Linear mixed-effects models using Eigen and S4 [Internet]. 19 Jul 2014 [cited 1 Dec 2014]. Available: http://cran.r-project.org/web/packages/lme4/index.html


25. Raaschou-Nielsen O, Andersen ZJ, Beelen R, Samoli E, Stafoggia M, Weinmayr G, et al. Air pollution and lung cancer incidence in 17 European cohorts: prospective analyses from the European Study of Cohorts for Air Pollution Effects (ESCAPE). Lancet Oncol. Elsevier Ltd; 2013;14: 813–822. doi:10.1016/S1470-2045(13)70279-1

26. Bluhm GL, Berglind N, Nordling E, Rosenlund M. Road traffic noise and hypertension. Occup Environ Med. 2007;64: 122–126. doi:10.1136/oem.2005.025866


28. Lederbogen F, Kirsch P, Haddad L, Streit F, Tost H, Schuch P, et al. City living and urban upbring-ing affect neural social stress processing in humans. Nature. 2011;474: 498–501. doi:10.1038/nature10190

29. Burgoine T, Forouhi NG, Griffin SJ, Wareham NJ, Monsivais P. Associations between exposure to takeaway food outlets, takeaway food consumption, and body weight in Cambridgeshire, UK: population based, cross sectional study. BMJ. 2014;348: g1464.

55


3

30. Saelens BE, Sallis JF, Black JB, Chen D. Neighborhood-Based Differences in Physical Activity: An Environment Scale Evaluation. Am J Public Health. American Public Health Association; 2003;93: 1552–1558. doi:10.2105/AJPH.93.9.1552

31. Maio S, Baldacci S, Carrozzi L, Polverino E, Angino A, Pistelli F, et al. Urban residence is as-sociated with bronchial hyperresponsiveness in Italian general population samples. Chest. 2009;135: 434–41. doi:10.1378/chest.08-0252

32. Wilcox S. Determinants of leisure time physical activity in rural compared with urban older and ethnically diverse women in the United States. J Epidemiol Community Heal. 2000;54: 667–672. doi:10.1136/jech.54.9.667

33. Mujahid MS, Roux AVD, Morenoff JD, Raghunathan TE, Cooper RS, Ni H, et al. Neighbor-hood characteristics and hypertension. Epidemiology. 2008;19: 590–8. doi:10.1097/EDE.0b013e3181772cd2

34. Mujahid MS, Diez Roux A V, Shen M, Gowda D, Sánchez B, Shea S, et al. Relation between neighborhood environments and obesity in the Multi-Ethnic Study of Atherosclerosis. Am J Epidemiol. 2008;167: 1349–57. doi:10.1093/aje/kwn047

35. Power MC, Kioumourtzoglou M-A, Hart JE, Okereke OI, Laden F, Weisskopf MG. The relation between past exposure to fine particulate air pollution and prevalent anxiety: observational cohort study. BMJ. 2015;350: h1111.


37. Edwards KL, Clarke GP, Ransley JK, Cade J. The neighbourhood matters: studying exposures relevant to childhood obesity and the policy implications in Leeds, UK. J Epidemiol Community Health. 2010;64: 194–201. doi:10.1136/jech.2009.088906

4The association of air pollution and depressed mood in 70,928 individuals from four European cohorts

Wilma L. Zijlema, Kathrin Wolf, Rebecca Emeny, Karl-Heinz Ladwig, Annette Peters, Håvard Kongsgård, Kristian Hveem, Kirsti Kvaløy, Tarja Yli-Tuomi, Timo Partonen, Timo Lanki, Marloes Eeftens, Kees de Hoogh, Bert Brunekreef, BioSHaRE, Ronald P. Stolk, Judith G.M. Rosmalen

International Journal of Hygiene and Environmental Health, forthcoming

Chapter 4

58

aBsTracT

Background

Exposure to ambient air pollution may be associated with impaired mental health, including depression. However, evidence originates mainly from animal studies and epidemiological studies in specific subgroups. We investigated the association between air pollution and depressed mood in four European general population cohorts.

methods

Data were obtained from LifeLines (the Netherlands), KORA (Germany), HUNT (Norway), and FINRISK (Finland). Residential exposure to particles (PM2.5, PM2.5absorbance, PM10) and nitrogen dioxide (NO2) was estimated using land use regression (LUR) models developed for the European Study of Cohorts for Air Pol-lution Effects (ESCAPE) and using European wide LUR models. Depressed mood was assessed with interviews and questionnaires. Logistic regression analyses were used to investigate the cohort specific associations between air pollution and depressed mood.

results

A total of 70,928 participants were included in our analyses. Depressed mood ranged from 1.6% (KORA) to 11.3% (FINRISK). Cohort specific associations of the air pollutants and depressed mood showed heterogeneous results. For example, positive associations were found for NO2 in LifeLines (odds ratio [OR]= 1.34; 95% CI: 1.17, 1.53 per 10 µg/m3 increase in NO2), whereas negative associations were found in HUNT (OR= 0.79; 95% CI: 0.66, 0.94 per 10 µg/m3 increase in NO2).

conclusions

Our analyses of four European general population cohorts found no consistent evidence for an association between ambient air pollution and depressed mood.

59

Air pollution and depression

4

inTrodUcTion

It is well established that exposure to air pollution can lead to a wide variety of adverse health effects [1]. Air pollution is for example associated with increased risks of pulmonary [2] and cardiovascular disease [3], and mortality [4,5]. Ex-posure to ambient air pollution has also been suggested to increase the risk of depressive symptoms, but few epidemiological studies have investigated these effects. So far, mainly short-term studies have been conducted and found relations with increased suicide risk [6] and depressive symptoms in Korean populations [7], and short-term increases in ambient air pollution were associated with emer-gency department visits for depression in Canada [8] and Korea [9]. However, no evidence for an association between both short- and long-term air pollution exposure and depressive symptoms could be seen in an US study [10]. Another US study among women reported that long-term and short-term exposure to air pollution was related to anxiety symptoms [11], which are often comorbid with depression [12].

Recently, studies on air pollution in relation to neuropsychological effects were reviewed. The authors concluded that the two are probably linked, but acknowledged that these results are not conclusive, as the number of studies was limited, and their sample sizes were small [13]. Another limitation in the current literature is the lack of studies investigating the possible confounding and synergistic effects of air pollution and noise on depressed mood. Individuals exposed to traffic related air pollution are probably also exposed to traffic related noise, and both exposures may be related to the pathogenesis of depression [13]. A systematic review of the effects of air pollution and ambient noise on different aspects of mental health concluded that both exposures may be associated with mood disorders [14]. The simultaneous analysis of air pollution and noise has not been undertaken extensively, and is required in future research [13,14].

In summary, most prior studies have provided limited evidence for the rela-tion between air pollution and depression, and did not analyze exposure to noise in addition to air pollution. Previous studies were mainly undertaken in Asian [6,7,9] and American populations [8,10], while relations have not yet been stud-ied in Europe. We investigated the association between air pollution exposure and depressed mood in four general population cohorts from Europe, while taking into account exposure to road traffic noise.

Chapter 4

60

meTHods

study population

This study is an analysis of cohort data obtained by BioSHaRE (Biobank Stan-dardisation and Harmonisation for Research Excellence in the European Union), a collaborative project that aims to facilitate harmonization and standardization of data, and the sharing and pooling of data across multiple biobanks and data-bases. The present study included the BioSHaRE cohorts with information about air pollution exposure and depression prevalence. The cohorts were: LifeLines (three Northern provinces of the Netherlands) [15,16], HUNT3 (Nord-Trøndelag area, Norway) [17], KORA (F3 and F4) (Augsburg area, Germany) [18], and FIN-RISK2007 (Helsinki, Vantaa and Turku areas, Finland) [19]. All cohorts are general population based. Air pollution exposure estimation and depressed mood assess-ments were undertaken within overlapping periods (LifeLines, HUNT), or with a few years in between (KORA, FINRISK). We assume that the spatial contrasts in the measured and modelled annual average levels were stable over these periods [20]. Additional information about the study designs and populations is provided in Table 1. Ethical approval was obtained from the local authorized institutional review boards and written informed consent was obtained from all participants.

Air pollution exposure assessment

Air pollution estimates for the participant’s address locations were derived from two types of land use regression (LUR) models. For LifeLines, KORA, and FINRISK, estimates of particulate matter (PM2.5, PM2.5 absorbance (reflectance on PM2.5

Table 1. Cohort characteristics

LifeLines KORA HUNT FINRISK

Study region Three Northern provinces of the Netherlands

Augsburg area, Germany

Nord-Trøndelag area, Norway

Helsinki, Vantaa and Turku, Finland

Period of study measurements

2007-2013 (baseline) 2004-2005 (F3) and 2006-2008 (F4)

2006-2008 (HUNT3) 2007 (FINRISK 2007)

Air pollution LUR model (measurement period)

ESCAPE (2009-2010) and EU-wide(2007)

ESCAPE (2008-2009) and EU-wide (2005-2007)

EU-wide (2006-2007) ESCAPE (2010-2011)

Depression measure MINI diagnostic interview

PHQ-9 interview version

HADS-D questionnaire

CES-D questionnaire

Abbreviations: ESCAPE = European Study of Cohorts for Air Pollution Effects; EU-wide = European wide; MINI = Mini-International Neuropsychiatric Interview; PHQ-9 = depression module of the patient health ques-tionnaire; HADS-D = depression subscale of the Hospital Anxiety Depression Scale; CES-D = Center for Epide-miological Studies Depression scale; LUR = land use regression

61


4

filters, i.e. a marker of black carbon), and PM10) and nitrogen dioxide (NO2) were calculated using LUR models that were previously developed in ESCAPE (Euro-pean Study of Cohorts for Air Pollution Effects) [21,22]. The HUNT study area (Nord-Trøndelag area, Norway) was not included in the ESCAPE project and hence no ESCAPE LUR model was available to be linked to the HUNT cohort. Therefore, for HUNT (and in addition for LifeLines and KORA), estimates of PM10 and NO2 were calculated using Western European-wide (EU-wide) LUR models enhanced with satellite-derived estimates of ground level air pollution [23]. Detailed de-scriptions of model development and validation can be found in Supplemental Digital Content, including tables S1-S2, and elsewhere [21–23]. Briefly, ESCAPE LUR models were developed for NO2, PM2.5, PM2.5 absorbance, and PM10 based on estimated annual average concentrations from intensive monitoring campaigns, taking place in each study area between October 2008 and April 2011 [24,25]. Measurements were undertaken in three two-week periods in the cold, warm and intermediate season. For each measurement site the annual average concentra-tion was calculated, with adjustment for temporal variation using measurements from centrally located reference sites with year-round measurement data. The air pollution concentrations obtained from the measurement campaign were then used as outcome variables for LUR model development for each of the ar-eas. Geographic Information System (GIS)-derived land use, road network, and other topographic data were used as predictors of the spatial variation in annual average air pollution levels. LUR models were developed locally, but followed a standardized protocol [21,22].

The EU-wide model incorporates GIS-derived land use, road network and topo-graphic data, as well as satellite-derived estimates of ground level concentrations for PM2.5 (as an indicator of PM10) and NO2. In these multiple linear regression equations, ambient concentrations of NO2 and PM10 (years 2005-2007) obtained from regulatory monitoring were used as dependent variables. Model develop-ment followed the ESCAPE procedure to construct the multiple linear regression equations for Western Europe (17 countries) [23]. The main difference between the ESCAPE and EU-wide models is that the ESCAPE models are region specific, while EU-wide models are developed for a much larger area. ESCAPE models were developed for specific European regions, while the EU-wide models were developed for 17 countries in Western Europe. In addition, monitoring data used in ESCAPE models originated from a monitoring campaign specifically conducted for the ESCAPE-project with monitoring sites selected for this purpose, whereas monitoring data for the EU-wide models were obtained from regulatory monitor-ing networks.

Chapter 4

62

depressed mood assessment

Depressed mood was assessed with standardized face-to-face interviews (Life-Lines and KORA) or with questionnaires (HUNT and FINRISK). LifeLines partici-pants were interviewed by trained medical professionals when they visited the research facilities. Depressed mood was assessed with a psychiatric interview (the Mini-International Neuropsychiatric Interview; MINI) [26]. The MINI is a brief structured interview for diagnosing psychiatric disorders as defined by the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV). Participants were asked to indicate whether they experienced symptoms of depression in the last two weeks (yes/no). These 9 symptoms were based on the DSM-IV criteria for the diagnosis of major depressive disorder (MDD), and a cut off of ≥5 symptoms, of which at least one of the key symptoms (depressed mood or anhedonia) was used in accordance with DSM-IV [26]. KORA participants were interviewed by trained medical professionals during their visit to the study center. Depressed mood was assessed with the interview version of the 9 item depression module of the Patient Health Questionnaire-9 (PHQ-9) [27]. Participants rated the fre-quency of symptoms of depression over the past two weeks on a scale ranging from not at all (0) to nearly every day (3). As with the MINI, the 9 items are based on the 9 DSM-IV criteria for diagnosis of MDD, and the same cut off was used. In HUNT, depressed mood was assessed with the depression subscale of the Hospital Anxiety Depression Scale (HADS) [28]. The HADS depression subscale is a self-administered questionnaire consisting of seven symptoms, each scored from not present (0) to highly present (3) in the previous week. A cut off of 10 or more points was chosen for the current study. This is different from the cut off of ≥8, which was most often used in previous studies [29]. We chose for ≥10 because this cut off was found to be more specifically related to a clinical diagnosis of MDD [30]. In FINRISK, the 20 item Center for Epidemiological Studies Depression Scale (CES-D) [31] was administered. The CES-D assesses feelings of depression in the previous week. Participants rated the frequency of these symptoms on a scale ranging from rarely or none of the time (0) to most or all of the time (3). A cut off of ≥16 is recommended and used by many studies. However, we chose a cut off of ≥21, as this cut off has shown to identify cases of severe depression [32], which is more related to MDD as measured by the MINI and the PHQ-9.

covariates

The covariates were chosen a priori based on prior knowledge [13,14,33–37]. Data on covariates including sex, age, level of education, household income, his-tory of myocardial infarction (MI), history of asthma, and current chronic obstruc-tive pulmonary disease (COPD) were available from questionnaires. Data were

63


4

harmonized across cohorts where possible according to the DataSHaPER method-ology [38]. Level of education and household income data were not available to us from HUNT, and could therefore not be used in the HUNT-specific analyses. COPD data were not available in KORA. Household equivalent income was calculated as net household income per month divided by the square root of the number of per-sons in the household. For FINRISK, data on household income before taxes was used in the analyses, since information of the tax rate for each household was not available. Household disposable income seems rather similar for the Netherlands, Germany, Norway, and Finland during the period of study measurements [39]. Urbanity was available for LifeLines, KORA and FINRISK, and was operational-ized as household density per square kilometer (KORA and FINRISK) or address density per square kilometer (LifeLines). Urbanity was not available for the HUNT cohort, and was therefore not included in the HUNT-specific analyses.

The noise exposure indicators were derived from recent noise maps for home addresses of study participants. Road traffic noise was operationalized as day-evening-night time (Lden) annual average in decibels A (dB(A)). Road traffic noise estimates for LifeLines and HUNT were derived from a new implementation of the Common Noise Assessment Methods in Europe (CNOSSOS-EU) noise model [40]. For KORA, road traffic noise was calculated by the interim calculation method for environmental noise at roads (VBUS) [41], which is based on the German standard RLS-90 (“Richtlinien für den Lärmschutz an Strassen”) [42]. In FINRISK, road traffic noise was estimated in accordance with the Environmental Noise Directive (2002/49/EC) and using the Nordic prediction method [43]. The road traffic noise models typically contain empirically derived equations to determine the initial noise level based on traffic flow and sound propagation based on known environmental factors and physical processes. Quantitative data of road networks, traffic flows, land cover, and building height were obtained from local sources.

statistical analyses

Data were analyzed on the cohort level following a common protocol (described below). Data analyses were undertaken at the University Medical Center Gronin-gen, the Netherlands for LifeLines, HUNT and FINRISK using SPSS (version 22). Analyses for KORA were done locally, using R (version 3.1.0). For each cohort, logistic regression analyses with depressed mood (yes/no) as dependent variable were used to analyze associations between air pollution and depressed mood. Separate regression models were constructed for each of the air pollutants and separately for air pollution estimates derived from the ESCAPE and EU-wide model. Models were firstly adjusted for sex, age, level of education, and household income (minimum confounder model); additionally for MI, asthma, COPD, and ur-

Chapter 4

64

banity (extended confounder model); and finally additionally for road traffic noise (main model). In addition, an alternative confounder model, with adjustment for age, sex, asthma, MI, COPD, and road traffic noise was fitted for the analyses with EU-wide air pollution estimates. This was decided because those analyses included HUNT data for which no data on socioeconomic status and urbanity was available. Data was analyzed for participants from whom we had complete data on all covariates that were available in the cohorts. Effect estimates are presented as odds ratios (OR) for depressed mood, with 95% confidence intervals (CI), per 10 µg/m³ for NO2 and PM10, 1 10-5/m for PM2.5 absorbance, and 5 µg/m³ for PM2.5. The fixed increments were chosen according to ESCAPE and are based on the exposure contrasts to enable broad comparisons between the pollutants [5].

resUlTs

A total of 70,928 participants was included in our analyses (LifeLines n=32,145; KORA n=5,314; HUNT n=32,102; and FINRISK n=1,367). Population characteris-tics are summarized in Table 2. Prevalence of depressed mood ranged from 1.6% (KORA) to 11.3% (FINRISK). Mean age was highest in KORA (55.3 years, standard deviation (SD)=12.3 years) and lowest in LifeLines (43.8 years, SD=11.7 years), and ranged from 18 to 96 years across the four cohorts. All cohorts included more women than men, ranging from 51.5% women in KORA, to 56.9% women in LifeLines (Table 2). Distributions of estimated annual average air pollution

Table 2. Population characteristics in LifeLines, KORA, HUNT and FINRISK

Characteristic LifeLines KORA HUNT FINRISK

N 32,145 5,314 32,102 1,367

Depressed mood (%) 681 (2.1) 87 (1.6) 1,226 (3.8) 155 (11.3)

Age, years (mean±SD) 43.8±11.7 55.3±12.3 54.7±15.3 51.9±14.0

Women (%) 18,276 (56.9) 2,736 (51.5) 17,875 (55.7) 771 (56.4)

Educational level (%)

Primary or secondary 7,743 (24.1) 597 (11.2) NA 673 (49.2)

Post-secondary, non-tertiary 21,745 (67.6) 3,885 (73.1) NA 282 (20.6)

Tertiary 2,657 (8.3) 832 (15.7) NA 412 (30.1)

Household equivalent income, euros/month (mean±SD)

1,551±540a 1,097±569a NA 2,388±1159b

Myocardial infarction (%) 252 (0.8) 128 (2.4) 1,094 (3.4) 22 (1.6)

Asthma (%) 2,607 (8.1) 421 (7.9) 3,745 (11.7) 58 (4.3)

Chronic obstructive pulmonary disease (%) 1,494 (4.6) NA 1,076 (3.4) 25 (1.8)

SD = standard deviation; NA = not available for the cohort.a after taxes b before taxes

65


4

levels, degree of urbanity, and road traffic noise are presented in Figure 1. Median concentrations of NO2 and PM2.5absorbance were highest in KORA (18.8 (inter-quartile range (IQR) 4.9) µg/m3 and 1.66 (IQR 0.21) 10-5/m respectively, based on ESCAPE models), while median PM10 and PM2.5 concentrations were highest

0 10 20 30 40 50 60 70 80 90

LifeLines (ESCAPE)

LifeLines (EU-wide)

KORA (ESCAPE)

KORA (EU-wide)

HUNT (EU-wide)

FINRISK (ESCAPE)

µg/m3

NO2

0 5 10 15 20 25 30 35 40

LifeLines (ESCAPE)

LifeLines (EU-wide)

KORA (ESCAPE)

KORA (EU-wide)

HUNT (EU-wide)

FINRISK (ESCAPE)

µg/m3

PM10

0 5 10 15 20 25

LifeLines (ESCAPE)

KORA (ESCAPE)

FINRISK (ESCAPE)

µg/m3

PM2.5

0 0,5 1 1,5 2 2,5 3

LifeLines (ESCAPE)

KORA (ESCAPE)

FINRISK (ESCAPE)

10-5/m

PM2.5 absorbance

0 2000 4000 6000 8000 10000 12000

LifeLines

KORA

FINRISK

Households/km2

Urbanity

0 10 20 30 40 50 60 70 80 90

LifeLines

KORA

HUNT

FINRISK

Decibel (A)

Road traffic noise (Lden)

figure 1. Distribution of estimated annual average air pollution levels, degree of urbanity, and estimated 24 hour average road traffic noise for LifeLines, KORA, HUNT and FINRISK. Median, 25th and 75th per-centiles are shown in the box, whiskers indicate minimum and maximum estimates. Urbanity is op-erationalized as household density per square kilometer (KORA and FINRISK) or address density per square kilometer (LifeLines).Abbreviations: ESCAPE = European Study of Cohorts for Air Pollution Effects; EU-wide = European wide; NO2 = nitrogen dioxide; PM10 = particulate matter with aerodynamic diameter ≤10 µm; PM2.5 = particulate mat-ter with aerodynamic diameter ≤2.5 µm; PM2.5absorbance = reflectance on PM2.5 filters, i.e. a marker of black carbon; Lden = day-evening-night time annual average road traffic noise.

Chapter 4

66

in LifeLines (23.95 (IQR 0.65) µg/m3 and 15.4 (IQR 0.16) µg/m3 respectively, based on ESCAPE models). Median levels of air pollution and noise were lowest in HUNT (median NO2 11.7 (IQR 5.0) µg/m3 and PM10 11.0 (IQR 1.5) µg/m3, based on EU-wide models; Lden 49.4 (IQR 5.9) dB(A)). As our data from FINRISK only included participants from urban areas, median degree of urbanity was highest for the FINRISK cohort. Correlations between air pollution levels, urbanity and road traffic noise within each cohort are presented in the Supplemental Digital Content, Table S3. For NO2 and PM10 in LifeLines and KORA, correlations between estimated levels from both the ESCAPE and EU-wide models were calculated. NO2 estimates were highly correlated (Spearman’s rho = 0.86 in LifeLines and 0.76 in KORA), while correlations between PM10 estimates were less strong (Spearman’s rho = 0.54 in LifeLines and 0.39 in KORA). Correlations between air pollution levels, urbanity, and road traffic noise ranged from moderate to high, depend-ing on the pollutant and the cohort. Correlations between road traffic noise and air pollution estimates in HUNT were low, probably because traffic data sets for HUNT were less detailed as for the other cohorts.

results from cohort specific logistic regression analyses

Cohort specific associations of air pollution and depression are presented in Table 3. Odds ratios for air pollution levels (ESCAPE and EU-wide models) and de-pressed mood, adjusted for age, sex, education, and household income (minimal confounder model), all indicated higher odds for depressed mood in the LifeLines cohort, except for PM2.5. When regression models were additionally adjusted for asthma, MI, COPD, and urbanity (extended confounder model), associations were no longer statistically significant, except the associations of EU-wide modelled NO2 (OR 1.31, 95% CI 1.06, 1.63) and PM10 (OR 2.92, 95% CI 1.47, 5.79) and depressed mood. Associations between air pollution levels from the EU-wide models and depressed mood in LifeLines remained positive and significant after additional adjustment for road traffic noise (main model), and only a small attenuation of the ORs was observed for NO2 (OR 1.31, 95% CI 1.04, 1.66) and PM10 (OR 2.89, 95% CI 1.43, 5.85) (Table 3). None of the associations in the KORA and FINRISK cohorts were statistically significant. Associations from the main model were negative for ESCAPE modelled NO2 (both cohorts), negative (KORA) and positive (FINRISK) for ESCAPE modelled PM10, and positive for PM2.5 and PM2.5absorbance (both co-horts). A positive (NO2) and a negative (PM10) association was found in KORA for the EU-wide modelled pollutants. In HUNT, NO2 and PM10 (EU-wide models) were significantly associated with lower odds for depressed mood. These associations remained statistically significant after adjustment for age, sex, asthma, myocar-

67


4

table 3. Cohort-specific associations of air pollution and depressed mood.Associations are presented in exposure increments of 10 µg/m3 (NO2 and PM10); 5 µg/m3 (PM2.5); or 1 10-5/m (PM2.5absorbance). LifeLines n=32,145; KORA n=5,314; HUNT n=32,102; FINRISK n=1,367.

Study

Odds ratio (95% confidence interval)

Minimal confounder model

Extended confounder model

Main model Alternative confounder model

NO2 (ESCAPE)

LifeLines 1.49 (1.23, 1.81) 1.09 (0.77, 1.56) 1.04 (0.69, 1.57) NA

KORA 1.18 (0.68, 2.06) 1.16 (0.67, 2.03) 0.76 (0.34, 1.70) NA

FINRISK 0.89 (0.61, 1.29) 0.93 (0.57, 1.52) 0.96 (0.57, 1.63) NA

PM10 (ESCAPE)

LifeLines 4.18 (1.45, 12.04) 0.70 (0.15, 3.31) 0.43 (0.07, 2.62) NA

KORA 1.08 (0.44, 2.64) 0.93 (0.36, 2.37) 0.81 (0.31, 2.13) NA

FINRISK 0.89 (0.51, 1.54) 1.05 (0.50, 2.21) 1.08 (0.50, 2.33) NA

PM2.5 (ESCAPE)

LifeLines 2.47 (0.94, 6.52) 1.20 (0.41, 3.54) 1.04 (0.32, 3.40) NA

KORA 1.60 (0.46, 5.54) 1.54 (0.44, 5.38) 1.06 (0.25, 4.51) NA

FINRISK 1.22 (0.59, 2.51) 1.32 (0.62, 2.80) 1.39 (0.64, 3.05) NA

PM2.5absorbance (ESCAPE)

LifeLines 1.91 (1.13, 3.25) 0.79 (0.37, 1.70) 0.56 (0.22, 1.44) NA

KORA 2.09 (0.62, 7.02) 2.05 (0.61, 6.89) 1.44 (0.33, 6.33) NA

FINRISK 1.18 (0.55, 2.54) 1.30 (0.58, 2.95) 1.44 (0.60, 3.46) NA

NO2 (EU-wide)

LifeLines 1.39 (1.21, 1.60) 1.31 (1.06, 1.63) 1.31 (1.04, 1.66) 1.34 (1.17, 1.53)

KORA 1.11 (0.76, 1.63) 1.23 (0.84, 1.79) 1.01 (0.54, 1.89) 1.01 (0.67, 1.54)

HUNT 0.78 (0.65, 0.93)a 0.77 (0.65, 0.93)b NA 0.79 (0.66, 0.94)

PM10 (EU-wide)

LifeLines 3.35 (2.14, 5.26) 2.92 (1.47, 5.79) 2.89 (1.43, 5.85) 2.66 (1.63, 4.35)

KORA 1.06 (0.26, 4.56) 1.41 (0.33, 6.09) 0.48 (0.06, 3.91) 0.74 (0.16, 3.47)

HUNT 0.38 (0.21, 0.68)a 0.39 (0.21, 0.70)b NA 0.36 (0.20, 0.66)

Abbreviations: ESCAPE = European Study of Cohorts for Air Pollution Effects; EU-wide = European wide; NO2 = nitrogen dioxide; PM10 = particulate matter with aerodynamic diameter ≤10 µm; PM2.5 = particulate mat-ter with aerodynamic diameter ≤2.5 µm; PM2.5absorbance = reflectance on PM2.5 filters, i.e. a marker of black carbon; NA=not available for cohorta Not adjusted for education and household income; b Not adjusted for education, household income, and urban-ityMinimal confounder model: age, sex, education, household incomeExtended confounder model: age, sex, education, household income, asthma, myocardial infarction, COPD, ur-banityMain model: age, sex, education, household income, asthma, myocardial infarction, COPD, urbanity, road traffic noiseAlternative confounder model: age, sex, asthma, myocardial infarction, COPD, road traffic noise

Chapter 4

68

dial infarction, COPD and road traffic noise (alternative confounder model: NO2

OR 0.79, 95% CI 0.66, 0.94; PM10 OR 0.36, 95% CI 0.20, 0.66, Table 3).In summary, statistically significant associations between EU-wide modelled

NO2 and PM10, and depressed mood were observed in LifeLines and HUNT. These pollutants were related to higher odds for depressed mood in LifeLines, but were related to lower odds for depressed mood in HUNT. Since urbanity and road traf-fic noise are exposures that might co-occur with air pollution, we evaluated the effect sizes of these covariates. In the main models, odds ratios for urbanity were consistently 1.000 and statistically non-significant, except for the main models with ESCAPE air pollution in LifeLines. In these models, odds ratios were also 1.000, but were statistically significant. None of the associations between road traffic noise and depressed mood were statistically significant, and odds ratios were close to 1. Small positive odds ratios were observed in LifeLines, HUNT, and KORA, and small negative odds ratios were observed in FINRISK (data not shown).

discUssion

We found no clear evidence for a relation between air pollution and depressed mood. Results were heterogeneous between and sometimes within cohorts, and in some cases sensitive to adjustment for urbanity and road traffic noise. Statisti-cally significant results from the cohort specific analyses were found for NO2 and PM10 (EU-wide models) in LifeLines and HUNT; these air pollutants were however associated with higher (LifeLines) and lower (HUNT) odds for depressed mood. Another notable contrast is that for LifeLines, significant relations with depressed mood were observed with air pollution estimates from the EU-wide models, but not with air pollution estimates from the ESCAPE models.

Strengths of our study include the very large sample size, the standardized analy-ses protocol, and the adjustment of our analyses for road traffic noise. Despite these strengths, our results fit into the literature of inconsistent findings on the association between air pollution and depression. One possible explanation for lack of clear associations is the air pollution level in our cohorts. Studies under-taken in Asia did find relations between short-term air pollution exposure and suicide risk [6], depressive symptoms [7], and emergency department visits for depression [9]. Ambient air pollution levels in Asia are generally much higher than in Europe, which may explain these conflicting results. For example, in one of these studies levels of PM10 were almost twice as high as the highest average concentration in our study (43.7 µg/m3 in Korea vs. 24.2 µg/m3 in the LifeLines

69


4

cohort from the Netherlands) [7]. Notably, the only cohort in which air pollution was significantly associated with increased depression prevalence was the Life-Lines cohort. The findings in the cohorts in which air pollution was not related to depression are consistent with one prior study undertaken in the United States [10]. No evidence for a relation between air pollution and depressive symptoms was found. The authors suggested that higher concentration levels would have been needed to detect am association between air pollution and depressive symptoms [10]. Several studies suggest that an effect of air pollution could be via neuroinflammation, a process proposed to be involved in depression [44,45]. Animal studies showed that inhaled ultrafine particles (<100 nm) and PM2.5 can induce neuroinflammation, either by entering the brain directly, or by inducing immune mediators which reach the brain [46–50]. Observations in human sub-jects highly exposed to air pollution showed that air pollution components may reach the brain and cause neuroinflammation in humans as well [51].

Our study did not provide a clear answer to the question regarding the asso-ciation between air pollution and depressed mood. However, the heterogeneous results have important implications for the interpretation of other studies, by showing how the same question approached with the same statistical analytical strategy can lead to different answers. One obvious source of this heterogene-ity is differences in the populations studied, including genetic variations, or other cohort- or region-specific differences that we did not assess in this study. However, we identified heterogeneity also within the cohort specific results, as-sociated with the exposure modeling. Associations between NO2 and PM10 from EU-wide models and depressed mood in LifeLines were consistently positive and significant, but associations between these same pollutants from ESCAPE models in LifeLines were not significant. The main difference between the ESCAPE and EU-wide models is that ESCAPE models were constructed locally, whereas the EU-wide models were developed for a much larger area in Western Europe. Fur-thermore, ESCAPE models used more local predictors (e.g. traffic counts) than the EU-wide models, but ESCAPE as well as EU-wide models have shown to explain a large fraction of the spatial variability in annual average air pollution concentra-tions [21–23]. However, NO2 and PM10 estimates from EU-wide models had larger variation among the LifeLines participants than NO2 and PM10 estimates from ESCAPE models. We speculate that the larger variation in the EU-wide estimated air pollution exposures may be an explanation for the discrepancy in results from the ESCAPE and EU-wide models in LifeLines.

