childhood predictors of adult obesity: a systematic review · lifestyle factors with a view to...

Childhood predictors of adult obesity:a systematic review

TJ Parsons1*, C Power2, S Logan1 and CD Summerbell3

1Systematic Reviews Training Unit, Department of Epidemiology and Public Health, Institute of Child Health, London; and 2Departmentof Epidemiology and Public Health, Institute of Child Health, London; and 3School of Health, University of Teesside, Teesside, UK

OBJECTIVE: To identify factors in childhood which might in¯uence the development of obesity in adulthood.BACKGROUND: The prevalence of obesity is increasing in the UK and other developed countries, in adults andchildren. The adverse health consequences of adult obesity are well documented, but are less certain for childhoodobesity. An association between fatness in adolescence and undesirable socio-economic consequences, such as lowereducational attainment and income, has been observed, particularly for women. Childhood factors implicated in thedevelopment of adult obesity therefore have far-reaching implications for costs to the health-services and economy.SEARCH STRATEGY: In order to identify relevant studies, electronic databasesÐMedline, Embase, CAB abstracts,Psyclit and Sport DiscusÐwere searched from the start date of the database to Spring 1998. The general searchstructure for electronic databases was (childhood or synonyms) AND (fatness or synonyms) AND (longitudinal orsynonyms). Further studies were identi®ed by citations in retrieved papers and by consultation with experts.INCLUSION CRITERIA: Longitudinal observational studies of healthy children which included measurement of a riskfactor in childhood (< 18 y), and outcome measure at least 1 y later. Any measure of fatness, leanness or change infatness or leanness was accepted. Measures of fat distribution were not included. Only studies with participants froman industrialized country were considered, and those concerning minority or special groups, e.g. Pima Indians orchildren born preterm, were excluded.FINDINGS:� Risk factors for obesity included parental fatness, social factors, birth weight, timing or rate of maturation, physicalactivity, dietary factors and other behavioural or psychological factors.

� Offspring of obese parent(s) were consistently seen to be at increased risk of fatness, although few studies havelooked at this relationship over longer periods of childhood and into adulthood. The relative contributions of genesand inherited lifestyle factors to the parent ± child fatness association remain largely unknown.

� No clear relationship is reported between socio-economic status (SES) in early life and childhood fatness. However,a strong consistent relationship is observed between low SES in early life and increased fatness in adulthood.Studies investigating SES were generally large but very few considered confounding by parental fatness. Womenwho change social class (social mobility) show the prevalence of obesity of the class they join, an association whichis not present in men. The in¯uence of other social factors such as family size, number of parents at home and child-care have been little researched.

� There is good evidence from large and reasonably long-term studies for an apparently clear relationship forincreased fatness with higher birth weight, but in studies which attempted to address potential confounding bygestational age, parental fatness, or social group, the relationship was less consistent.

� The relationship between earlier maturation and greater subsequent fatness was investigated in predominantlysmaller, but also a few large studies. Again, this relationship appeared to be consistent, but in general, the studieshad not investigated whether there was confounding by other factors, including parental fatness, SES, earlierfatness in childhood, or dietary or activity behaviours.

� Studies investigating the role of diet or activity were generally small, and included diverse methods of risk factormeasurement. There was almost no evidence for an in¯uence of activity in infancy on later fatness, and inconsistentbut suggestive evidence for a protective effect of activity in childhood on later fatness. No clear evidence for aneffect of infant feeding on later fatness emerged, but follow-up to adulthood was rare, with only one studymeasuring fatness after 7 y. Studies investigating diet in childhood were limited and inconclusive. Again,confounding variables were seldom accounted for.

� A few, diverse studies investigated associations between behaviour or psychological factors and fatness, butmechanisms through which energy balance might be in¯uenced were rarely addressed.

CONCLUSIONS AND RESEARCH PRIORITIES: The major research gap identi®ed by the current review is the lack oflong-term follow-up data spanning the childhood to adulthood period. This gap could in part be ®lled by: (i) follow-upof existing groups on whom good quality baseline data have already been collected; and (ii) further exploitation ofexisting longitudinal datasets.Many of the risk factors investigated are related, and may operate on the same causal pathways. Inherent problems inde®ning and measuring these risk factors make controlling for confounding, and attempts to disentangle relationshipsmore dif®cult. A given risk factor may modify the effect of another, and cumulative effects on the development ofobesity are likely, both over time for speci®c risks, or at any particular time over a range of risk factors. An additionalapproach to addressing these issues may be to use large samples on whom more basic measures of risk factors havebeen collected.

*Correspondence: TJ Parsons, Systematic Reviews TrainingUnit, Department of Epidemiology and Public Health, Institute ofChild Health, 30 Guildford Street, London WC1N 1EH, UKReceived 7 June 1999, accepted 19 August 1999.

International Journal of Obesity (1999) 23, Suppl 8, S1±S107ß 1999 Stockton Press All rights reserved 0307±0565/99 $15.00

http://www.stockton-press.co.uk/ijo

Speci®c issues that remain unresolved include:� the mechanism by which SES in early life in¯uences obesity in adulthood;� whether the relationships between birth weight and maturation and later obesity persist after accounting forconfounding factors;

� whether any relationships between dietary factors and activity and later fatness are due to a direct effect, or totracking in dietary or activity behaviour;

� how psychological factors and behaviours in¯uence energy balance, and therefore fatness.

A further neglected area of research is the identi®cation of factors predicting the maintenance of a healthy relativeweight, which may or may not be the opposite of predictors of obesity.The challenge to future research remains to discern which are the important and modi®able factors, or clustering offactors, and effects over time, on the causal pathway to the development of obesity.

Keywords: systematic review; obesity; child; adult; predictors

Introduction

Objective

The aim of this systematic review was to identifyfactors in childhood which might in¯uence the devel-opment of obesity in adulthood.

Background

The prevalence of obesity is increasing in adults1 andchildren in the UK;2 British children are gettingfatter.3 A similar rise in prevalence of obesity, relatedto increasing af¯uence, is reported from many devel-oped countries throughout the world.4 The associationbetween adult obesity and adverse health outcomes,including diabetes, coronary heart disease, cancer andrespiratory problems, is well documented,4 but there islimited evidence for an association between adoles-cent obesity and increased risks of adult morbidity5

and mortality.6 More immediate effects of developingobesity include psychosocial outcomes, with socialisolation and peer problems more common in fatterchildren.7 Overweight adolescent women have lowereducational attainment, lower incomes and are lesslikely to marry than those not overweight.8 If theserelationships are indeed causal, then they imply far-reaching consequences for costs to the health services,and the economy. The limited data available suggestthe direct medical costs might amount to 4 ± 5% oftotal health care.9 In the USA, an annual cost of US$4billion has been attributed to loss of productivity dueto obesity.9 The relationship between relative weightand morbidity and mortality may not be linear, but J-shaped, with an increased risk of adverse health out-comes reported at the lower extremes of fatness inmany studies.4 The extent is unclear to which thisrelationship between underweight and an increasedrisk of adverse outcome re¯ects reverse causality,whereby smoking and pre-existing disease causeboth underweight and increased morbidity andmortality.4

So far treatment of established obesity has beenlargely ineffective and therefore prevention is prefer-able.10 ± 12 Existing data suggest that children may bemore responsive to prevention than adults,13 and thatrisk factors identi®ed in children may be more amen-

able to change. Despite the attractiveness of preven-tion strategies aimed at children, a cautious approachis required to avoid the possibility of unforeseenadverse consequences, particularly in respect of inter-ruption of normal growth pattern in early childhoodand distorted body image, excessive dieting andanorexia in adolescence. For instance, the recommend-ation for children over the age of 2 y in the USA thattotal fat intake amount to 30% of total calories orless14 has in some observational studies been asso-ciated with a greater risk of inadequate intake ofseveral vitamins.15

The childhood period is also important for adultobesity because tracking of overweight, albeit mod-erate, is observed between childhood and adulthood.This topic has recently been reviewed.5,16 The ®guresvary according to the de®nition of obesity and lengthof follow-up, but fat children have a high risk of goingon to become fat adults; for example in the 1958British birth cohort, 38% of boys and 44% of girlsabove the 95th BMI centile at age 7 were obese at age33.17 Even so, only a small proportion of fat adultswere fat in childhood. It is likely that there are factorsoperating in early adulthood which promote obesity,but there may also be factors operating in childhoodthat promote adult obesity. It is still a matter of debatewhether there are particular stages in childhood,during which physiological alterations increase therisk of later obesity. These stages are termed criticalperiods, and may include the prenatal period, theadiposity rebound (second rise in adiposity occurringat about 6 y), and puberty.18

In addition to the observed tracking of adiposityfrom childhood to adulthood, it has been suggestedthat lifestyle habits such as diet and activity levelsmay also track during childhood and into adulthood.There is some evidence that such tracking occurs,19± 21

although relationships are modest. If lifestyle patternsestablished in childhood persist into adulthood, child-hood becomes a target period for modi®cation oflifestyle factors with a view to preventing obesity.

Obesity is a multi-factorial condition with wide-ranging causes including genetic, social, cultural andbehavioural factors, all of which may interact. Thecurrent review takes a public health perspective, andtherefore focuses on potentially modi®able factors.