We adjusted our analyses for urbanity and road traffic noise because these exposures may co-occur with air pollution. However, overfitting of the statistical models may occur, especially for urbanity which is often used as a predictor vari-

Chapter 4

70

able in LUR models for estimation of air pollution concentrations. LUR models that used population density as a predictor were NO2 (ESCAPE: LifeLines and KORA) and PM10 (ESCAPE: LifeLines, FINRISK; and EU-wide). However, inspec-tion of the variance inflation factors indicated no multicollinearity. Although to a lesser extent, the same may be applicable to road traffic noise.

A general limitation that applies to all air pollution models is that they provide estimates of exposure at each participant’s residential address. Since we have no data on the participants daily mobility and workplace exposure, this could have led to bias due to exposure misclassification. Such bias may have underestimated the relation between air pollution and depressed mood. Our study may not have detected potential short-term effects, since long-term (annual average) air pollu-tion was modelled. Short-term effects could be studied with time-series or panel studies investigating short- or intermediate-term associations. An additional source of heterogeneity is related to the outcome measure. The cohorts included various depression measures, which was probably reflected in the varying de-pression prevalence, with 1.6% in KORA and 11.3% in FINRISK. Although it is conceivable that depression prevalence differs between countries [52], compar-ing face-to-face interviews with questionnaires must have undoubtedly played a role in these varying prevalence rates. Outcome misclassification may be present in our study, for example due to the cut offs used to determine depressed mood in this study. We used higher cut offs than usual for the HADS and CES-D in order to harmonize with the other depression measures in our study. Sensitivity analyses were undertaken in HUNT and FINRISK to investigate whether different cut offs of the HADS and CES-D changed the results. Using the cut off 8 for HADS in HUNT and cut off 16 for CES-D in FINRISK did not change the overall conclusions (results available upon request). Apart from the main outcome, heterogeneity might also be related with the fact that not all covariates were available for all cohorts, and also the available data were heterogeneous. We did not have access to data about socioeconomic status and degree of urbanity from the HUNT cohort, making it not possible to adjust the HUNT analyses for these factors. For FINRISK, we only had data on household income before taxes. A limitation that applies to all cohorts is that we did not take into account the use of antidepressants. Participants that use antidepressants, may have less symptoms of depression, and may not be classified as having depressed mood because they have a lower score for the depression as-sessment. This might have led to an underestimation of the association between air pollution exposure and depressed mood.

In this large multi-cohort study, we found no consistent evidence for an associa-tion between ambient air pollution and depressed mood. Regardless of whether

71


4

there is a true effect of air pollution on depressed mood, our study highlights the importance of multi-cohort studies, where results from multiple cohorts are used for answering research questions. The heterogeneous results observed in this study illustrate how various study methods can lead to various results. Had the relation of air pollution and depressed mood been studied in only one cohort, conclusions might have been different. Moreover, inclusion of data from additional cohorts might have changed our conclusions. Given these contradicting results, investigation of the relation between air pollution and depressed mood in cohort studies from other geographical areas is needed.

Chapter 4

72

references


2. Gehring U, Gruzieva O, Agius RM, Beelen R, Custovic A, Cyrys J, et al. Air Pollution Exposure and Lung Function in Children: The ESCAPE Project. Environ Health Perspect. 2013; doi:10.1289/ehp.1306770

3. Cesaroni G, Forastiere F, Stafoggia M, Andersen ZJ, Badaloni C, Beelen R, et al. Long term expo-sure to ambient air pollution and incidence of acute coronary events: prospective cohort study and meta-analysis in 11 European cohorts from the ESCAPE Project. BMJ. 2014;348: f7412. doi:10.1136/bmj.f7412

4. Fischer PH, Marra M, Ameling CB, Hoek G, Beelen R, de Hoogh K, et al. Air Pollution and Mortal-ity in Seven Million Adults: The Dutch Environmental Longitudinal Study (DUELS). Environ Health Perspect. 2015; doi:10.1289/ehp.1408254


6. Kim C, Jung SH, Kang DR, Kim HC, Moon KT, Hur NW, et al. Ambient particulate matter as a risk fac-tor for suicide. Am J Psychiatry. 2010;167: 1100–1107. doi:10.1176/appi.ajp.2010.09050706; 10.1176/appi.ajp.2010.09050706


8. Szyszkowicz M, Rowe BH, Colman I. Air pollution and daily emergency department visits for depression. Int J Occup Med Environ Health. 2009;22: 355–362. doi:10.2478/v10001-009-0031-6

9. Cho J, Choi YJ, Suh M, Sohn J, Kim H, Cho S-K, et al. Air pollution as a risk factor for depressive episode in patients with cardiovascular disease, diabetes mellitus, or asthma. J Affect Disord. 2014;157: 45–51. doi:http://dx.doi.org/10.1016/j.jad.2014.01.002

10. Wang Y, Eliot MN, Koutrakis P, Gryparis A, Schwartz JD, Coull BA, et al. Ambient Air Pollution and Depressive Symptoms in Older Adults: Results from the MOBILIZE Boston Study. Environ Health Perspect. 2014;122: 553–8. doi:10.1289/ehp.1205909


12. Lamers F, van Oppen P, Comijs HC, Smit JH, Spinhoven P, van Balkom AJLM, et al. Comorbid-ity patterns of anxiety and depressive disorders in a large cohort study: the Netherlands Study of Depression and Anxiety (NESDA). J Clin Psychiatry. 2011;72: 341–8. doi:10.4088/JCP.10m06176blu

13. Guxens M, Sunyer J. A review of epidemiological studies on neuropsychological effects of air pollution. Swiss Med Wkly. 2012;141: w13322. doi:10.4414/smw.2011.13322


73


4



17. Krokstad S, Langhammer A, Hveem K, Holmen TL, Midthjell K, Stene TR, et al. Cohort Profile: the HUNT Study, Norway. Int J Epidemiol. 2013;42: 968–77. doi:10.1093/ije/dys095

18. Holle R, Happich M, Löwel H, Wichmann HE. KORA—a research platform for population based health research. Gesundheitswesen. 2005;67 Suppl 1: S19–25. doi:10.1055/s-2005-858235

19. Konttinen H, Männistö S, Sarlio-Lähteenkorva S, Silventoinen K, Haukkala A. Emotional eating, depressive symptoms and self-reported food consumption. A population-based study. Appe-tite. 2010;54: 473–9. doi:10.1016/j.appet.2010.01.014

20. Eeftens M, Beelen R, Fischer P, Brunekreef B, Meliefste K, Hoek G. Stability of measured and modelled spatial contrasts in NO(2) over time. Occup Environ Med. 2011;68: 765–770. doi:10.1136/oem.2010.061135; 10.1136/oem.2010.061135




24. Cyrys J, Eeftens M, Heinrich J, Ampe C, Armengaud A, Beelen R, et al. Variation of NO2 and NOx concentrations between and within 36 European study areas: Results from the ESCAPE study. Atmos Environ. 2012;62: 374–390. doi:10.1016/j.atmosenv.2012.07.080

25. Eeftens M, Tsai M-Y, Ampe C, Anwander B, Beelen R, Bellander T, et al. Spatial variation of PM2.5, PM10, PM2.5 absorbance and PMcoarse concentrations between and within 20 European study areas and the relationship with NO2 – Results of the ESCAPE project. Atmos Environ. 2012;62: 303–317. doi:10.1016/j.atmosenv.2012.08.038

26. Sheehan D V, Lecrubier Y, Sheehan KH, Amorim P, Janavs J, Weiller E, et al. The Mini-Interna-tional Neuropsychiatric Interview (M.I.N.I.): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. J Clin Psychiatry. 1998;59 Suppl 2: 22–33;quiz 34–57.

27. Kroenke K, Spitzer RL, Williams JB, Lowe B. The Patient Health Questionnaire Somatic, Anxiety, and Depressive Symptom Scales: a systematic review. Gen Hosp Psychiatry. 2010;32: 345–359. doi:10.1016/j.genhosppsych.2010.03.006

28. Mykletun A, Stordal E, Dahl AA. Hospital Anxiety and Depression (HAD) scale: factor structure, item analyses and internal consistency in a large population. Br J Psychiatry. 2001;179: 540–4.

29. Bjelland I, Dahl AA, Haug TT, Neckelmann D. The validity of the Hospital Anxiety and Depres-sion Scale. An updated literature review. J Psychosom Res. 2002;52: 69–77.

Chapter 4

74

30. Crawford JR, Henry JD, Crombie C, Taylor EP. Normative data for the HADS from a large non-clinical sample. Br J Clin Psychol. 2001;40: 429–34.

31. Radloff LS. The CES-D Scale: A Self-Report Depression Scale for Research in the General Popula-tion. Appl Psychol Meas. 1977;1: 385–401. doi:10.1177/014662167700100306

32. Cloninger CR, Svrakic DM, Przybeck TR. Can personality assessment predict future depression? A twelve-month follow-up of 631 subjects. J Affect Disord. 2006;92: 35–44. doi:10.1016/j.jad.2005.12.034

33. Cesaroni G, Forastiere F, Stafoggia M, Andersen ZJ, Badaloni C, Beelen R, et al. Long term expo-sure to ambient air pollution and incidence of acute coronary events: prospective cohort study and meta-analysis in 11 European cohorts from the ESCAPE Project. BMJ. 2014;348: f7412.

34. Jerrett M, Shankardass K, Berhane K, Gauderman WJ, Künzli N, Avol E, et al. Traffic-related air pollution and asthma onset in children: a prospective cohort study with individual exposure measurement. Environ Health Perspect. 2008;116: 1433–8. doi:10.1289/ehp.10968

35. Schikowski T, Mills IC, Anderson HR, Cohen A, Hansell A, Kauffmann F, et al. Ambient air pollu-tion: a cause of COPD? Eur Respir J. 2014;43: 250–63. doi:10.1183/09031936.00100112

36. Evans DL, Charney DS, Lewis L, Golden RN, Gorman JM, Krishnan KRR, et al. Mood disorders in the medically ill: scientific review and recommendations. Biol Psychiatry. 2005;58: 175–89. doi:10.1016/j.biopsych.2005.05.001



39. OECD. Household disposable income (indicator). In: OECD Data [Internet]. 2015 [cited 30 Oct 2015]. doi:10.1787/dd50eddd-en


41. Bundesanzeiger. Bekanntmachung der Vorläufigen Berechnungsverfahren für den Umgebung-slärm nach § 5 Abs. 1 der Verordnung über die Lärmkartierung (34. BImSchV). 2006.

42. RLS90. RLS90 (Richtlinien für Lärmschutz an Straßen-RLS 90). In: Amtsblatt des Bundesminis-ters für Verkehr der Bundesrepublik Deutschland (VkBl.) [Internet]. 1990 [cited 16 Oct 2015]. Available: http://www.gesetze-im-internet.de/bimsch v_16/BJNR010360990.html

43. Bendtsen H. The Nordic prediction method for road traffic noise. Sci Total Environ. 1999;235: 331–338. doi:10.1016/S0048-9697(99)00216-8

44. Hurley LL, Tizabi Y. Neuroinflammation, neurodegeneration, and depression. Neurotox Res. 2013;23: 131–144. doi:10.1007/s12640-012-9348-1

45. Dantzer R, O’Connor JC, Freund GG, Johnson RW, Kelley KW. From inflammation to sickness and depression: when the immune system subjugates the brain. Nat Rev Neurosci. 2008;9: 46–56. doi:10.1038/nrn2297

46. Fonken LK, Xu X, Weil ZM, Chen G, Sun Q, Rajagopalan S, et al. Air pollution impairs cognition, provokes depressive-like behaviors and alters hippocampal cytokine expression and morphol-ogy. Mol Psychiatry. 2011;16: 973,987–995. doi:10.1038/mp.2011.76

75


4

47. Oberdorster G, Sharp Z, Atudorei V, Elder A, Gelein R, Kreyling W, et al. Transloca-tion of inhaled ultrafine particles to the brain. Inhal Toxicol. 2004;16: 437–445. doi:10.1080/08958370490439597

48. Peters A, Veronesi B, Calderon-Garciduenas L, Gehr P, Chen LC, Geiser M, et al. Translocation and potential neurological effects of fine and ultrafine particles a critical update. Part Fibre Toxicol. 2006;3: 13. doi:10.1186/1743-8977-3-13

49. Block ML, Calderón-Garcidueñas L. Air pollution: mechanisms of neuroinflammation and CNS disease. Trends Neurosci. 2009;32: 506–516. doi:10.1016/j.tins.2009.05.009

50. Terzano C, Di Stefano F, Conti V, Graziani E, Petroianni A. Air pollution ultrafine particles: toxic-ity beyond the lung. Eur Rev Med Pharmacol Sci. 2010;14: 809–821.

51. Calderon-Garciduenas L, Solt AC, Henriquez-Roldan C, Torres-Jardon R, Nuse B, Herritt L, et al. Long-term air pollution exposure is associated with neuroinflammation, an altered innate immune response, disruption of the blood-brain barrier, ultrafine particulate deposition, and accumulation of amyloid beta-42 and alpha-synuclein in children and youn. Toxicol Pathol. 2008;36: 289–310. doi:10.1177/0192623307313011

52. Wittchen H-U, Jacobi F. Size and burden of mental disorders in Europe—a critical review and appraisal of 27 studies. Eur Neuropsychopharmacol. 2005;15: 357–76. doi:10.1016/j.euro-neuro.2005.04.012

Chapter 4

76

chAPter 4 suPPleMentAl MAteriAl

description of escAPe land use regression (lur) models for ambient air pollution

Exposure to ambient air pollution was estimated using land use regression (LUR) models developed for the European Study of Cohorts for Air Pollution Effects (ESCAPE). Within the ESCAPE project, LUR models for 38 European study areas were developed using a standardized approach. ESCAPE LUR models were devel-oped for NO2 (nitrogen dioxide), NO2 background, PM2.5 (particulate matter with a diameter ≤2.5 µm), PM2.5 absorbance (reflectance on PM2.5 filters, i.e. a marker of black carbon), and PM10 (particulate matter with a diameter ≤10 µm). A har-monized measurement campaign was conducted in each of the study areas dur-ing 2008-2011. In each study area 40 NOx and 20 additional PM measurements were conducted. Measurements were undertaken in three two-week periods in the cold, warm and intermediate season. For each measurement site the annual average concentration was calculated, with adjustment for temporal variation using measurements from centrally located reference sites with year-round measurement data. The air pollution concentrations obtained from the measure-ment campaign were then used as outcome variables for LUR model development for each of the areas. GIS (geographic information system) derived variables (e.g. distance to nearest road, traffic intensity, built-up land, population density, altitude) were used as predictor variables, to explain these measured concentra-tions. LUR models were developed locally, but followed a standardized protocol. A supervised forward stepwise procedure was used with pre-specified predictor variables. First, univariate regression models with all potential predictor variables were conducted, and the predictor with the highest adjusted explained variance was selected to be included in the model. Subsequently, all remaining predictor variables were added to the model consecutively, while the change in adjusted explained variances was recorded. The variable with the highest increase in adjusted explained variance was retained if the associated regression coefficient was in the pre-specified direction of the effect; if the additional predictor vari-ables increased the adjusted explained variance by >1%; if the direction of the regression coefficient for the already included predictors remained the same. The addition of variables was repeated until there were no variables that added more than 1% to the adjusted explained variance of the previous model. Finally, vari-ables with p-values larger than 0.1 were sequentially removed from the model. For the final models, diagnostic tests were applied to assess model fit (Variance Inflation Factors, Cook’s D, Moran’s I). Leave-one-out cross validation (LOOCV) was used to evaluate model performance. The final model was fitted to all sites,

77


4

but one, while leaving the included variables constant. The predicted air pollution concentrations from the model were compared with the concentrations from the measurement campaign at the left-out site. The procedure was repeated for all sites, and the adjusted explained variance between the predicted and measured concentrations was calculated as for evaluation of model performance. NO2, PM2.5, and PM2.5absorbance LUR models for the models included in the current study included at least one traffic variable (e.g. traffic intensity). NO2 and PM10 LUR models included indicators of population density, while these were not included in the PM2.5, and PM2.5absorbance LUR models (Supplemental Material Table S1). Adjusted explained variances for the LUR models included in the current study ranged between 83-86% (NO2), 67-83% (PM10), 67-88% (PM2.5), and 65-92% (PM2.5absorbance) [21,22].

description of eu-wide land use regression (lur) model for ambient air pollution

Exposure to ambient air pollution was estimated using European (EU)-wide LUR models enhanced with satellite derived air pollution estimates. These EU-wide models incorporate GIS-derived land use, road network, and topographic data, as well as satellite-derived estimates of ground level concentrations for PM2.5 (as a proxy for PM10 because PM10 satellite measurements were not available) and NO2

[23]. Model development follows the ESCAPE procedure to construct the multiple linear regression equations. In these multiple linear regression equations, annual mean ambient concentrations of NO2 and PM10 (years 2005-2007) obtained from regulatory monitoring were used as dependent variables. Independent variables (predictors), including GIS-derived land use, road network, and topographic data, as well as satellite-derived estimates of ground level concentrations for PM2.5 and NO2 were used to construct the multiple linear regression equations. Models were developed according to the ESCAPE supervised stepwise selection of predictor variables. Models were evaluated against measured PM10 and NO2 concentrations at an independent subset of 20% sites that were reserved for this purpose. For NO2, all models included satellite-derived surface NO2, the length of minor roads, length of major roads, percentage total built up land, and percentage semi natural land (Supplement Material, Table S2). NO2 LUR models explained 55-60% of the variation in ambient NO2 concentrations based on regulatory monitoring. For PM10, all models included satellite-derived PM2.5, the Y coordinate (because of a general decreasing trend in PM10 concentrations from south to north), and the length of major roads. PM10 LUR models explained 38-47% of the variation in ambient PM10 concentrations based on regulatory monitoring [23].

Chapter 4

78

Tab

le s

1. P

redi

ctor

var

iabl

es fo

r ESC

APE

land

use

regr

essi

on m

odel

s (ba

sed

on [2

1,22

]).

NO

2PM

10PM

2.5

PM2.

5abs

orba

nce

Life

Line

sKO

RAFI

NRI

SKLi

feLi

nes

KORA

FIN

RISK

Life

Line

sKO

RAFI

NRI

SKLi

feLi

nes

KORA

FIN

RISK

Traffi

c lo

ad_2

5x

Traffi

c lo

ad_2

5_10

0x

Traffi

c lo

ad_5

0x

xx

x

Traffi

c lo

ad_5

00x

Hea

vy tr

affic

load

_25

x

Hea

vy tr

affic

load

_25_

500

x

Hea

vy tr

affic

load

_50

x

Traffi

c m

ajor

load

_50

xx

Traffi

c m

ajor

load

_500

x

Traffi

c m

ajor

load

_100

0x

xx

Dis

tinvn

earc

1x

Intm

ajor

invd

isx

Road

leng

th_2

5x

Road

leng

th_2

5_30

0x

Road

leng

th_5

0x

xx

Road

leng

th_3

00x

Road

leng

th_1

000

x

Maj

or ro

ad le

ngth

_50

xx

xx

xx

Urb

an g

reen

_500

x

Urb

an g

reen

_500

0x

Nat

ural

_100

x

Nat

ural

_500

xx

x

Popu

latio

n_50

00x

xx

Hou

seho

lds_

100

x

Hld

_res

_500

x

Hld

_res

_500

0x

Regi

onal

est

imat

e N

O2

x

Regi

onal

est

imat

e PM

2.5

x

Regi

onal

est

imat

e PM

2.5 a

bsor

banc

ex

Expl

anat

ion

of th

e va

riab

le n

ames

(som

e va

riab

les a

re b

uffe

rs w

ith _X

indi

catin

g th

e ra

dius

of t

he b

uffe

r in

met

ers)

:Th

e sur

face

area

(m2 ) o

f all

resi

dent

ial l

and

(hld

_res

_x),

urba

n gr

een

spac

e (ur

ban

gree

n_X)

, nat

ural

land

(nat

ural

_X);

popu

latio

n (N

) (po

pula

tion_

X) o

r num

ber (

N) o

f hou

seho

lds (

hous

ehol

ds_X

); a

regi

onal

con

cent

ratio

n es

timat

e (μ

g/m

3 or

10−

5m−1

); to

tal l

engt

h (m

) of

all

road

and

all

maj

or r

oad

segm

ents

(ro

ad le

ngth

_x, m

ajor

roa

d le

ngth

_x);

inve

rse

dist

ance

(m

−1)

to th

e ne

ar-

est r

oad

of th

e ce

ntra

l roa

d ne

twor

k (d

istin

vnea

rc1)

; the

pro

duct

of i

nver

se/i

nver

se s

quar

ed d

ista

nce

to th

e ne

ares

t maj

or r

oad

and

the

traf

fic in

tens

ity o

n th

is r

oad

(veh

icle

s·da

y−1m

−1/

vehi

cles

·day

−1m

−2) (

intm

ajor

invd

ist)

; the

sum

of (

traf

fic in

tens

ity ×

the

leng

th o

f all

road

seg

men

ts) w

ithin

a b

uffe

r (v

ehic

les·

day−

1·m

) for

all

road

s (t

raffi

c lo

ad_x

), fo

r m

ajor

road

s (t

raffi

c m

ajor

load

_X),

and

for h

eavy

traf

fic (h

eavy

traf

fic lo

ad_X

).

79


4

Table s2. Predictor variables for EU-wide land use regression models (based on [23])

NO2 PM10

2005 2006 2007 2005-2007 2005 2006 2007 2005-2007

Minor road length_200 x

Minor road length_200-2500 x

Minor road length_1500 x x x

Minor road length_1800m x

Minor road length_1800-10000 x

Minor road length_1500-10000 x x x

Major road length_100 x x x x x x

Major road length x x

Population_1800 x x

Impervious surface_1000 (% area) x

Total built up land_300 x x x x

Total built up land_400 x

Total built up land_600 x

Semi-natural land_500 x

Semi-natural land_600 x x x



Tree canopy 100m (% area) x

Tree canopy 500m (% area) x

Satellite-derived surface NO2 2005 x



Satellite-derived surface NO2 2005-2007 x

Satellite-derived surface PM2.5 2001 x x x x

Altitude x

Distance to sea x x x

Y coordinate x x x x

Explanation of the variable names (some variables are buffers with _X indicating the radius of the buffer in meters):Total built up land (% area): residential areas, industrial areas, ports, transport infrastructure, airports, mines, dumps and construction sites.

Chapter 4

80

tab

le s

3. S

pear

man

’s co

rrel

atio

ns b

etw

een

air p

ollu

tion

leve

ls, u

rban

ity a

nd ro

ad tr

affic

noi

se

Spea

rman

’s rh

o

NO

2 ESC

APE

NO

2 EU

-wid

ePM

10 E

SCA

PEPM

10 E

U-w

ide

PM2.

5 ESC

APE

PM2.

5abs

ESC

APE

L den

LLKO

FRLL

KOH

ULL

KOFR

LLKO

HU

LLKO

FRLL

KOFR

LLKO

FR

NO

2 EU

-wid

e0.

860.

76N

A

PM10

ESC

APE

0.74

0.67

0.67

0.67

0.38

PM10

EU

-wid

e0.

770.

68N

A0.

770.

850.

780.

540.

39N

A

PM2.

5 ES

CAPE

0.52

0.44

0.45

0.53

0.33

NA

0.71

0.43

0.69

0.51

0.29

PM2.

5abs

ESC

APE

0.79

0.67

0.49

0.71

0.35

NA

0.93

0.66

0.72

0.58

0.34

NA

0.67

0.48

0.99

L den

0.56

0.43

0.36

0.43

0.37

-0.0

60.

570.

290.

210.

380.

310.

060.

460.

380.

240.

610.

460.

31

Urb

anit

y0.

880.

690.

660.

920.

86N

A0.

700.

340.

670.

780.

76N

A0.

520.

260.

210.

710.

280.

250.

420.

260.

16

Abbr

evia

tions

: ESC

APE

= Eu

rope

an S

tudy

of C

ohor

ts fo

r Air

Pol

lutio

n Ef

fect

s; E

U-w

ide

= Eu

rope

an w

ide;

NO 2

= n

itrog

en d

ioxi

de; P

M10

= p

artic

ulat

e m

atte

r with

aer

odyn

amic

di

amet

er ≤

10 µ

m; P

M2.

5 = p

artic

ulat

e m

atte

r with

aer

odyn

amic

dia

met

er ≤

2.5

µm; P

M2.

5abs

= re

flect

ance

on

PM2.

5 fil

ters

, i.e

. a m

arke

r of b

lack

car

bon;

Lde

n = 2

4 ho

ur a

vera

ge

road

traf

fic n

oise

; LL

= Li

feLi

nes;

KO

= KO

RA; H

U =

HUN

T; F

R =

FIN

RISK

; NA

= no

t ava

ilabl

e fo

r coh

ort.

5How to assess common somatic symptoms in large-scale studies: a systematic review of questionnaires

Wilma L. Zijlema, Ronald P. Stolk, Bernd Löwe, Winfried Rief, BioSHaRE, Peter D. White, Judith G.M. Rosmalen

Journal of Psychosomatic Research 2013 Jun;74(6):459-68

Chapter 5

84

aBsTracT

objective

Many questionnaires for assessment of common somatic symptoms or functional somatic symptoms are available and their use differs greatly among studies. The prevalence and incidence of symptoms are partially determined by the methods used to assess them. As a result, comparison across studies is difficult. This article describes a systematic review of self-report questionnaires for somatic symptoms for use in large-scale studies and recommends two questionnaires for use in such studies.

methods

A literature search was performed in the databases Medline, PsycINFO and EMBASE. Articles that reported the development, evaluation, or review of a self-report somatic symptom measure were included. Instrument evaluation was based on validity and reliability, and their fitness for purpose in large scale stud-ies, according to the PhenX criteria.

results

The literature search identified 40 questionnaires. The number of items within the questionnaires ranged from 5 to 78 items. In 70% of the questionnaires, headaches were included, followed by nausea/upset stomach (65%), shortness of breath/breathing trouble (58%), dizziness (55%), and (low) back pain/backaches (55%). Data on validity and reliability were reported and used for evaluation.

conclusion

Questionnaires varied regarding usability and burden to participants, and rel-evance to a variety of populations and regions. Based on our criteria, the Patient Health Questionnaire-15 and the Symptom Checklist-90 somatization scale seem the most fit for purpose for use in large-scale studies. These two questionnaires have well-established psychometric properties, contain relevant symptoms, are relatively short, and are available in multiple languages.

85

Systematic review of symptom questionnaires

5

inTrodUcTion

A symptom is a self-reported bodily sensation or mental experience that is per-ceived by a person as a change from normal health [1]. A British study showed that the mean number of symptoms reported in the general population is between three and four in the last two weeks [2] and a recent Norwegian study found a strong association between the number of symptoms and functional status [3]. Symptoms may be mediated by a change in bodily function or may be associated with disease. It is commonly the case that a symptom cannot be conclusively explained by organic pathology. These symptoms are referred to as functional somatic symptoms (FSS). The more symptoms reported, the more likely symp-toms are functional in nature [4,5]. Patients who frequently complain of physical symptoms that either lack a demonstrable organic basis, or that are judged to be in excess of what would be expected based on medical findings, are thought to be suffering from the process of somatization. This view is qualified by the knowledge that future research may find medical explanations for some of these FSS. Somatization refers to a tendency to experience and communicate somatic distress in response to psychosocial stress and seek medical help for it [6]. Func-tional somatic symptoms are common [7-9], disabling [10-12], costly [13], and patients often feel misunderstood, guilty and even ashamed [14].

The occurrence of symptoms is often assessed by using a self-report symp-tom questionnaire. Many different questionnaires are available and the use of these questionnaires differs greatly among current studies, complicating the comparison of studies. Firstly, the questionnaires differ greatly in the number of symptoms questioned. Secondly, there is a large variety in the type of symptoms included in the questionnaires. Previous studies have suggested that certain types of symptoms cluster together [15-17]. Although not found in all studies, the following four clusters are commonly reported: cardiopulmonary (including autonomic symptoms), gastrointestinal, musculoskeletal, and general symptoms. Thirdly, some of these questionnaires assess symptoms in general, while oth-ers focus on medically unexplained symptoms. Being certain that a symptom is medically unexplained can be difficult, has low inter-rater reliability [18,19], and would be impossible to assess in large-scale studies. Fourthly, the time frame of assessment varies largely. Some questionnaires are based on life-time symptoms; but several researchers suggest that recall of lifetime symptoms is unreliable and inconsistent [20-22]. Others address time frames of between a week and a month. Fifthly, some questionnaires enquire only about the categorical presence of symp-toms, while others inquire about symptom severity; both symptom diversity and severity may be important [9].

Chapter 5

86

Given the heterogeneity of scales with respect to content, scaling, and di-mensionality, severity scores from different FSS or somatic symptom scales are incomparable. Using cut-off-scores, each scale identifies a unique subgroup that differs in various aspects from subgroups identified by other scales. It would be very useful to have a gold standard measure for the assessment of both FSS and somatic symptoms in general in large-scale studies. Agreement regarding the use of such an instrument can facilitate systematic comparisons or meta-analytical studies and thereby contribute to the understanding of the etiology of functional somatic symptoms.

To the best of our knowledge, an overview of the currently available symptom questionnaires has not been reported. This article describes a systematic review of self-report questionnaires for common somatic symptoms for use in large-scale studies. It will conclude with a recommendation for which symptom question-naires are the best to use. This recommendation will be based on validity and reliability on the one hand, and applicability in large-scale studies on the other hand. The first aspects will be evaluated in terms of number and types of symp-toms included, the response scale, time frame covered, and data on validity and reliability. The second aspects will be evaluated using the PhenX (Phenotypes and eXposures) criteria, according to which a measure should be well-established; easy in its use; of low burden to participants; relevant for future use; and ap-plicable to a variety of populations and regions [23].

meTHods

search strategy

A literature search was performed in the databases Medline, EMBASE and PsycIN-FO on the 15th of October 2012. A search term was formulated for searching the databases, which contained a combination of somatoform disorder or synonyms and questionnaire or synonyms and symptoms. For Medline, the following search term was used: (“somatoform disorders/classification”[MeSH Major Topic] OR “somatoform disorders/diagnosis”[MeSH Major Topic] OR “somatoform dis-orders/epidemiology”[MeSH Major Topic] OR “functional somatic symptoms” [Title/Abstract]) AND (questionnaire[Title/Abstract] OR screen*[Title/Abstract] OR “self report”[Title/Abstract] OR “index”[Title/Abstract]) AND symptoms. For EMBASE and PsycINFO comparable search terms were used. The search was conducted without language restrictions. Personal files of the authors were also reviewed for relevant articles. Additional searches were performed using the

87


5

search engine Google and the authors of the questionnaires were contacted to obtain supplemental information of the symptom questionnaires, if needed.

screening and selection procedure

The titles and abstracts of the retrieved articles were screened by two indepen-dent researchers. The articles were chosen for the development, evaluation, or review of somatic symptom or somatization questionnaires. Additionally, the measure had to be a self-report symptom checklist. Interviews were excluded, as these are not suitable for use in large studies. Furthermore, the questionnaires chosen had to include symptoms from more than one symptom cluster; not just symptoms of the gastrointestinal tract or cardiopulmonary system. When the symptom questionnaire was a sub-scale derived from a larger questionnaire, the symptom subscale had to have been separately validated and used. There were no criteria for the target population of the questionnaire. Discrepancies between the two researchers were resolved by consensus. Full articles were then obtained for all included studies. Based on the full text, articles that still fulfilled the inclusion criteria were included in the review.

data extraction

Authors, year of publication, name of questionnaire, purpose of questionnaire, questionnaire instructions, list of symptoms, answering scale, and language of questionnaire, were extracted for every questionnaire. Data extraction from papers describing the validation of a questionnaire also included validity data, characteristics about the population used (clinical or general, gender and age dis-tribution, nationality or race), and number of participants. Data extraction from articles written in languages other than English was done by native speakers.

instrument evaluation

The questionnaires were evaluated according to the following criteria; the first set of criteria concerned the validity and reliability of the instrument. Firstly, we examined the type of symptoms included. We assumed that the proportion of questionnaires including a specific symptom reflected expert knowledge on the importance of that specific symptom for the underlying construct. To ensure that the questionnaire was not too restrictive with regard to the type of symptom, we evaluated whether the questionnaires included at least one symptom from each of the following symptom clusters identified in previous studies [15-17]: cardiopulmonary (including autonomic symptoms), gastrointestinal, musculosk-eletal, and general symptoms. Since being certain that a symptom is medically unexplained can be difficult, has low inter-rater reliability [18,19], and would be

Chapter 5

88

impossible to assess in large-scale studies, we proposed that self-report symptom questionnaires should preferably question symptoms in general, as opposed to medically unexplained symptoms. Secondly, we noted the time frame covered by the questionnaire. Studies have shown that the recall of lifetime symptoms is unreliable [20-22]. We therefore do not recommend the use of lifetime as a recall period. Thirdly, we assessed the response scale. We looked at the format of the questionnaires’ responses: severity, frequency, and the number of response categories. Fourthly, we recorded the psychometric characteristics. The internal consistency was assessed, which reflects whether items in a questionnaire are correlated, thus measuring the same concept. The factor structure of a question-naire indicates which symptom clusters are present within the questionnaire. Correlations of the symptom questionnaire with other related constructs, for example health anxiety and illness behavior, also inform the validity of the ques-tionnaire. In addition, the test-retest reliability of a questionnaire is an important indicator for the stability of the questionnaire. It is likely that symptom reporting fluctuates over time. Therefore, short time intervals for test-retest reliability would be most appropriate, and an interval of 3-4 weeks is commonly used [24]. We therefore chose to report the test-retest reliability for time intervals no longer than 1 month.

The second set of criteria concerns the applicability to large-scale studies, which we based on the PhenX (Phenotypes and eXposures) criteria [23]. PhenX is an initiative that aims to reach consensus about which measures should be used in large-scale genomic studies, but also applies to other large-scale studies. Within the PhenX initiative several criteria for the selection of well-established, low-burden, high-quality measures were developed. Firstly, the questionnaire should be a well-established measure (used repeatedly over time and coming from highly regarded sources). Secondly, the questionnaire can be used by scientists who do not have expertise in that specific domain. Thirdly, the questionnaire should be a relatively low burden to participants and investigators. Fourthly, the question-naire should be relevant for the near future. Fifthly, the questionnaire should be applicable to a variety of populations (different ethnic groups, ages and gender); and sixthly, the questionnaire should not be constricted to a particular country/region [23].

89


5

resUlTs

search results

The search retrieved a total of 1214 articles. 1136 articles were excluded, of which 127 were duplicates; 928 were not about FSS, somatization or were not self-report questionnaires; 5 were about questionnaires with a relevant subscale, but the subscale was not separately validated and used; and 76 were about question-naires already included, but contained no validity data. Additionally, 21 articles were added from our personal files. Thus, 99 articles were included, describing 40 questionnaires (figure 1).

Part 1: validity and reliability

Evaluation of questionnaires

Table 1 shows the types of questionnaires that were reported within the 99 in-cluded articles. The 40 questionnaires showed considerable overlap in their aims

figure 1. Flow chart of systematic review process

Chapter 5

90

and sometimes even in their abbreviations (e.g., the PHQ and the BSI each refer to two different instruments). Furthermore, some of the questionnaires are adapted versions of other questionnaires. For example, the SOMS has several versions which differ in the number of items and the time frame for symptom recall, and the BSI-6 is a shortened version of the SCL-90 SOM. The majority of questionnaires assessed the frequency of symptoms (57%). Symptom severity was assessed by 28% of the questionnaires, and 15% assessed both frequency and severity.

Table 1. Overview of the 40 symptom questionnaires and their propertiesa

Questionnaire Items Domain assessed Scale Time frame Target populationMultiple languages

4DSQ [25] 16b somatization 5 categories:no to very often or constantly

past week primary care patients yes

ASR [26] 11b somatic complaints 3 categories:not true to very true or often true

past six months

adults yes

BDS Checklist [27]

25 BDS; pattern of symptoms rather than a simple symptom count (based on SCAN interview)

5 categories:not at all to a lot

past month patients yes

BSI-6 [28] 6b somatization 5 categories:not at all to always

past week adolescents and adults

yes

BSI [29,30] 44 somatic symptoms associated with anxiety and depression

3 categories:symptom absent to present on more than 15 days during the past month

past month patients yes

Cambodian SSI [31]

23 somatic symptoms and cultural syndromes; with a 12-item somatic subscale and an 11-item syndrome subscale

5 categories:not at all to extremely

past month traumatized Cambodian refugees

yes

C-PSC [32] 12 psychosomatic symptoms

frequency, 5 categories: not a problem to every day; severity, 5 categories: not a problem to very, very bad

children no

CSI [33,34] 36 intensity of somatic complaints

4 categories:not at all to a whole lot

past two weeks

children no

FBL [35] 78 somatic complaints frequency, 5 categories: almost every day to almost never; intensity, 5 categories: very strongly to insensitive

‘lately’ yes

91


5

Table 1. Overview of the 40 symptom questionnaires and their propertiesa (continued)


GBB-24 [36] 24 physical complaints 5 categories:never to severe

patients and general population

yes

GSL [37] 37 psychosomatic stress symptoms

4 categories:never to constantly

no

HEALTH-49 [38] 7b somatoform complaints

5 categories:not at all to verymuch

past two weeks

primary care patients yes

ICD-10 Symptom List [39,40]

14 somatization disorder

yes/ no past two years patients yes

Kellner’s SQ [41] 17b somatic symptoms yes/ no or true/ false past week to day

patients and general population

yes

Malaise Inventory [42]

8 psychiatric morbidity

yes/ no ‘no specific time frame is defined, but the focus on recent state is evident’

no

Manu [43] 5 somatization disorder

yes/ no no

MSPQ [44] 13heightened somatic and autonomic awareness

4 categories:not at all to extremely, could not have been worse

past week designed specifically for chronic backache patients

no

NSS [45] 6 nonspecific symptoms for nonpsychotic morbidity

present/ not present at least three months

patients yes

Othmer & DeSouza [46]

7 somatization disorder

yes/ no lifetime general population yes

PHQ [47,48] 14 somatic symptoms items 1-11, 7 categories: not at all to all of the time; items 12-13, 7 categories: 0 times to 7+ times; item 14, 7 categories: 1 day to 7+ days

no

PHQ-15 [15,20] 15 probable somatoform disorder

3 categories:not at all to bothered a lot

past month primary care patients yes

PILL [49] 54 common physical symptoms and sensations

5 categories:never or almost never to more than once every week

lifetime no

PSC-17 (SUNYA) [50]

17 psychosomatic symptoms

frequency, 5 categories: daily to not a problem; intensity, 5 categories: extremely bothersome to not a problem

frequency scores

general and clinical populations; initially developed for muscle contractions headache patients

no

Chapter 5

92



PSC-51 [51] 51 or 55

somatization 4 categories:not at all to most of the time


PSS [52] 35 psychosomaticsymptoms

frequency, 4 categories: never to almost every day; disturbance, 3 categories: none to strong

past three months

children and adolescents

no

PVPS [53] 14b somatization 3 categories:never occurred to frequently occurred

past month people of Vietnamese origin

yes

RPSQ [54] 26 somatization in irritable bowel syndrome patients

5 categories:never or only once to every day

past month IBS patients no

R-SOMS-2 [55] 29 somatization yes/ no past two years primary care patients yes

SCI [56] 22 various physical symptoms

frequency, 5 categories: never to daily; intensity, 5 categories: no problems to extremely troublesome

past month general population no

SCL [57] 11 common somatic complaints

5 categories:almost never to quite often

past month children yes

SCL-90 SOM [58]

12b somatization 5 categories:not at all to extremely

past week psychiatric medical out-patients/general population

yes

SHC [59] 29 subjective health complaints

severity, 4 categories:not at all to serious; duration: number of days

past month general population yes

SOMS-7[60,61]

53 intervention effects in somatoform disorders

5 categories:not at all to very severe


SSI [62] 35 somatization yes/ no lifetime primary care patients yes

SQ-48 [63] 7b somatization 5 categories:never to very often

past week clinical and non-clinical populations

yes

Swartz [64] 11 symptoms that potentially predict a diagnosis of DIS/DSM-3 somatization disorder

yes/ no lifetime general population no

Syrian Symptom Checklist [65]

19 psychosomatic symptoms; diagnose individuals, follow up treatment, evaluate treatment intervention

4 categories:never to always

past few weeks

no

93


5

Questionnaire content

Almost 500 descriptions of symptoms were used in the 40 symptom question-naires. After combining descriptions referring to the same symptoms, we found 31 symptoms that were present in at least six questionnaires (table 2). The symp-tom of headache was most often included in the symptom questionnaires (70%). Other frequent symptoms included nausea/upset stomach (65%), shortness of breath/breathing trouble (58%), dizziness (55%), and (low) back pain/backache (55%). We grouped the symptoms into the above-mentioned clusters cardiopul-monary (including autonomic symptoms), gastrointestinal, musculoskeletal, and general symptoms. Remaining symptoms that did not fit into one of the clusters were classified under the category other symptoms. The latter category included symptoms of the sensory organs (e.g. blurred vision), neurological symptoms (e.g. numbness, amnesia), and symptoms of the urinary tract or sexual organs (e.g. pain during urination, pain during intercourse). Nine out of 40 questionnaires did not include at least one symptom from each symptom cluster (Othmer & DeSouza questionnaire, SQ-48, ASR, Manu questionnaire, NSS, Syrian Symptom Checklist, ICD-10 symptom list, RPSQ, HEALTH-49; figure 2).



WHO-SSD [66] 12 somatoform disorders

yes/ no past six months

general population yes

YSR [67] 9b somatic complaints 3 categories:not true to very true or often true

past six months

11-18 yr. olds yes

von Zerssen [68]

24 somatic complaints 4 categories:not at all to strong

no

aAccording to original questionnaire publication. bSubscale of a larger questionnaireAbbreviations: 4DSQ: Four-Dimensional Symptom Questionnaire; ASR: Adult Self Report; BDS Checklist: Bodily Distress Syndrome Checklist; BSI: Bradford Somatic Inventory; BSI-6: Brief Symptom Inventory; Cambodian SSI: Cambodian Somatic Symptom and Syndrome Inventory; C-PSC: Children’s Psychosomatic Symptom Check-list; CSI: Children’s Somatization Inventory; FBL: Freiburger Beschwerden Liste (Freiburg Complaint List); GBB-24: Giessener Beschwerdebogen (Giessen Subjective Complaints List); GSL: Goldberg Symptom List; HEALTH-49: Hamburger Module zur Erfassung allgemeiner Aspekte psychosozialer Gesundheit für die thera-peutische Praxis (Hamburger modules to measure general aspects of psychosocial health for therapeutic prac-tice); ICD-10 Symptom List: International Classification of Diseases-10 Symptom List; MSPQ: Modified Somatic Perception Questionnaire; NSS: Nonspecific Symptom Screen; PHQ: Physical Health Questionnaire; PHQ-15: Patient Health Questionnaire; PILL: Pennebaker Inventory of Limbic Languidness; PSC-17 (SUNYA): Psycho-somatic Symptom Checklist; PSC-51: Physical Symptom Checklist; PSS: Upitnika Psihosomatskih Simptoma (Psychosomatic Symptoms questionnaire); PVPS: Phan Vietnamese Psychiatric Scale; RPSQ: Recent Physical Symptoms Questionnaire; R-SOMS-2: Revised Screening for Somatoform Symptoms; SCI: Somatic Symptom Checklist Instrument; SCL: Somatic Complaint List; SCL-90 SOM: Symptom Checklist-90 somatization scale; SHC: Subjective Health Complaints Inventory; SOMS-7: Screening for Somatoform Symptoms; SSC: Syrian Symptom Checklist; SSI: Somatic Symptom Index; SQ-48: Symptom Questionnaire-48; WHO-SSD: World Health Organization-Screener for Somatoform Disorders; YSR: Youth Self Report

Chapter 5

94

Table 2. Overview of most included symptoms in the 40 symptom questionnaires. Symptom types that were included in ≥6 questionnaires are shown.

SYMPTOMS %

Cardiopulmonary/autonomic

Shortness of breath; breathing trouble; breathing difficulties 58

Chest pain; pains in heart or chest 43

Palpitations; heart palpitations; palpitations (heart pounding) 38

Sweating; sweating a lot; excessive perspiration 28

Gastrointestinal

Nausea or upset stomach; nausea; nausea, feel sick; nauseated (sick to stomach) 65

Constipation; diarrhea (constipation); diarrhea; constipated (diarrhea); frequent diarrhea 55

Abdominal pain, stomachaches, stomach pain; pain in stomach 53

Bloating; bloating (gassy); belching 35

Vomiting; vomiting, throwing up; vomiting spells 30

Trouble swallowing; difficulty swallowing; difficulty swallowing/lump in throat 28

Heartburn; heartburn or acid regurgitation 23

Loss of appetite, poor appetite 23

Dry mouth 20

General symptoms

Headaches 70

Dizziness 55

Fatigue; fatigability; feeling tired/little energy; tired; overtired 43

Fainting 25

Weakness; feeling weak 18

Musculoskeletal

Back pain; pains in lower back; backache; low back trouble 55

Pain in extremities 33

Neck pain; neck or shoulder pain; pain or tension in neck and shoulders; neck/shoulder muscle ache; neck soreness; shoulder pain

28

Painful muscles; soreness of muscles 23

Joint pain 23

Heaviness in arms/legs 15

Other

Numbness/tingling; numbness/tingling in individual body parts; numbness/tingling; numbness 43

Blurred vision; double/blurred vision; double vision 28

Irregular menstrual periods; painful menstruation; menstrual cramps/other problems with period 23

Pain during urination; difficulty urinating 18

Lump in throat 18

Amnesia; memory loss 18

Pain during intercourse; sexual problems; pain/problems sexual intercourse 15

Symptom clusters are based on previous studies [15-17].

95


5

The number of items within the questionnaires ranged from 5 to 78 items. Nineteen of the 40 (48%) questionnaires consisted of 15 items or fewer, as shown in figure 2. The time frames for symptom recall varied between lifetime and the previous week. However, most of the questionnaires inquired about symptoms of the past week or past month (figure 2). Eight out of 40 (20%) questionnaires required the symptoms to be medically unexplained.

figure 2. Number of items and time frames for symptom recall. Black bars indicate questionnaires that do not contain symptoms from each symptom cluster. *No information about time frames were avail-able from these questionnaires. Abbreviations: 4DSQ: Four-Dimensional Symptom Questionnaire; ASR: Adult Self Report; BDS Checklist: Bodily Distress Syndrome Checklist; BSI: Bradford Somatic Inventory; BSI-6: Brief Symptom Inventory; Cambodian SSI: Cambodian Somatic Symptom and Syndrome Inven-tory; C-PSC: Children’s Psychosomatic Symptom Checklist; CSI: Children’s Somatization Inventory; FBL: Freiburger Beschwerden Liste (Freiburg Complaint List); GBB-24: Giessener Beschwerdebogen (Giessen Subjective Complaints List); GSL: Goldberg Symptom List; HEALTH-49: Hamburger Module zur Erfas-sung allgemeiner Aspekte psychosozialer Gesundheit für die therapeutische Praxis (Hamburger mod-ules to measure general aspects of psychosocial health for therapeutic practice); ICD-10 Symptom List: International Classification of Diseases-10 Symptom List; MSPQ: Modified Somatic Perception Question-naire; NSS: Nonspecific Symptom Screen; PHQ: Physical Health Questionnaire; PHQ-15: Patient Health Questionnaire; PILL: Pennebaker Inventory of Limbic Languidness; PSC-17 (SUNYA): Psychosomatic Symptom Checklist; PSC-51: Physical Symptom Checklist; PSS: Upitnika Psihosomatskih Simptoma (Psy-chosomatic Symptoms questionnaire); PVPS: Phan Vietnamese Psychiatric Scale; RPSQ: Recent Physical Symptoms Questionnaire; R-SOMS-2: Revised Screening for Somatoform Symptoms; SCI: Somatic Symp-tom Checklist Instrument; SCL: Somatic Complaint List; SCL-90 SOM: Symptom Checklist-90 somatiza-tion scale; SHC: Subjective Health Complaints Inventory; SOMS-7: Screening for Somatoform Symptoms; SSC: Syrian Symptom Checklist; SSI: Somatic Symptom Index; SQ-48: Symptom Questionnaire-48; WHO-SSD: World Health Organization-Screener for Somatoform Disorders; YSR: Youth Self Report

Chapter 5

96

internal consistency

Most questionnaires had an acceptable internal consistency; except for the BSI-6 and the Malaise Inventory, for which an internal consistency of <0.70 was found (table 3).

factor structure

The factor structure was investigated for some questionnaires, which referred to whether the items within a questionnaire clustered into meaningful factors. Factor analysis of the PHQ-15 revealed three factors: cardiopulmonary, gastroin-testinal, and general pain/fatigue; which explained 46% of the variance. Factor analysis of the PSC-17 (SUNYA) resulted in one factor, explaining 67% of the vari-ance; however another study reported a five factor solution with a total variance of 36%. Overall, most factor analyses of the questionnaires resulted in multiple factors, often including factors with symptoms originating from the same organ system (data not shown).

test-retest reliability

The test-retest reliability was reported for a few questionnaires, and referred to the variation of measurements at different time points, for the same individual (table 3). Stability within a short time interval (≤ two weeks) was found for the PVPS, RPSQ, Syrian Symptom Checklist, PHQ-15, SCL-90 SOM, PSC-17 (SUNYA), C-PSC, and the SOMS-2.

validity

Some of the questionnaires were correlated with other instruments that aimed to measure symptoms, FSS or somatization. Sometimes correlations with structured interviews were found, ranging from 0.52 to 0.80 (table 3). Correlations with the SCL-90 SOM were often investigated. For example, the correlation of the SOMS (time frame of two years) with the SCL-90 SOM was 0.44; however investigating another version of the SOMS in a much larger sample, the SOMS-7 (time frame of seven days, which is equal to the SCL-90 SOM) showed a correlation of 0.76. Furthermore, the SCL-90 SOM and 4DSQ showed a correlation of 0.82, and the correlation of the SCL-90 SOM and the PHQ-15 was 0.38 (table 3). Correlations of the symptom questionnaires with related concepts such as health anxiety and illness behavior were also investigated (table 3). Strong correlations with health anxiety were found for the PHQ-15 and SOMS-7; and less strong for the BSI. For illness behavior, significant correlations were found for the RPSQ, SCL-90 SOM, and the PHQ-15. As found by our literature search, the questionnaires that were most often validated were the PHQ-15 and SCL-90 SOM.

97


5

tab

le 3

. Int

erna

l con

sist

ency

, tes

t-re

test

relia

bilit

y an

d va

lidity

of 2

8a sym

ptom

que

stio

nnai

res

Que

stio

nnai

re

Inte

rnal

con

sist

ency

(Cro

nbac

h’s

α)Te

st-r

etes

tre

liabi

lity

Valid

ity

over

all s

cale

subs

cale

s

4DSQ

[25]

0.80

-0.8

4 [2

5]co

rrel

atio

n w

ith S

CL-9

0 SO

M =

0.82

(p<0

.001

) [25

]

BDS

Chec

klis

t [27

]0.

82-0

.86

[27]

BSI-6

[28]

=0.6

2 [6

9]

BSI [

29]

corr

elat

ion

with

GH

Q s

omat

ic s

ympt

oms

=0.6

4 (B

ritis

h sa

mpl

e) a

nd =

0.55

(Pak

ista

ni s

ampl

e) 2

9; W

I (h

ypoc

hond

ria) n

o si

gnifi

cant

ass

ocia

tion

[70]

C-PS

C [3

2]=0

.83

[32]

at 1

wee

k in

terv

al: P

ears

on p

rodu

ct-

mom

ent c

orre

latio

n co

effici

ent =

0.90

[32]

CSI [

33]

0.90

-0.9

2 [3

4,71

,71]

FBL

[35]

0.95

-0.9

7 [3

5,72

]

GBB

-24

[36]

=0.9

3 [3

6]0.

80-0

.85

[36]

HEA

LTH

-49

[38]

=0.8

2 [3

8]

Mal

aise

Inve

ntor

y [4

2]0.

47-0

.53

[42]

MSP

Q [4

4]0.

78-0

.85

[44]

PHQ

[47,

48]

0.60

-0.8

8 [4

8]

PHQ

-15

[15,

20]

0.78

-0.8

7 [2

0,73

-76]

at 2

wee

ks in

terv

al: d

icho

tom

ized

PH

Q

scor

e (≥

3) κ

coe

ffici

ent =

0.60

; PH

Q m

eans

of

sco

res

of 2

(sev

ere

som

atic

sym

ptom

s)

intr

acla

ss c

orre

latio

n co

effici

ent =

0.83

[7

4]; a

t 2 w

eeks

inte

rval

cor

rela

tion

=0.6

5 (p

<0.0

01) [

76];

at 1

mon

th in

terv

al

corr

elat

ion

=0.5

4 [7

7]

corr

elat

ion

with

SCL

-90

SOM

=0.

38 (p

<0.0

01) [

77];

prop

ortio

n of

var

ianc

e R²

exp

lain

ed b

y PH

Q-1

5 fo

r SF

-20

scal

es: c

linic

vis

its =

7.8;

dis

abili

ty d

ays

=1.4

[2

0]; c

orre

latio

n w

ith C

IDI s

ympt

om c

ount

=0.

52

[73]

; cor

rela

tion

WI (

hypo

chon

dria

) =0.

50 (m

ales

), =0

.55

(fem

ales

), (b

oth

p<0.

001)

[78]

, =0.

33 (p

<0.0

01)

[77]

; cor

rela

tion

with

nr.

of s

ympt

oms

repo

rted

in

inte

rvie

w =

0.63

(p≤0

.001

); nr

. of m

edic

ally

un

expl

aine

d sy

mpt

oms

asse

ssed

by

gene

ral

prac

titio

ner =

0.63

(p≤0

.001

) [4]

tab

le 3

. Int

erna

l con

sist

ency

, tes

t-re

test

relia

bilit

y an

d va

lidity

of 2

8a sym

ptom

que

stio

nnai

res (

cont

inue

d)

Que

stio

nnai

re

Inte

rnal

con

sist

ency

(Cro

nbac

h’s

α)Te

st-r

etes

tre

liabi

lity

Valid

ity

over

all s

cale

subs

cale

s

PILL

[49]

0.88

-0.9

1 [4

9]co

rrel

atio

n w

ith C

orne

ll M

edic

al In

dex

(inte

rvie

w

sym

ptom

cou

nt) =

0.57

[49]

PSC-

17 (S

UN

YA) [

50]

0.74

-0.8

1 [7

9,80

]at

1 w

eek

inte

rval

cor

rela

tion

=0.8

8 (p

<0.0

001)

; at 4

wee

ks in

terv

al =

0.84

(p

<0.0

001)

[50]

PSC-

51 [5

1]=0

.88

[81]

PSS

[52]

0.90

-0.9

3 [5

2]

PVPS

[53]

=0.9

0 [5

3]at

4 d

ays

inte

rval

: cor

rela

tion

=0.8

4 [5

3]

RPSQ

[54]

=0.8

6 [5

4]at

2 w

eeks

inte

rval

: =0.

88 [5

4]co

rrel

atio

ns w

ith C

orne

ll M

edic

al In

dex

(inte

rvie

w

sym

ptom

cou

nt) =

0.80

; doc

tor v

isits

(sel

f-rep

ort)

=0

.50;

sel

f-rat

ed h

ealth

=0.

31; B

SI-1

8 (s

omat

izat

ion

subs

cale

) =0.

70 [5

4]

R-SO

MS-

2 [5

5]=0

.83

[55]

; 19-

item

s ve

rsio

n: K

R-20

=0.

88

[82]

Corr

elat

ion

with

clin

ical

sym

ptom

eva

luat

ion

=0.6

3 [5

5]

SCL

[57]

=0.8

3 [5

7]

SCL-

90 S

OM

[58]

0.70

-0.9

0 [5

8,79

,83-

87]

at 1

wee

k in

terv

al: t

est-

rete

st c

oeffi

cien

t =0

.82

[58]

corr

elat

ion

high

ly s

truc

ture

d in

terv

iew

by

trai

ned

clin

icia

ns =

0.73

[58]

;D

isea

se c

onvi

ctio

n sc

ale

=0.6

7 [8

8]; c

orre

latio

n w

ith n

r. of

prim

ary

care

con

sulta

tions

=0.

27 (y

r. be

fore

bas

elin

e) a

nd =

0.16

(yr.

afte

r bas

elin

e), (

both

p<

0.00

1) [8

9]

SHC

[59]

0.75

-0.8

2 [9

0]0.

49-0

.77

[90]

SOM

S-2

[91]

for i

tem

s 1-

42, 0

.87-

0.89

; for

item

s 1-

35,

0.84

-0.8

7 [9

1]

35-it

ems

vers

ion:

at 7

2 ho

urs

inte

rval

co

rrel

atio

n =0

.85;

=0.

71 (m

ales

); =0

.93

(fem

ales

); =0

.97

(tin

nitu

s pa

tient

s); =

0.91

(a

nxie

ty a

nd d

epre

ssio

n pa

tient

s); =

0.75

(p

ain

patie

nts)

[92]

grou

p w

ith s

omat

izat

ion

diso

rder

(acc

ordi

ng to

SO

MS)

had

hig

hest

sco

res

on W

I (hy

poch

ondr

ia) [

93];

corr

elat

ion

with

SCL

-90

SOM

=0.

44 [9

1]

99


5

tab

le 3

. Int

erna

l con

sist

ency

, tes

t-re

test

relia

bilit

y an

d va

lidity

of 2

8a sym

ptom

que

stio

nnai

res (

cont

inue

d)

Que

stio

nnai

re

Inte

rnal

con

sist

ency

(Cro

nbac

h’s

α)Te

st-r

etes

tre

liabi

lity

Valid

ity

over

all s

cale

subs

cale

s

SOM

S-7

[60]

=0.9

2 (g

ende

r-un

spec

ific

item

s) [6

0]co

rrel

atio

n w

ith D

SM-4

som

atiz

atio

n =0

.66;

SCL

-90

SOM

=0.

76; W

I (hy

poch

ondr

ia) =

0.46

; Mar

k’s

scal

e (d

isab

ility

) =0.

40 (a

ll p<

0.00

1) [6

0]

SQ-4

8 [6

3]=0

.89

[63]

Syri

an S

ympt

om C

heck

list

[65]

0.56

-0.8

5 [6

5]at

7-1

0 da

ys in

terv

al: P

ears

on’s

corr

elat

ion

coeffi

cien

t =0.

89 (m

ales

), =0

.91

(fem

ales

) [6

5]

YSR

[67]

0.76

-0.7

7 [9

4]a Da

ta w

ere

not a

vaila

ble

for

all 4

0 qu

estio

nnai

res.

The

full

tabl

e w

ith v

alid

ity d

ata

incl

udin

g in

form

atio

n ab

out s

tudy

pop

ulat

ion

and

othe

r ps

ycho

met

ric

prop

ertie

s is

ava

il-ab

le fr

om th

e au

thor

s. Ab

brev

iatio

ns: 4

DSQ:

Fou

r-Di

men

sion

al S

ympt

om Q

uest

ionn

aire

; BDS

Che

cklis

t: Bo

dily

Dis

tres

s Syn

drom

e Ch

eckl

ist;

BSI:

Brad

ford

Som

atic

Inve

ntor

y;

BSI-6

: Bri

ef S

ympt

om In

vent

ory;

BSI

-18:

Bri

ef S

ympt

om In

vent

ory;

CID

I: Co

mpo

site

Inte

rnat

iona

l Dia

gnos

tic In

terv

iew

; C-P

SC: C

hild

ren’

s Psy

chos

omat

ic S

ympt

om C

heck

list;

CSI:

Child

ren’

s So

mat

izat

ion

Inve

ntor

y; F

BL: F

reib

urge

r Be

schw

erde

n Li

ste

(Fre

ibur

g Co

mpl

aint

Lis

t); G

BB-2

4: G

iess

ener

Bes

chw

erde

boge

n (G

iess

en S

ubje

ctiv

e Co

mpl

aint

s Li

st);

GHQ:

Gen

eral

Hea

lth Q

uest

ionn

aire

; HEA

LTH

-49:

Ham

burg

er M

odul

e zu

r Erf

assu

ng a

llgem

eine

r Asp

ekte

psy

chos

ozia

ler G

esun

dhei

t für

die

ther

apeu

tisch

e Pr

axis

(Ham

-bu

rger

mod

ules

to m

easu

re g

ener

al a

spec

ts o

f psy

chos

ocia

l hea

lth fo

r th

erap

eutic

pra

ctic

e; M

MPI

: Min

neso

ta M

ultip

hasi

c Pe

rson

ality

Inve

ntor

y; M

SPQ:

Mod

ified

Som

atic

Pe

rcep

tion

Ques

tionn

aire

; PH

Q: P

hysi

cal H

ealth

Que

stio

nnai

re; P

HQ-

15: P

atie

nt H

ealth

Que

stio

nnai

re; P

ILL:

Pen

neba

ker

Inve

ntor

y of

Lim

bic

Lang

uidn

ess;

PSC

-17

(SUN

YA):

Psyc

hoso

mat

ic S

ympt

om C

heck

list;

PSC-

51: P

hysi

cal S

ympt

om C

heck

list;

PSS:

Upi

tnik

a Ps

ihos

omat

skih

Sim

ptom

a (P

sych

osom

atic

Sym

ptom

s qu

estio

nnai

re);

PVPS

: Pha

n Vi

etna

mes

e Ps

ychi

atri

c Sca

le; R

PSQ:

Rec

ent P

hysi

cal S

ympt

oms Q

uest

ionn

aire

; R-S

OMS-

2: R

evis

ed S

cree

ning

for S

omat

ofor

m S

ympt

oms;

SCL

: Som

atic

Com

plai

nt L

ist;

SCL-

90

SOM

: Sym

ptom

Che

cklis

t-90

som

atiz

atio

n sc

ale;

SF-

20: S

hort

-For

m G

ener

al H

ealth

sur

vey-

20; S

HC:

Sub

ject

ive

Hea

lth C

ompl

aint

s In

vent

ory;

SOM

S-7:

Scr

eeni

ng fo

r So

mat

o-fo

rm S

ympt

oms;

SSC

: Syr

ian

Sym

ptom

Che

cklis

t; SS

PS: S

halli

ng S

ifneo

s Pe

rson

ality

Sca

le; S

Q-48

: Sym

ptom

Que

stio

nnai

re-4

8; T

AS: T

oron

to A

lexi

thym

ia S

cale

; WI:

Whi

tely

In

dex;

YSR

: You

th S

elf R

epor

t

Chapter 5

100

Part 2: Phenx criteria

Usability and burden to participants

Questionnaires should be of relatively low burden to participants and investiga-tors, and should therefore be both as short as possible and understandable to a wide range of persons. As shown in figure 2, the number of items within question-naires ranged from 5 to 78, with 48% of the questionnaires being 15 items or shorter. All questionnaires assessed could be used by scientists with no specific expertise in the assessment of somatic symptoms, FSS or somatization, and no training is needed.

Relevance for future studies and to a variety of populations and regions

We can assume that most symptoms included in these questionnaires will be relevant for future studies. As shown in table 1, some questionnaires are available in multiple languages, contributing to their relevance to a variety of regions. Most questionnaires are available in English; however, we were not able to find English versions for four questionnaires: the Syrian Symptom Checklist, von Zerssen questionnaire, SCI, and FBL. The Cambodian SSI is a questionnaire that is very region specific, with symptoms that will probably not be recognized in all regions of the world, such as “Khyal attacks.” Some of the questionnaires were developed for a specific population; the CSI, YSR, SCL, C-PSC, and the PSS were all developed for children or adolescents. The other questionnaires were either applicable to all age categories or the target population was not explicitly mentioned (table 1). The RPSQ is a questionnaire developed to investigate somatization in irritable bowel syndrome patients and therefore included no symptoms from the gastro-intestinal cluster, because these symptoms are assumed to be present in these patients. Finally, it is worth mentioning that there are seven questionnaires that have gender specific or reproduction related items (e.g. painful menstruation or impotence). These are the Goldberg Symptom List, Othmer & DeSouza question-naire, PHQ-15, PSC-51, SOMS-7, ICD-10 symptom list, and the SSI. In summary, there were five questionnaires that fulfilled all PhenX criteria. These were the ASR, BSI-6, PHQ-15, SCL-90 SOM, and the SQ-48.

discUssion

Of the 40 questionnaires included in this review, we believe many are unsuitable for use in large-scale studies. Many of these questionnaires did not contain the most relevant symptoms, were too lengthy, used a recall period that is too long,

101


5

or were inapplicable to a variety of populations and regions. In addition, some of the questionnaires were more extensively validated than others. Based on our criteria, there were two questionnaires that seem reasonably suitable for the use in large-scale studies. These were the PHQ-15 [15,20] and the SCL-90 SOM [58]. Both are not restricted to medically unexplained symptoms but concern symp-toms in general, and both measure symptom severity. With the validity of these questionnaires well-established, these instruments would be suitable for use in future studies. These questionnaires are also relatively short, which makes them easy to use and of little burden to participants. No specific training is needed, making them usable by scientists who do not have expertise in this specific do-main. Furthermore, they are available in multiple languages and contain relevant symptoms.