Childhood predictors of adult obesityTJ Parsons et al

S2

There is an extensive literature concerning geneticfactors and obesity, which at the gene level, isbeyond the scope of this review. Moreover, asobserved by Prentice,22 most obesities today must becaused by environmental and lifestyle factors inmodern life, since the large increases in prevalenceof obesity are occurring within a relatively constantgene pool. This does not mean that genetic effects areunimportant, but that the balance of genetic andenvironmental in¯uences is changing. Genetic make-up undoubtedly contributes to obesity, but it is alsohighly likely that behaviours or lifestyle factors, whichare potentially modi®able, contribute to the apparentheritability of obesity. The inheritance of phenotypeparental fatness as a predictor of adult fatness, istherefore included in this review.The de®nition of obesity is not without problems,

particularly in childhood, during growth and matura-tion. Many different methods are currently in use toestimate body fatness or relative weight, and for eachmethod various cut-off levels are used to describeoverweight or obesity. The pros and cons of eachmethod have been discussed elsewhere5,23 and thesedetails are beyond the scope of the current review. Inadults, BMI is the standard obesity assessment,24,25

although it is important to recognize that is only anindicator of obesity. For a given BMI, there is a largerange of percentage body fat and thus the sensitivity(proportion of truly obese subjects detected by themeasure), may be low.26

Body weight (or BMI) can vary considerably inadult life, for example, by 10 kg (3 kg=m2) over 18 y.27

The use of an arbitrary dichotomous (or categorical)classi®cation of obesity (or grades of) will inevitablyresult in substantial numbers of individuals entering orleaving the `obese' group over time. This is a dis-advantage of using a cut-off to de®ne obesity. Inchildren, the validity of several convenient methodsfor assessing fatness, (skinfolds, relative weight, BMI)has been compared with underwater weighing.28 Thesensitivity (as above) and speci®city (proportion oftruly non-fat subjects detected by the measure) variedfor the different methods and cut-off points used tode®ne obesity. Despite reservations that BMI does notidentify obese subjects as well as other methods, forexample skinfold measurements,28 the recent consen-sus report on obesity evaluation and treatment inchildren29 concluded that BMI was a straightforwardand useful measure of adiposity. To encourage con-sistency in de®ning fatness, the International TaskForce on Obesity is developing a provisional inter-national reference population and BMI standards toclassify overweight and obesity, using age and sexspeci®c curves.30,31 There may however be a goodcase for surveys to collect skinfold or body circum-ference data as well.5

To investigate the in¯uence of factors in childhoodon obesity in adulthood, several sources of evidencemay be drawn on: cross-sectional, longitudinal andsecular trend data, and intervention studies in both

children and adults. The strongest evidence however,comes from longitudinal studies, and it is these thatform the basis for the current review. Where possible,studies starting in childhood and following throughinto adulthood have been used, but those longitudinalstudies examining the associations between exposureof interest and fatness later in childhood have alsobeen included.

Methodology and results

Methodology

The best evidence for assessing the impact of child-hood predictors on adult obesity is provided by long-itudinal studies. This review therefore aimed toidentify all such studies.

Criteria for considering studies for this review

Inclusion criteria. (1) Types of studies: longitudinalobservational studies in healthy children, of at least1 y duration, in which a baseline measurement of aproposed risk factor was made in childhood, de®nedas under 18 y of age were included.(2) Population: only studies where participants

were from a developed, industrialized country wereconsidered.(3) Predictors: predictors of obesity (diet, physical

activity etc.) were not pre-de®ned so that pre-con-ceived ideas about what may or may not be a riskfactor for obesity were not imposed on the searchstrategy, leading to a biased group of studies beingidenti®ed in the literature. Therefore predictors werenot searched for speci®cally in the search strategy(Appendix II). All predictors addressed in the litera-ture, which were relevant from a public health per-spective and potentially amenable to change, wereincluded.(4) Outcome: any measure of fatness, leanness or

relative weight, or change in fatness, leanness orrelative weight was included. Absolute change inweight was considered only if it occurred whollywithin adulthood (� 18 y), since height is assumedto be constant.

Exclusion criteria. (1) Type of study (quality): stu-dies were excluded if they were less than 1 y induration, or if basic information, such as numbersof participants, the ages at which they were measured,or suf®cient de®nition of risk factor or outcome waslacking, or not referenced to a previous publication.(2) Population: studies concerning minority

groups, e.g. Pima Indians, vegans, children bornpreterm or to diabetic mothers, or other specialgroups, were excluded.(3) Predictors: predictors such as smoking by the

child, contraception, pregnancy occurring under the


S3

age of 18, which usually concern older children, werenot included.(4) Outcome: studies including only a measure of

fat distribution rather than overall fatness wereexcluded.

Search strategy. The following sources are includedin the search process:

Medline 1966 ±April 1998EMBASE 1980 ±May 1998CAB Abstracts 1973 ±May 1998PSYCLIT 1996 ±March 1998Sport Discus 1975 ±March 1998conference abstractsconsultation with experts.

The electronic search strategies incorporated threeconcepts: childhood, fatness and longitudinal. Appro-priate synonyms, misspellings and truncations wereincluded. In addition, potentially relevant longitudinalstudies known about at the time of searching wereincluded speci®cally within the strategy. Search strat-egies used for each database are included in AppendixII. A list of experts consulted is given in Appendix III.Proceedings of the North American Association forthe Study of Obesity Meeting, Cancun, 9 ± 13 Novem-ber, 1997,32 and the 8th International Congress onObesity, Paris, 29 August ± 3 September, 199833 werehand-searched.Additional existing reviews relevant to the broader

subject area, secular trend and ecological studies wereconsulted in order to provide a context for the studiesidenti®ed for the current review. These studies werenot identi®ed by a systematic process. Studies pub-lished too recently to be identi®ed by the searchstrategy, but identi®ed by the review group in thecourse of reading, or brought to our attention byexperts consulted, were included.

Process of study selection. Titles and abstracts ofstudies identi®ed by electronic searches were exam-ined on screen. Those thought to possibly meet inclu-sion criteria were obtained, along with papersidenti®ed by consulting experts. Full copies of thesepapers were examined to determine whether they metthe study inclusion criteria.

Methodological quality of included studies. Apartfrom determination of whether studies were at least1 y in duration and included adequate descriptions ofstudy populations and data, methodological qualitywas not formally assessed. Although there is a body ofresearch around quality issues for trials,34 there is verylittle relating to observational data. In particular, thereis little empirical research on the magnitude of theeffect of different aspects of study design on thevalidity of observational studies. Theoretically it

would be possible to develop a quality scoringsystem based on factors like participant selection,sample size, length of follow-up, proportion offollow-up, and methods of measurement of outcome(objective=subjective). However, further complica-tions arise due to the heterogeneity in de®nition ofoutcome, statistical methods employed to analyseresults, and adjustment for a variety of confoundingvariables. Formal assessment of quality was thereforenot carried out, due to dif®culties in developingquality criteria for such a heterogeneous group ofstudies, and time constraints in doing so. Instead,limitations of studies identi®ed are discussed withineach section of the review.

Data extraction. A data extraction form wasdesigned to include information on the following:

� study design, prospective or retrospective;� study setting, population characteristics, subject

recruitment;� numbers of participants, invited, eligible, enrolled;� numbers of participants at each follow-up;� ages at which baseline and follow-up measures

were made;� de®nition and measurement of predictors and out-

come;� method of analysis;� main ®ndings.

This information is included in the summary tables ofstudies, presented in Appendix I, Tables A1 ± 11.

Analysis. Formal statistical combination of results,even within subject headings, was clearly inappropri-ate due to the marked heterogeneity of all aspects ofidenti®ed studies. Although it would be useful to beable to estimate the magnitude of the effect of eachpredictor, effect size varies depending on the qualityissues discussed above, the de®nition of predictorvariables, and on whether adjustments have beenmade for covariates. Study results are therefore pre-sented in tabular form, grouped by categories of riskfactor.

Steering group and collaborators. An advisorygroup of experts in aspects of obesity, systematicreviews and=or statistics was formed and met inDecember 1997 for discussion. Subsequently, mem-bers were consulted individually about sections of thereview within their area of expertise. Members of thesteering group and collaborators are given in Appen-dix III.

Results

The electronic database search produced 22,359 hits:Medline 10,212Embase 10,169CAB Abstracts 661


S4

PsycLIT 535Sport Discus 782

Numbers of studies identi®ed are given within eachsection of the review. Several studies were publishedand identi®ed after the search strategy was run andincluded in the review. Five non-English publications(Italian, Czech, French, Dutch, Japanese)35 ± 39 wereidenti®ed that potentially contained relevant informa-tion and were translated. Only one was included in thereview.39

GeneticÐ inheritance of phenotype

Introduction

It is widely acknowledged that genetic factors play arole in obesity, and that fatness is, to some extent`heritable'. This topic has formed the subject ofseveral reviews in the last decade.40 How muchvariation in fatness is explained by genetic factors,rather than environmental factors, is far more con-troversial. There is little agreement about the propor-tion of heritability due to genes and that due toinherited lifestyle habits.Essentially three types of studies address the pheno-

typic inheritance of fatness: family, twin and adoptionstudies. Family studies may be further divided intothose investigating parent ± offspring fatness relation-ships and those investigating similarity between sib-lings, although obviously many family studiesexamine both. Studies of all types are largely cross-sectional, and therefore estimate heritability at asingle time-point. For the current review, our primaryfocus was the longitudinal relationship between par-ental and offspring fatness, but since such studies arescarce, we also considered cross-sectional data fromfamily, twin and adoption studies coincidentally iden-ti®ed by the search strategy.

Methodology

From the results of the search strategy, we identi®edstudies reporting longitudinal data, where data con-cerning fatness of offspring were measured at multipletime-points, and parental fatness was assessed at leastonce. However, we also considered cross-sectionaldata from family, twin and adoption studies identi®edby the search strategy. It should be emphasized thatthese latter studies were not searched for system-atically and thus we were not attempting to identifyall such studies in the literature.