However, several questions remain. The first concerns which symptoms should be included in a questionnaire. We decided that questionnaires should not be re-stricted to medically unexplained symptoms, given the difficulties in establishing the causes of symptoms, especially in large-scale studies. Inquiring into symptoms in general is also in line with the new proposal for the psychiatric classification of patients with somatoform disorder, which abandons the distinction between symptoms that are medically explained and those that are not [95]. Furthermore, research has shown that a high total somatic symptom count, regardless of whether these are medically explained or unexplained, are important predictors of decreased health status and health-related quality of life [96].

Another aspect is which specific symptoms should be assessed. We reasoned that the proportion of instruments including a particular symptom would reflect expert knowledge and empirical data on the importance of that symptom for these questionnaires. Based on this reasoning, important symptoms are headaches, nausea/upset stomach, shortness of breath/breathing trouble, dizziness, and (low) back pain/backache. Both the PHQ-15 and the SCL-90 SOM contain these five symptoms. The importance and relevance of these symptoms for the process of somatization is underlined by a study showing which symptoms often remain medically unexplained in internal medicine outpatients. These were especially headaches, dizziness, constipation, back pain, abdominal pain, chest pain, numb-ness and impotence (≥75% of these symptoms remained unexplained) [5]. Of the questionnaires with 15 items or less, the PHQ-15, the SCL-90 SOM, and the Swartz questionnaire inquired about at least five of these often unexplained symptoms. These questionnaires also included symptoms from all relevant symptom clus-ters, and thus covered an appropriate diversity of symptoms, while retaining a reasonable number of items.

Chapter 5

102

The PHQ-15 and the SCL-90 SOM also differ in some aspects. A significant, but not very high correlation was found between the two questionnaires. Subse-quently, only six symptoms overlap. This is probably due to the differences in the conceptual background of the symptom questionnaires. The PHQ-15 consists of the most prevalent symptoms in the US primary care [15], whereas the SCL-90 SOM consists of a subset of somatic symptoms associated with psychopathology [58]. Studies focused at somatization might thus preferably include the SCL-90 SOM, whereas studies interested in common somatic symptoms might preferably include the PHQ-15. Although the recent DSM-5 field trials to assess somatic symptoms used a shortened version of PHQ-15 97, showing that this question-naire is also commonly used to assess somatization.

Another difference is that the PHQ-15 contains items related to reproduction, in contrast to the SCL-90 SOM. These reproduction related items (menstruation problems and pain or problems during sexual intercourse) would make the PHQ-15 less applicable to children. Furthermore, although the PHQ-15 is well validated in both sexes, ‘menstruation problems’ should be ignored by men when filling out the questionnaire; and a negative answer to ‘pain or problems during sexual intercourse’ can have multiple interpretations.

A second remaining question concerns the timeframe. Most of the question-naires enquired about symptoms in the last month or a shorter time period, with the PHQ-15 covering the past month and the SCL-90 SOM covering the past week. Answers to questionnaires that enquire about symptoms in a longer time period (e.g. the ASR with six months or the Swartz questionnaire with lifetime) are probably prone to recall bias [21], which could result in a less reliable instru-ment. On the other hand, a short timeframe might detect fluctuations in symptom levels that are not meaningful for symptom measurement. The ideal time frame for symptom questionnaires remains to be investigated, based on an appropriate balance between the risk of excessive recall bias and the detection of meaningful fluctuations. A related issue concerns the answer modalities, which are typically based on frequency for the questionnaires covering longer time windows and both frequency and severity for the shorter time windows. Severity may be a more appropriate and reliable answering scale, compared to frequency of symptoms. The PHQ-15 and SCL-90 SOM both measure symptom severity with respectively a 3-points scale and a 5-points scale. Further research could determine the optimal number of response categories for these questionnaires.

A third question concerns the relevance to a variety of age groups. A question-naire should be preferably suitable for all age groups. It is not known whether the PHQ-15 and SCL-90 SOM are also suitable for children or adolescents. The PHQ-15 has four symptoms overlapping with the YSR, a questionnaire specific for

103


5

children. As mentioned earlier, the PHQ-15 contains two symptoms about repro-ductive function which would be inapplicable to children. The SCL-90 SOM has three symptoms overlapping with the YSR, but no items related to reproduction.

internal consistency, factor structure, test-retest reliability and validity

Both the PHQ-15 and SCL-90 SOM showed acceptable internal consistency and factor analyses resulted in meaningful factors. Results of test-retest reliability indicated that the results of the PHQ-15 and SCL-90 SOM are relatively stable within a short time period. Intercorrelations of the PHQ-15 and SCL-90 SOM with other measures of somatization showed some overlap with these, indicat-ing they measure a comparable construct. Additionally, the PHQ-15 and SCL-90 SOM showed some overlap with other related constructs, such as health anxiety and illness behavior. Correlations with health anxiety and illness behavior are especially relevant, as these constructs are part of the new DSM proposal for the diagnosis of somatic symptom disorder [98]. Although some overlap is expected, high correlations of symptom questionnaires with other constructs can indicate a lack of discrimination.

In addition to these psychometric properties, other areas could be further as-sessed. Sensitivity to change of the questionnaires could be evaluated, in particular if the questionnaire will be used in longitudinal studies. Especially in anticipation of the DSM-5, it would be important to assess the sensitivity and specificity of the questionnaires for the new diagnostic criteria for somatic symptom disorder. Assessments from an item response theory (IRT) approach are also important to evaluate. For example, with IRT it is possible to investigate whether test scores have the same meaning across different groups. This examination of differential item functioning (DIF) for subgroups can be useful in questionnaire evaluation of, for example, general and clinical populations or different ethnic populations [99]. Another example is that IRT can be used to translate scores of one instrument into the other, using a set of anchor items, which are similar items in different ques-tionnaires, that function similarly across questionnaires. Questionnaire scores are then equated, making the individual scores comparable [100]. This could enhance the comparability between studies, even if different questionnaires are used.

limitations

Although many efforts were taken to identify all symptom questionnaires cur-rently available, there is still a chance that some questionnaires were missed. Furthermore, it is possible that not all validation studies of the questionnaires were included in this study. The main goal of this study was to identify all the relevant questionnaires and not to identify all the validation studies of these

Chapter 5

104

questionnaires. Therefore, the search term used for the literature search in this review was not constructed to identify all the available validation studies. Another limitation concerning the validity of the questionnaires must be acknowledged: in order to evaluate validity, a gold standard is required. For the assessment of symptoms, a gold standard is not available and the questionnaires’ validity can therefore not be compared to a gold standard. Additionally, some characteristics of specific symptom questionnaires were unavailable. Some questionnaires may also be available in more languages than reported here, as our search was not targeted at identifying all available versions of the questionnaires. It was not possible to establish the well-establishment (i.e. used repeatedly over time and coming from highly regarded sources) of the questionnaires, as we could not determine whether every questionnaire was often used or came from a highly regarded source.

conclUsion

We believe this is the first study that systematically reviewed somatic symptom questionnaires and evaluated them for use in large-scale studies. We provided an overview of which symptom questionnaires are available and suitable. By providing this review, we hope to end the use of different symptom instruments in different studies, which makes comparisons between studies difficult, if not impossible. Although the recommendation of a symptom list was mainly intended for large-scale studies, this could also apply to other types of studies. Ultimately, this will contribute to the harmonization of studies and thereby contribute to knowledge of the prevalence of somatic symptoms or FSS.

105


5

references

1. Rhodes VA, Watson PM. Symptom distress—The concept: Past and present. Semin Oncol Nurs. 1987;3(4):242-247.

2. McAteer A, Elliott AM, Hannaford PC. Ascertaining the size of the symptom iceberg in a UK-wide community-based survey. Br J Gen Pract. 2011;61(582):e1-11.

3. Bruusgaard D, Tschudi-Madsen H, Ihlebaek C, Kamaleri Y, Natvig B. Symptom load and func-tional status: Results from the ullensaker population study. BMC Public Health. 2012;12:1085.

4. Körber S, Frieser D, Steinbrecher N, Hiller W. Classification characteristics of the patient health questionnaire-15 for screening somatoform disorders in a primary care setting. J Psychosom Res. 2011;71(3):142-147.

5. Kroenke K, Mangelsdorff AD. Common symptoms in ambulatory care: Incidence, evaluation, therapy, and outcome. Am J Med. 1989;86(3):262-266.

6. Lipowski ZJ. Somatization: The concept and its clinical application. Am J Psychiatry. 1988;145(11):1358-1368.

7. Steinbrecher N, Koerber S, Frieser D, Hiller W. The prevalence of medically unexplained symp-toms in primary care. Psychosomatics. 2011;52(3):263-271.

8. Jackson JL, Passamonti M. The outcomes among patients presenting in primary care with a physical symptom at 5 years. J Gen Intern Med. 2005;20(11):1032-1037.

9. Rosmalen JGM. The way forward: A case for longitudinal population-based studies in the field of functional somatic syndromes. J Psychosom Res. 2010;68(5):399-401.

10. Gureje O, Simon GE, Ustun TB, Goldberg DP. Somatization in cross-cultural perspective: A world health organization study in primary care. Am J Psychiatry. 1997;154(7):989-995.

11. Barsky AJ, Borus JF. Functional somatic syndromes. Ann Intern Med. 1999;130(11):910-921. 12. Kroenke K, Spitzer RL, deGruy FV, et al. Multisomatoform disorder - an alternative to undiffer-

entiated somatoform disorder for the somatizing patient in primary care. Arch Gen Psychiatry. 1997;54(4):352-358.

13. Barsky AJ, Orav EJ, Bates DW. Somatization increases medical utilization and costs independent of psychiatric and medical comorbidity. Arch Gen Psychiatry. 2005;62(8):903-910.

14. Nettleton S. ‘I just want permission to be ill’: Towards a sociology of medically unexplained symptoms. Soc Sci Med. 2006;62(5):1167-1178.

15. Kroenke K, Spitzer RL, deGruy FV,3rd, Swindle R. A symptom checklist to screen for somato-form disorders in primary care. Psychosomatics. 1998;39(3):263-272.

16. Fink P, Toft T, Hansen MS, Ornbol E, Olesen F. Symptoms and syndromes of bodily distress: An exploratory study of 978 internal medical, neurological, and primary care patients. Psychosom Med. 2007;69(1):30-39.

17. Rosmalen JGM, Tak LM, de Jonge P. Empirical foundations for the diagnosis of somatization: Implications for DSM-5. Psychol Med. 2010:1-10.

18. Kroenke K. Physical symptom disorder: A simpler diagnostic category for somatization-spectrum conditions. J Psychosom Res. 2006;60(4):335-339.

19. Rief W, Mewes R, Martin A, Glaesmer H, Braehler E. Evaluating new proposals for the psychiatric classification of patients with multiple somatic symptoms. Psychosom Med. 2011;73(9):760-768.

20. Kroenke K, Spitzer RL, Williams JB. The PHQ-15: Validity of a new measure for evaluating the severity of somatic symptoms. Psychosom Med. 2002;64(2):258-266.

Chapter 5

106

21. Leiknes KA, Finset A, Moum T, Sandanger I. Methodological issues concerning lifetime medi-cally unexplained and medically explained symptoms of the composite international diagnostic interview: A prospective 11-year follow-up study. J Psychosom Res. 2006;61(2):169-179.

22. Simon GE, Gureje O. Stability of somatization disorder and somatization symptoms among primary care patients. Arch Gen Psychiatry. 1999;56(1):90-95.

23. Hamilton CM, Strader LC, Pratt JG, et al. The PhenX toolkit: Get the most from your measures. Am J Epidemiol. 2011;174(3):253-260.

24. Hayes AF. Assessing and quantifying reliability. In: Statistical methods for communication sci-ence. Mahwah, NJ: Lawrence Erlbaum Associates.; 2005:103.

25. Terluin B, van Marwijk HW, Ader HJ, et al. The four-dimensional symptom questionnaire (4DSQ): A validation study of a multidimensional self-report questionnaire to assess distress, depression, anxiety and somatization. BMC Psychiatry. 2006;6:34.

26. Achenbach TM, Rescorla LA. Manual for the ASEBA adult forms & profiles. Burlington, VT: University of Vermont, Research Center for Children, Youth, & Families.; 2003.

27. Budtz-Lilly A, Fink P, Ørnbøl E, et al. Measuring bodily distress in a primary care population: The use of a new patient questionnaire (submitted). 2012.

28. Derogatis LR, Melisaratos N. The brief symptom inventory: An introductory report. Psychol Med. 1983;13(3):595-605.

29. Mumford DB. Somatic sensations and psychological distress among students in britain and pakistan. Soc Psychiatry Psychiatr Epidemiol. 1989;24(6):321-326.

30. Mumford DB, Bavington JT, Bhatnagar KS, Hussain Y, Mirza S, Naraghi MM. The bradford somatic inventory - a multiethnic inventory of somatic symptoms reported by anxious and depressed-patients in britain and the indo-pakistan subcontinent. British Journal of Psychia-try. 1991;158:379-386.

31. Hinton DE, Kredlow MA, Bui E, Pollack MH, Hofmann SG. Treatment change of somatic symp-toms and cultural syndromes among cambodian refugees with PTSD. Depression Anxiety. 2011;4(2).

32. Wisniewski JJ, Naglieri JA, Mulick JA. Psychometric properties of a children’s psychosomatic symptom checklist. J Behav Med. 1988;11(5):497-507.

33. Walker LS, Garber J, Greene JW. Somatization symptoms in pediatric abdominal pain patients: Relation to chronicity of abdominal pain and parent somatization. J Abnorm Child Psychol. 1991;19(4):379-394.

34. Garber J, Walker LS, Zeman J. Somatization symptoms in a community sample of children and adolescents: Further validation of the children’s somatization inventory. Psychological Assess-ment: A Journal of Consulting and Clinical Psychology. 1991;3(4):588-595.

35. Fahrenberg J. Somatic complaints in the german population. J Psychosom Res. 1995;39(7):809-817.

36. Brahler E, Schumacher J, Brahler C. First all-germany standardization of the brief form of the gissen complaints questionnaire GBB-24. Psychother Psychosom Med Psychol. 2000;50(1):14-21.

37. Herman SL, Lester D. Physical symptoms of stress, depression, and suicidal ideation in high school students. Adolescence. 1994;29(115):639-641.

38. Rabung S, Harfst T, Kawski S, Koch U, Wittchen HU, Schulz H. Psychometric analysis of a short form of the “hamburg modules for the assessment of psychosocial health” (HEALTH-49). Z Psychosom Med Psychother. 2009;55(2):162-179.

107


5

39. World Health Organization. Somatisation disorder. in: The ICD-10 classification of mental and behavioural disorders. diagnostic criteria for research. Geneva: World Health Organization; 1993.

40. Khoo EM, Mathers NJ, McCarthy SA, Low WY. Somatisation disorder and its associated factors in multiethnic primary care clinic attenders. Int J Behav Med. 2012;19(2):165-173.

41. Kellner R. A symptom questionnaire. J Clin Psychiatry. 1987;48(7):268-274. 42. Rodgers B, Pickles A, Power C, Collishaw S, Maughan B. Validity of the malaise inventory in

general population samples. Soc Psychiatry Psychiatr Epidemiol. 1999;34(6):333-341. 43. Manu P, Lane TJ, Matthews DA, Escobar JI. Screening for somatization disorder in patients with

chronic fatigue. Gen Hosp Psychiatry. 1989;11(4):294-297. 44. Main CJ. The modified somatic perception questionnaire (MSPQ). J Psychosom Res.

1983;27(6):503-514. 45. Srinivasan TN, Suresh TR. The nonspecific symptom screening method. detection of nonpsy-

chotic morbidity based on nonspecific symptoms. Gen Hosp Psychiatry. 1991;13(2):106-114. 46. Othmer E, DeSouza C. A screening test for somatization disorder (hysteria). Am J Psychiatry.

1985;142(10):1146-1149. 47. Spence JT, Helmreich RL, Pred RS. Impatience versus achievement strivings in the type A

pattern: Differential effects on students’ health and academic achievement. J Appl Psychol. 1987;72(4):522-528.

48. Schat AC, Kelloway EK, Desmarais S. The physical health questionnaire (PHQ): Construct validation of a self-report scale of somatic symptoms. J Occup Health Psychol. 2005;10(4):363-381.

49. Pennebaker JW. The psychology of physical symptoms. New York: Springer-Verlag; 1982. 50. Attanasio V, Andrasik F, Blanchard EB, Arena JG. Psychometric properties of the SUNYA revision

of the psychosomatic symptom checklist. J Behav Med. 1984;7(2):247-257. 51. de Waal MWM, Arnold IA, Spinhoven P, Eekhof JAH, van Hemert AM. The reporting of specific

physical symptoms for mental distress in general practice. J Psychosom Res. 2005;59(2):89-95. 52. Vulić-Prtorić A. Upitnik psihosomatskih simptoma za djecu i adolescente - procjena valjanosti.

Suvremena Psihologija. 2005;8(2):211-227. 53. Phan T, Steel Z, Silove D. An ethnographically derived measure of anxiety, depression and soma-

tization: The phan vietnamese psychiatric scale. Transcult Psychiatry. 2004;41(2):200-232. 54. MacLean EW, Palsson OS, Turner MJ, Whitehead WE. Development and validation of new

disease-specific measures of somatization and comorbidity in IBS. J Psychosom Res. 2012. 55. Fabiao C, Silva MC, Barbosa A, Fleming M, Rief W. Assessing medically unexplained symptoms:

Evaluation of a shortened version of the SOMS for use in primary care. BMC Psychiatry. 2010;10:34.

56. Bohman H, Jonsson U, Paaren A, von Knorring L, Olsson G, von Knorring AL. Prognostic significance of functional somatic symptoms in adolescence: A 15-year community-based follow-up study of adolescents with depression compared with healthy peers. BMC Psychiatry. 2012;12:90.

57. Jellesma FC, Rieffe C, Terwogt MM. The somatic complaint list: Validation of a self-report ques-tionnaire assessing somatic complaints in children. J Psychosom Res. 2007;63(4):399-401.

58. Derogatis LR, Lipman RS, Rickels K, Uhlenhuth EH, Covi L. The hopkins symptom checklist (HSCL): A self-report symptom inventory. Behav Sci. 1974;19(1):1-15.

59. Ursin H, Endresen IM, Ursin G. Psychological factors and self-reports of muscle pain. Eur J Appl Physiol Occup Physiol. 1988;57(3):282-290.

Chapter 5

108

60. Rief W, Hiller W. A new approach to the assessment of the treatment effects of somatoform disorders. Psychosomatics. 2003;44(6):492-498.

61. Rief W, Hiller W. Das screening für somatoforme störungen SOMS [the screening for somato-form symptoms]. manual zum fragebogen. 2nd revised edition ed. Bern: Huber-Verlag; 2008.

62. Escobar JI, Rubio-Stipec M, Canino G, Karno M. Somatic symptom index (SSI): A new and abridged somatization construct. prevalence and epidemiological correlates in two large com-munity samples. J Nerv Ment Dis. 1989;177(3):140-146.

63. Carlier I, Schulte-Van Maaren Y, Wardenaar K, et al. Development and validation of the 48-item symptom questionnaire (SQ-48) in patients with depressive, anxiety and somatoform disor-ders. Psychiatry Res. 2012;30;200(2-3):904-10. doi: 10.1016/j.psychres.2012.07.035.

64. Swartz M, Hughes D, George L, Blazer D, Landerman R, Bucholz K. Developing a screening index for community studies of somatization disorder. J Psychiatr Res. 1986;20(4):335-343.

65. Rudwan SJ. A checklist of the syrian psychological psychosomatic symptoms. Journal of the Social Sciences. 2000;28(4):113-138.

66. Phillips KA, Fallon BA, King S. Somatoform and factitious disorders and malingering measures. In: Rush AJ,Jr., First MB, Blacker D, Rush AJ,Jr., First MB, Blacker D, eds. Handbook of psychiatric measures (2nd ed.). Arlington, VA US: American Psychiatric Publishing, Inc; 2008:559-585.

67. Gledhill J, Garralda ME. Functional symptoms and somatoform disorders in children and adolescents: The role of standardised measures in assessment. Child and Adolescent Mental Health. 2006;11(4):208-214.

68. Ladwig KH, Marten-Mittag B, Lacruz ME, Henningsen P, Creed F, MONICA KORA Investigators. Screening for multiple somatic complaints in a population-based survey: Does excessive symptom reporting capture the concept of somatic symptom disorders? findings from the MONICA-KORA cohort study. J Psychosom Res. 2010;68(5):427-437.

69. Galdon MJ, Dura E, Andreu Y, et al. Psychometric properties of the brief symptom inventory-18 in a spanish breast cancer sample. J Psychosom Res. 2008;65(6):533-539.

70. Chakraborty K, Avasthi A, Kumar S, Grover S. Psychological and clinical correlates of functional somatic complaints in depression. Int J Soc Psychiatry. 2012;58(1):87-95.

71. Meesters C, Muris P, Ghys A, Reumerman T, Rooijmans M. The children’s somatization inven-tory: Further evidence for its reliability and validity in a pediatric and a community sample of dutch children and adolescents. J Pediatr Psychol. 2003;28(6):413-422.

72. Jensen H, Salazar Z, Aguilar I. Adaptación castellana de la ‘lista de síntomas orgánicos de friburgo.’ Avances en Psicologia Clinica Latinoamericana. 1986;4:97-113.

73. Interian A, Allen LA, Gara MA, Escobar JI, Diaz-Martinez AM. Somatic complaints in primary care: Further examining the validity of the patient health questionnaire (PHQ-15). Psychoso-matics. 2006;47(5):392-398.

74. van Ravesteijn H, Wittkampf K, Lucassen P, et al. Detecting somatoform disorders in primary care with the PHQ-15. Ann Fam Med. 2009;7(3):232-238.

75. Ros Montalban S, Comas Vives A, Garcia-Garcia M. Validation of the spanish version of the PHQ-15 questionnaire for the evaluation of physical symptoms in patients with depression and/or anxiety disorders: DEPRE-SOMA study. Actas Esp Psiquiatr. 2010;38(6):345-357.

76. Han C, Pae CU, Patkar AA, et al. Psychometric properties of the patient health questionnaire-15 (PHQ-15) for measuring the somatic symptoms of psychiatric outpatients. Psychosomatics. 2009;50(6):580-585.

109


5

77. Gulec MY, Gulec H, Simsek G, Turhan M, Sunbul EA. Psychometric properties of the turkish version of the patient health questionnaire-somatic, anxiety, and depressive symptoms. Compr Psychiatry. 2012;53(5):623-629.

78. Lee S, Ng KL, Ma YL, Tsang A, Kwok KP. A general population study of the chinese whiteley-7 index in hong kong. J Psychosom Res. 2011;71(6):387-391.

79. Bagby RM, Taylor GJ, Atkinson L. Alexithymia: A comparative study of three self-report mea-sures. J Psychosom Res. 1988;32(1):107-116.

80. Chibnall JT, Tait RC. The psychosomatic symptom checklist revisited: Reliability and validity in a chronic pain population. J Behav Med. 1989;12(3):297-307.

81. de Waal MW, Arnold IA, Spinhoven P, Eekhof JA, Assendelft WJ, van Hemert AM. The role of comorbidity in the detection of psychiatric disorders with checklists for mental and physical symptoms in primary care. Soc Psychiatry Psychiatr Epidemiol. 2009;44(1):78-85.

82. Fabiao C, Barbosa A, Fleming M, Silva C. Screening of somatization disorder in primary health care. results from a pilot study. Acta Med Port. 2008;21(4):319-328.

83. Holi MM, Sammallahti PR, Aalberg VA. A finnish validation study of the SCL-90. Acta Psychiatr Scand. 1998;97(1):42-46.

84. Schmitz N, Hartkamp N, Brinschwitz C, Michalek S. Computerized administration of the symp-tom checklist (SCL-90-R) and the inventory of interpersonal problems (IIP-C) in psychosomatic outpatients. Psychiatry Res. 1999;87(2-3):217-221.

85. Schmitz N, Hartkamp N, Brinschwitz C, Michalek S, Tress W. Comparison of the standard and the computerized versions of the symptom check list (SCL-90-R): A randomized trial. Acta Psychiatr Scand. 2000;102(2):147-152.

86. Schmitz N, Hartkamp N, Kiuse J, Franke GH, Reister G, Tress W. The symptom check-list-90—R (SCL-90—R): A german validation study. Quality of Life Research: An International Journal of Quality of Life Aspects of Treatment, Care & Rehabilitation. 2000;9(2):185-193.

87. Van Kamp I, Stoop LA. De hopkins symptom checklist bij epidemiologisch onderzoek. Ned Tijdschr Psychol. 1986;41(8):369-373.

88. Ciccone DS, Just N, Bandilla EB. Non-organic symptom reporting in patients with chronic non-malignant pain. Pain. 1996;68(2-3):329-341.

89. Tomenson B, McBeth J, Chew-Graham CA, et al. Somatization and health anxiety as predictors of health care use. Psychosom Med. 2012;74(6):656-664.

90. Eriksen HR, Ihlebaek C, Ursin H. A scoring system for subjective health complaints (SHC). Scand J Public Health. 1999;27(1):63-72.

91. Rief W, Schäfer S, Fichter MM. SOMS: Ein screening-verfahren zur identifizierung von personen mit somatoformen störungen. Diagnostica. 1992;38(3):228-241.

92. Rief W, Hiller W, Goebel G, Fichter MM. Zur zeitlichen stabilität der angaben hypochondrischer und somatoformer symptome. Diagnostica. 1995;41(2):172-180.

93. Hiller W, Rief W, Fichter MM. Further evidence for a broader concept of somatization disorder using the somatic symptom index. Psychosomatics. 1995;36(3):285-294.

94. Janssens KA, Oldehinkel AJ, Rosmalen JGM. Parental overprotection predicts the development of functional somatic symptoms in young adolescents. J Pediatr. 2009;154(6):918-23.e1.

95. American Psychiatric Association. APA DSM-5 development. http://www.dsm5.org. Updated 2012. Accessed November/09, 2012.

96. Creed FH, Davies I, Jackson J, et al. The epidemiology of multiple somatic symptoms. J Psycho-som Res. 2012;72(4):311-317.

Chapter 5

110

97. Narrow WE, Clarke DE, Kuramoto SJ, et al. DSM-5 field trials in the united states and canada, part ill: Development and reliability testing of a cross-cutting symptom assessment for DSM-5. Am J Psychiatry. 2013;170(1):71-82.

98. Dimsdale J, Creed F, DSM-V Workgroup on Somatic Symptom Disorders. The proposed diag-nosis of somatic symptom disorders in DSM-V to replace somatoform disorders in DSM-IV - a preliminary report. J Psychosom Res. 2009;66(6):473-476.

99. Reise SP, Waller NG. Item response theory and clinical measurement. Annual Review of Clinical Psychology. 2009;5:27-48.

100. Fischer HF, Tritt K, Klapp BF, Fliege H. How to compare scores from different depression scales: Equating the patient health questionnaire (PHQ) and the ICD-10-symptom rating (ISR) using item response theory. Int J Methods Psychiatr Res. 2011;20(4):203-214.

111


5

aPPendix Abbreviations questionnaires

4DSQ Four-Dimensional Symptom Questionnaire

ASR Adult Self Report

BDS Checklist Bodily Distress Syndrome Checklist

BSI Bradford Somatic Inventory

BSI-6 Brief Symptom Inventory

Cambodian SSI Cambodian Somatic Symptom and Syndrome Inventory

C-PSC Children’s Psychosomatic SymptomChecklist

CSI Children’s Somatization Inventory

FBL Freiburger Beschwerden Liste (Freiburg Complaint List)

GBB-24 Giessener Beschwerdebogen (Giessen Subjective Complaints List)

GSL Goldberg Symptom List

HEALTH-49 Hamburger Module zur Erfassung allgemeiner Aspekte psychosozialer Gesundheit für die therapeutische Praxis (Hamburger modules to measure general aspects of psychosocial health for therapeutic practice)

ICD-10 Symptom List International Classification of Diseases-10 Symptom List

MSPQ Modified Somatic Perception Questionnaire

NSS Nonspecific Symptom Screen

PHQ Physical Health Questionnaire

PHQ-15 Patient Health Questionnaire

PILL Pennebaker Inventory of Limbic Languidness

PSC-17 (SUNYA) Psychosomatic Symptom Checklist

PSC-51 Physical Symptom Checklist

PSS Upitnika Psihosomatskih Simptoma (Psychosomatic Symptoms questionnaire)

PVPS Phan Vietnamese Psychiatric Scale

RPSQ Recent Physical Symptoms Questionnaire

R-SOMS-2 Revised Screening for Somatoform Symptoms

SCI Somatic Symptom Checklist Instrument

SCL Somatic Complaint List

SCL-90 SOM Symptom Checklist-90 somatization scale

SHC Subjective Health Complaints Inventory

SOMS-7 Screening for Somatoform Symptoms

SSI Somatic Symptom Index

SQ-48 Symptom Questionnaire-48

WHO-SSD World Health Organization-Screener for Somatoform Disorders

YSR Youth Self Report

6Noise and somatic symptoms: a role for personality traits?

Wilma L. Zijlema, David Morley, Ronald P. Stolk, Judith G.M. Rosmalen

International Journal of Hygiene and Environmental Health 2015 Aug;218(6):543-9

Chapter 6

114

aBsTracT

objectives

We investigated the role of a stress-sensitive personality on relations between noise, noise annoyance and somatic symptom reporting. Firstly, we investigated the cross-sectional association of road traffic noise exposure and somatic symp-toms, and its modification by hostility and vulnerability to stress. Secondly, we in-vestigated the cross-sectional association of noise annoyance from eight sources (e.g. road traffic, aircraft, neighbors) and somatic symptoms, and it’s confounding by hostility and vulnerability to stress.

methods

Data were obtained from LifeLines, a general population cohort from the Neth-erlands. Road traffic noise was estimated using the Common Noise Assessment Methods in Europe (CNOSSOS-EU) noise model. Noise annoyance, hostility, vulnerability to stress, and somatic symptoms were assessed with validated ques-tionnaires.

results

Poisson regression models adjusted for demographic and socioeconomic variables indicated no association of noise exposure and somatic symptoms (incidence rate ratio (IRR) 1.001; 95% confidence interval (CI) 1.000-1.001; n=56,937). Interac-tions of noise exposure and hostility and vulnerability to stress were not statisti-cally significant. Small positive associations were found for noise annoyance from each of the eight sources and somatic symptoms, when adjusted for demographic and socioeconomic variables (e.g. for road traffic noise annoyance IRR 1.014, 95% CI 1.011-1.018; n=6,177). Additional adjustment for hostility and vulnerability to stress resulted in small decreases of the IRRs for noise annoyance from each of the eight sources, but the associations remained statistically significant.

conclusions

Personality facets hostility and vulnerability to stress did not modify the relation between road traffic noise exposure and somatic symptom reporting, or confound relations between noise annoyance and symptoms.