Results

Limitations of the data. The main problem in inves-tigating the in¯uence of parental fatness is that itrepresents the cumulative result, not only of genetic

make-up, but of lifestyle effects and behaviours whichmay also be inherited. It is important that this issue isacknowledged when, as seen in subsequent sections ofthe review, it might be desirable to adjust for apotential confounding effect of parental fatness.Three studies were identi®ed which reported multi-

ple measures of fatness from childhood into adult-hood,41 ± 43 and a further ®ve studies which reportedmultiple measures of fatness within childhood.44 ± 48

Findings. Longitudinal studiesÐchildhood to adult-hood: two of the three studies which followedthrough into adulthood included a single measure ofparental fatness,41,42 the third included several mea-sures over the study course.43 The ®ndings are sum-marized in Appendix I, Table A1; all studiesobserved that offspring fatness increased with paren-tal fatness, and at all ages, from childhood intoadulthood.Data from Kaplowitz et al suggested that this

relationship may be stronger between mothers andtheir offspring than fathers and offspring, and that themother ± offspring relationship strengthens overtime.41 Whitaker et al found that parental obesitywas a more important predictor of offspring obesityearlier in childhood (< 6 y), becoming less importantwith increasing age of the child.43 Data from Lake etal showed that parental obesity in¯uences tracking ofthe offspring's own obesity, which is much stronger ifboth parents are obese.42

Other types of studies: all other studies identi®ed,family and sibling,39,44 ± 63 twin64,65 and adoption stu-dies,66± 68 support the view that genetic factors contri-bute to fatness level, although the estimated effects varybetween studies (Appendix I, Tables A2± 4).

Summary

The presence of a genetic in¯uence on obesity hasbeen well researched, although the size of the geneticcontribution and whether it differs during childhood,has not been extensively investigated. Existing stu-dies, however, show consistency in that parent obesityis an important factor in predicting adult obesity ofoffspring, and that offspring of obese parents whothemselves are fatter in childhood, may be at parti-cular risk.

Impressions

There would seem to be little doubt that the develop-ment of fatness has a genetic component. The eightlongitudinal studies we identi®ed which met ourinclusion criteria demonstrated a positive relationshipbetween parent and offspring fatness, which wasuniversally supported by the additional range ofstudies considered. What is far less clear, and isimportant from a public health perspective, is howmuch of the variation in fatness is due to the geneticcomponent, and how much to the environment. Also


S5

important is the in¯uence of the environment on thegenetic component, i.e. gene ± environment interac-tions. For example, a genetic predisposition to selecthigh fat foods, or to perform little physical activitywill be dependent on the availability of such foods orthe necessity of performing activity. The strong rela-tionship between parental and offspring fatness islikely to represent both genetic factors and cumulativeenvironmental in¯uences including learnt lifestylebehaviours. Adoption and twin studies are morepowerful tools than family studies in resolving geneticand environmental effects, although those we identi-®ed were cross-sectional and therefore did not meetour original inclusion criteria. As seen in this reviewand others,40 these studies tend to give high estimatesof heritability, and argue that little variation in fatnessis due to the environment.68 Each design however, hasits disadvantages, and incorporates assumptions thatmay well be violated, with implications for thestrength of the ®ndings.40 An integrated approach,incorporating the best aspects of each design, may addto current understanding of this nature ± nurturedebate.40 These questions are largely beyond thescope of the current systematic review, and havebeen more comprehensively discussed elsewhere.40

One of the major disadvantages of family studies, isthat most include only one time-point and thereforeparents and offspring are measured at different ages. Ifgenetic and environmental in¯uences vary with age ordifferent genetic=environmental factors operate atdifferent ages, observed relationships between parentsand children will be weakened. The longitudinalstudies with follow-up data in adulthood identi®edfor this review partly address this issue. AlthoughKaplowitz et al 41 found that mother ± child correla-tions increased with age of the child, Lake et al 42

found that mother ± child correlations did not increaseover 11 y, and Whitaker et al 43 observed a decrease inlikelihood of a child of an obese mother being obese,with increasing age of the child. The study of Kaplo-witz41 was comparatively small, and the outcomemeasure was based on skinfold thicknesses, whereasthe other studies used BMI, which might contribute tothese differences.Findings from both studies of Lake et al 42 and

Whitaker et al 43 suggest that, in addition to parentobesity being a risk factor for adult obesity in theiroffspring, parental obesity is particularly important forthe development of fatness in childhood. In the studyof Whitaker et al, below 6 y, parent obesity is espe-cially predictive of adult obesity in the offspring, butabove 9 y, fatness of the child becomes more impor-tant.43 The child of obese parent(s) is therefore atincreased risk of becoming fat early in life, and oncerelatively fat, he=she is more likely to be so later inadulthood. Data from the British 1958 birth cohort42

show that obese children of obese parents are muchmore likely to be obese in adulthood than obesechildren of normal weight parents. As the prevalenceof obesity in the population increases, more children

will show stronger tracking of fatness into adulthood,although the contributions of genetic and modi®ablelifestyle factors to the tracking of fatness remainlargely unknown.

Social factors

Introduction

One of the most striking facts about obesity is thenegative relationship in adults between obesity andsocio-economic status (SES) in the developed world,especially among women.69 About 10 y ago, Sobaland Stunkard published a comprehensive review ofSES and obesity,69 which included 144 studies fromall over the world, of men, women and=or children.Despite the fact that these studies used a range ofmeasures of SES and fatness, and spanned a period of45 y, a strong negative relationship between SES andfatness was obvious in women across the developedworld, such that fatness increased from higher socialgroups (professional and management) to lower socialgroups (unskilled and manual). Of 54 studies ofwomen, 85% demonstrated a negative relationship,13% showed no relationship, and only 2% a positiverelationship (greater fatness among higher socialgroups). Among men and children, however, therelationship was far less consistent, with studiesmore evenly spread over positive, negative and norelationships. In stark contrast, in developing coun-tries, not a single study found a negative relationshipbetween SES and fatness in women. Indeed, the vastmajority reported a positive relationship, in both menand women, and in children.

UK data concerning SES and BMI among adults iscollected regularly as part of the Health Survey forEngland. Both in the ®rst survey in 1991,70 and in1996 (most recent published data for adults)71 a lowermean BMI was demonstrated in women from non-manual, rather than manual households, but the oppo-site was found in men. Both surveys demonstrated aclear social gradient in the prevalence of obesity inwomen, but no clear pattern in men.

Since 1995, children have been included in theHealth Survey. Combined data from 1995 ± 199763

show no pattern in social class and prevalence ofoverweight in males aged 2 ± 15 y or 16 ± 24 y, or infemales aged 2 ± 15 y. In females aged 16 ± 24 y,prevalence of obesity was higher in manual thannon-manual social classes. In addition, a recentstudy of social factors and fatness in children foundno consistent in¯uence of social class, parent educa-tion, number of parents, hours worked by motheroutside the home, free school meals, or family sizeon fatness of children aged 5 ± 11 y.72

However, these data are cross-sectional, as weremost other studies in Sobal and Stunkard's review,and are therefore unable to determine long-term


S6

relationships between SES in childhood and adultobesity. Associations between low SES and greaterfatness in studies of adults have been attributed to twomain causal models: low SES may promote thedevelopment of fatness or, alternatively, greater fat-ness may lead to downward social mobility andthereby lower SES. There is some evidence fromlongitudinal studies to support both models, but thecurrent systematic review examines evidence on theformer, that low SES in childhood promotes thedevelopment of fatness later in life. Establishing thedirection of association with SES is likely to provideclues on the aetiology of adult obesity, especially ifrelationships are investigated for several life stages,since this might suggest the likely timing of riskfactors.SES has not been the only social measure proposed

to be related to the development of obesity, and othersocial factors reported in the literature, such as familystructure and functioning, age of mother at birth ofchild, rearing area, and scholarly abilities in childhoodwere included in this review. Studies concerning thesefactors (except mother's age) had to ful®l our initialcriteria and be of at least 1 y in duration.

Methodology

All longitudinal, observational studies identi®ed bythe search strategy, which reported an assessment ofsocial group or other social measure during childhood(< 18 y), and fatness at least 1 y later, were includedin this section of the review. Some cross-sectionalstudies were also included, for reasons describedbelow.

Results

Limitations of the data. SES is recognized to be acrude summary measure of a range of circumstances,making it dif®cult to determine what it is about socialgroup that relates to obesity. Social class cannot itself`cause' obesity, but might in¯uence energy balancevia characteristics of socio-economic groups related tomaterial circumstances, behaviour or knowledge.Consideration of some of these measures might helpdisentangle and inform on the association betweenSES and obesity.Since the prevalence of overweight in women is

greater in lower social groups, maternal, if not alsopaternal fatness is likely to be a confounding variablein the relationship between SES and subsequent fat-ness of offspring. Only two studies included a mea-sure of parent relative weight in analyses.73,74

Certain problems were encountered with identify-ing studies for this section of the review. In severallongitudinal studies it was dif®cult to elicit when SEShad been assessed. Both those studies in which it wasclearly stated, and those in which it seemed highlylikely that SES was measured at baseline, wereincluded. Further questions arose concerning the

stability of parental occupation, income and educationover time, and the importance of these factors beingassessed at least 1 y before offspring fatness. Weconsidered parent education likely to be reasonablyconstant, and so whether it was measured at least ayear prior to offspring fatness unimportant. We there-fore considered additional studies identi®ed by oursearch, which measured parent education level andfatness in offspring in child- or adulthood, cross-sectionally. It should be emphasized that these studieswere not searched for systematically, but were coin-cidentally identi®ed by the strategy, and will form(probably a small) subset of all published cross-sec-tional studies investigating this relationship.Thirty papers and one abstract were identi®ed from

21 studies investigating the effect of social factors inchildhood on subsequent fatness (Tables 1 and 2, andAppendix I, Table A5). Twelve studies (18 papers andone abstract)75 ± 93 spanned the period from childhoodto adulthood; one paper and one abstract were fromthe 1946 UK birth cohort study,77,78 two papers werefrom the Tecumseh Community Health study,80,89

three papers were from the same study by Lissauet al,82 ± 84 and four papers were from the 1958 UKbirth cohort study.90 ± 93 Nine studies (12 papers)61,73,74,94 ± 102 spanned a period within childhood; twopapers were from the same study by Agras et al,73,94

and three papers were from the same study by Krameret al.74,96,97 Papers from the 1958 UK birth cohort90 ± 93 and the Tecumseh Community Health study,89

Table 1 Childhood to adulthoodÐtwelve studies

Numbers of studies

Relationship Women Men Women�men

combined

Negative 4.77,81,89,93 8.76,77,81,85,86,88,89,93 4.78,79,82,87a

None 1.75 1.75 0.Positive 0. 0. 0.

Total 5. 9. 4.

Negative relationship: lower risk of fatness associated withhigher social group. Positive relationship: higher risk offatness associated with higher social group.aThe study ofLissau et al 82 is included here, but this study reportednegative or no relationship depending on SES variable used,and therefore the SES± fatness relationship could also beclassi®ed as none.

Table 2 Within childhoodÐnine studiesa

Numbers of studies

Relationship Girls Boys Girls�boys

combined

Negative 3.98 ± 100 1.98 3.61,73,101

None 0. 1.100 2.74,95

Positive 0. 1.99 0.

Total 3. 3. 5.

Negative relationship: lower risk of fatness associated withhigher social group. Positive relationship: higher risk of fatnessassociated with higher social group.aStudies of Garn et al 89 andLasker et al 90 (interim reports of childhood to adulthood studies)not included in table. See text for relationships.


S7

reported follow-up measurement at several time-points, in childhood as well as adulthood. Wheremore than one paper from a study was identi®ed,only the most relevant, or where more than one paperis relevant, only the most recent paper is cited whenthe study is referred to. Studies were generally large,with over half of sample sizes more than 1500.

Findings. (i) Childhood to adulthood: a total of 12studies75 ± 93 measured the in¯uence of social factorsin childhood on fatness in adulthood, with a durationof between 10 and � 55 y (Table 1 and Appendix I,Table A5). All studies examined the in¯uence of SES,based on parent(s) occupation,76,79,82,86 ± 88,93 parent(s)education,80,82,85 a composite of the two,77,81 or familyincome.75,89

Of ®ve studies of women, four found a negativerelationship,77,81,89,93 and one found no relationship75

between childhood SES and subsequent fatness. Ninestudies included men: eight observed a negativerelationship76,77,81,85,86,88,89,93 and one no relation-ship.75 Four studies analysed data from men andwomen combined: three showed a negative relation-ship,78,79,87 the other showed either a negative or norelationship depending on the SES variable in ques-tion.82 Three studies looked at social mobility(between class of origin and own class) and adultobesity in men, all ®nding no relationship.76,77,81 Twostudies included women, and both reported that theprevalence of overweight was lower in the upwardlymobile than in the downwardly mobile groups.77,81

Additional social factors were investigated in sev-eral European studies which showed scholarly dif®-culties=special education,83 parental support, poorhousing, child's hygiene,84 single parent families85

and geographical rearing area85,87 to be related tolater fatness, although some of these factors were nolonger predictors after adjustment for other covariates.The in¯uence of family size on fatness seems to beinconsistent, with greater fatness associated with bothincreasing and decreasing number of children in thefamily.85,86 Two studies found no consistent in¯uenceof birth order on later fatness.79,86

(ii) Within childhood: follow-up measurementsof 11 studies fell within childhood: nine studies basedwholly within childhood61,73,74,95,98 ± 102 and two child-

to adulthood studies with follow-up measurementsin childhood.89,90 Once again all studies assessedSES, based on either parent(s)'s occupation,61,90,98,100

parent(s)'s education,73,80,98,99,101 a combination ofthe two,74,95,98 or family income.102 Study durationwas 2 ± 17 y.

Of the nine childhood studies, (Table 2 and Appen-dix I, Table A5) six studies combined data from boysand girls61,73,74,95,101,102 and three observed a negativerelationship between SES and fatness at 6 ±17 y.61,73,101 Two studies found no associationbetween SES and fatness,74,95 and in one study therelationship was not explicit.102 One study observed anegative relationship in boys and girls separately,98

and a further study found a negative relationship ingirls but not boys.100 The ninth study found a negativerelationship in girls, although many were missingfrom follow-up (Table A5), and that the proportionof overweight and severely overweight boys com-bined, was less in those whose parents had lowerlevels of schooling compared with the whole studypopulation.99 This ®nding is probably due to theparticular comparison made, since prevalence of over-weight and severe overweight was lower in not onlythe lowest, but also the highest of three parentaleducation groups, than in the middle group.

In the two studies which included parental fatnessin analysis, the negative relationships between parenteducation and childhood BMI persisted after adjust-ment for parental BMI,73 whereas in the other study,SES was not a signi®cant predictor of child obesity ina multivariate model.74

The two studies with several repeat follow-upmeasurements89,90 found no, or even tending towardspositive, relationships between SES and fatness inyounger children, which became negative relation-ships in older children, strengthening with age intoadulthood. Within childhood, one study looked atsocial mobility,90 but found no in¯uence on BMI atage 7, 11 or 16 y in boys and girls combined. A Britishstudy found no in¯uence of the state of the home orcare of children in the ®rst 3 y of life on obesity at12 y, and that more obese than non-obese children,had mothers over 35 y at birth.61

(iii) Cross-sectional studies: seven studies (eightpapers)8,103 ± 109 investigated the cross-sectional rela-

Table 3 Seven cross-sectional studies

Numbers of studies

Girls Boys Girls�boys combined

Relationship Black White Mixed race Black White Mixed race Black White

Negative 2106,109 28,107 1109 0a 2103,108

None 2106,109 1109 18b 1104

PositiveU-shaped 1104

Negative relationship: lower risk of fatness associated with higher social group. Positive relationship: higher risk offatness associated with higher social group. aThis becomes one study if Khoury et al 107 are included in this group.bThis becomes two studies if Khoury etal 107 are included in this group. Khoury etal 107 found either a negative or norelationships depending on outcome variable used.


S8

tionship between parental education level and fatnessin their offspring (Table 3 and Appendix I, Table A6);two papers were from the NGHS of NHLBIstudy.105,106 Parental education level was assessedwhile their offspring were children in ®ve stu-dies,103 ± 108 when their offspring were adults in onestudy,109 and during a transitional child-to-adultperiod in one study.8 Table 3 shows that some studiespresent data from White and Black participants sepa-rately, since studies in the USA have demonstratedethnic differences in the prevalence of, and relation-ships between, risk factors and obesity.110 Amongstudies which analysed girls=women separately, twostudies8,107 found a negative relationship between SESand fatness, and two studies which also analysed byrace found a negative relationship in White girls butnot in Black girls.106,109 A study of 16 ± 24-y-oldmales found no effect of SES on fatness,8 whereas astudy in boys found a negative relationship with BMI,but none with triceps skinfold.107 Three studies reportresults for girls and boys combined, one found a U-shaped relationship in Blacks and no relation inWhites,104 the other two studies found negative rela-tionships.103,108

Summary

The longitudinal studies identi®ed for the present reviewshow that both in males and females, there is a remark-ably consistent negative relationship between SES inchildhood and fatness in adulthood. This negative rela-tionship is not quite so consistent when fatness ismeasured in childhood, or cross-sectionally. Socialmobility, the change between social class of origin andown social class later in life, appears to in¯uence obesityonly in women, in whom prevalence of overweight isless in the upwardly than the downwardly mobile. Thein¯uence of social factors other than social class, such asfamily size, number of parents at home and child-care,has not been well researched.

Impressions

The longitudinal evidence identi®ed for the currentreview therefore contrasts with that from cross-sec-tional studies in showing a strong relationship withSES among both sexes and not, as the latter show, aneffect that is largely con®ned to women.69 Of a totalof 21 longitudinal studies, 18 found a negative rela-tionship between some measure of SES and laterfatness. Three studies found no relationship betweenSES and fatness in childhood, but only one studyfound no relationship between SES and fatness inadulthood.75 In the latter study of Arnesen et al,75 SESwas self-reported retrospectively as whether economicconditions in childhood were very dif®cult, dif®cult,good or very good, rather than based on a moreobjective measure, which might contribute to thelack of relationship. In the present review, the con-sistency of the negative relationship was mirrored bythose cross-sectional studies of parent education leveland offspring fatness identi®ed by the search strategy.