115

Noise and somatic symptoms

6

inTrodUcTion

Around 140 decibels (dB), sound exposure passes the pain threshold [1]. But does sound have to be that loud to hurt? Evidence showing that sound at lower levels can have effects on health already exists. Environmental noise has been associated with a variety of adverse health effects, including hearing loss, cardiovascular dis-ease [1] and impaired neurocognitive function [2]. Not only the actual exposure to noise , but also an individual’s annoyance from noise and noise sensitivity con-tribute to adverse health effects [3]. Some studies suggest it may not be the noise itself that is associated with adverse health for certain outcomes, but instead the individual’s annoyance from noise. This is demonstrated by a recent study showing that noise annoyance was strongly associated with somatic symptoms, such as headaches and fatigue, while modelled road traffic noise exposure was not [4]. Similar findings were also reported in studies from Norway and Sweden [5,6]. A strong predictor of noise annoyance is noise sensitivity, which refers to an increased reaction to noise. Individuals that are noise sensitive pay more attention to sound, are more likely to evaluate it negatively, and have stronger emotional reactions to noise [7]. Noise sensitivity has been associated with lower health-related quality of life [8], and depressive symptoms [9]. In addition, noise sensitivity is related to other environmental sensitivities, including environmen-tal chemosensory responsivity [10], to susceptibility to stress in general [11], and to personality traits such as neuroticism [12]. These results lead to the hypothesis that noise sensitivity reflects a more general susceptibility to stressors [13]. If noise sensitivity is part of a more general tendency to be susceptible to stressors, how is such a trait influencing the relation between noise and somatic symptoms? We aim to investigate the role a stress-sensitive personality has on the relations between noise, noise annoyance and somatic symptom reporting. Our research questions are as follows: Firstly, is there a relationship between noise and somatic symptoms in persons who are vulnerable to stressors? We hypothesize that noise exposure may be related to somatic symptoms, but only in a subgroup who are vulnerable to stress. Secondly, is there a relationship between noise annoyance and somatic symptoms because persons vulnerable to stressors report both more noise annoyance and more symptoms? In other words, can this relationship be explained by a confounding effect of a general trait of vulnerability to stress? We hypothesize that sensitive persons will report more noise annoyance [7] and also more symptoms as a result [14].

The general trait of vulnerability to stress is captured in the personality trait neuroticism. Neuroticism can be described as the tendency to experience nega-tive and distressing emotions, and is positively correlated with noise annoyance

Chapter 6

116

[15,16], and noise sensitivity [17]. Neuroticism is considered to be a broad personality trait composed of six facets: anxiety, hostility, depression, self-con-sciousness, impulsiveness, and vulnerability to stress. Some argue that research is needed on the level of neuroticism’s facets instead of the trait in general [18]. When investigating noise, noise annoyance and somatic symptoms, the neuroti-cism facets hostility and vulnerability to stress seem relevant to these relations, while the remaining facets might be less appropriate to study. Studying facets of neuroticism that are theoretically more related to the relationship tested here, may provide a better insight in the relation of noise, noise annoyance, somatic symptoms and personality. We tested these associations in LifeLines, a large population based cohort from the Netherlands [19].

meTHods

study design and participants

LifeLines is a multi-disciplinary prospective population-based cohort study exam-ining in a unique three-generation design the health and health-related behaviors of 167,729 persons living in the North East region of The Netherlands. It employs a broad range of investigative procedures in assessing the biomedical, socio-demographic, behavioural, physical and psychological factors which contribute to the health and disease of the general population, with a special focus on multi-morbidity and complex genetics [20]. Inclusion of study participants began in 2006 via general practitioners and also self-enrollment. All participants provided written informed consent. The study protocol was carried out in accordance to the Declaration of Helsinki, and was approved by the medical ethical review com-mittee of the University Medical Center Groningen. A detailed description of the LifeLines Cohort Study has been published elsewhere [19].

Baseline measurements were performed between 2006 and 2013 and ap-proximately three years thereafter a follow up questionnaire was sent out. The present study included baseline data with road traffic noise estimates available for 75,304 participants, aged between 18 and 92 years. At the time of our study follow up measurements were still ongoing, and follow up data were released for 61,967 participants. Data obtained during baseline and follow up measure-ments were used in this study, resulting in different sample sizes for the main constructs. Modelled road traffic noise was estimated for home addresses at the time of baseline measurements. Noise annoyance was assessed during the follow up measurements. Baseline data (n=56,937) were used to investigate the association between road traffic noise exposure and somatic symptoms, and the

117


6

modification of this association by hostility and vulnerability to stress. Follow up data (n=46,558) were used for evaluation of the association between noise annoy-ance and somatic symptoms and its confounding by hostility and vulnerability to stress. Participants with incomplete data regarding somatic symptoms (baseline n=2,365; follow up n=1,372), hostility (baseline n= 5,402; follow up n=5,544), vulnerability to stress (baseline n=5,114; follow up n=5,318), and household equivalent income (baseline n= 9,904; follow up n=9,343) were excluded. The sample size at follow up differed for the various analyses depending on the num-ber of missing data for the noise annoyance questions.

road traffic noise

Road traffic noise was estimated using a new implementation of the CNOSSOS-EU noise modelling framework [21]. Briefly, the noise level is estimated on road segments within 500 meters of a receptor. Noise propagation to the receptor is assessed with a consideration of possible attenuation due to refractions on build-ings, absorption by the atmosphere and interactions with reflective or absorbent land cover surfaces. The CNOSSOS-EU framework contains empirically derived equations to determine both the initial noise level based on traffic flow and also the sound attenuation based on known environmental factors and physical pro-cesses. To estimate source noise on road segments in the Netherlands, information is used of hourly flow of passenger cars, heavy goods vehicles and their average speeds. The sound propagation model is based on the CORINE landcover dataset that has a European wide coverage. Traffic data originated from year 2009 and landcover data from 2006. The final sound level is expressed as the day-evening-night time (Lden) annual average in A weighted decibels (dB(A))[22]. Lden is the average A-weighted noise level, estimated over a 24 hour period, with a 10 dB(A) penalty added to the night (23.00–07.00 hours), and a 5 dB(A) penalty added to the evening period (19.00–23.00 hours) noise level. The penalties are added to indicate people’s extra sensitivity to noise during the night and evening.

noise annoyance

Noise annoyance from eight different sources was assessed using a standardized questionnaire [23]. Participants were asked whether the noises were audible in their homes, and if so, to what extent they were bothered, disturbed or annoyed by the noise. The sources of annoyance include noise from road traffic, railroad, aircraft, industrial sources, wind turbines, construction and demolition activities, shops and restaurants, and neighbors. Annoyance could be indicated on a scale ranging from “0: not bothered” to “10: extremely bothered” [23]. When par-ticipants indicated that the noise from a specific source was not audible in their

Chapter 6

118

home, the value for annoyance was set to zero. Noise annoyance was assessed during follow up measurements. The item regarding road traffic noise annoyance was implemented in the questionnaire at a later stage than the other noise annoy-ance questions, and was therefore available for a smaller sample (n=6,162).

somatic symptoms

The Symptom Checklist-90 somatization scale (SCL-90 SOM) questionnaire was used for the assessment of 12 common somatic symptoms [24]. The SCL-90 SOM is a widely used and validated self-report symptom checklist [25] that assesses headaches, faintness or dizziness, pains in heart or chest, pains in lower back, nausea or upset stomach, soreness of your muscles, numbness or tingling in your body, hot or cold spells, feeling weak in parts of your body, heavy feelings in arms or legs, a lump in your throat, and trouble getting your breath [24]. The question-naire assesses to what extent participants have been limited by these symptoms in the past seven days. Items were scored on a 5-point scale ranging from (1) “Not at all” to (5) “Extremely”. A sum score of all twelve items, ranging from 12 to 60, was calculated. The Cronbach’s alpha was 0.80 at baseline and follow up, indicat-ing acceptable internal consistency of the SCL-90 SOM. We used the sum score as a measure of symptom reporting in general, instead of individual symptoms. The reason for this is that we hypothesized that noise, noise annoyance, and certain personality aspects may be related to symptom reporting in general, and not to any symptoms in particular. The SCL-90 SOM questionnaire was assessed at both baseline and follow up.

hostility and vulnerability to stress

Hostility and vulnerability to stress are two facets of the personality trait neuroti-cism. These facets were measured using the neuroticism subscale of the Revised NEO Personality Inventory (NEO-PI-R) [26]. The authorized Dutch translation of this scale has good reliability and validity [27]. Hostility and vulnerability to stress were both assessed with 8 items, which were answered on a 5-point Likert-type scale that ranged from strongly disagree to strongly agree. Sum scores for both hostility and vulnerability to stress were calculated by adding all items belonging to that particular facet, ranging from 8 to 40. In our study, Cronbach’s alpha was 0.74 for hostility and 0.81 for vulnerability to stress, indicating accept-able internal consistency of the scales. Hostility and vulnerability to stress were assessed during baseline measurements.

119


6

demographic variables

Details on sex, age, highest obtained educational degree, household income and number of persons in the household were obtained with a questionnaire at base-line. Household equivalent income was calculated as net household income per month divided by the square root of the number of persons in the household [28].


Descriptive statistics were used to describe the characteristics of the study population. Because of the non-normal distributions of road traffic noise and the SCL-90 SOM sum score, medians and interquartile ranges (IQRs) were calculated. For other variables, means with standard deviations (SD) and percentages were calculated. Correlations between modelled road traffic noise, noise annoyance, hostility, vulnerability to stress, and somatic symptoms were calculated with Spearman’s rho. Incidence rate ratios (IRR) of somatic symptoms were estimated with Poisson regression. Poisson regression is used to model count variables. Although somatic symptom score is not a classical example of a count variable, it can be interpreted as a count of symptom severity and its distribution resembles the Poisson distribution. The “robust option” was used for obtaining robust standard errors for the parameter estimates to control for overdispersion of the data [29]. For the association between road traffic noise and somatic symptoms, a model adjusted for age, sex, education and household equivalent income was analyzed using the baseline sample. In the second model, hostility and vulnerabil-ity to stress and an interaction term of road traffic noise exposure and hostility and vulnerability to stress were added, for the investigation of modification by hostility and vulnerability to stress. For the association between noise annoyance and somatic symptoms, models adjusted for age, sex, education, and household equivalent income were fitted using the follow up sample. Subsequently, models additionally adjusted for hostility and vulnerability to stress were fitted, to in-vestigate confounding of the association between noise annoyance and somatic symptoms. Model fit was deemed acceptable when the p-value of the goodness-of-fit chi-squared test was >.05, and when the p-value of the Omnibus test was <.05. Associations were considered statistically significant if the 95% confidence intervals (CI) for the IRRs did not include one. All analyses were performed with IBM SPSS Statistics.

Chapter 6

120

resUlTs

Population characteristics

Of the 56,937 participants in our baseline sample, 43.5% were male, mean age was 42.4 years, and median road traffic noise exposure was 55.9 Lden dB(A). Median sum score on the SCL-90 SOM was 15 (Table 1a). Muscle soreness (47%), lower back pain (45%) and headaches (43%) were the most often reported symp-toms in the study population. Population characteristics of the follow up sample (n=46,558) were comparable to the baseline sample (Table 1b).

Noise from road traffic and neighbors was most often audible in participant’s homes (72.1% and 61.1% respectively; Table 2). If audible in their homes, partici-pants were mostly annoyed by noise from neighbors (15.0%) and construction and demolition activities (11.8%) (Table 2). Correlations between the dependent and independent variables are presented in the Supplementary Table. For ex-ample, the correlation between modelled road traffic noise and road traffic noise annoyance was low (Spearman’s rho .223, p<.001), and correlations between annoyance from other noise sources and modelled road traffic noise were even lower.

Table 1a. Population characteristics of baseline sample (n=56,937)

Males, % 43.5

Age (years), mean (SD) 42.4 (10.3)

Highest obtained educational degree, %

No education or primary education 1.6

Lower or preparatory vocational education 11.1

Lower general secondary education 12.4

Intermediate vocational education or apprenticeship 31.9

Higher general secondary education or pre-university secondary education 9.4

Higher vocational education or university 33.6

Household equivalent income per month (%)

<€1000 17.8

€1000 - €1300 22.1

€1300 - €1600 16.7

€1600 - €1900 22.2

>€1900 21.1

Road traffic noise (Lden), median (IQR) 54.6 (4.2)

SCL-90 SOM sum score, median (IQR) 15 (5)

Hostility, mean (SD) 22.7 (2.6)

Vulnerability to stress, mean (SD) 25.7 (2.1)

Abbreviations: SD: standard deviation; IQR: interquartile range; Lden: annual average day-evening-night time road traffic noise; SCL-90 SOM: Symptom Checklist-90 somatization scale

121


6

Associations of road traffic noise and somatic symptoms, and modification by hostility and vulnerability to stress

Poisson regression was used for the investigation of the association of road traffic noise and somatic symptoms, and the modification by hostility and vulnerability to stress (Table 3). First, a model adjusted for age, sex, education, and household equivalent income showed no association between road traffic noise and somatic symptoms (IRR 1.001; 95% CI 1.000-1.001). In the second model, interaction ef-fects of noise and hostility and vulnerability to stress were analyzed to investigate

Table 1b. Population characteristics of follow up sample (n=46,558)

Males, % 41.2

Age (years), mean (SD) 46.6 (10.6)

Highest obtained educational degree, %

No education or primary education 1.4

Lower or preparatory vocational education 10.9

Lower general secondary education 12.6

Intermediate vocational education or apprenticeship 31.2

Higher general secondary education or pre-university secondary education 9.5

Higher vocational education or university 34.5

Household equivalent income per month (%)

<€1000 16.1

€1000 - €1300 22.2

€1300 - €1600 17.1

€1600 - €1900 21.4

>€1900 23.1

SCL-90 SOM sum score median (IQR) 15 (6)

Hostility, mean (SD) 18.5 (4.3)

Vulnerability to stress, mean (SD) 18.1 (4.1)

Abbreviations: SD: standard deviation; IQR: interquartile range; SCL-90 SOM: Symptom Checklist-90 somatiza-tion scale

Table 2. Noise annoyance from eight sources in the follow up sample.Participants indicated whether noise from eight specific sources was audible in their homes. If so, they rated the noise annoyance on a scale from 0 (not bothered) to 10 (extremely bothered).

N Audible in home, % Noise annoyance score >4, %

Road traffic 6,162 72.1 8.0

Railroad 46,440 16.3 1.6

Aircraft 46,323 26.8 5.6

Industrial sources 46,413 8.1 6.4

Wind turbines 46,558 1.6 2.3

Construction and demolition activities 46,310 14.2 11.8

Shops and restaurants 46,449 6.7 8.1

Neighbors 45,886 61.1 15.0

Chapter 6

122

modification by hostility and vulnerability to stress. Both interaction terms were not statistically significant, suggesting no modification by hostility and vulner-ability to stress.

We used multiple imputation methods to account for the missing values for somatic symptoms, hostility and vulnerability to stress, education and household equivalent income at baseline. Analyses with imputed values did not result in dif-ferent conclusions as with the original data. This indicates that the missing values probably had no effect on our data, and results based on the original dataset were reported.

Associations of noise annoyance and somatic symptoms, and confounding by hostility and vulnerability to stress

Positive associations of noise annoyance and somatic symptoms were found in the Poisson regression models conducted in the follow up sample (Table 4). The models adjusted for age, sex, education level, and household equivalent income resulted in IRRs ranging from 1.014 (95% CI 1.011-1.018; road traffic noise an-noyance) to 1.022 (95% CI 1.018-1.026; railroad noise annoyance). This means that, for example, each additional unit on the railroad noise annoyance scale is associated with an estimated 2.2% increase on the somatic symptoms scale. The models were subsequently adjusted for hostility and vulnerability to stress to investigate confounding of the noise annoyance-somatic symptoms relationship. Here, all associations of noise annoyance and somatic symptom score remained positive and statistically significant. Generally, adjustment for hostility and vulner-ability to stress did result in small decreases of the IRRs for noise annoyance with respect to the models without hostility and vulnerability to stress. The largest associations with somatic symptom score were found for noise annoyance from

table 3. Associations between road traffic noise (Lden) and somatic symptoms (baseline sample n=56,937) estimated from Poisson regression analyses.

SCL-90 SOM sum score

IRR 95% CI

Model 1 Road traffic noise 1.001 1.000;1.001

Model 2 Road traffic noise 0.999 0.997;1.002

Hostility 1.007 1.000;1.014

Vulnerability 1.009 1.002;1.017

Road traffic noise*hostility 1.000 1.000;1.000

Road traffic noise*vulnerability 1.000 1.000;1.000

Abbreviations: SCL-90 SOM: Symptom Checklist-90 somatization scale; IRR: incidence rate ratio; 95% CI: 95% confidence interval; vulnerability: vulnerability to stressModels were adjusted for sex, age, educational level and household equivalent income

123


6

industrial sources (IRR 1.018, 95% CI 1.014-1.021) and railroad noise annoyance (IRR 1.017 95% CI 1.013-1.022), while the smallest association was found for an-noyance from road traffic noise (IRR 1.011, 95% CI 1.007-1.014). In each of these

table 4. Associations between noise annoyance and somatic symptoms (follow up sample) estimated from Poisson regression analyses.

SCL-90 SOM sum score

IRR 95% CI

Model 1a Road traffic (n=6,177) 1.014 1.011;1.018

Model 1b Railroad (n=46,485) 1.022 1.018;1.026

Model 1c Aircraft (n=46,376) 1.016 1.014;1.019

Model 1d Industrial sources (n=46,451) 1.021 1.017;1.024

Model 1e Wind turbines (n=46,558) 1.012 1.005;1.020

Model 1f Construction and demolition activities (n=46,346) 1.020 1.018;1.022

Model 1g Shops and restaurants (n=46,494) 1.018 1.015;1.021

Model 1h Neighbors (n=45,967) 1.016 1.015;1.017

Model 2a Road traffic (n=6,177) 1.011 1.007;1.014

Hostility 1.010 1.009;1.012


Model 2b Rail road (n=46,485) 1.017 1.013;1.022

Hostility 1.011 1.010;1.012


Model 2c Aircraft (n=46,376) 1.013 1.011;1.016

Hostility 1.011 1.010;1.012


Model 2d Industrial sources (n=46,451) 1.018 1.014;1.021

Hostility 1.011 1.010;1.012


Model 2e Wind turbines (n=46,558) 1.010 1.003;1.017

Hostility 1.011 1.010;1.012


Model 2f Construction and demolition activities (n=46,346) 1.016 1.014;1.018

Hostility 1.011 1.010;1.011


Model 2g Shops and restaurants (n=46,494) 1.016 1.013;1.019

Hostility 1.011 1.010;1.012


Model 2h Neighbors (n=45,967) 1.012 1.011;1.013

Hostility 1.010 1.010;1.011


Abbreviations: SCL-90 SOM: Symptom Checklist-90 somatization scale; IRR: incidence rate ratio; 95% CI: 95% confidence interval; vulnerability: vulnerability to stressModels 1a-1h were adjusted for sex, age, educational level and household equivalent incomeModels 2a-2h were adjusted for sex, age, educational level and household equivalent income, hostility and vul-nerability to stress

Chapter 6

124

models, hostility and vulnerability to stress were also positively associated with somatic symptom reporting. We also used multiple imputation methods to ac-count for the missing values for somatic symptoms, hostility and vulnerability to stress, education and household equivalent income at follow up. Again, analyses with imputed values did not result in different conclusions as with the original data. Therefore results based on the original dataset were reported.

discUssion

We found no indications for a relationship of road traffic noise exposure on so-matic symptom reporting, both in the entire group and in the subgroup character-ized by hostility or vulnerability to stress. In contrast, noise annoyance from road traffic and seven other noise sources were positively associated with somatic symptom reporting. This association could not entirely be explained by hostility or vulnerability to stress. These findings are comparable to previous studies. Two other studies, that also used modelled noise estimates and comparable symptom questionnaires also did not find a relation between road traffic noise and somatic symptoms [4,5]. Comparable conclusions were also drawn by a study that used measured traffic noise [6]. We did find positive associations between self-report-ed noise annoyance from several sources and somatic symptoms. A recent study undertaken in Switzerland also found that increased noise annoyance from road traffic, aircraft, railroad, industry and neighbors was related to higher symptom reporting [4]. Discrepancies between actual and perceived exposures in relation to somatic symptoms were also reported in a recent study examining electromag-netic fields and symptom reporting [30]. Related to this, the correlation of road traffic noise annoyance and modelled road traffic noise in our study was rather low, and was similar to a previous study [13].

Hostility and vulnerability to stress did not modify the relation between noise and symptoms and they explained only a small part of the effect of noise annoy-ance on symptom reporting. From this, it seems that in the current study, hostility and vulnerability to stress did not play a large role in the relationship between somatic symptoms and exposure to noise or noise annoyance. A previous study [31] showed that neuroticism, of which hostility and vulnerability to stress are components of, accounted for the relationship found for noise sensitivity and mental health complaints, but similar to our study, not for somatic complaints. The authors argued that neuroticism, noise sensitivity and mental health complaints are highly related to each other, while for somatic complaints, these relations are smaller [31]. Fyhri and Klæboe (2009) found that the relation between noise

125


6

sensitivity and somatic symptom reporting was far stronger than the relation be-tween noise annoyance and symptoms. This indicates that noise sensitivity plays an important role in the noise and symptom relationship. They argued that noise sensitivity is an important factor in the noise-symptom relationship, but they also proposed that a third variable, which they described as “susceptibility”, is involved in both increased noise sensitivity and increased symptom reporting [5].

Schreckenberg and colleagues hypothesized that noise sensitivity partly reflects a general environmental sensitivity and is associated with a decreased evaluation of subjective health. In their study, noise sensitivity was related to subjective physical health, and noise sensitivity was related to noise annoyance from aircraft, but not to noise annoyance from road traffic. In addition, noise sensitivity was generally not associated with residential satisfaction. As a conclusion, they state that noise sensitivity is more specifically related to noise responses rather than a predictor of perception of environmental quality [13]. Although not investigated here, noise sensitivity may be a more important factor than the personality facets hostility and vulnerability to stress in relations between noise exposure, noise annoyance and somatic symptoms.

Several strengths and limitations of this study need to be taken into consider-ation when interpreting the results. One limitation of our study is the amount of missing data in the dataset, especially for household equivalent income. Despite this missing data, a large number of subjects (n=56,937 and n=46,558) were still available for this study. Analyses with imputed values did not result in different conclusions as with the original dataset. This suggests that the missing values had little or no effect on our results. The sample size for the analyses of noise annoy-ance from road traffic noise was smaller compared to the sample size for analyses on the other noise annoyance sources. The road traffic noise annoyance question was implemented in the LifeLines questionnaire at a later stage, and therefore not filled out by all participants. Nevertheless, results from road traffic noise an-noyance were comparable to noise annoyance results from other sources with a larger sample size. Furthermore, only the association with road traffic noise and the interaction with hostility and vulnerability to stress was investigated. Noise from other sources (aircraft, railroad or neighbors) could have effects on somatic symptom reporting, as some sources of noise might provoke different reactions or effects than others [32]. However, in the LifeLines study area, the flight frequency is quite low and only a small airport is situated there. Also, the rail network is not very extensive. Our data on noise annoyance from aircraft and railroads showed that this was audible by 27% and 16% of the participants, ver-sus 72% for road traffic noise. Also, noise annoyance from aircraft and railroads was relatively low. Noise from neighbors was often heard in participants’ houses

Chapter 6

126

(61%) and was found to be relatively annoying. Strengths of this study include the use of standardized assessments for road traffic noise [22], noise annoyance [23], somatic symptoms [24], and neuroticism facets hostility and vulnerability to stress [26] enabling comparisons with other studies. Previous studies used, for example, other measures for the assessment of noise exposure, for example road traffic volume [31], which is a rather crude measure of road traffic noise. Some specific symptoms, such as sleeping problems, fatigue and headaches, are more likely to be related to noise and noise annoyance than other symptoms [33]. We used a composite index of symptom severity, based on the 12 symptoms of the SCL-90 SOM. Although not each of these symptoms may related to noise and noise annoyance, we hypothesized that noise, noise annoyance, and certain personality aspects are related to symptom reporting in general, and not to any symptoms in particular. High internal consistency of the SCL-90 SOM (Cronbach’s alpha of 0.80) also underlines that it measures a general construct. A similar composite symptom score has also been used in previous studies [4,31]. Another strength is that we investigated two facets of neuroticism: hostility and vulnerability to stress. Studying these two facets of neuroticism may be more relevant to the noise and somatic symptom relationship, instead of neuroticism in general that also entails less appropriate facets such as self-consciousness and impulsiveness.

conclusions

In the LifeLines Cohort Study, a large population cohort from the Netherlands, no evidence was found for a relationship between road traffic noise and somatic symptoms in a subgroup that is hostile and vulnerable to stress. Increases in noise annoyance from several noise sources were associated with increases in the reporting of somatic symptoms, and this was independent of hostility and vulner-ability to stress. The role of the personality facets hostility and vulnerability to stress in the relations of noise, annoyance and symptoms appears small.

127


6

references



3. Van Kamp I, Davies H. Noise and health in vulnerable groups: a review. Noise Health. 2013;15: 153–9. doi:10.4103/1463-1741.112361



6. Öhrström E, Skånberg A, Svensson H, Gidlöf-Gunnarsson A. Effects of road traffic noise and the benefit of access to quietness. J Sound Vib. 2006;295: 40–59. doi:10.1016/j.jsv.2005.11.034

7. Stansfeld SA. Noise, noise sensitivity and psychiatric disorder: epidemiological and psychophys-iological studies. Psychol Med Monogr Suppl. 1992;22: 1. doi:10.1017/S0264180100001119

8. Shepherd D, Welch D, Dirks KN, Mathews R. Exploring the relationship between noise sensitiv-ity, annoyance and health-related quality of life in a sample of adults exposed to environmental noise. Int J Environ Res Public Health. 2010;7: 3579–94. doi:10.3390/ijerph7103580

9. Stansfeld SA, Shipley M. Noise sensitivity and future risk of illness and mortality. Sci Total Environ. 2015;520: 114–119. doi:10.1016/j.scitotenv.2015.03.053

10. Karnekull SC, Jonsson FU, Larsson M, Olofsson JK. Affected by smells? Environmental chemical responsivity predicts odor perception. Chem Senses. 2011;36: 641–648. doi:10.1093/chemse/bjr028

11. Nordin S, Ljungberg JK, Claeson A-S, Neely G. Stress and odor sensitivity in persons with noise sensitivity. Noise Health. 2013;15: 173–7. doi:10.4103/1463-1741.112366

12. Belojevic G, Jakovljevic B. Factors influencing subjective noise sensitivity in an urban popula-tion. Noise Health. 2001;4: 17–24.

13. Schreckenberg D, Griefahn B, Meis M. The associations between noise sensitivity, reported physical and mental health, perceived environmental quality, and noise annoyance. Noise Health. 2010;12: 7–16. doi:10.4103/1463-1741.59995

14. Rosmalen JGM, Neeleman J, Gans RO, de Jonge P. The association between neuroticism and self-reported common somatic symptoms in a population cohort. J Psychosom Res. 2007;62: 305–311. doi:10.1016/j.jpsychores.2006.10.014

15. Öhrström E, Björkman M, Rylander R. Noise annoyance with regard to neurophysiological sen-sitivity, subjective noise sensitivity and personality variables. Psychol Med. 1988;18: 605–13.

16. Thomas JR, Jones DM. Individual differences in noise annoyance and the uncomfortable loudness level. J Sound Vib. 1982;82: 289–304. doi:http://dx.doi.org/10.1016/0022-460X(82)90536-3

17. Stansfeld S, Clark C, Jenkins L, Tarnopolsky A. Sensitivity to noise in a community sample: I. Measurement of psychiatric disorder and personality. Psychol Med. 1985;15: 243–254.

18. Ormel J, Jeronimus BF, Kotov R, Riese H, Bos EH, Hankin B, et al. Neuroticism and common mental disorders: meaning and utility of a complex relationship. Clin Psychol Rev. 2013;33: 686–97. doi:10.1016/j.cpr.2013.04.003

Chapter 6

128



21. Kephalopoulos S, Paviotti M, Anfosso-Lédée F. Common Noise Assessment Methods in Europe (CNOSSOS-EU). 2012.


23. International Organization for Standardization (ISO). Acoustics – Assessment of Noise Annoy-ance by Means of Social and Socio-acoustic Surveys; ISO/TS 15666. Geneva, Switzerland; 2003.

24. Derogatis L, Lipman R, Rickels K, Uhlenhuth E, Covi L. The Hopkins Symptom Checklist (HSCL): A self-report symptom inventory. Behav Sci. 1974;19: 1–15. doi:10.1002/bs.3830190102

25. Zijlema WL, Stolk RP, Löwe B, Rief W, White PD, Rosmalen JGM. How to assess common somatic symptoms in large-scale studies: A systematic review of questionnaires. J Psychosom Res. 2013;74: 459–468. doi:10.1016/j.jpsychores.2013.03.093

26. Costa P, McCrae R. Revised NEO Personality Inventory (NEO PI-R) and the Five Factor Inven-tory (NEO-FFI): professional manual. Odessa, FL; 1992.

27. Hoekstra HA, Ormel J, De Fruyt F. NEO-PI-R Handleiding. Hogrefe Uitgevers BV Amsterdam, Netherlands. 2007.

28. Buhrman B, Rainwater L, Schmaus G, Smeeding T. Sensitivity estimates across ten countries using the luxembourg income study (LIS) data base. Rev Income Wealth. 1988;2: 115–42.

29. Cameron A, Trivedi P. Microeconometrics using stata. Stata Press, College Station TX; 2009. 30. Baliatsas C, Bolte J, Yzermans J, Kelfkens G, Hooiveld M, Lebret E, et al. Actual and perceived

exposure to electromagnetic fields and non-specific physical symptoms: An epidemiological study based on self-reported data and electronic medical records. Int J Hyg Environ Health. 2015; doi:10.1016/j.ijheh.2015.02.001

31. Hill EM, Billington R, Krägeloh C. Noise sensitivity and diminished health: testing mod-erators and mediators of the relationship. Noise Health. 2014;16: 47–56. doi:10.4103/1463-1741.127855

32. Floud S, Vigna-Taglianti F, Hansell A, Blangiardo M, Houthuijs D, Breugelmans O, et al. Medica-tion use in relation to noise from aircraft and road traffic in six European countries: results of the HYENA study. Occup Environ Med. 2011;68: 518–24. doi:10.1136/oem.2010.058586

33. Huss A, Küchenhoff J, Bircher A, Heller P, Kuster H, Niederer M, et al. Symptoms attributed to the environment – a systematic, interdisciplinary assessment. Int J Hyg Environ Health. 2004;207: 245–254. doi:http://dx.doi.org/10.1078/1438-4639-00286

129


6

sup

ple

men

tal t

able

. Spe

arm

an’s

corr

elat

ions

bet

wee

n ro

ad tr

affic

noi

se, n

oise

ann

oyan

ce, s

omat

ic sy

mpt

oms,

host

ility

and

vul

nera

bilit

y to

stre

ss.

Road

tr

affic

nois

e (L

den)

NA

road

tr

affic

NA

ra

ilroa

dN

A

airc

raft

NA

in

dust

rial

NA

win

d tu

rbin

esN

A

cons

truc

tion

NA

sh

ops

NA

ne

ighb

ors

SCL-

90

SOM

BL

SCL-

90

SOM

FU

Hos

tilit

yVu

lner

abili

ty

Road

traffi

c no

ise

(Lde

n)-

--

--

--

--

--

--

NA

road

traffi

c.2

23*

--

--

--

--

--

--

NA

railr

oad

.064

*.2

05*

--

--

--

--

--

-

NA

airc

raft

.035

*.2

23*

.139

*-

--

--

--

--

-

NA

indu

stria

l.0

06.1

83*

.080

*.1

11*

--

--

--

--

-

NA

win

d tu

rbin

es-.0

40*

.044

*.0

42*

.058

*0.

71*

--

--

--

--

NA

con

stru

ctio

n.1

01*

.220

*.0

99*

.153

*.1

07*

.013

*-

--

--

--

NA

sho

ps.0

90*

.198

*.0

51*

.066

*.1

24*

.005

.131

*-

--

--

-

NA

nei

ghbo

rs.1

01*

.297

*.1

37*

.167

*.0

66*

-.004

.203

*.0

93*

--

--

-

SCL-

90 S

OM

bas

elin

e.0

06.0

94*

.029

*.0

45*

.045

*.0

11*

.069

*.0

37*

.101

*-

--

-

SCL-

90 S

OM

follo

w u

p.0

11*

.097

*.0

39*

.049

*.0

49*

.013

*.0

79*

.041

*.1

13*

0.55

7*-

--

Hos

tility

.003

.080

*.0

20*

.025

*.0

25*

.005

.045

*.0

17*

.094

*0.

263*

0.23

8*-

-

Vuln

erab

ility

.013

.091

*.0

18*

.034

*0.

13*

.005

.058

*.0

20*

.092

*0.

267*

0.23

1*0.