These cross-sectional studies were included in thereview on the basis that adult `social class'=occupation=income might change quite considerablyover a number of years, whereas education is likely tobe more constant. We recognize however, that thisassumption may differ across countries, and is possi-bly dependent on factors such as access and cost ofeducation. For example, the USA has a history ofpayment for further education and a larger proportionof mature students than the UK, where until recentlyfurther education has been free, and largely taken upon ®nishing school. If however, it is assumed thatparent education is reasonably constant, it becomesirrelevant whether it is measured at least a year priorto offspring fatness, and therefore cross-sectionalstudies coincidentally identi®ed by the search strategyand concerning this relationship were included.Studies investigating social mobility showed no

in¯uence of social mobility on fatness in men,whereas women show the prevalence of obesity ofthe social group they join. Further evidence for socialmobility effects in women but not men is available intwo large additional studies.8,111 Data from the British1958 birth cohort demonstrate that, at 23 y, womenwho had been obese at 16 y had signi®cantly feweryears of schooling and lower earnings than their non-obese peers.111 These relationships were not seen inmen. Similar ®ndings have been reported from theUSA.8 Thus in women, a cycle is created where lowerSES leads to greater fatness, which in turn tends todecrease SES. In men, the former part of the cycleappears to be weaker and the latter not to exist.Negative attitudes to obesity are more prevalent ingirls than boys.112 Dornbusch et al 113 found that girlswere dissatis®ed with their weight because it repre-sented fat, but boys were satis®ed because, to them,weight represented muscularity. In the same study, thedesire to be thin among girls increased with socialclass, whereas it differed little with social class forboys. It has been suggested that girls are highly awareof the value society places on attractiveness inwomen, and see thinness and attractiveness as synony-mous.114 Furthermore, insecurity in adolescent girlsmay increase their compliance with societies' stan-dards. Thus, as Sobal and Stunkard summarize,69 thecircumstances and values of life in the SES into whichone is born may transmit pressures for thinness toadolescent girls during their socialization.Studies to date therefore demonstrate a consistent

pattern regarding social mobility and fatness inwomen. Even so, social mobility cannot account forSES differences in obesity, since these differenceshave now been shown in several studies to pre-datethe mobility. Importantly, strong associations havebeen found between fatness and social class oforigin. Thus, in developed countries, the longitudinalrelationship between childhood SES and child or adultfatness, and the cross-sectional relationship inwomen69 would appear to be robust, whereas cross-sectional relationships in men and children, and even


S9

longitudinal relationships in young children seem tobe less so. Interestingly, this pattern does not seem todepend on the measure of SES, and holds acrossseveral decades, during which social patterns havechanged considerably. Although education, occupa-tion and income are related, it could be argued thateach represents quite different aspects of SES. Overthe last 50 y, the proportion of the UK populationcompleting school has increased greatly, as has thepercentage of those entering further education. At thesame time, as reported in a recent government pub-lication,115 the nature and organization of paid workhas undergone major transformations. The economyhas shifted from heavy industry and manufacturingtowards service sector employment. Women's parti-cipation in the workforce has increased markedly,part-time working for both men and women hasincreased and full-time working decreased. A greaterproportion of the population is now outside the labourforce, due to unemployment, early retirement andeducation. Meanwhile, the average household dispo-sable income has grown, yet has not been evenlydistributed across the population. The gap betweenhigh and low incomes has increased and the propor-tion of households living on low incomes doubledbetween 1961 and 1990. These trends are not uniqueto Britain: other developed countries also show per-sisting or widening socio-economic inequalities.115

Other social factors also show changing patterns.Households have been getting smaller, with thenumber of UK households increasing by 44% between1961 and 1995, while the size of the populationincreased by only 12%. The number of children perfamily has decreased, and the average age of themother at ®rst birth has increased. The percentage ofhouseholds headed by lone parents with dependentchildren has risen from 2% in 1961 to 7% in 1995 ±1996.It is intriguing therefore, that while social patterns

have been shifting in the developed world, a consis-tent relationship between SES and fatness hasemerged, although exactly what it is about SES thatpromotes the development of greater fatness remainslargely unknown. Being born into a particular socialclass cannot itself `cause' obesity, but characteristicsof socio-economic groups related to material circum-stances and behaviour or knowledge, which ultimatelyin¯uence energy balance, might. Attitudes towardsobesity in developed societies however, may providesome explanation of the relationship to SES. Obesityis far more severely stigmatized in women than inmen, both socially116 and economically.

Intra-uterine growth

Introduction

Birth weight is a crude indicator of growth in utero,and an association between birth weight and adult

obesity might therefore implicate foetal environmentin the development of obesity in adult life. There aremany factors which operate to determine the devel-opment of the foetus, and birth weight has beenobserved to be related to, amongst other factors,socio-economic status,115 gestational age,117 maternalsmoking,118 gestational diabetes119 and famine.120

Such factors could potentially confound the observedrelationship between birth weight and later fatness.There have been several recent suggestions formechanisms which may contribute to a birthweight ± later fatness relationship. Higher birthweights and greater levels of obesity in offspring ofmothers with gestational diabetes suggest that devel-opment of the hypothalamus=pituitary=adrenal axis, inparticular cortisol and insulin secretions, may alterfoetal programming and development. Metabolicresponses of adipocytes, muscles and liver tissue,and ef®ciency of macronutrient oxidation, may alsobe determined, to some extent, in utero.

Secular increases in birth weight in developedcountries, including the UK,121 suggest that if greaterweight at birth is positively correlated with fatness inadulthood, this may partly underlie the trend inincreasing obesity.

Methodology

All longitudinal, observational studies identi®ed bythe search strategy, which reported birth weight and ameasure of fatness at least 1 y later, were included inthis section of the review. Both studies reportingmeasured birth weights, and birth weights reportedby parents=guardians were included.

Results

Limitations of the data. The relationship betweenbirth weight and later fatness was investigated in 19study populations, reported in 20 publications (Table4 and Appendix I, Table A7). Eight studies (sevenpapers) spanned from childhood into adult-hood;77,79,85,122 ± 125 one paper by Curhan et al 123

included data from the Nurses Health Studies I andII, the remainder falling within the childhood period(< 18 y).47,54,57,63,74,96,97,99,126 ± 130 Three papers werefrom the same study by Kramer et al 74,96,97 Unfortun-ately, although gestational age is a major determinantof birth weight, only three studies47,85,125 attempted todeal with gestational age in the analyses. Five studiesreported on parental fatness47,54,57,74,123 (NursesHealth Study I in the paper by Curhan et al 123).Two studies made some attempt to investigate theeffect of parental fatness on the birth weight ± laterfatness relationship.57,123 Guillame et al 54 observedthat maternal (but not paternal) BMI was positivelyrelated to birth weight but did not make an adjustmentin analyses. O'Callaghan et al 47 did not includeparental fatness and birth weight in the same multi-


S10

variate model, and Kramer et al 74 reported a positiverelationship between birth weight and later fatnessdespite adjusting for parental fatness. Five studiesconsidered social group,47,74,99,125,126 but only fourof these included birth weight and social group inthe same multivariate model.74,99,125,126

Findings. (i) Childhood to adulthood: seven out ofeight studies investigating the in¯uence of birthweight on adult fatness used BMI as a measure offatness,77,85,122 ± 125 the eighth using weight for heightand age.79 Five studies observed a positive relation-ship, with adult BMI increasing with increasing birthweight,85,122,123,125 and one study found no associa-tion.124 This last study is the smallest, including only137 subjects, while four of the other ®ve are largestudies including between 4062 and 92,940 subjects.The study of Braddon et al,77 with a total sample sizeof 3249 men and women combined, found a positiverelationship in men, but not in women.(ii) Within childhood: thirteen publications from 11

studies examined the relationship between birthweight and fatness later in childhood, all ofwhich included BMI as an outcome measure.47,54,57,63,74,96,97,99,126 ± 130 Nine of 11 studies47,54,57,63,74,99,126,129,130 found positive relationshipsbetween birth weight and BMI, although in the smal-lest of these studies (n� 216) the relationship was ofborderline signi®cance.126 Six of the studiesreporting signi®cant positive relationships are rela-tively large, reporting results from between 1028 and20,7467 subjects.47,54,57,63,99,129 Both positive and no

signi®cant relationships between birth weight andBMI were seen in one study,127 depending ongender and infant feeding method. This was a rela-tively small study (432 subjects with follow-up data)and the data were analysed in subgroups by mode ofinfant feeding and gender. The other study128 report-ing no signi®cant relationship between birth weightand BMI was again a small study (188 subjects).Three studies examined associations between birthweight and skinfold measurements (in addition toBMI), and results were inconsistent; a negative rela-tionship was observed with subscapular skinfold,126

and positive relationships with triceps126 or sum oftwo130 or three skinfolds.74

Summary

On the whole, the association between birth weightand later BMI, as reported in the longitudinal studiesidenti®ed for this review, would appear to be consis-tent and positive. Study sample sizes were generallylarge, and interestingly, those studies observing anegative, or no relationship, were those with thesmallest sample sizes either in total,124,126,128 or dueto analyses by sub-group.127 Despite the apparentconsistency in the larger studies, in those whichattempted to take account of potential confounders,the relationship was less clear.

Impressions

Birth weight is strongly related to gestational age,117

yet only two studies attempted to account for this inanalyses. Sorensen et al 125 included an adjustment for

Table 4 Relationship between birth weight and subsequent fatness

Author Age at outcome Outcomemeasure Relationship

Outcome in adulthood

Allison et al 122 28 ±52y BMI PositiveBraddon et al 77 36y BMI Positive (males)

None (females)Charney et al 79 20 ±30y Overweight based on weight for height and age PositiveCurhan et al 123a 58y (NHS I) BMI Positive (females)Curhan et al 123a 36.5 y (NHS II) BMI Positive (females)Hulman et al 124 28y Obesity based on BMI None (sexes combined)Rasmussen and Johansson85 18.2 y Overweight based on BMI Positive (for higher birth weights only)Sorensen et al 125 18 ±26y BMI Positive (males)Outcome in childhood

Barker et al 126 14 ±16y BMI Positive (P� 0.08)Subscapular skinfold Negative (females)Triceps skinfold None

Fomon et al 127 8y BMI Positive in formula fed malesNone in formula fed femalesNone in breast fed males or females

Guillaume et al 54 6 ± 13y BMI Positive (sexes combined)Harland et al 128 12 ±16y BMI None (sexes combined)Kramer et al 74,96,97 2y BMI Positive at 1 and 2y

Sum triceps� subscapular� suprailiac skinfolds Positive at 1 and 2y (sexes combined)Maffeis et al 57 12y Obesity based on BMI Positive (males and females)O'Callaghan et al 47 5y Obesity based on BMI Positive (sexes combined)Seidman et al 99 17y Overweight based on BMI Positive (males)Taylor et al 129 8 ± 11y BMI Positive (males)Zive et al 130 4y BMI Positive

Sum triceps� subscapular skinfolds PositiveHealth Survey for England63 2 ± 15y BMI Positive

aTwo studies reported in one publication.