467*

-

*p<0

.05

Abbr

evia

tions

: Lde

n: a

nnua

l ave

rage

day

-eve

ning

-nig

ht ti

me

road

traf

fic n

oise

; NA:

noi

se a

nnoy

ance

; SCL

-90

SOM

BL:

Sym

ptom

Che

cklis

t-90

som

atiz

atio

n sc

ale

base

line

mea

-su

rem

ent;

SCL-

90 S

OM F

U: S

ympt

om C

heck

list-

90 so

mat

izat

ion

scal

e fo

llow

up

mea

sure

men

t; vu

lner

abili

ty: v

ulne

rabi

lity

to st

ress

7Road traffic noise, blood pressure and heart rate: Pooled analyses of harmonized data from 91,718 participants from LifeLines, EPIC-Oxford, and HUNT3

Wilma L. Zijlema, Yutong Cai, Dany Doiron, Isabel Fortier, Kees de Hoogh, David Morley, John Gulliver, Susan Hodgson, Paul Elliott, BioSHaRE, Timothy Key, Havard Kongsgard, Kristian Hveem, Amadou Gaye, Paul Burton, Anna Hansell, Ronald P. Stolk, Judith G.M. Rosmalen

Submitted

Chapter 7

132

aBsTracT

Background

Exposure to road traffic noise may increase blood pressure and heart rate. It is unclear to what extent exposure to air pollution may influence this relationship.

objectives

We investigated associations between noise, blood pressure and heart rate, with harmonized data from three European cohorts, while taking into account expo-sure to air pollution.

methods

Road traffic noise exposure was assessed using a European noise model based on the Common Noise Assessment Methods in Europe framework (CNOSSOS-EU). Exposure to air pollution was estimated using a European-wide land use regres-sion model. Blood pressure and heart rate were obtained by trained clinical pro-fessionals. Pooled cross-sectional analyses of harmonized data were conducted at the individual level and with random-effects meta-analyses.

results

We analyzed data from 91,718 participants, across the three participating cohorts (mean age 47.0 (±13.9) years). Each 10 dB(A) increase in noise was associated with a 0.95 (95% CI 0.78;1.12) bpm increase in heart rate, and a decrease in blood pressure of 0.07 (95% CI -0.30;0.17) mmHg for systolic and 0.42 (95% CI -0.57;-0.27) mmHg for diastolic blood pressure. Adjustments for PM10 or NO2 attenuated the associations, but remained significant for DBP and HR. Results for BP (but not HR) differed by cohort, with negative associations with noise in Lifelines, no significant associations with EPIC-Oxford and positive associations of noise >60 dB(A) in HUNT3.

conclusion

Our study suggests that road traffic noise may be related to increased heart rate. No consistent evidence for a relation between noise and blood pressure was found.

133

Noise, blood pressure, and heart rate

7

inTrodUcTion

Environmental noise poses a major public health problem [1]. Persistent noise exposure has been associated with increased risks for cardiovascular diseases [2–4]. Noise is generally believed to provoke stress through perceived discomfort [5], and also to result in subconscious activation of stress systems [6,7]. Direct or indirect activation of the sympathetic and endocrine systems is followed by increases in heart rate, blood pressure, and release of stress hormones [8]. Sleep disturbance may also disrupt secretion of stress hormones, affecting metabolism and the cardiovascular system, ultimately resulting in cardiovascular disease [9]. Some studies have found associations between traffic noise and hypertension [10,11], and elevated heart rate [12,13], but others have not [14–16], or only observed associations in specific subgroups (e.g. diabetes patients) [17]. These inconsistencies may arise because of differences between populations, differ-ences in the assignment of noise exposure, the outcome variables or covariates, or differences between statistical approaches used [10].

Another unresolved question is the extent to which exposure to air pollution is involved in the relation between noise and cardiovascular outcomes. Exposure to ambient air pollution, especially particulate matter (PM), is also associated with hypertension [18,19] and cardiovascular morbidity and mortality [20]. As road traffic is the main common source for both noise and air pollution, it is important to distinguish the cardiovascular effects of each. Until recently, studies did not take into account the contemporaneous exposures to traffic-related noise and air pol-lution. A recent systematic review identified nine studies that have assessed the confounding effects of noise and air pollution on cardiovascular health outcomes [21]. Based on these studies, the authors concluded that noise and air pollution probably have independent effects on cardiovascular diseases. However, due to the heterogeneity of the studies that were included and because of the sparsity of the current literature, a final conclusion on the respective roles of air pollution and noise in cardiovascular morbidity cannot yet be drawn. Additional research is needed where effects of traffic-related noise and air pollution on cardiovascular health are investigated simultaneously.

The purpose of this cross-sectional study was to investigate associations between road traffic noise, blood pressure and heart rate, while taking into account expo-sure to ambient air pollution. We investigated these associations with data from three large European cohorts: LifeLines, EPIC-Oxford and HUNT3. We used a har-

Chapter 7

134

monized approach and federated data analyzes to enable comparison of results between different studies and regions.

meTHods

study populations

This study was undertaken within the Biobank Standardisation and Harmonisa-tion for Research Excellence in the European Union (BioSHaRE-EU) project, using a harmonized approach to noise and air pollution exposure, health data and potential confounders. Within BioSHaRE, tools were developed for data harmoni-zation and federated data analyses [22]. Data were obtained from three European cohorts: LifeLines (the Netherlands) [23], EPIC-Oxford (United Kingdom) [24], and HUNT3 (Norway) [25]. LifeLines is a multi-disciplinary prospective popula-tion based cohort study examining the health and health-related behaviors of per-sons living in the North East region of the Netherlands. EPIC-Oxford (the Oxford cohort of the European Prospective Investigation into Cancer and Nutrition) is a nationwide study in the UK aimed at investigating how diet influences the risk for cancers and other chronic diseases. In addition to inclusion of members of the general population, recruitment was focused on including participants with a wide range of dietary habits as well as vegetarians and vegans. HUNT3 (the 3rd survey of the Nord-Trøndelag Health Study) is a prospective population based study from the Nord-Trøndelag County in Norway, examining health related life-style, prevalence and incidence of somatic and mental illness and disease, health determinants, and associations between disease phenotypes and genotypes. All individual participants provided written informed consent and study protocols were approved by the local ethical committees.

exposure assessment

Road traffic noise exposure at individual home addresses was assessed using a European noise model based on Common Noise Assessment Methods in Europe (CNOSSOS-EU) [26]. The CNOSSOS-EU noise modelling framework was developed as a common methodology for noise modelling across Europe [27]. The CNOSSOS-EU framework contains empirically derived equations to determine average noise level based on traffic flow, and sound propagation based on known environmental factors and physical processes. Quantitative data of road networks, traffic flows, land cover, building height, and meteorology were obtained from local sources. Propagation effects such as distance from receiver to the noise source, land cover type, building obstruction, and meteorological conditions are included as sound

135


7

propagation parameters in the model. Traffic data originated from year 2009 and landcover data from 2006. Model discrimination between noisier and quieter areas gave Spearman’s rank = 0.75; p <0.001, but the predicted noise levels have relatively large errors (root mean square error (RMSE) = 4.46 dB(A)) [26]. We used Lden (day-evening-night time period of 24 hours) as an indicator of road traf-fic noise in the current study. Lden is the average A-weighted noise level, estimated over a 24 hour period, with a 10 dB penalty added to the night (23.00–07.00 hours), and a 5 dB penalty added to the evening period (19.00–23.00 hours). The penalties are added to indicate people’s extra sensitivity to noise during the night and evening.

Exposure to particulate matter with a diameter ≤10 µm (PM10) and nitrogen di-oxide (NO2) at individual home addresses was estimated using a land use regres-sion model for Western Europe [28]. Air pollution estimates were based on input data for the years 2006-2007: land use, road network, and other topographic data, and satellite-derived estimates of air pollutants ground level concentrations for PM2.5 (as an indicator of PM10) and NO2 [28].

Blood pressure and heart rate measurements

Systolic and diastolic blood pressure and heart rate measurements were conduct-ed by trained clinical professionals within each participating cohort. In LifeLines, measurements took place at the research facilities according to a standardized protocol. Blood pressure and heart rate were measured 10 times within a period of 10 minutes, using an automated Dinamap Monitor (GE Healthcare, Freiburg, Germany). The final two accurate recordings were averaged for systolic and dia-stolic blood pressure and heart rate. In EPIC-Oxford, blood pressure and heart rate measurements took place at the participant’s general practice. No standardized method of blood pressure measurement was used and only a single measurement was taken [29]. In HUNT3, blood pressure and heart rate were measured using a Dinamap 845XT (Critikon, Tample, FL, USA), and measurements took place at the research facilities. Measurements were carried out three times, and the average of the second and third recording was used in the analyses.

covariates

Data on age, sex, educational level, alcohol intake, and medication use were har-monized according to the DataSHaPER (DataSchema and Harmonization Platform for Epidemiological Research) methodology [30,31]. DataSHaPER provides a tem-plate to facilitate harmonization and pooling of data between studies. First, ‘target variables’ needed to answer our research question were identified. Relevant data were then identified within all studies, and the potential for harmonization was

Chapter 7

136

evaluated. When harmonization was deemed possible, cohort-specific data were transformed into a common, harmonized format in order to be co-analyzed.

Educational level was defined as highest level of education completed by the participant. Categories were adapted from the UNESCO Revision of the Interna-tional Standard Classification of Education. Education data were obtained from the question “What is your highest completed education?” in LifeLines and EPIC-Oxford. Educational level was only questioned in HUNT2 (the 2nd survey of the Nord-Trøndelag Health Study) but not in HUNT3; we used such data for the same participants who answered this question in HUNT2. Alcohol use was defined as grams of alcohol consumed on average per week. If information about serving size for different drinks was not mentioned, then the average serving size was imputed to allow derivation of grams of alcohol. Antihypertensive medication use (yes/no) was either self-reported by a question targeting the use of antihypertensive medication (HUNT3) or extracted from a list of medications using the following Anatomical Therapeutic Chemical (ATC) Codes (C02, C03, C04, C07, C08, C09) or equivalent in other classifications (LifeLines). Data on medication use were avail-able in LifeLines and HUNT3, not in EPIC-Oxford. Address history was available from municipal registries, and were available in LifeLines and HUNT3.


We investigated the associations between road traffic noise and the following three cardiovascular outcomes: systolic blood pressure (SBP), diastolic blood pressure (DBP), and heart rate (HR). The associations were investigated with two statistical approaches:

1. Pooled linear regression analyses of harmonized data for the three cohorts were conducted at the individual level using the DataSHIELD approach [32]. DataSHIELD enables federated analysis of multiple studies by analyzing individ-ual-level harmonized data from each cohort without physically pooling the data, which stay behind the firewalls of the cohort’s host computers. As a result, indi-vidual data can be simultaneously analyzed without transferring data externally. DataSHIELD offers a solution to practical, ethical, and legal issues associated with data sharing and analysis [33]. We used the ds.glm function in DataSHIELD (ver-sion 4.1.0) (http://cran.datashield.org/), which is comparable to the glm function in the R statistical environment [34].

2. Cohort-specific linear regression analyses were also undertaken using the ds.glm function in DataSHIELD (version 4.1.0). Subsequently, study specific as-sociations were combined using random effects meta-analyses (using the DerSi-monian-Laird estimator) to account for heterogeneity between the cohorts [35]. The I2 statistic [36] was used to quantify heterogeneity of cohort specific results.

137


7

Meta-analyses were performed in R (version 3.1.2) using the rma function from the metaphor package [37].

Associations with 24-hour (Lden) road traffic noise were initially adjusted for age, sex, educational level, alcohol use, and additionally for PM10, or NO2. The pooled linear regression analyses (approach 1) were also adjusted for cohort, to account for study-specific effects. Effect estimates are presented as regression coefficients with 95% confidence intervals (CI) per 10 dB(A). In addition, pooled and cohort-specific associations between road traffic noise, BP, and HR were analyzed with noise classified into categories of <55 dB(A), 55-60 dB(A), and ≥60 dB(A).

We performed sensitivity analyses for evaluation of effect modification of the associations by sex, age, and use of antihypertensive medication. First, the main effects (e.g. Lden and antihypertensive medication use) and their interaction were included in the models. If statistical significance (p <0.05) of the interaction term was observed, analyses were stratified for that variable. We also performed sen-sitivity analyses to explore the impact of residential mobility by restricting the analyses to participants that lived at their current address for ≥2 and ≥5 years. Sensitivity analyses for medication use and residential mobility were undertaken in LifeLines and HUNT3, since data on medication use and address history were not available in EPIC-Oxford.

resUlTs

Data from 91,718 participants, with a mean age of 47.0 (±13.9) years, were avail-able for this study. Pooled and cohort specific population characteristics are sum-marized in Table 1. Median level of road traffic noise (Lden) exposure was highest in EPIC-Oxford (54.9 dB(A)) and lowest in HUNT3 (49.4 dB(A)). Median levels of PM10 varied between 11.1 (HUNT3) and 23.6 (LifeLines) µg/m3, and NO2 levels varied between 11.7 (HUNT3) and 25.6 (EPIC-Oxford) µg/m3 (Table 1 and Fig-ure 1). Spearman rank correlations between road traffic noise and PM10, and road traffic noise and NO2, were r=0.06 and 0.06 respectively in EPIC-Oxford, r=0.07 and -0.05 in HUNT3, and r=0.40 and 0.46 in LifeLines (all p < 0.001). Mean levels of SBP ranged from 124.4 to 132.8 mmHg, with highest mean level observed in HUNT3. Mean levels of DBP and HR were reasonably similar between the cohorts (Table 1). Use of antihypertensive medication was 11.7% in LifeLines and 26.0% in HUNT3.

Chapter 7

138

results from pooled linear regression analyses, cohort-specific analyses, and meta-analyses

Pooled linear regression analyses, adjusted for age, sex, education, alcohol use, and cohort, showed significant associations between noise (per 10dB(A)) and DBP (β= -0.42; 95% CI -0.57; -0.27 mmHg), and HR (β= 0.95; 95% CI 0.78; 1.12 bpm) but not SBP (β= -0.07; 95% CI -0.30; 0.17 mmHg) (Table 2). Additional adjustment for PM10 or NO2 resulted in an attenuation of the associations, but associations remained significant for DBP and HR (Table 2). Cohort-specific linear regression analyses showed heterogeneous results. In LifeLines, models unadjusted for air pol-lution yielded significant negative associations between noise and SBP (Figure 2), whereas in EPIC-Oxford and HUNT3 positive associations were found, although not statistically significant. For DBP, results for LifeLines and EPIC-Oxford showed similar negative associations, while these were null for HUNT3 (Figure 2). Associa-tions between noise and HR were consistently positive in all three cohorts, although not statistically significant in EPIC-Oxford (Figure 2). Confounding by air pollution was strongest in LifeLines, and less pronounced in EPIC-Oxford and HUNT3. When we combined the cohort-specific estimates using random-effects meta-analyses, associations were generally smaller and less precise, as observed from the wider confidence intervals, if compared to the results from the pooled individual data regression analyses. In contrast with the individual-level regression analyses, pooled estimates for the association between noise and DBP were not statistically

Table 1. Pooled and cohort specific population characteristics.

Pooled LifeLines EPIC-Oxford HUNT3

n 91,718 59,226 11,475 21,017

Age (years); mean (SD) 47.0 (13.9) 43.6 (12.3) 46.0 (14.2) 56.9 (16.1)

Females, % 59.3 57.9 76.0 54.3

Educational level

primary and secondary, % 67.1 67.5 50.2 75.5

tertiary, % 32.9 32.5 49.8 24.5

Alcohol use (grams/week); median (IQR) 31.6 (71.2) 36.1 (81.6) 40.9 (74.6) 14.0 (40.0)

SBP (mmHg); mean (SD) 126.2 (16.8) 124.4 (15.2) 123.4 (19.0) 132.8 (18.7)

DBP (mmHg); mean (SD) 74.0 (10.1) 73.4 (9.3) 76.1 (11.0) 74.5 (11.3)

HR (bpm); mean (SD) 71.2 (11.1) 71.7 (10.9) 72.0 (10.8) 69.4 (11.5)

Antihypertensive medication, % 15.9 11.7 NA 26.0

Lden (dB(A)); median (IQR) 53.5 (4.6) 54.7 (4.3) 54.9 (7.2) 49.4 (6.0)

PM10 (µg/m3); median (IQR) 20.5 (2.3) 23.6 (2.4) 21.9 (3.0) 11.1 (1.5)

NO2 (µg/m3); median (IQR) 19.2 (8.0) 20.6 (8.8) 25.6 (9.5) 11.7 (5.0)

Length of residence (years); median (IQR) 11.6 (14.4) 10.0 (12.0) NA 16.0 (21.0)

Abbreviations: SD=standard deviation; SBP=systolic blood pressure; DBP=diastolic blood pressure; HR=heart rate; bpm=beats per minute; Lden=24 hour noise estimate; dB(A)=decibels A; IQR=interquartile range; PM10= particulate matter with a diameter ≤10 µm; NO2=nitrogen dioxide; NA=not available for cohort.

139


7

figure 1 Distribution of estimated annual average road traffic noise (Lden) and air pollution levels for LifeLines, EPIC-Oxford, and HUNT3. Median, 25th and 75th percentiles are shown in the box, whiskers indicate 5% and 95% limits. Abbreviations: Lden = day-evening-night time annual average road traffic noise; PM10 = particulate matter with aerodynamic diameter ≤10 µm; NO2 = nitrogen dioxide.

Table 2. Estimated associations between road traffic noise (Lden) per 10 dB(A) and systolic blood pres-sure (mmHg), diastolic blood pressure (mmHg), and heart rate (bpm). Pooled associations were esti-mated with DataSHIELD, n=91,718. Models were adjusted for age, sex, cohort, educational level, and alcohol use. Models were additionally adjusted for PM10 or NO2 as specified in the table.

β per 10 dB(A) (95% confidence interval)

unadjusted for PM10 or NO2 adjusted for PM10 adjusted for NO2

SBP -0.07 (-0.30; 0.17) -0.06 (-0.31; 0.18) 0.07 (-0.17;0.32)

DBP -0.42 (-0.57;-0.27) -0.37 (-0.52; -0.22) -0.27 (-0.43; -0.12)

HR 0.95 (0.78; 1.12) 0.71 (0.54; 0.89) 0.64 (0.46; 0.82)

Abbreviations: SBP=systolic blood pressure; DBP=diastolic blood pressure; HR=heart rate; bpm=beats per minute; Lden=24 hour noise estimate; dB(A)=decibels A; PM10= particulate matter with a diameter ≤10 µm; NO2=nitrogen dioxide.

Chapter 7

140

Cohort Study Effect 95% CI weight size

SBP


SBP


SBP

figure 2

141


7


DBP


DBP


DBP

figure 2 (continued)

Chapter 7

142


HR


HR


HR

figure 2 (continued)

143


7

significant in the random-effects meta-analyses (Table 2 and Figure 2). I2 values indicated moderate to high heterogeneity among cohort-specific results for SBP (78.5% unadjusted for air pollution, 80.9% adjusted for NO2, and 64.0% adjusted for PM10) and DBP (68.2% unadjusted for air pollution, 65.6% adjusted for NO2, and 19.8% adjusted for PM10). Heterogeneity of the results for HR ranged from 84.0% unadjusted for air pollution, to 0% adjusted for NO2 or PM10.

Pooled analyses with the categorical noise variable and SBP, DBP, and HR resulted in similar conclusions as with the analyses with noise as a continuous variable (Table 3), with the exception that in HUNT3, the highest noise category (≥60 dB(A)) was associated with statistically significantly higher SBP (β= 4.69; 95% CI 2.44; 6.94 mmHg) and DBP (β= 1.73; 95% CI 0.36; 3.09 mmHg) compared with the lowest noise category (<55 dB(A)) (Table 3).

figure 2 Cohort specific and pooled associations between road traffic noise (Lden) and systolic blood pressure (SBP, in mmHg), diastolic blood pressure (DBP, in mmHg), and heart rate (HR, in beats per minute) per 10 dB(A). Pooled associations were estimated with random effects meta-analyses. Total n=91,718; LifeLines n=59,226; EPIC-Oxford n=11,475; HUNT3 n=21,017. Models were adjusted for age, sex, cohort, educational level, and alcohol use. Models were additionally adjusted for NO2 or PM10 if specified.

table 3. Estimated associations between road traffic noise (Lden) categories (reference category is Lden <55 dB(A)) and systolic blood pressure (mmHg), diastolic blood pressure (mmHg), and heart rate (bpm). Associations were estimated with DataSHIELD. Models were adjusted for age, sex, cohort (only pooled analyses), educational level, and alcohol use.

Pooled (n=91,718)SBP DBP HR

Lden 55-60 dB(A) (n=25,126) -0.43 (-0.67; -0.20) -0.39 (-0.54; -0.25) 0.56 (0.38; 0.73)

Lden >=60 dB(A) (n=9,614) -0.19 (-0.52; 0.14) -0.62 (-0.83; -0.41) 0.89 (0.63; 1.14)

LifeLines (n=59,226)SBP DBP HR

Lden 55-60 dB(A) (n=19,464) -0.46 (-0.70; -0.22) -0.39 (-0.55; -0.24) 0.73 (0.53; 0.92)

Lden >=60 dB(A) (n=8,099) -0.58 (-0.91; -0.24) -0.58 (-0.79; -0.37) 1.25 (0.99; 1.52)

EPIC-Oxford (n=11,475)SBP DBP HR

Lden 55-60 dB(A) (n=4,241) -0.47 (-1.10; 0.17) -0.23 (-0.63; 0.18) 0.39 (-0.03; 0.81)

Lden >=60 dB(A) (n=1,279) 0.03 (-0.95; 1.00) -0.41 (-1.03; 0.21) 0.28 (-0.37; 0.92)

HUNT3 (n=21,017)SBP DBP HR

Lden 55-60 dB(A) (n=1,421) -0.52 (-1.47; 0.42) -0.07 (-0.64; 0.50) 0.59 (-0.03; 1.20)

Lden >=60 dB(A) (n=236) 4.69 (2.44; 6.94) 1.73 (0.36; 3.09) 1.11 (-0.35; 2.58)

Abbreviations: SBP=systolic blood pressure; DBP=diastolic blood pressure; HR=heart rate; bpm=beats per min-ute; Lden=24 hour noise estimate; dB(A)=decibels A.

Chapter 7

144

sensitivity analyses

Since we observed significant interactions with sex and age (SBP, DBP, HR), and use of antihypertensive medication (DBP, HR), we carried out stratified analyses by these variables. Results were stronger for males than females, and for partici-pants aged 65 years and older (n=9,360), road traffic noise was non-significantly associated with both higher SBP and DBP (Table 4). Comparing participants

table 4. Estimated associations between road traffic noise (Lden) per 10 dB(A) and systolic blood pres-sure (mmHg), diastolic blood pressure (mmHg), and heart rate (bpm), stratified for age, sex, medica-tion use, and length of residence. Pooled associations were estimated with DataSHIELD. Models were adjusted for age, sex, cohort, educational level, and alcohol use (except when stratified for that variable).

β per 10 dB(A) (95% confidence interval)

SBP males (n=37,304)a -0.36 (-0.72; -0.005)

females (n=54,414)a 0.16 (-0.14; 0.47)

<45 years (n=42,992)a -0.66 (-0.96; -0.37)

45-55 years (n=28,292)a -0.33 (-0.80; 0.13)

55-65 years (n=14,049)a -0.62 (-1.32; 0.07)

≥65 years (n=9,360)a 0.75 (-0.20; 1.70)

length of residence ≥2 years (n=75,573)b -0.12 (-0.38; 0.14)


DBP males (n=37,304)a -0.55 (-0.79; -0.32)

females (n=54,414)a -0.33 (-0.52; -0.14)

<45 years (n=42,992)a -0.26 (-0.45; -0.06)

45-55 years (n=28,292)a -0.15 (-0.45; 0.15)

55-65 years (n=14,049)a -0.27 (-0.68; 0.14)

≥65 years (n=9,360)a 0.13 (-0.39; 0.65)

no AHT (n=68,738)b -0.38 (-0.55; -0.22)

AHT (n=11,503)b -0.08 (-0.55; 0.39)

length of residence ≥2 years (n=75,573)b -0.33 (-0.49; -0.17)


HR males (n=37,304)a 1.15 (0.87; 1.43)

females (n=54,414)a 0.82 (0.61; 1.04)

<45 years (n=42,992)a 0.83 (0.58; 1.07)

45-55 years (n=28,292)a 1.12 (0.85; 1.51)

55-65 years (n=14,049)a 1.02 (0.56; 1.47)

≥65 years (n=9,360)a 0.81 (0.25; 1.36)

no AHT (n=68,738)b 1.04 (0.85; 1.24)

AHT (n=11,503)b 0.82 (0.29; 1.36)

length of residence ≥2 years (n=75,573)b 1.04 (0.84; 1.23)

length of residence ≥5 years (n=62,296)b 1.04 (0.82; 1.26)

Abbreviations: SBP=systolic blood pressure; DBP=diastolic blood pressure; HR=heart rate; bpm=beats per min-ute; Lden=24 hour noise estimate; dB(A)=decibels A; AHT=antihypertensive treatment.a Pooled analyses for LifeLines, EPIC-Oxford, and HUNT3b Pooled analyses for LifeLines and HUNT3

145


7

without (n= 68,738 LifeLines and HUNT3) and with (n=11,503) antihypertensive medication use, associations were more pronounced in non-users (Table 4). As-sociations remained in the same directions as in the total sample for participants that lived ≥2 and ≥5 years at their current home address (n= 75,573 and 62,296 respectively in LifeLines and HUNT3 combined), but the association between road traffic noise and DBP was no longer statistically significant for participants living longer than 5 years at their current address (Table 4).

discUssion

In our study of three European population-based cohort studies including 91,718 participants, exposure to road traffic noise was related to an elevated resting HR. No consistent evidence for a relation between road traffic noise and BP was found. Sensitivity analyses indicated that associations between road traffic noise and BP and HR may be more pronounced in men, and in individuals older than 65 years.

Unlike some previous studies we did not find an association between road traf-fic noise and elevated BP. We found no evidence for an association between road traffic noise and SBP, and we observed negative associations between road traffic noise and DBP. The negative association with DBP was observed in younger age groups, those who are not on antihypertensive medication, and those who are less likely to have lived in their house for five years. These results are unexpected, but we observed marked heterogeneity between cohorts with negative associations appearing to be driven by associations in the LifeLines Cohort Study, a younger cohort from the Netherlands.

One possible explanation for lack of clear associations with BP is exposure mis-classification. In a study in Spain by Foraster et al. (2014), nighttime traffic noise was estimated for both outdoor and indoor exposure. While indoor traffic noise was associated with hypertension and increased SBP, associations with outdoor noise were less consistent. Indoor noise estimates may be a better reflection of an individual’s ‘true’ exposure. We modelled at the national scale, which resulted in a less detailed noise exposure estimation than might have been possible at the local scale, where higher resolution input data are generally available and result in higher resolution model output [26]. A study investigating associations between road traffic noise, PM2.5, and hypertension in a German cohort, found no associations in their study population from the Greater Augsburg region, but did find associations for those living in the city of Augsburg, for which they had a more detailed noise exposure assessment [18]. However, we did observe associations between road traffic noise and HR suggesting that the model was

Chapter 7

146

sufficient to allow us to discriminate exposure contrasts across our population. Van Kempen and Babisch (2012) reported a positive association between road traffic noise and hypertension based on their meta-analysis of 24 studies [10]. The pooled effect estimates increased with increasing years of residence in the van Kempen and Babisch meta-analysis (2012), while we observed that the as-sociation between road traffic noise and DBP in the subgroup that lived ≥5 years at their current home address was no longer statistically significant. This may be related to decreased statistical power due to a smaller sample size, or may indicate potential unmeasured confounding. They observed stronger associations in studies investigating men, which is similar to our findings.

Most of the previous studies that have investigated relations between noise and heart rate, have done so in experimental settings [12,13,38] or in children [14,15,39]. Some of these studies also reported inconsistent results for BP and HR [14,39]. Additional general population studies are needed to replicate our results for noise and heart rate.

We found that additional adjustment for PM10 or NO2 generally resulted in at-tenuation of the associations between noise, BP, and HR. However, associations with DBP and HR remained statistically significant, indicating that these asso-ciations cannot be entirely explained by exposure to air pollution. Foraster et al. (2014) reported that the associations between indoor noise and SBP were not confounded by NO2, while there was confounding by NO2 in the analyses with outdoor noise, and Babisch et al. (2014) reported that inclusion of air pollutants in their analyses only slightly diminished the associations between noise and hypertension. In another study from the Netherlands, no evidence for confound-ing by air pollution was observed [40]. Exposure to air pollution may have some effect on the relation between noise, BP, and HR, but more studies are needed to disentangle these effects.

Our study has several strengths and limitations. A major strength of this study is its sample size. With data from 91,718 participants, our study is one of the largest studies examining noise and cardiovascular effects to date. The use of DataSHIELD to conduct virtual pooling increased power further and allowed us to examine interactions in a number of subgroups. Data originated from three different study regions, enabling comparison across different areas. Another strength of the study was harmonization of both exposure estimation and outcomes. However, there was still large heterogeneity between cohorts so we have also reported results for each cohort separately. At the time of our study, we were not able to use mixed models that take into account heterogeneous multi-cohort data in the

147


7

virtually pooled analyses, as this statistical technique was not yet implemented in DataSHIELD. Selection of participants in each cohort may have affected our results – those in Lifelines were younger, while recruitment in EPIC-Oxford was focused on including vegetarians and vegans. Because of this selection for po-tentially ‘health conscious’ persons, participants in EPIC-Oxford may be healthier than participants in the other cohorts [41,42]. Associations for SBP and HR were smaller in EPIC-Oxford than for the other cohorts.

Although exposure ranges of road traffic noise and ambient air pollution of the three cohorts combined were relatively large, median levels were relatively low, if compared to other European cohorts (e.g.[43]). It is therefore unclear to what extent the conclusions of our study can be generalized to regions with higher levels of noise and air pollution. Furthermore, although large efforts were made to harmonize the study data, blood pressure and heart rate data could not be completely harmonized. Measurement procedures differed between the cohorts, and could have resulted in variations in outcome data and measurement bias. Blood pressure and heart rate measurements in EPIC-Oxford were conducted only once, while they are known to be highly variable. In addition, measurements were standardized in LifeLines and HUNT3, but not in EPIC-Oxford. Measurement errors reduce the power to detect an association, and will bias associations to-ward the null. Furthermore, data harmonization of education level resulted in a harmonized variable with only two categories (primary and secondary education; tertiary education), resulting in information loss of the education data. We were not able to use any other data to account for socioeconomic status, and residual confounding by socioeconomic status may therefore remain.

In summary, pooling harmonized data from multiple European cohorts allowed the use of large sample sizes with a wide noise exposure range. In our study with 91,718 participants from the Netherlands, the UK, and Norway, road traffic noise was associated with elevated resting heart rate. No consistent evidence for a rela-tion between road traffic noise and blood pressure was observed and findings were heterogeneous between cohorts.

Chapter 7

148

references

1. WHO. Burden of disease from environmental noise. Quantification of healthy life years lost in Europe. WHO European Centre for Environment Health, Bonn Office; 2011.

2. Halonen JI, Hansell AL, Gulliver J, Morley D, Blangiardo M, Fecht D, et al. Road traffic noise is associated with increased cardiovascular morbidity and mortality and all-cause mortality in London. Eur Heart J. 2015; doi:10.1093/eurheartj/ehv216

3. Van Kempen E, Kruize H, Boshuizen HC, Ameling CB, Staatsen BAM, de Hollander AEM. The association between noise exposure and blood pressure and ischemic heart disease: A meta-analysis. Environ Health Perspect. 2002;110: 307–317.

4. Vienneau D, Schindler C, Perez L, Probst-Hensch N, Röösli M. The relationship between trans-portation noise exposure and ischemic heart disease: A meta-analysis. Environ Res. 2015;138: 372–380. doi:10.1016/j.envres.2015.02.023

5. Bluhm G, Nordling E, Berglind N. Road traffic noise and annoyance—an increasing environ-mental health problem. Noise Health. 2004;6: 43–9.