S11

gestational age, and O'Callaghan et al 147 adjusted forsize for gestational age. Both studies reported positiverelationships between birth weight and adult BMIafter these adjustments, suggesting that the relation-ship may be independent of gestational age.Although the positive relationship between birth

weight and subsequent fatness appears consistent,the picture is less clear when other confoundingvariables, such as parental fatness and socio-economicstatus (SES) are taken into account. Results from twostudies suggest that maternal fatness in¯uences therelationship.57,123 One study reported on the variationin the relationship between birth weight and later BMIwith reported mother's ®gure at age 50 y.123 Only formothers with a medium ®gure (rather than thin orheavy), was there an increased likelihood of being inthe highest vs the lowest BMI quintile with increasingbirth weight. A second study reported57 that within the1st and 2nd quartiles of maternal BMI, prevalence ofobesity increased with birth weight > 3500 g, but thisrelationship did not hold in the 3rd and 4th quartiles ofmaternal BMI. Since only two studies investigated thein¯uence of parental fatness and ®ndings were incon-sistent, this issue remains unresolved.The relationship between birth weight and fatness

might also be confounded by SES, and the relation-ships between the three variables are not all whatmight be expected. Babies born to parents of lowersocial groups tend to have lower birth weights.115

Adult fatness, particularly in women, is morecommon among those of lower social group oforigin. One might therefore expect lower birthweight to be associated with greater subsequent fat-ness, which the studies identi®ed for this review showis generally not the case. SES would be expected toact as a confounding variable, decreasing the like-lihood of observing a positive relationship betweenbirth weight and adult fatness. Unfortunately none ofthe studies which considered SES, reported on therelationship between birth weight and fatness, beforeand after adjusting for SES.47,74,99,125,126,131 Four stu-dies reported adjusting for SES.74,99,125,126 In thesefour studies, birth weight remained a positive predic-tor of BMI after adjustment for SES.In the study of Fomon et al,127 the presence of a

relationship varied according to infant feedingmethod, suggesting that the birth weight ± later fatnessassociation may be moderated by other factors whichalso potentially in¯uence later fatness, althoughsample sizes in this study were relatively small.The in¯uence of birth weight on BMI does not

seem to be uniform across the range of birth weight.In those studies which reported later BMI by birthweight, or the effect of birth weight relative to areference group (mid-range of the birth weight dis-tribution), a J-shaped relationship was observed.Lower birth weights therefore were also observed tobe associated with an increased risk of greater BMIlater in life, although the effect was generally notas great as that of high birth weights.47,99,123,124

Unfortunately in the studies identi®ed, it was notpossible to determine whether this observation wasconfounded by socio-economic characteristics.

Allison et al reported results from a twin study, inwhich they too, found birth weight correlated withadult BMI. They went on to attempt to investigatewhether this relationship persisted after controlling forgenetic factors. They argued that if intra-pair differ-ences in birth weight in monozygotic twins wererelated to differences in adult relative weight, thengenetic factors could not be solely responsible for therelationship, and the intra-uterine environment mustbe important. They did not ®nd that birth weightdifferences were related to adult differences in relativeweight, and therefore suggested that, after allowingfor genetic factors, the intra-uterine environment,summarized by birth weight, was not a major con-tributor to adult fatness. Studies of infants whosemothers smoked during pregnancy,118,132 or wereexposed to famine120 seem to refute this argument.Birth weight is lower in babies born to mothers whosmoke,118 although `catch up' is evident by the ®rstyear,118 and fatness at 5 y is greater than in children ofnon-smoking mothers.132 Clearly this relationshipmay be confounded by social class which was notaccounted for in these studies. Data from the Dutchfamine study120 suggest that exposure to famine earlyin pregnancy leads to greater prevalence of obesity inoffspring in adulthood, whereas exposure late inpregnancy reduces the risk of obesity. However,famine exposure resulted in an excess of prematuredeliveries, infants of very low birth weight, perinataldeaths and malformations of the nervous system,making it dif®cult to generalize the role of intra-uterine environment to healthy populations.

There is good evidence for a relationship betweenbirth weight and adult fatness, as shown by generallylarge and reasonably long-term studies. Despite thisconsistency, in studies which attempted to takeaccount of potential confounders such as maternalfatness57,123 the relationship between birth weightand later fatness was inconsistent. It is unclear theextent to which the observed relationship is due to adirect effect of the intra-uterine environment, or togenetic or other factors.

Maturation

Introduction

Maturity has long been noted to be associated withvariation in body size; those more advanced in biolo-gical maturation tend to be taller, heavier and fatter,than average and later maturing individuals of thesame chronological age.133,134 Improvement in envir-onmental conditions, primarily nutrition and exposureto infectious disease, are major causes of earliermaturation. Menarche has been the most commonly


S12

used marker of maturation, and in developed countriesthere has been a trend towards earlier menarche overthe last hundred years or so, with age of menarche atabout 15 y in 1860 in the south of the UK, falling tobelow 13 y by 1970.135 At the same time, increaseshave occurred in height and weight. However, overthe last two decades, there has been a marked slowingof the earlier maturation trend in Europe, NorthAmerica and Japan, with stabilization in most better-off countries.135 Most datasets have shown that girlsfrom better off families reach menarche earlier thanthose from worse-off, and urban girls earlier thanrural.135 However, within the last 20 y, upper socialclasses in developed countries have been found toshow a slightly later age of menarche,136,137 indicatingthat the relationship between socio-economic status(SES) and menarche might have changed. The laterage of menarche in higher social groups has beenattributed to the improvement in social conditions,removing the primary causes for late maturation inlower social groups, and to a lower weight for heightin the higher social groups, ®rst observed to beassociated with age of menarche at least 20 y ago.138

In women, both too little and excess body fat havebeen associated with disrupted reproductive function,leading to the hypothesis that there might be a mini-mum or threshold weight-for-height for menarche andmaintenance of regular ovulatory menstrual cycles tooccur.138

Additional factors relate to age of menarche andinclude mother's age of menarche (positive relation-ship), percentage dietary energy from fat (acceleratesmenarche)139 physical training (delays menarche)139

and eating disorders (delays menarche).The suggested relationships between maturation

and body size have not been con®ned to girls, andin both boys and girls, skeletal maturity and peakheight velocity have been used as markers of biolo-gical maturation.In addition to the measures of maturation described

above which occur during puberty, the timing of anearlier period during growth, termed the `adiposityrebound' has been reported to be associated withfatness.18 Body mass index and skinfold measure-ments increase during the ®rst year of life, subse-quently decrease and then increase again at about 6 yof age. This second rise in childhood adiposity hasbeen de®ned as the adiposity rebound.

Methodology

All longitudinal, observational studies identi®ed bythe search strategy, which reported a measure of time,or rate, of maturation during childhood (< 18 y), andfatness at least 1 y later, were included in this sectionof the review.

Results

Limitations of the data. Twenty-one papers describ-ing results from 18 study populations investigating the

relationship between maturation and later fatness wereidenti®ed (Table 5 and Appendix I, Table A8). Threepapers were from the 1958 UK birth cohort,17,140,141

two papers were from the Zurich Growth Study,142,143

two papers from the same study by Rolland-Cacheraet al,144,145 and two papers from the AmsterdamGrowth and Health Study.146,147 In addition, twopapers were from the Fels study and may haveincluded some of the same individuals.148,149 Onepaper by Garn et al described two studies,150 andanother by Wellens et al described three cohorts.151

Sixteen of 18 studies spanned from childhood intoadulthood. One of these 16 studies did not measurematuration as such, but reported on the pattern ofvelocity of change in BMI,142,143 which best ®t intothis section of the review. Follow-up measurements intwo studies fell within childhood (< 18 y).152,153 Themajority of studies were small, with sample sizes of100 ± 300, although three studies were larger, withover 1000 individuals.17,154,155 The childhood to adult-hood period however, was well documented, with 16out of 18 study populations reporting outcome mea-sures in adulthood. Factors that might confound therelationship between maturation and fatness includeparental fatness, socio-economic status (SES), earlierfatness in childhood, diet and activity.