6. Babisch W, Fromme H, Beyer A, Ising H. Increased catecholamine levels in urine in subjects ex-posed to road traffic noise: The role of stress hormones in noise research. Environ Int. 2001;26: 475–481. doi:10.1016/S0160-4120(01)00030-7

7. Selander J, Bluhm G, Theorell T, Pershagen G, Babisch W, Seiffert I, et al. Saliva cortisol and exposure to aircraft noise in six European countries. Environ Health Perspect. 2009;117: 1713–1717. doi:10.1289/ehp.0900933

8. Chrousos GP, Gold PW. The concepts of stress and stress system disorders. Overview of physical and behavioral homeostasis. JAMA. 1992;267: 1244–52.

9. Münzel T, Gori T, Babisch W, Basner M. Cardiovascular effects of environmental noise exposure. Eur Heart J. 2014;35: 829–36. doi:10.1093/eurheartj/ehu030

10. Van Kempen E, Babisch W. The quantitative relationship between road traffic noise and hypertension: a meta-analysis. J Hypertens. 2012;30: 1075–1086. doi:10.1097/HJH.0b013e328352ac54; 10.1097/HJH.0b013e328352ac54

11. Foraster M, Künzli N, Aguilera I, Rivera M, Agis D, Vila J, et al. High blood pressure and long-term exposure to indoor noise and air pollution from road traffic. Environ Health Perspect. 2014;122: 1193–200. doi:10.1289/ehp.1307156

12. Holand S, Girard A, Laude D, Meyer-Bisch C, Elghozi JL. Effects of an auditory startle stimulus on blood pressure and heart rate in humans. J Hypertens. 1999;17: 1893–7.

13. Raggam RB, Cik M, Höldrich RR, Fallast K, Gallasch E, Fend M, et al. Personal noise ranking of road traffic: subjective estimation versus physiological parameters under laboratory condi-tions. Int J Hyg Environ Health. 2007;210: 97–105. doi:10.1016/j.ijheh.2006.08.007

14. Belojevic G, Jakovljevic B, Stojanov V, Paunovic K, Ilic J. Urban road-traffic noise and blood pressure and heart rate in preschool children. Environ Int. 2008;34: 226–31. doi:10.1016/j.envint.2007.08.003

15. Van Kempen E, van Kamp I, Fischer P, Davies H, Houthuijs D, Stellato R, et al. Noise exposure and children’s blood pressure and heart rate: the RANCH project. Occup Environ Med. 2006;63: 632–639. doi:10.1136/oem.2006.026831

16. Paunovic K, Stansfeld S, Clark C, Belojevic G. Epidemiological studies on noise and blood pressure in children: Observations and suggestions. Environ Int. 2011;37: 1030–1041. doi:10.1016/j.envint.2011.03.017

149


7

17. Dratva J, Phuleria HC, Foraster M, Gaspoz J-M, Keidel D, Kuenzli N, et al. Transportation Noise and Blood Pressure in a Population-Based Sample of Adults. Environ Health Perspect. 2012;120: 50–55. doi:10.1289/ehp.1103448

18. Babisch W, Wolf K, Petz M, Heinrich J, Cyrys J, Peters A. Associations between traffic noise, particulate air pollution, hypertension, and isolated systolic hypertension in adults: the KORA study. Environ Health Perspect. 2014;122: 492–8. doi:10.1289/ehp.1306981

19. Schwartz J, Alexeeff SE, Mordukhovich I, Gryparis A, Vokonas P, Suh H, et al. Association between long-term exposure to traffic particles and blood pressure in the Veterans Administration Nor-mative Aging Study. Occup Environ Med. 2012;69: 422–7. doi:10.1136/oemed-2011-100268


21. Tétreault L-F, Perron S, Smargiassi A. Cardiovascular health, traffic-related air pollution and noise: are associations mutually confounded? A systematic review. Int J Public Health. 2013;58: 649–66. doi:10.1007/s00038-013-0489-7

22. Doiron D, Burton P, Marcon Y, Gaye A, Wolffenbuttel BH, Perola M, et al. Data harmonization and federated analysis of population-based studies: the BioSHaRE project. Emerg Themes Epidemiol. 2013;10: 12. doi:10.1186/1742-7622-10-12


24. Davey GK, Spencer EA, Appleby PN, Allen NE, Knox KH, Key TJ. EPIC-Oxford: lifestyle character-istics and nutrient intakes in a cohort of 33 883 meat-eaters and 31 546 non meat-eaters in the UK. Public Health Nutr. 2003;6: 259–69. doi:10.1079/PHN2002430

25. Krokstad S, Langhammer A, Hveem K, Holmen TL, Midthjell K, Stene TR, et al. Cohort Profile: the HUNT Study, Norway. Int J Epidemiol. 2013;42: 968–77. doi:10.1093/ije/dys095


27. Kephalopoulos S, Paviotti M, Anfosso-Lédée F, Van Maercke D, Shilton S, Jones N. Advances in the development of common noise assessment methods in Europe: The CNOSSOS-EU frame-work for strategic environmental noise mapping. Sci Total Environ. 2014;482-483: 400–10. doi:10.1016/j.scitotenv.2014.02.031


29. Appleby PN, Davey GK, Key TJ. Hypertension and blood pressure among meat eaters, fish eat-ers, vegetarians and vegans in EPIC-Oxford. Public Health Nutr. 2002;5: 645–654. doi:10.1079/PHN2002332



Chapter 7

150

32. Wolfson M, Wallace SE, Masca N, Rowe G, Sheehan NA, Ferretti V, et al. DataSHIELD: resolving a conflict in contemporary bioscience-performing a pooled analysis of individual-level data without sharing the data. Int J Epidemiol. 2010;39: 1372–1382. doi:10.1093/ije/dyq111


34. R Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2013.

35. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials. 1986;7: 177–88. 36. Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses.

BMJ. 2003;327: 557–60. doi:10.1136/bmj.327.7414.557 37. Viechtbauer W. Conducting meta-analyses in R with the metafor package. J Stat Softw. 2010;36:

1–48. 38. Goyal S, Gupta V, Walia L. Effect of noise stress on autonomic function tests. Noise Health. 12:

182–6. doi:10.4103/1463-1741.64976 39. Regecová V, Kellerová E. Effects of urban noise pollution on blood pressure and heart rate in

preschool children. J Hypertens. 1995;13: 405–12. 40. De Kluizenaar Y, Gansevoort RT, Miedema HM, de Jong PE. Hypertension and road traffic noise

exposure. J Occup Environ Med. 2007;49: 484–492. doi:10.1097/JOM.0b013e318058a9ff 41. Crowe FL, Appleby PN, Allen NE, Key TJ. Diet and risk of diverticular disease in Oxford cohort

of European Prospective Investigation into Cancer and Nutrition (EPIC): prospective study of British vegetarians and non-vegetarians. BMJ. 2011;343: d4131.

42. Crowe FL, Appleby PN, Travis RC, Key TJ. Risk of hospitalization or death from ischemic heart disease among British vegetarians and nonvegetarians: results from the EPIC-Oxford cohort study. Am J Clin Nutr. 2013;97: 597–603. doi:10.3945/ajcn.112.044073


8General Discussion

155

General Discussion

8

The aim of this dissertation was to investigate whether the living environment, and in particular traffic-related noise and air pollution, adversely affects health. A part of this research was undertaken in multiple European cohort studies, from which data were harmonized and pooled. Our secondary aim was to evaluate the advantages and disadvantages of such combining of multiple cohort studies for environmental epidemiology.In this chapter we will discuss the main findings of this dissertation, the meth-odological strengths and limitations, the implications of our study findings, and directions for future research.

sUmmary of main findings

relations between the environment and health

The title of this dissertation is “(un)healthy in the city: adverse health effects of traffic-related noise and air pollution.” We cannot simply assume that city living is associated with increased exposure to environmental stressors, and therefore adversely affects the health of urban residents. This was underlined by our study described in Chapter 3, where we investigated the associations of urbanity and four disorders with a large burden of disease. Our results suggest a differential health impact of urbanity according to type of disease. Living in an urban envi-ronment appears to be associated with better cardiometabolic health, but with worse respiratory function and mental health. These findings were in agreement with previous research [1–4]. The question that followed was how these urban-rural health differences could be explained. We postulated what underlying mechanisms could help explain the differential associations between health and degree of urbanity. Previous research has given numerous indications of harm-ful environmental exposures associated with city living. Health differences may arise because of environmental factors to which people are directly exposed [5]. LifeLines participants living in urban neighborhoods had highest exposure to air pollution (nitrogen dioxide; Figure 1) and road traffic noise (Figure 2), compared to participants in neighborhoods of lower degree of urbanity. Differences in lifestyle may also underlie urban-rural health differences, and may occur when one environment is, for example, more suitable for physical activity than another [6]. Furthermore, selective migration, for instance when specific environments attract or repel persons with certain health-related characteristics, has also been proposed as a mechanism explaining urban-rural health differences [5].

Chapter 8

156

0

5

10

15

20

25


semi-rural rural

µg/m

3

figure 1. Median exposure to ambient nitrogen dioxide (NO2) according to degree of urbanity within the LifeLines Cohort Study.

50

51

52

53

54

55

56

57

58

59

60

61


semi-rural rural

dB(A

)

figure 2. Median exposure to road traffic noise (Lden; average A-weighted noise level estimated over a 24 hour period) according to degree of urbanity within the LifeLines Cohort Study.

157

General Discussion

8

In Chapter 3, it was found that living in more urbanized areas was related to a decreased lung function and impaired mental health. In line with the differential lifestyle hypothesis, we found that individuals from urban areas were more often current smokers, and this may have influenced the association between urbanity and lung function. Our findings from Chapter 3 may also be explained by increased exposure to air pollution in urban areas. It is well established that air pollution can lead to a variety of adverse health effects [7], including decreased respiratory health [8,9]. Recently, research has provided some indications that exposure to air pollution can adversely affect the brain [10–16], and that it may be associated with impaired mental health [17–20]. We studied the association between expo-sure to air pollution and depressed mood (Chapter 4), but found no consistent evidence to support this hypothesis. Differential exposure to air pollution may explain the urban-rural differences in respiratory health, but our findings do not provide support for such an explanation for the urban-rural differences in depres-sion. Next to air pollution, environmental noise may also play a role in impaired neurocognitive function, mood disorders, and neurodegenerative disease [21]. We did not study road traffic noise as a primary risk factor for depressed mood, but we did adjust our analyses for air pollution and depressed mood for road traffic noise. The adjusted associations between road traffic noise and depressed mood were mostly positive, but not statistically significant. There may be a link between noise and mood disorders, but this cannot be concluded with certainty from our study.

In contrast to decreased respiratory function and worsened mental health, urban living was related to a lower risk for cardiometabolic disease (Chapter 3). However, we observed that road traffic noise was related to elevated resting heart rate (Chapter 7), suggesting that individuals from urban areas, that are more highly exposed to noise, may be at risk for cardiovascular disease. Road traffic noise was, however, not associated with elevated blood pressure in the included cohorts. Exposure to air pollution may also be related to adverse cardiovascular effects, including hypertension and cardiovascular disease [22,23]. We therefore adjusted our analyses with road traffic noise, blood pressure, and heart rate for air pollution exposure. However, associations between exposure to noise, blood pressure and heart rate were only partly explained by exposure to air pollution. The adjusted associations between air pollution and blood pressure were nega-tive and mostly statistically significant, while positive and statistically significant associations between air pollution and heart rate were observed. These findings suggest a possible link between exposure to air pollution and cardiovascular out-comes, but since we found opposite relations for blood pressure and heart rate, these findings should be interpreted with caution. Other risk factors, for example

Chapter 8

158

lifestyle factors, that are more strongly related to cardiometabolic health, are probably more likely to explain urban-rural differences in cardiometabolic health. However, smoking and the number of days on which participants were physically active could not explain the association between urbanity and metabolic syn-drome, suggesting that these lifestyle differences were not responsible for the urban-rural differences in metabolic syndrome. Also, while findings in Chapter 7 relate to cardiovascular outcomes, findings from Chapter 3 comprise a broader health outcome: the metabolic syndrome. This is a clustering of risk factors as-sociated with atherosclerosis and coronary heart disease, including increased waist circumference, elevated triglycerides, low high density lipoprotein (HDL) cholesterol, hypertension, and elevated fasting glucose levels [24]. Environmental risk factors may have differential impacts on each of the components of metabolic syndrome.

Common somatic symptoms have also been related to various environmental exposures [25,26]. In Chapter 6, we found that exposure to road traffic noise and the reporting of common somatic symptoms, such as headache, chest pain, and breathing trouble, were not related. Annoyance from road traffic noise, however, was related to increased symptom reporting. We found that this relationship was only for a small part influenced by the personality facets hostility and vulner-ability to stress, suggesting that the simultaneous reporting of increased annoy-ance and symptoms is not due to a general tendency to be vulnerable to stress [27]. The correlation between modelled road traffic noise and road traffic noise annoyance was low. This indicated that individuals highly exposed to road traf-fic noise, are not necessarily annoyed by traffic noise, and vice versa. Previous studies also reported discrepancies between actual and perceived exposures in relation to somatic symptom reporting [28,29], indicating that the perception of being exposed to a potentially harmful environmental source is sometimes more strongly associated with diminished health than the actual exposure.

Taken together, these findings suggest that the living environment is associated with adverse health effects, but this is only partly explained by environmental noise or air pollution.

Using mUlTiPle coHorT sTUdies in environmenTal ePidemiology

This dissertation was written within the framework of the European harmoni-zation initiative BioSHaRE (Biobank Standardisation and Harmonisation for Research Excellence in the European Union; www.bioshare.eu). Within BioSHaRE,

159

General Discussion

8

tools were developed for data harmonization [30,31] and federated data analyses [32,33]. In Chapters 4 and 7, we reported on studies undertaken with harmonized data from multiple European studies. In Chapter 4, data from a total of 70,928 participants from four European cohort studies were used. In Chapter 7, data from a total of 91,718 participants were obtained from three European cohort studies. We used harmonized exposure assessment methods to estimate levels of road traffic noise [34] and air pollution [35–37] for large geographical areas. With these data combined, we had a very large and rich dataset for our research. The advantages of such an approach are widespread. We were able to make use of very large sample sizes, so that we could detect the small effect sizes that are common in environmental epidemiology. Using data from large geographical areas resulted in wide exposure ranges of road traffic noise and air pollution. Furthermore, re-sults could be compared across regions, in which not only the exposures, but also genetic, social and cultural factors vary. This makes the generalizability of our re-sults larger. Being able to generalize study results across large regions in Europe is important for establishing European exposure norms and guidelines. Although exposure ranges were large, median levels of road traffic noise and air pollution were relatively low, if compared to other cohorts from Southern Europe [22] or Asia [19]. It is therefore unclear to what extent the conclusions in this dissertation can be generalized to regions with higher levels of noise and air pollution. An-other advantage of our study population was that they all originated from general population based cohorts. Many of these cohorts were established with the aim of investigating a broad range of research questions (e.g. LifeLines, HUNT, KORA), or with more specific aims, for example investigating the relation between diet and cancer (EPIC-Oxford). However, none of these cohorts were specifically set up to investigate relations between environmental exposures and adverse health outcomes. Chances of self-selection are probably higher for cohorts that were specifically set up to investigate effects of a particular environmental exposure. Risk of bias due to self-selection is probably not very high in our research.

The findings from this dissertation also illustrate some of the difficulties that arise with multi-cohort studies. In our study on air pollution and depressed mood (Chapter 4), we used two different harmonized air pollution exposure models, but results from these two models differed. Associations of NO2 and PM10 from EU-wide models and depressed mood in LifeLines were consistently positive and significant, while associations of NO2 and PM10 from ESCAPE models in LifeLines were initially positive and significant, but no longer significant and sometimes negative when adjusted for additional covariates. The main difference between the ESCAPE and EU-wide exposure models is that ESCAPE models were constructed

Chapter 8

160

for the Netherlands and Belgium, whereas the EU-wide models were developed for a much larger area in Western Europe. As a result, ESCAPE models may per-form better on a regional scale, while EU-wide models may make comparisons across multiple study areas easier. Especially for NO2, the EU-wide models may explain less of the spatial variation, if compared to the ESCAPE models [38]. The use of harmonized exposure models for large regions may be at the expense of the ability to detect spatial variation within those regions. Further assessment of the performance of these harmonized exposure models and replication our study findings is needed.

Also in Chapter 4, we reported opposite associations between air pollution and depressed mood in LifeLines and HUNT. NO2 and PM10, estimated with the EU-wide model, were positively associated with depressed mood in LifeLines, while negative associations were found in HUNT. Associations in the other two studies (KORA and FINRISK) were not statistically significant. Regardless of whether there is a true effect of air pollution on depressed mood, these findings underline the importance of multi-cohort studies. If the association between air pollution and depressed mood was studied in only one cohort, conclusions could have been different. Moreover, if we could have used data from additional cohort studies, results from these additional cohorts might have changed our conclusions. Given these contradicting results, replication of these findings in cohort studies from other geographical areas is needed.

In our study about noise, blood pressure and heart rate (Chapter 7), we pooled individual level data from three studies, enabling us to undertake our analyses within one large dataset. Analyzing data from 91,718 participants resulted in high statistical power. We could not use this approach to our data analysis in Chapter 4, where we studied the association between air pollution and depressed mood. The reason for this was that the outcome assessment of depressed mood was largely heterogeneous between the studies, and pooling of these heterogeneous data could not be justified.

As discussed earlier, variability in exposure ranges and disease outcomes is required to establish small effects in environmental epidemiology, and increases generalizability of the results. But variability in methods for exposure and outcome assessment is undesirable, since it makes comparisons between stud-ies more difficult, and introduces heterogeneity when data are pooled. Using a large study area also increases risk of residual confounding, when heterogeneity of these areas is large [39]. Moreover, additional heterogeneity is introduced in the likely event that the different cohort studies used different methodologies for participant recruitment. With the pooling of individual data originating from multiple studies, suitable statistical techniques should be in place to analyze

161

General Discussion

8

these data. In Chapter 7, we therefore used two statistical methods to address this heterogeneity. Pooled regression analyses, adjusted for cohort, yielded larger effect estimates than the random-effects meta-analyses. Furthermore, results from random-effects meta-analyses were less precise. This is probably due to the heterogeneity among the cohort-specific results. Random-effects meta-analyses provide more conservative pooled estimates and confidence intervals when data is heterogeneous [40], which is probably the reason for the different results as compared to the individual-level pooled analyses. More appropriate statistical techniques, for example mixed models, that take into account heterogeneous multi-cohort data, should be used in future research.

With the harmonization and pooling of multi-cohort data, the researcher is strong-ly dependent on the data that is available at that stage. We aimed to investigate relations between road traffic noise and common somatic symptoms in multiple European studies, but did not succeed in finding sufficient studies that assessed common somatic symptoms. Prospective harmonization (i.e. implementing stan-dardized and harmonized measures) of symptom questionnaires may offer a solu-tion to this by recommending a standard assessment method. If the recommended standard is implemented in future cohort studies, this will largely contribute to collaboration between cohort studies. As a first step for prospective harmoniza-tion, we performed a systematic review of symptom questionnaires (Chapter 5). We identified 40 symptom questionnaires and evaluated them based on validity and reliability, and their fitness for purpose in large scale studies. Based on these criteria, we recommended two symptom questionnaires that seemed most fit for use in large scale studies: the Patient Health Questionnaire-15 (PHQ-15) and the Symptom Checklist-90 Somatization Scale (SCL-90 SOM) [41]. We stressed that there is a need for use of standardized and harmonized symptom questionnaires. The standardized and harmonized assessment of common somatic symptoms in large-scale population studies could greatly benefit research on environmental stressors and common somatic symptoms, by enabling the comparisons between studies, and the sharing and pooling of data.

Another important challenge is that with the pooling of heterogeneous data, it is almost inevitable that data is lost. Detailed information in one study may be lost if it is to be harmonized with less detailed information from another study. For example, in Chapter 7, the harmonization of data on educational degree was challenging because the cohorts all used different questions to assess educational degree. We eventually used a categorical variable with only two categories, be-cause in some studies only little information was available, while in other studies much more detailed information on educational degree was available. When using

Chapter 8

162

data from multiple studies that are retrospectively harmonized, it is important to find a balance between the gains and losses: increasing the sample size, exposure range, while not losing too much detailed data due to harmonization.

Finally, in practical terms, difficulties with data sharing between studies may arise. Firstly, obtaining the data can be challenging since cohorts have different procedures for data access. Harmonization of data access procedures would greatly simplify the sharing of cohort data. Data sharing may lead to data own-ership issues. Furthermore, geocoded data, that is needed to estimate environ-mental exposures, is highly privacy sensitive, because it can possibly lead back to a participant’s home address. To protect participant privacy, addresses and corresponding geocodes are stored in a file separately from all other participant data. Subsequently, environmental exposure estimates are linked to the geocodes, after which only the exposure estimates are linked to the main database, with all other anonymized participant data. With this procedure, addresses and geocodes can never be associated with other participant data, and privacy protection is en-sured. Still, scientific boards of cohort studies may be hesitant to share this type of data. Because of these legal and ethical concerns, agreement on data sharing took additional time, or sometimes agreement on data sharing was not reached at all.

In summary, harmonization and pooling of data from multiple cohort studies can largely contribute to environmental epidemiological research. But it is important to invest in high quality harmonization, take into account heterogeneity of data, and be aware of ethical and legal challenges associated with the sharing of spatial data.

meTHodological limiTaTions

In addition to some of the difficulties with multi-cohort studies, as described above, our research also has some other limitations. One of the main challenges in environmental epidemiology is the exposure assessment. Although this is true for all epidemiologic research (e.g. recall and reporting bias in research on smoking habits, food intake, physical activity), exposure ascertainment of environmental sources includes additional challenges. We used indirect exposure assessments of road traffic noise and air pollution. These exposures were estimated at the home addresses of participants, assuming this to be a good indicator of personal expo-sure to road traffic noise and air pollution. There are a number of factors that are not taken into account when estimating exposure at home addresses. Daily mobil-

163

General Discussion

8

ity (e.g. traveling to work or school) and workplace exposure may be sources of exposure misclassification. For example, exposure misclassification arises when an individual lives in an area with low air pollution levels, but travels through ar-eas with high pollution, and works in a place where levels may also be higher than in their residential area. Even when travel activities account for only a small part of daily life, they can account for a large portion of an individual´s total exposure to air pollution [42]. Exposure at the workplace can contribute largely to differ-ential exposure, since individuals often spend a large amount of time at work. Combining exposures from both home and work locations will result in better estimations of an individual’s exposure [43]. Ideally, personal monitoring of loca-tions and exposures would be used, but this is nearly impossible for large-scale studies. However, the introduction of new technologies, including smartphone applications and small personal sensors, may be promising in providing better personal estimates of exposure to air pollution and other environmental hazards [44]. Another limitation of our exposure assessments is that road traffic noise and air pollution levels are estimated at the outdoor level, instead of indoor. For noise, a number of modifying factors may apply. Housing insulation (type of win-dows) and (bed)room orientation are characteristics that influence indoor noise exposure, and possibly modify relationships with disease outcomes. Other noise modifying factors include use of noise reducing remedies, hearing problems, and window opening habits [45]. Some of these factors may also apply to indoor air pollution exposure. We did not account for these factors in our studies, and this might have contributed to exposure misclassification of our study population. The exposure models used in our studies have proved to be sufficient in estimating the spatial variability in exposure to air pollution [35,36,38] and noise [34].

A general difficulty in environmental epidemiology includes the long latency period between exposure and disease, which complicates the identification of harmful exposures and their health risks. Long-term prospective studies or the retrospective assessment of exposures are needed to determine effects of long-term exposure to environmental agents. Our research was cross-sectional, and the observed relations cannot be interpreted as causal effects. The observed as-sociations in our cross-sectional research are particularly useful for formulating hypotheses for future studies. Our findings should therefore be interpreted with caution, because reverse causality cannot be excluded. We cannot claim that these individuals have a certain disease because they live in a particular environment. Instead, individuals with a certain disease may have a preference to live in a cer-tain environment. In our research on road traffic noise, blood pressure, and heart rate (Chapter 7), we did find similar associations if we restricted our analyses to a subgroup that lived at least 2 or 5 years at the same address. Although we are not

Chapter 8

164

sure whether the road traffic noise exposure remained the same in that period, it gives some indication for little influence of length of residence on the association between road traffic noise, blood pressure and heart rate Nonetheless, account-ing for length of residence still does not give indications for causality. It merely provides insights in possible mechanisms of noise effects, for example effects of habituation or conversely a long latency period.

In addition to traffic-related noise and air pollution exposure, there are clearly many other environmental exposures that may be related to urban-rural health differences, or to spatial variations in health in general. Other environmental risk factors with known public health impacts, high individual risks, and public con-cern are for example benzene, dioxins, second-hand smoke, formaldehyde, lead, ozone, and radon. Studying these risk factors were outside the scope of this dis-sertation, but are nonetheless important environmental risk factors for a variety of diseases [46].

imPlicaTions of sTUdy findings and direcTions for fUTUre researcH

environmental policies and society

Policy makers and urban planners should pay attention to making urban areas more healthy. Prevention on a macro level, in addition to the individual level, could improve health of many individuals. By structurally changing people’s en-vironments and reducing risk factors for disease, effects on public health can be substantial [47]. Our findings provide further support for preventive policies aimed at reducing air pollution and traffic noise. Estimated effects of environ-mental exposures on disease may be small, also in our studies. Health impact assessments show small, but potentially wide spread health benefits of the reduc-tion of environmental noise and air pollution. For example, a simulation study of speed limit reduction and traffic re-allocation showed that a 2% decrease in population exposure to NO2, and a 1 to 4% decrease in noise exposure, resulted in small decreases of disease burden of <5% [48]. However, even small effects are relevant for public health, given the fact that so many individuals are exposed to air pollution and noise [46]. This was also underlined by a review focusing on the health impacts of policies promoting active travel, which encourage people to go walking and cycling instead of using their cars. These policies resulted in large individual health benefits for those who became more physically active. Health benefits from the reduction of noise and air pollution were smaller, but may be more wide-spread, since large populations are affected [49]. Living environments

165

General Discussion

8

that support active travel, and subsequently increase physical activity and air quality, have the potential to contribute to public health [6,50]. Another prom-ising approach to improve public health is contact with nature. Living close to green spaces or other natural environments has been related to positive effects on health [51–53]. Health benefits have been observed with regard to life expectancy [51], birth outcomes [54], mental health [52,55], and self-reported general health [56]. This may occur through multiple mechanisms [53], one of them being that natural environments help reduce air pollution and noise [57,58].

It is clear that there are great opportunities for improving public health through the design of healthy living environments. Environmental epidemiology can make great contributions to this by informing policy makers and urban planners. Both individual-level and population-level benefits will contribute to public health.

directions for future research

Future research should focus on investigating adverse health effects of environ-mental exposures in multiple international studies with broad exposure ranges. To this end, we call upon the use of standardized and harmonized exposure and outcome measures, enabling high quality research in multiple cohort studies. In two studies presented in this dissertation, we were able to use data from three (Chapter 7) or four (Chapter 4) cohort studies. Including data from additional cohort studies would have contributed to the generalizability of our results. Also, using data from additional studies could have contributed to more firm conclu-sions, since the contradicting results in our study complicated that. To accom-plish collaboration between cohort studies, the research community should invest in possibilities for the sharing, harmonization, and pooling of data. Tools have been developed within the BioSHaRE project (e.g. DataSHaPER [30,31], and DataSHIELD [32,33,59]), but the availability of tools alone is not enough to ensure collaboration. What currently is needed, is the willingness to share data. A promising development in medical science in general, is the introduction of new journal policies that require authors to share data from clinical trials [60]. A similar initiative, not limited to clinical trials, was initiated by PLOS journals, that “requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception” [61]. These developments will hopefully lead to the increase of data sharing.

As mentioned as an important limitation, our research was based on cross-sec-tional data. Impacts of many environmental exposures often only manifest after a longer time period. Therefore, prospective studies should be initiated, so that long-term adverse health effects of environmental exposures can be detected.

Chapter 8

166

Future studies should also take into account residential history of participants, and if available, link these addresses to historical exposure data. This enables the study of adverse health effects of environmental exposures in a life course per-spective. Such an approach needs additional information on motivations for home relocation. Motives for relocation, for example major life events, are important for the investigation of selective migration.

Disentangling the adverse health effects of traffic-related noise and air pollution is a major challenge in environmental epidemiology. Although it is increasingly acknowledged in environmental epidemiology [21,62], research should take into account the effects of correlated environmental exposures to which individuals are simultaneously subjected. Otherwise, health risks may be falsely attributed to an environmental exposure that is highly correlated with another [63]. Further attention should be given to exposure assessment for noise and air pollution. As discussed as one of the methodological limitations, in addition to exposure at the residence, future studies should also take into account exposures at schools, the workplace, and during commuting. In LifeLines, workplace addresses are cur-rently being collected and geocoded, allowing for outdoor exposure estimation at both home and work location. Furthermore, for noise and air pollution exposure it would be advisable to estimate exposures indoors, in addition to outdoors. In addition, the research community should extensively investigate adverse health effects of environmental exposures in vulnerable subgroups. Children, patients, and groups with low socioeconomic status may be more vulnerable to environ-mental stressors [64]. Individuals who perceive they are exposed [28], who are disturbed by the exposure [27], and who have concerns about the health risks associated with the exposure may also have an increased risk for adverse health effects [65]. This too calls for large sample sizes, which may be obtained through the combined analyses of multiple studies. However, efforts should be made to ensure high quality data harmonization.

Lastly, research should explore practical solutions for making our living envi-ronments more healthy environments. For example, technical innovations could make motorized traffic more quiet and less polluting. In addition, we should think of how we can effectively modify our living environment, in order to make residents more physically active, improve their diets, and reduce their alcohol and tobacco use. The development of adequate and efficient prevention strate-gies requires involvement of professionals from multiple disciplines, including epidemiologists, policy makers, and urban planners, and may also benefit from collaboration with communities [47]. Future research could combine efforts from multiple disciplines, and contribute to the prevention of non-communicable diseases through the design of healthy living environments.

167

General Discussion

8

conclUding remarks

This dissertation illustrates that the living environment may be associated with adverse health effects, but this is only partly explained by environmental noise or air pollution. The use of data from multiple cohort studies provides excellent opportunities for studying harmful environmental exposures. The scientific com-munity should invest in the sharing and harmonization of their cohort data, sup-porting environmental epidemiological research on a large geographical scale.

Chapter 8

168

references

1. Broms K, Norback D, Eriksson M, Sundelin C, Svardsudd K. Effect of degree of urbanisation on age and sex-specific asthma prevalence in Swedish preschool children. BMC Public Health. 2009;9: 303. doi:10.1186/1471-2458-9-303

2. Maio S, Baldacci S, Carrozzi L, Polverino E, Angino A, Pistelli F, et al. Urban residence is as-sociated with bronchial hyperresponsiveness in Italian general population samples. Chest. 2009;135: 434–41. doi:10.1378/chest.08-0252


4. Trivedi T, Liu J, Probst JC, Martin AB. The metabolic syndrome: are rural residents at increased risk? J Rural Health. 2013;29: 188–97. doi:10.1111/j.1748-0361.2012.00422.x


6. Lovasi GS, Grady S, Rundle A. Steps Forward: Review and Recommendations for Research on Walkability, Physical Activity and Cardiovascular Health. Public Health Rev. 2012;33: 484–506.


8. Adam M, Schikowski T, Carsin AE, Cai Y, Jacquemin B, Sanchez M, et al. Adult lung function and long-term air pollution exposure. ESCAPE: a multicentre cohort study and meta-analysis. Eur Respir J. 2015;45: 38–50. doi:10.1183/09031936.00130014


10. Genc S, Zadeoglulari Z, Fuss SH, Genc K. The adverse effects of air pollution on the nervous system. J Toxicol. 2012;2012: 782462. doi:10.1155/2012/782462

11. Fonken LK, Xu X, Weil ZM, Chen G, Sun Q, Rajagopalan S, et al. Air pollution impairs cognition, provokes depressive-like behaviors and alters hippocampal cytokine expression and morphol-ogy. Mol Psychiatry. 2011;16: 973,987–995. doi:10.1038/mp.2011.76

12. Oberdorster G, Sharp Z, Atudorei V, Elder A, Gelein R, Kreyling W, et al. Translocation of inhaled ultrafine particles to the brain. Inhal Toxicol. 2004;16: 437–445. doi:10.1080/ 08958370490439597

13. Peters A, Veronesi B, Calderon-Garciduenas L, Gehr P, Chen LC, Geiser M, et al. Translocation and potential neurological effects of fine and ultrafine particles a critical update. Part Fibre Toxicol. 2006;3: 13. doi:10.1186/1743-8977-3-13

14. Block ML, Calderón-Garcidueñas L. Air pollution: mechanisms of neuroinflammation and CNS disease. Trends Neurosci. 2009;32: 506–516. doi:10.1016/j.tins.2009.05.009

15. Terzano C, Di Stefano F, Conti V, Graziani E, Petroianni A. Air pollution ultrafine particles: toxic-ity beyond the lung. Eur Rev Med Pharmacol Sci. 2010;14: 809–821.