Findings. (i) Childhood to adulthood: ®fteen studiesexamined the relationship between time or rate ofmaturation and adult fatness, 13 of which includedBMI as an outcome measure. Ten studies observed anegative relationship between maturation and adultBMI.17,144,146,149 ± 151,154 ± 157 where the earlier or morerapid the rate of maturation, the greater BMI inadulthood. Three studies found no relationshipbetween timing of maturation and adult BMI.151,158

One study used skinfold measurements to de®nefatness150 ®nding a negative relationship betweenmaturation and each of several skinfolds.150 Threestudies included skinfold measurements in addition toBMI, reporting negative associations with some, andnone with others.146,147,151,158 One study reported thatmore rapid maturation was related to greater subse-quent percentage body fat148 and another that earliermaturation was related to greater weight-for-height.159

Where the relationship between maturation and laterfatness was not statistically signi®cant, it was negativein direction in all but one study, in which the directionof the non-signi®cant relationship was not given.149

An additional study investigated the pattern of changein velocity of BMI and fat and lean areas of the arm,noting that the peaks and troughs in velocity, espe-cially at puberty, were greater in those with higherBMI in early adulthood.142,143

(ii) Within childhood: one study investigated theassociation between age at menarche and BMI at 14 y,and observed a negative relationship,152 the otherlooked at gynaecologic age and change in skinfold


S13

measures, ®nding no relationship.153 In this latterstudy, although the relationship was positive in direc-tion (not signi®cant), gynaecologic age was a predic-tor in a complex model which included many

variables, one of which was also a measure of matura-tion. Two interim papers of childhood to adulthoodstudies, also reported negative relationships betweentiming of maturation and fatness at 16 y.141,145

Table 5 Relationships between maturation in childhood and subsequent fatness

Study Age at outcome Measure ofmaturity Outcomemeasure Relationship

Outcome in adulthoodBeunen et al 158 30y PHVage BMI None

Triceps NoneSubscapular NegativeSuprailiac NegativeMedial calf skinfolds None=negative

(depending on test)Burke et al 154 18 ± 30y Age at menarche BMI NegativeGarn et al 150a (NCPP) 20 ± 35y Age at menarche BMI NegativeGarn et al 150a (Tecumseh) 20 ± 35y Age at menarche Triceps, subscapular, iliac,

abdominal skinfoldsNegative for allskinfolds

Gasser142 17 ± 25y Pattern of velocity of changein BMI throughout childhood

BMI velocitySum triceps�biceps, sumsubscapular� suprailiacskinfolds

Cannot classify asother studies:patterning of velocitydiffered, refer tomore detailed table(Appendix I).

Gasser et al 143 17 ± 25y Pattern of velocity of changein fat and lean areas of upperarm throughout childhood

Fat area of upper armLean area of upper arm

Cannot classify asother studies:patterning of velocitydiffered, refer tomore detailed table(Appendix I).

Guo et al 148 23y Rate of maturation % body fat NegativeSiervogel et al 149 18y Age at max. BMI BMI Positive

Age at max. velocity BMI NoneAge at min. BMI Negative

Miller et al 159 22y Age at menarche Weight-for-height NegativePower et al 17 33y Tanner stage BMI Negative=none

(females=males)Axillary hair (males) NegativeAge at menarche Negative

Freeman et al 140 23y Comparison of relativeheight at 7 and 16y

BMI Negative

Stark et al 141 16y Age at menarche Relative weight NegativeProkopec and Bellisle156 18y Adiposity rebound BMI NegativeRolland-Cachera et al 144 21y Adiposity rebound BMI Boys: negative

Girls: negative=nonedepending oncomparison

Rolland-Cachera et al 145 16y Adiposity rebound BMI NegativeSt. George et al 157 18y Age at menarche BMI NegativeSherman et al 155 19 ± 73y Age at menarche BMI Negativevan Lenthe et al 146 27y Skeletal age BMI Skeletal age: negative

boys and girlsPHVage (boys only)Age at menarche (girls)

Sum skinfolds (biceps, triceps,subscapular, suprailiac)

PHVage: noneAge at menarche: negative(all apply to BMI andskinfolds)

Post and Kemper147 21y Skeletal age Sum skinfolds (as above) Negative (boys and girls)Wellens et al 151b 19.2 y Age of menarche BMI, triceps skinfold NoneWellens et al 151b 18.8 y Age of menarche BMI, triceps skinfold NoneWellens et al 151b 18.3 y Age of menarche BMI, triceps skinfold NegativeOutcome in childhoodHediger et al 153 15 ± 19y Gynaecologic age D triceps skinfold None

D subscapular skinfold NoneKnishkowy et al 152 14y Age at menarche BMI Negative

aTwo studies reported in one publication. bData from three cohorts reported in one publication. Negative relationshipÐearlier=rapidmaturation associated with greater fatness. Relationship: negative if reported as such with no statistical test mentioned, or signi®canceat P<0.05 stated; none if relationship reported as not signi®cant. Adiposity rebound� second rise in adiposity, usually occurs at �6 yon BMI and skinfold charts. PHVage�age at peak height velocity.


S14

Summary

The relationship between early or rapid maturationand greater subsequent fatness appears to be remark-ably consistent, but potentially confounding factorswere seldom accounted for. Social group was reportedin two studies,150 dietary intake in another,147 and twostudies included earlier fatness as a covariate.17,153

Impressions

The relationship between maturation in childhood andadult fatness is consistent with early or more rapidmaturers demonstrating greater fatness in later life.This relationship holds true in both males andfemales, and for a variety of markers of maturation:adiposity rebound, occurring at about 6 y of age, andpeak height velocity, pubertal stages or age atmenarche, all measured during adolescence. Fewstudies adjusted for possible confounding factors intheir analyses. In one of the studies reported by Garnet al (NCPP)150 the observed negative relationshippersisted if the sample was restricted on the basis ofeducation, and in the other (Tecumseh),150 a negativerelationship was observed in a group of subjects ofsimilar SES. Post and Kemper147 looked at differencesin diet and activity between late and early maturersand observed that activity levels, energy intake andprotein intake were higher between 13 and 21 y inboth boys and girls who were late maturers. Whetherthe effect of maturation timing on later fatness isindependent of diet and activity is not reported, butthe potential correlations between diet, activity andmaturation are demonstrated.It is evident, from several studies,17,140,141,144,

156,157,160 that a relationship between fatness earlierin childhood and maturation also exists. In general,this relationship would appear to be negative, suchthat greater fatness early in life is associated withearlier maturation, but may well depend on how earlyin childhood fatness is measured. Data from the 1958UK birth cohort, show a negative relationship betweenfatness from 7 y onwards and timing of matura-tion.17,140,141 Data from the Fels160 and Dunedin157

studies show similar associations, although one studyfound the relationship at 6 y was not statisticallysigni®cant.152 French data, which are more detailedin early childhood, also demonstrate this negativerelationship between early fatness and timing ofmaturation, but only above 3 ± 4 y of age.144 Belowthis, the relationship seems to be less consistent, withgreater142 or lower144,156 early fatness=birth weight inthe early compared to the later maturers, or nodifference.17,144 Although the relationship betweenfatness in early childhood and age of maturation isunclear, that between fatness at 5 y until puberty andage of maturation appears consistent. Since fatnesstends to track, the relationship between maturationand later fatness may be largely due to the associationbetween earlier and later fatness. One study attemptedto address this,17 ®nding that once an adjustment was

made for BMI at 11 y, the relationship between pub-erty and BMI at 33 y was markedly reduced. Inanother study which included initial fatness in amultivariate analysis, no relationship was seenbetween maturation and later fatness.153

There is good evidence for a consistent associationbetween early maturation and greater fatness in adult-hood, which suggests that the generally accepted idealof faster and greater growth in childhood may notalways be appropriate. However, there is also evi-dence for a relationship between greater pre-pubertalfatness and earlier maturation, which would seem tolargely account for the relationship between earlymaturation and greater fatness. Other lifestyle factorssuch as diet and activity might contribute to thisrelationship which were hardly investigated by thestudies we identi®ed.

Physical activity

Introduction

Physical activity and obesity are linked by the energybalance equation, in which obesity is the result ofaccumulated excess energy, and physical activity is acomponent of energy expenditure. Energy expendedduring activity varies enormously, from close to basalmetabolic rate (BMR) while at rest in bed, to morethan 10-fold the BMR during vigorous exercise. Thisvariation would tend to implicate a sedentary lifestylein the development of obesity, but the relationshipsbetween activity or energy expenditure and fatness,both within and across populations, are not as clear asmight be expected. As discussed by Ferro-Luzzi andMartino,161 some surveys have found increased phy-sical activity to be associated with lower fatness,whereas others have not. They also point out thatindividuals in developing countries show the lowestbody mass index (BMI), athletes are not very differ-ent, and those from developed countries have a BMIsomewhat higher. However, although energy expendi-ture=BMR (essentially energy expenditure adjustedfor body size) was highest in the athletes, there wasno difference between subjects from developed anddeveloping countries. This apparent contradictionmight be due to engagement in very low cost activityamong those in developing countries after completionof their essential work and domestic activities. Evi-dently, differences in energy intake will also contri-bute to the differences in BMI between developingand developed countries. As the relationship betweendiet and fatness may be confounded by energy expen-diture, so that between energy expenditure and fatnessmay be confounded by dietary intake.While the prevalence of obesity has been increasing

in the developed world, it is believed that levels ofactivity have been declining. Although activity pat-terns have changed considerably, activity and energy