16. Calderon-Garciduenas L, Solt AC, Henriquez-Roldan C, Torres-Jardon R, Nuse B, Herritt L, et al. Long-term air pollution exposure is associated with neuroinflammation, an altered innate immune response, disruption of the blood-brain barrier, ultrafine particulate deposition, and accumulation of amyloid beta-42 and alpha-synuclein in children and youn. Toxicol Pathol. 2008;36: 289–310. doi:10.1177/0192623307313011

169

General Discussion

8


18. Wang Y, Eliot MN, Koutrakis P, Gryparis A, Schwartz JD, Coull BA, et al. Ambient Air Pollution and Depressive Symptoms in Older Adults: Results from the MOBILIZE Boston Study. Environ Health Perspect. 2014;122: 553–8. doi:10.1289/ehp.1205909


20. Kim C, Jung SH, Kang DR, Kim HC, Moon KT, Hur NW, et al. Ambient particulate matter as a risk fac-tor for suicide. Am J Psychiatry. 2010;167: 1100–1107. doi:10.1176/appi.ajp.2010.09050706; 10.1176/appi.ajp.2010.09050706




24. Alberti KG, Eckel RH, Grundy SM, Zimmet PZ, Cleeman JI, Donato KA, et al. Harmonizing the metabolic syndrome: a joint interim statement of the International Diabetes Federation Task Force on Epidemiology and Prevention; National Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International. Circulation. 2009;120: 1640–1645. doi:10.1161/CIRCULATIONAHA.109.192644

25. Herr CEW, Zur Nieden A, Kopka I, Rethage T, Gieler U, Eikmann TF, et al. Assessment of somatic complaints in environmental health. Int J Hyg Environ Health. 2009;212: 27–36. doi:10.1016/j.ijheh.2008.01.003

26. Kipen HM, Fiedler N. Environmental factors in medically unexplained symptoms and related syndromes: The evidence and the challenge. Environ Health Perspect. 2002;110: 597–599.






Chapter 8

170


33. Doiron D, Burton P, Marcon Y, Gaye A, Wolffenbuttel BH, Perola M, et al. Data harmonization and federated analysis of population-based studies: the BioSHaRE project. Emerg Themes Epidemiol. 2013;10: 12. doi:10.1186/1742-7622-10-12




37. Vienneau D, de Hoogh K, Bechle MJ, Beelen R, van Donkelaar A, Martin R V, et al. Western European Land Use Regression Incorporating Satellite- and Ground-Based Measurements of NO and PM. Environ Sci Technol. 2013; doi:10.1021/es403089q


39. Pedersen M, Siroux V, Pin I, Charles MA, Forhan A, Hulin A, et al. Does consideration of larger study areas yield more accurate estimates of air pollution health effects? An illustra-tion of the bias-variance trade-off in air pollution epidemiology. Environ Int. 2013;60: 23–30. doi:10.1016/j.envint.2013.07.005

40. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials. 1986;7: 177–88. 41. Zijlema WL, Stolk RP, Löwe B, Rief W, White PD, Rosmalen JGM. How to assess common somatic

symptoms in large-scale studies: A systematic review of questionnaires. J Psychosom Res. 2013;74: 459–468. doi:10.1016/j.jpsychores.2013.03.093

42. De Nazelle A, Seto E, Donaire-Gonzalez D, Mendez M, Matamala J, Nieuwenhuijsen MJ, et al. Im-proving estimates of air pollution exposure through ubiquitous sensing technologies. Environ Pollut. 2013;176: 92–9. doi:10.1016/j.envpol.2012.12.032

43. Ragettli MS, Phuleria HC, Tsai M-Y, Schindler C, de Nazelle A, Ducret-Stich RE, et al. The rel-evance of commuter and work/school exposure in an epidemiological study on traffic-related air pollution. J Expo Sci Environ Epidemiol. 2014; doi:10.1038/jes.2014.83

44. Nieuwenhuijsen MJ, Donaire-Gonzalez D, Rivas I, de Castro M, Cirach M, Hoek G, et al. Variability in and agreement between modeled and personal continuously measured black carbon levels using novel smartphone and sensor technologies. Environ Sci Technol. 2015;49: 2977–82. doi:10.1021/es505362x

45. Babisch W, Swart W, Houthuijs D, Selander J, Bluhm G, Pershagen G, et al. Exposure modifiers of the relationships of transportation noise with high blood pressure and noise annoyance. J Acoust Soc Am. 2012;132: 3788–808. doi:10.1121/1.4764881

171

General Discussion

8

46. Hanninen O, Knol AB, Jantunen M, Lim TA, Conrad A, Rappolder M, et al. Environmental Burden of Disease in Europe: Assessing Nine Risk Factors in Six Countries. Environ Health Perspect. 2014; doi:10.1289/ehp.1206154

47. Franco M, Bilal U, Diez-Roux A V. Preventing non-communicable diseases through structural changes in urban environments. J Epidemiol Community Health. 2014; jech–2014–203865–. doi:10.1136/jech-2014-203865

48. Schram-Bijkerk D, van Kempen E, Knol AB, Kruize H, Staatsen B, van Kamp I. Quantitative health impact assessment of transport policies: two simulations related to speed limit reduction and traffic re-allocation in the Netherlands. Occup Environ Med. 2009;66: 691–8. doi:10.1136/oem.2008.041046

49. De Nazelle A, Nieuwenhuijsen MJ, Antó JM, Brauer M, Briggs D, Braun-Fahrlander C, et al. Improving health through policies that promote active travel: a review of evidence to sup-port integrated health impact assessment. Environ Int. 2011;37: 766–77. doi:10.1016/j.envint.2011.02.003

50. Rojas-Rueda D, de Nazelle A, Tainio M, Nieuwenhuijsen MJ. The health risks and benefits of cycling in urban environments compared with car use: health impact assessment study. BMJ.; 2011;343: d4521. doi:10.1136/bmj.d4521

51. Mitchell R, Popham F. Effect of exposure to natural environment on health inequalities: an obser-vational population study. Lancet. 2008;372: 1655–60. doi:10.1016/S0140-6736(08)61689-X

52. Triguero-Mas M, Dadvand P, Cirach M, Martínez D, Medina A, Mompart A, et al. Natural outdoor environments and mental and physical health: Relationships and mechanisms. Environ Int. 2015;77: 35–41. doi:10.1016/j.envint.2015.01.012

53. Hartig T, Mitchell R, de Vries S, Frumkin H. Nature and health. Annu Rev Public Health. 2014;35: 207–28. doi:10.1146/annurev-publhealth-032013-182443

54. Dadvand P, Sunyer J, Basagaña X, Ballester F, Lertxundi A, Fernández-Somoano A, et al. Sur-rounding greenness and pregnancy outcomes in four Spanish birth cohorts. Environ Health Perspect. 2012;120: 1481–7. doi:10.1289/ehp.1205244

55. Gascon M, Triguero-Mas M, Martínez D, Dadvand P, Forns J, Plasència A, et al. Mental Health Benefits of Long-Term Exposure to Residential Green and Blue Spaces: A Systematic Review. Int J Environ Res Public Health. 2015;12: 4354–4379. doi:10.3390/ijerph120404354

56. Maas J, Verheij RA, Groenewegen PP, de Vries S, Spreeuwenberg P. Green space, urbanity, and health: how strong is the relation? J Epidemiol Community Health. 2006;60: 587–592. doi:10.1136/jech.2005.043125

57. Dadvand P, Rivas I, Basagaña X, Alvarez-Pedrerol M, Su J, De Castro Pascual M, et al. The as-sociation between greenness and traffic-related air pollution at schools. Sci Total Environ. 2015;523: 59–63. doi:10.1016/j.scitotenv.2015.03.103

58. Dzhambov AM, Dimitrova DD. Urban green spaces’ effectiveness as a psychological buffer for the negative health impact of noise pollution: a systematic review. Noise Health. 2014;16: 157–65. doi:10.4103/1463-1741.134916

59. Wolfson M, Wallace SE, Masca N, Rowe G, Sheehan NA, Ferretti V, et al. DataSHIELD: resolving a conflict in contemporary bioscience-performing a pooled analysis of individual-level data without sharing the data. Int J Epidemiol. 2010;39: 1372–1382. doi:10.1093/ije/dyq111

60. Godlee F, Groves T. The new BMJ policy on sharing data from drug and device trials. Br Med J. 2012;345: e7888–e7888. doi:10.1136/bmj.e7888

61. Bloom T, Ganley E, Winker M. Data Access for the Open Access Literature: PLOS’s Data Policy. PLoS Med. 2014;11: e1001607. doi:10.1371/journal.pmed.1001607

Chapter 8

172

62. Foraster M. Is it traffic-related air pollution or road traffic noise, or both? Key questions not yet settled! Int J Public Health. 2013; doi:10.1007/s00038-013-0500-3

63. Keatinge W. Commentary: Mortality from environmental factors, but which ones? Int J Epide-miol. 2003;32: 398–399. doi:10.1093/ije/dyg079

64. Schwartz J, Bellinger D, Glass T. Exploring potential sources of differential vulnerability and susceptibility in risk from environmental hazards to expand the scope of risk assessment. Am J Public Health. 2011;101 Suppl: S94–101. doi:10.2105/AJPH.2011.300272

65. Porsius JT, Claassen L, Smid T, Woudenberg F, Petrie KJ, Timmermans DRM. Symptom reporting after the introduction of a new high-voltage power line: A prospective field study. Environ Res. 2015;138C: 112–117. doi:10.1016/j.envres.2015.02.009

Summary

177

Summary

sUmmary

Many diseases can be attributed to environmental factors. One of the major chal-lenges in environmental epidemiology is detecting the small risks associated with environmental exposures. Studies with large numbers of participants and broad exposure ranges are needed to investigate the complex interaction between the environment and the development of multifactorial chronic diseases. To this end, we used harmonized data from multiple European cohort studies to study harm-ful health effects of the environment. More specifically, the aim of this disserta-tion was to examine whether the living environment, and in particular urbanity, traffic-related noise and air pollution, are associated with adverse health effects.

Firstly, we reported on the associations between urbanity and respiratory func-tion, metabolic syndrome, anxiety, and depression in the LifeLines Cohort Study (Chapter 3). Our results suggest a differential health impact of urbanity according to type of disease. Living in an urban environment appears to be beneficial for cardiometabolic health, but seems to have an adverse impact on respiratory func-tion and mental health. Smoking and physical activity only explained part of these urban-rural health differences. Other factors may be important in explaining health disparities in urban and rural areas. Examples are exposure to traffic noise and air pollution, which were found to be increased in urban areas (Chapter 2).

As a next step, we studied the association between exposure to air pollution and depressed mood (Chapter 4). We analyzed data from 70,928 participants from four European studies: LifeLines (the Netherlands), KORA (Germany), HUNT (Norway), and FINRISK (Finland). We used harmonized air pollution estimates originating from two different exposure models (described in Chapter 2). De-pressed mood was assessed with interviews or questionnaires, which were dif-ferent among the four studies. We found no consistent evidence for an association between air pollution and depressed mood. We observed heterogeneous results between the different cohort studies. An interesting finding was that both posi-tive and negative associations between air pollution and depressed mood were observed with the two different air pollution exposure models in LifeLines.

In Chapter 6, we reported on the associations between road traffic noise, noise annoyance, and common somatic symptoms in the LifeLines Cohort Study. In a sample of 56,937 participants, we found that exposure to road traffic noise and the reporting of common somatic symptoms, such as headache, chest pain, and breathing trouble, were not related. Annoyance from road traffic noise, however, was related to increased symptom reporting. We found that this relationship was only for a small part influenced by the personality facets hostility and vulnerabil-

178

Summary

ity to stress, suggesting that the simultaneous reporting of increased annoyance and symptoms is not due to a general tendency to be vulnerable to stress. The correlation between modelled road traffic noise and road traffic noise annoyance was low. This indicated that individuals highly exposed to road traffic noise, are not necessarily annoyed by traffic noise, and vice versa.

We initially aimed to investigate relations between road traffic noise and com-mon somatic symptoms in multiple European studies, but did not succeed in finding sufficient studies that assessed common somatic symptoms. Prospective harmonization (i.e. implementing standardized and harmonized measures) of symptom questionnaires may offer a solution to this by recommending a standard assessment method. As a first step for prospective harmonization, we performed a systematic review of symptom questionnaires (Chapter 5). We identified 40 symptom questionnaires and evaluated them based on validity and reliability, and their fitness for purpose in large-scale studies. Based on these criteria, we recommended two symptom questionnaires that seemed most fit for use in large scale studies: the Patient Health Questionnaire-15 (PHQ-15) and the Symptom Checklist-90 Somatization Scale (SCL-90 SOM). The standardized and harmo-nized assessment of common somatic symptoms in large-scale population studies could greatly benefit research on environmental stressors and common somatic symptoms, by enabling the comparisons between studies, and the sharing and pooling of data.

We also studied relations between road traffic noise, air pollution, blood pres-sure, and heart rate with harmonized data from three European studies (Chapter 7). Data from 91,718 participants were obtained from LifeLines (the Netherlands), EPIC-Oxford (United Kingdom), and HUNT (Norway). Road traffic noise and air pollution levels were estimated with harmonized exposure models (described in Chapter 2), and cohort data was harmonized according to the DataSHaPER meth-odology. The data were analyzed using a new statistical software developed for pooling data from different studies (DataSHIELD).We found that exposure to road traffic noise was associated with an increased heart rate in all three studies, which was in line with our hypothesis. We also observed that noise was non-significantly associated with a small decrease in systolic blood pressure, and with a statisti-cally significant decrease in diastolic blood pressure. Additional adjustment for air pollution resulted in an attenuation of the associations, but associations remained significant for diastolic blood pressure and heart rate. Similar to our findings in Chapter 4, cohort-specific results for noise and blood pressure were heterogeneous. When cohort-specific associations were summarized with the more conservative random-effects meta-analyses, only significant associations between road traffic noise and heart rate were observed.

179

Summary

Based on our studies, we concluded that the living environment may be associ-ated with adverse health effects. The main advantages of using multiple cohort studies were the very large sample sizes, broad exposure ranges, and the ability to make comparisons across cohorts. Our findings also illustrate that heterogene-ity between cohort studies can be challenging for multi-cohort research. In our study populations, exposure to traffic-related noise and air pollution was not consistently associated with our study outcomes. These findings are meaningful because they show us to be cautious with interpreting results from a single cohort study. The use of data from multiple cohort studies provides excellent opportuni-ties for studying harmful environmental exposures, but the scientific community should invest in the sharing and harmonization of their cohort data, supporting research on large geographical scales.

Nederlandse samenvatting

183

Nederlandse Samenvatting

nederlandse samenvaTTing

Veel aandoeningen zijn gerelateerd aan blootstellingen uit de omgeving. In dit proefschrift onderzoeken wij de nadelige gezondheidseffecten van het leven in de stad ten opzichte van het platteland. Ook kijken we naar het effect van blootstel-ling aan van verkeersgeluid en luchtverontreiniging. We doen deze studies met behulp van geharmoniseerde gegevens afkomstig uit verschillende Europese bevolkingsonderzoeken. Onderzoek bij een groot aantal deelnemers met een brede mate van blootstelling kan inzicht geven in de complexe relaties tussen de omgeving en multifactoriële aandoeningen.

Allereerst beschrijven wij de relatie tussen de stedelijkheid van de woonomgev-ing en een aantal aandoeningen (hoofdstuk 3). Uit onze resultaten blijkt dat een verslechterde longfunctie, angst, en depressie vaker voorkomen bij individuen die wonen in een stedelijke omgeving dan bij individuen die wonen in meer landelijke gebieden. Het metabool syndroom komt juist vaker voor bij individuen uit landeli-jke gebieden dan bij individuen uit stedelijke gebieden. Deze verschillen konden niet geheel verklaard worden door verschillen in demografische kenmerken, rookgedrag of lichamelijke activiteit. Mogelijk zijn andere factoren van belang bij het verklaren van de gezondheidsverschillen in stedelijke en landelijke gebieden. Voorbeelden hiervan zijn blootstelling aan verkeersgeluid en luchtverontreinig-ing, waarvan we vonden dat beide meer voorkomen in stedelijke gebieden (hoofdstuk 2).

Vervolgens onderzochten we de relatie tussen de blootstelling aan luchtveron-treiniging en depressieve stemming (hoofdstuk 4). We analyseerden gegevens van 70,928 deelnemers van vier Europese bevolkingsonderzoeken: LifeLines (Nederland), KORA (Duitsland), HUNT (Noorwegen) en FINRISK (Finland). Blootstelling aan luchtverontreiniging werd geschat aan de hand van twee ver-schillende voorspellingsmodellen (beschreven in hoofdstuk 2), en depressieve stemming werd vastgesteld met interviews en vragenlijsten. We vonden geen duidelijke consistente aanwijzingen voor een relatie tussen luchtverontreiniging en depressieve stemming. Resultaten in LifeLines wezen op een mogelijk hoger risico op depressieve klachten door blootstelling aan luchtverontreiniging, terwijl in HUNT juist een mogelijk lager risico op depressieve klachten werd gevonden. Een andere opmerkelijke bevinding was dat in LifeLines, deze relatie alleen werd gevonden met een van de twee voorspellingsmodellen voor luchtverontreiniging.

In hoofdstuk 6 beschrijven we de relatie tussen verkeersgeluid, geluidshin-der en het hebben van veelvoorkomende lichamelijke klachten. Blootstelling aan verkeersgeluid werd geschat aan de hand van een voorspellingsmodel en

184


geluidshinder, de subjectieve ervaring van mensen dat geluid als hinderlijk ervaren wordt, werd gemeten met een vragenlijst (beschreven in hoofdstuk 2). We bestudeerden de gegevens van 56,937 LifeLines deelnemers en vonden geen relatie tussen gemodelleerd verkeersgeluid en veelvoorkomende klachten, zoals hoofdpijn, buikpijn en pijn in de hartstreek. Geluidshinder was wel gerelateerd aan het hebben van meer lichamelijke klachten. Vervolgens onderzochten we of deze relatie beïnvloed werd door de karaktereigenschappen hostiliteit en kwets-baarheid voor stress. Deze invloed bleek klein, waarna we concludeerden dat het tegelijkertijd hinder ondervinden van geluid en het rapporteren van lichamelijke klachten waarschijnlijk niet verklaard kan worden door de karaktereigenschap-pen hostiliteit en kwetsbaarheid voor stress.

Het bovenbeschreven onderzoek naar geluid en lichamelijke klachten werd beoogd uitgevoerd te worden in meerdere Europese bevolkingsonderzoeken. Wij merkten echter op dat weinig andere bevolkingsonderzoeken deze klachten sys-tematisch uitvragen. Het prospectief harmoniseren (het implementeren van ges-tandaardiseerde en geharmoniseerde meetmethoden) van lichamelijke klachten vragenlijsten zou hier een oplossing kunnen bieden. Wij bekeken daarom alle lichamelijke klachten vragenlijsten die beschikbaar waren in de wetenschap-pelijke literatuur (hoofdstuk 5). We vonden 40 vragenlijsten die voldeden aan onze criteria, en evalueerden deze vragenlijsten op basis van de validiteit en betrouwbaarheid, en de geschiktheid voor het gebruik in grootschalig bevolk-ingsonderzoek. Op basis van deze criteria, deden we een aanbeveling voor twee vragenlijsten die het meest geschikt lijken voor het gebruik in grootschalig bevolk-ingsonderzoek: de PHQ-15 (Patient Health Questionnaire-15) en de SCL-90 SOM (Symptom Checklist-90 Somatization scale). Het gebruik van gestandaardiseerde en geharmoniseerde vragenlijsten voor veelvoorkomende lichamelijke klachten kan mogelijk een grote bijdrage leveren aan onderzoek naar de relatie tussen omgevingsblootstellingen en lichamelijke klachten, doordat resultaten tussen studies vergelijkbaar worden, en doordat gegevens makkelijker samengevoegd kunnen worden.

Tenslotte onderzochten wij de relatie tussen verkeersgeluid, bloeddruk en hartfrequentie. Deze relatie wordt mogelijk beïnvloed door blootstelling aan luchtverontreiniging. Daarom onderzochten we ook of de relatie tussen geluid en bloeddruk en hartfrequentie (deels) verklaard zou kunnen worden door bloot-stelling aan luchtverontreiniging. We onderzochten deze relaties met behulp van gegevens van 91,718 deelnemers van LifeLines (Nederland), EPIC-Oxford (Verenigd Koninkrijk) en HUNT (Noorwegen). Blootstelling aan verkeersgeluid en luchtverontreiniging werd geschat aan de hand van geharmoniseerde voor-spellingsmodellen (beschreven in hoofdstuk 2), en de studiegegevens werden

185


geharmoniseerd volgens gestandaardiseerde methodologie (DataSHaPER). De gegevens werden geanalyseerd met nieuwe statistische software dat ontwik-keld is voor het samenvoegen van gegevens uit verschillende onderzoeken (DataSHIELD). De resultaten van deze studie geven indicatie voor een consistent verband tussen verkeersgeluid en een verhoogde hartfrequentie. We vonden ook dat geluid geassocieerd was met een kleine, niet significante afname in systolische bloeddruk (bovendruk), en met een significante afname in diastolische bloeddruk (onderdruk). Blootstelling aan luchtverontreiniging had een kleine invloed op de relatie tussen geluid, bloeddruk en hartfrequentie. Net als in hoofdstuk 4, waren de resultaten voor geluid en bloeddruk verschillend per bevolkingsonderzoek. Wanneer we een meer conservatieve statische methode gebruikten, die meer rekening houdt met de heterogeniteit van de verschillende onderzoeken, vonden we alleen een significante relatie tussen geluid en hartfrequentie.

De bevindingen in dit proefschrift tonen aan dat omgevingsblootstellingen mogelijk de gezondheid nadelig beïnvloeden. Voordelen van het gebruiken van gegevens uit meerdere bevolkingsonderzoeken waren het grote aantal deelne-mers, de brede mate van blootstelling, en het vergelijken van resultaten tussen de bevolkingsonderzoeken. Onze bevindingen geven echter ook weer dat de heterogeniteit tussen de bevolkingsonderzoeken dit soort onderzoek kan be-moeilijken. In onze studiepopulaties lijkt blootstelling aan verkeersgeluid en luchtverontreiniging niet consistent van invloed te zijn op de aandoeningen die wij onderzocht hebben. Deze bevindingen zijn van belang omdat het ons leert voorzichtig om te gaan met onderzoeksresultaten afkomstig uit één cohort. Het gebruik van gegevens uit meerdere bevolkingsonderzoeken biedt uitstekende mogelijkheden voor het bestuderen van schadelijke omgevingsblootstellingen, maar de wetenschappelijke gemeenschap zou moeten investeren in het delen en harmoniseren van onderzoeksgegevens, zodat onderzoek op grote geografische schaal mogelijk is.

187

Dankwoord

dankwoord

Veel mensen hebben bijgedragen aan een zeer leerzame en leuke promotietijd. Ik wil graag iedereen bedanken, maar ik kan niet iedereen noemen. Daarom een kort dankwoord:

Beste Judith en Ronald, mijn promotores. Bedankt voor jullie vertrouwen en dat jullie altijd achter mij stonden. Ik heb veel van jullie geleerd en heb onze samenwerking al-tijd als erg prettig beschouwd. Ik zal onze overleggen en buitenlandse tripjes missen.

Dear coauthors, thank you for all your help with our papers. I learned a lot from each of you, and I think we did a nice job.

Dear BioSHaRE colleagues, thank you for our collaboration on the BioSHaRE proj-ect. Kees, bedankt voor dat jij mij wegwijs maakte in de wereld van de exposure science. Susan, many thanks for our collaboration in the BioSHaRE project. I really enjoyed my time in London. Thank you for all your help. Samuel, thank you for working together on the often not so easy projects in BioSHaRE. It was really help-ful to combine our efforts and solutions! Lisette, mede dankzij jou is BioSHaRE een prachtig project geworden. Bedankt voor al jouw hulp!

Beste Epidemiologie en ICPE collega’s, bedankt voor alle leerzame momenten en de gezelligheid.

Lieve kamergenootjes, promotieonderzoek met jullie aan mijn zij had nog wel veel langer mogen duren. Anne, wat heerlijk dat jij de ‘minder uitende versie’ bent van mij. Ik kijk met veel plezier terug op ons begrip voor elkaar, jouw gezelligheid en gebrom. Mara, mijn Brabantse schone. Ik heb veel ontzag voor jouw doorzettings-vermogen, relativeringsvermogen, en jouw prachtige accent. Hanna, hoe kan het toch dat jij zo bescheiden bent, terwijl jij één van de slimste, knapste en leukste bent? Maria, ik hou van jouw nuchterheid en directheid, die zijn onmisbaar (en verfrissend) in ons vak! Irma, bedankt dat ik dankzij jouw vrolijkheid niet de-pressief ben geraakt tijdens mijn eindstrijd. Maaike, mijn heerlijke, gezellige, en ‘oudste’ collega die altijd in is voor een praatje, een grapje, of een biertje. Te gek, dat jij mij ook op de allerlaatste dag van mijn promotietraject bij zal staan! Allen, en ook alle ex-kamergenootjes, bedankt!

Brenda, mi hermana para siempre! Ik ben trots dat ik jou mijn paranimf mag noemen.

Lieve families Zijlema en Ali, bedankt voor dat jullie er altijd zijn!

188

Curriculum Vitae & List of publications

cUrricUlUm viTae

Wilma Louise Zijlema (1986) was born in Groningen, the Netherlands. After completing secondary education (HAVO, Wessel Gansfort College, Groningen), she studied Biology and Medical Laboratory Research at the Hanze University of Applied Sciences. She received her Bachelor’s degree in 2007, after which she moved to Amsterdam to study Health Sciences at the VU University. She received her second Bachelor’s degree in 2010, and her Master’s degree in Prevention and Public Health in 2011. From 2011 through 2015, she worked as a PhD candidate at the Department of Epidemiology of the University Medical Center Groningen. Wilma worked for the BioSHaRE-EU (Biobank Standardisation and Harmonisa-tion for Research Excellence in the European Union) project. During her PhD research, Wilma harmonized, pooled, and analyzed data from multiple European biobanks for her investigation of the adverse health effects of traffic-related noise and air pollution. In September 2013, Wilma visited the MRC-PHE Centre for Environment and Health, Imperial College London for one month, to work on her PhD project, under the supervision of dr. Kees de Hoogh, dr. Susan Hodgson, and dr. Anna Hansell. She further developed her skills as a researcher by presenting at national and international conferences and by completing her training and supervision plan in the Graduate School of Medical Sciences SHARE.

lisT of PUBlicaTions

Zijlema wl, Wolf K, Emeny R, Ladwig K, Peters A, Kongsgård H, Hveem K, Kvaløy K, Yli-Tuomi T, Partonen T, Lanki T, Eeftens M, de Hoogh K, Brunekreef B, BioSHaRE, Stolk RP, Rosmalen JGM. The association of air pollution and depressed mood in 70,928 individuals from four European cohorts. Int J Hyg Environ Health, forthcoming

Zijlema wl, Klijs B, Stolk RP, Rosmalen JGM. (Un)Healthy in the City: respi-ratory, metabolic and mental health associated with urbanicity. PLoS One 2015;10(12):e0143910.

Zijlema wl, Morley D, Stolk RP, Rosmalen JGM. Noise and somatic symptoms: A role for personality traits? Int J Hyg Environ Health 2015;218(6):543-9.

189

Curriculum Vitae & List of publications

Janssens KAM, Zijlema wl, Joustra ML, Rosmalen, JGM. Mood and anxiety dis-orders in chronic fatigue syndrome, fibromyalgia and irritable bowel syndrome: Results from the LifeLines Cohort Study. Psychosom Med 2015;77(4):449-57.

Zijlema wl, Stolk RP, Löwe B, How to assess common somatic symptoms in large-scale studies: A systematic review of questionnaires. J Psychosom Res 2013;74:459–68.

submitted for publication

Zijlema wl, Cai Y, Doiron D, de Hoogh K, Morley D, Gulliver J, Hodgson S, BioSHaRE, Key T, Kongsgård H, Hveem K, Gaye A, Burton P, Hansell A, Stolk RP, Rosmalen JGM. Road traffic noise, blood pressure and heart rate: Pooled analyses of harmonized data from 91,718 participants from LifeLines, EPIC-Oxford, and HUNT3.

Zijlema wl, Smidt N, Klijs B, Morley D, Gulliver J, de Hoogh K, Scholtens S, Rosmalen JGM, Stolk RP. The LifeLines Cohort Study: a resource providing new opportunities for environmental epidemiology.

Cai Y, Zijlema wl, Doiron D, Burton P, Gaye P, Blangiardo M, Morley D, Gulliver J, de Hoogh K, Hveem K, Fortier I, Stolk R, Elliott P, Hansell A, Hodgson S. Road traf-fic air pollution, noise and adult prevalent asthma in three European biobanks: a harmonised approach from BioSHaRE project.

Joustra ML, Zijlema wl, Rosmalen JGM, Janssens KAM. Lifestyle factors in chronic fatigue syndrome and fibromyalgia. Results from LifeLines.

Peters LL, van Houwelingen AH, Boter H, Zijlema wl, Gussekloo J, de Craen AJM, Westendorp RGJ, Blom JW, Slaets JPJ, Buskens E, Burger H, den Elzen WPJ. Prediction modeling and risk assessment using individual characteristics and the Groningen Frailty Indicator in the oldest old population: the Leiden 85-plus Study.

190

SHARE

researcH insTiTUTe sHare

This thesis is published within the research institute shAre (Science in Healthy Ageing and healthcaRE) of the University Medical Center Groningen / University of Groningen.Further information regarding the institute and its research can be obtained from our internetsite: http://www.share.umcg.nl/

More recent theses can be found in the list below.((co-) supervisors are between brackets)

otten eIntroducing eHealth and other innovative options into clinical genetic patient care in view of increase efficiency and maintenance of quality of care; patients’ and providers’ perspectives(prof I van Langen, prof AV Ranchor, dr E Birnie)

Pouwels KSelf-controlled designs to control confounding(prof E Hak)

voerman aeLiving with prostate cancer; psychosocial problems, supportive care needs and social support groups(prof R Sanderman, prof M Hagedoorn, dr APh Visser)

janse mThe art of adjustment(prof AV Ranchor, prof MAG Sprangers, dr J Fleer)

nigatu YtObesity and depression; an intertwined public health challenge(prof U Bültmann, prof SA Reijneveld)

Aris-Meijer JlStormy clouds in seventh heaven; a study on anxiety and depression around childbirth(prof CLH Bockting, prof RP Stolk, dr H Burger)

191

SHARE

Kluitenberg BThe NLstart2run study; running related injuries in novice runners(prof RL Diercks, dr H van der Worp, dr M van Middelkoop)

Benjaminse aMotor learning in ACL injury prevention(prof E Otten, prof KAPM Lemmink, prof RL Diecks)

Potijk mrModerate prematurity, socioeconomic status, and neurodevelopment in early childhood; a life course perspective(prof SA Reijneveld, prof AF Bos, dr AF de Winter)

Beyer arThe human dimension in the assessment of medicines; perception, preferences, and decision making in the European regulatory environment(prof JL Hillege, prof PA de Graeff, prof B Fasolo)

alferink mPsychological factors related to Buruli ulcer and tuberculosis in Sub-saharan Africa(prof AV Ranchor, prof TS van der Werf, dr Y Stienstra)

Bennik ecEvery dark cloud has a colored lining; the relation between positive and negative affect and reactivity to positive and negative events(prof AJ Oldehinkel, prof J Ormel, dr E Nederhof, dr JACJ Bastiaansen)

stallinga HaHuman functioning in health care; application of the International Classification of Functioning, Disability and Health (ICF)(prof RF Roodbol, prof PU Dijkstra, dr G Jansen)

For more 2015 and earlier theses visit our website.

university of groningen (un)healthy in the city zijlema, wilma...example is the london smog of 1952....

Documents