S15

expenditure are not synonymous, and direct evidenceof decreasing rates of energy expenditure are lacking.Evidence to support changing patterns of activity isavailable from surveys of secular changes in howmuch time people devote to various activities; timespent in paid work by men, and in domestic activitiesby women, has decreased, while leisure time hasincreased.161 Most manual jobs have become mechan-ized, and data from the Allied Dunbar Health andFitness Survey (1990)162 show that only 20% of menand 10% of women are now employed in activeoccupations, which almost certainly differs from thesituation, say, 50 y ago. The same survey reports that,of young adults aged 16 ± 24, about a third take noregular activity (an average of three or more times perweek).Data on changing levels of activity in children is

particularly poor, but further crude evidence is gainedfrom use of transport data or by using markers ofinactivity or time spent in passive activity such aswatching television or reading. Trend data collectedby the Department of Transport demonstrates that, inthe UK between 1985 and 1992, walking and cyclingby children decreased substantially.163 During thisperiod, the average distance walked in a year by achild aged 0 ± 14 y declined by 20% and that cycledfell by 26%. In contrast, the average distance thatchildren travelled in a car in a year increased by 40%.Current ®gures for television viewing show that boysand girls aged 4 ± 15 y currently watch 17 ± 18 h oftelevision per week,164 and secular trends in televisionviewing show that adults watch twice as much tele-vision today compared with the 1960s.Physical activity could be considered a behaviour,

but since it is of particular interest in relation toobesity and has been more widely addressed in theliterature, it is discussed separately in this section.Physical activity is a complex exposure and certain

aspects of activity may be more important in preven-tion of obesity than others. Classically, physicalactivity is composed of four elements, as de®ned bythe FITT principle, frequency, intensity, time andtype. However, these elements relate more to adult®tness training, and it may be inappropriate to usethem to describe children's activity, which is highlysporadic and spontaneous in nature. It might be moreappropriate to quantify children's activity in terms ofamount or volume, in line with the shift in beliefregarding the type and amount of activity required toconfer a bene®t to health.165 Over the last decade therehas been a move away from an emphasis on exercisetraining, to promotion of a general increase in mod-erate activity, as re¯ected in the UK's recent HealthEducation Authority `More People, More Active,More Often' campaign.Even so, the problems of measuring physical activ-

ity, and de®ning the dimension under assessmentremain. In epidemiological studies, for reasons ofpracticality and cost, activity is frequently measuredby questionnaire. Questionnaires themselves vary

widely, but tend to focus on one or a few dimensionsof activity and will be most effective where activity iseasily recalled. Thus questionnaires may be appropri-ate for measuring vigorous activity which is likely tobe structured, but are unlikely to give a reliableestimate of moderate activity, which may be incorpo-rated into daily living activities. This is particularlytrue in children, for whom much activity may be play.Measures of inactivity such as television watchingmay be reasonably estimated by questionnaire, butfurther work is required to explain the relationshipsbetween television viewing and obesity. Televisionviewing may be a marker of decreased play or ®dget-ing activity, decreased vigorous activity and=orincreased energy intake. Objective methods of mea-suring activity include doubly-labelled water, wholebody calorimetry, movement sensors and heart ratemonitoring, each with its advantages and disadvan-tages. Doubly-labelled water is regarded as the goldstandard method for estimating total energy expendi-ture in free-living individuals, but is very expensiveand cannot measure patterns of activity. Whole bodycalorimetry is an alternative gold standard method forestimating total energy expenditure, but cannot mea-sure the individual in the free-living state and istherefore of limited use in large scale epidemiologicalstudies. Heart rate monitoring has the advantage ofbeing an assessment of both energy expenditure andactivity patterns, and with the advent of small andunobtrusive monitors, is an acceptable method in thefree-living situation. Heart rate monitoring compareswell with calorimetry and doubly-labelled water.Movement sensors vary quite considerably,166 buthave been used successfully to obtain activity datafrom children.167 Even if time and cost were noobject, since the dimension of physical activity mostimportant for protection against obesity is stillunclear, it is dif®cult to specify a measurementmethod of choice at present.

Methodology

To be included, the study must report a measurementof physical activity or inactivity (e.g. televisionwatching) or energy expenditure in childhood(< 18 y), and a follow-up measurement of fatness atleast 1 y later.

Results

Studies identi®ed fell clearly into two groups, thosewhere the baseline measurement was within the ®rstyear of life, or `pre-walking', and those where base-line measurement is after the ®rst year, or `post-walking'. Only in the post-walking group were out-comes measured in adulthood.

Nineteen publications from 16 studies were identi-®ed, describing relationships between activity, inac-tivity or energy expenditure and later fatness. Fivestudies (®ve papers and one abstract) assessed infantactivity;47,95,168 ± 171 one paper168 and the abstract169


S16

were from the same study. Twelve studies (14 papers)measured childhood activity;20,44 ± 46,167,170,172 ± 179

three papers from the Amsterdam Growth andHealth Study were identi®ed.172 ± 174 One study mea-sured both infant and childhood activity170 and isincluded in both groups.

1. Pre-walking activity

Limitations of the data. Five studies investigated thein¯uence of infant activity on fatness later in child-hood,47,95,168 ± 171 with outcome measurements con-®ned to the earlier years of childhood, up to 8 y(Table 6 and Appendix I, Table A9). Sample sizeswere generally small, except for one very largestudy.47 The measurement of activity in the latterstudy however, was very crude, with the motherasked to report on a questionnaire the frequency ofòveractivity' of the child. No de®nition or speci®cexplanation of òveractivity' was given. Activity andenergy expenditure in infancy might be associatedwith behaviour and temperament (see behaviour sec-tion), relationships on which there is very little infor-mation.

Findings. A variety of outcome measures were usedby the ®ve studies. No relationships between infantactivity and BMI, percentage body fat, subscapularskinfold, or sum of triceps and subscapular skinfolds,were found between 2 and 8 y. The only exceptionwas the study of Berkowitz et al 95 where, althoughthere was no association between frequency or inten-sity of infant activity and BMI at 6.5 y, there was anegative relationship between intensity (but not fre-quency) of infant activity and triceps skinfold.

Summary. There is very little evidence for an in¯u-ence of infant activity or energy expenditure on laterfatness. Studies however, were few, all but one weresmall in sample size, and reported on childhoodfatness only up to 8 y.

2. Post-walking activity

Limitations of the data. Two studies looked at thein¯uence of childhood activity on adult fatness, one ina sample of 181 individuals,172 ± 174 while the othercompared consistently active and inactive subjectsfrom a larger sample (n� 961)20 (Table 7 and Appen-dix I, Table A9). The remaining 10 studies reportedchildhood activity at 1 ± 12 y with study duration of2 ± 8 y. Sample sizes were generally small, only threestudies with more than 250 participants at follow-up.175,176,178 One of these, however, was a very largestudy, of just over 2000 children.176 Potential con-founders of the relationship between activity andfatness include parental fatness, socio-economicstatus (SES) and dietary factors, which were consid-ered by a few studies. Overall, interpretation is com-plicated by the wide variety of methods employed toassess activity, and the different dimensions of activ-ity under investigation.

Findings. (i) Childhood to adulthood: The Amster-dam Growth and Health Study172 ± 174 and the Cardio-vascular Risk in Young Finns Study20 investigated thein¯uence of childhood activity on adult fatness. In theformer study, using two slightly different methods ofanalysis, lower total activity was associated withhigher skinfold thicknesses at 27 and 29 y. In thelatter study,20 those active at each of three measure-ments during the study had lower subscapular skin-folds than those consistently sedentary, but no groupdifferences in BMI were found.(ii) Within childhood: Ten studies examined the

effect of activity, inactivity or energy expenditure, onfatness at ages between 6 and 17 y.44 ± 46,167,170,175 ± 179

Activity: six studies addressed various dimensionsof activity; both negative and no relationships werefound with later fatness.45,46,167,170,175,177 Results werenot consistent even within studies, differing accordingto dimension of activity,45 gender,170 or type ofanalysis.167

Inactivity: inactivity, as measured by time spentwatching television, was investigated in three

Table 6 Relationship between pre-walking physical activity and subsequent fatness

Study Age at outcome Description of activity Outcome Relationship

Berkowitz et al 95 6.5 y Neonatal activity;frequency and intensity

Log BMI None (frequency=intensity)

Triceps skinfold None (frequency)Negative (intensity)

Davies et al 168a 2y Total energy expenditure BMI NoneSum triceps� subscapular skinfolds None

Davies et al 169a 6.7 y As above Fat mass (adjusted for FFM) NoneKu et al 170 8y Total activity % body fat (UWW) NoneO'Callaghan et al 47 5y Òveractivity' Obesity based on BMI NoneWells et al 171 2 ± 3.5 y Total energy expenditure BMI None

Resting energy % body fat (deuterium dilution) NoneExpenditure Triceps skinfold None

Subscapular skinfold None

aOne paper168 and one abstract169 from same study. FFM� fat free mass, UWW�underwater weighing.


S17

studies.46,176,178 The two smaller studies observed noeffect of inactivity on later fatness.46,178 The third,much larger study, found a positive relationshipbetween time spent watching television and fatness6 y later in a sample of over 2000 children.176 Despitemarginal levels of signi®cance, these results are mademore convincing by the fact that baseline fatness andsocio-economic characteristics were adjusted for. Anhourly increment in television viewing per day wasassociated with a 2% increase in obesity prevalence inthis population.Energy expenditure: only one small study examined

the in¯uence of energy expenditure, both resting andtotal, on later childhood fatness.44 Resting energyexpenditure had no effect, whereas total energyexpenditure was positively related to increase in fat-ness.

Summary

The evidence for an in¯uence of childhood activity onlater fatness is inconsistent, with studies roughlydivided between ®nding either no effect or a protec-tive effect of activity. Due to different methods ofanalysis and the variety of methods used to measurephysical activity, it is dif®cult to estimate the size ofeffect that activity may have on the risk of developingobesity. Sample sizes were generally small, and these

studies have taken very different approaches to dataanalysis, making comparisons complex. Six studiesincluded potentially confounding variables in ana-lyses. Four studies included parental fatness as acovariate in analyses,45,46,167,178 two studies foundthat higher activity levels were associated withdecreased risk of fatness after adjustment,45,167 andtwo studies did not.46,178 Both studies investigatingthe effect of television watching on fatness includedSES in analyses,176,178 Dietz and Gortmaker foundthat longer duration of television viewing was asso-ciated with greater fatness,176 Robinson et al didnot.178 Klesges et al 45 included dietary variables intheir analyses, observing that higher activity wasassociated with lower risk of fatness, as did someanalyses from the Amsterdam Growth and HealthStudy.174 Maffeis et al 46 also included energy andmacronutrient intake in analyses but found no in¯u-ence of physical activity. Despite using a question-naire to gauge hours of television viewing per day,which is a relatively crude method of measuring(in)activity, the largest study identi®ed demonstrateda small, but adverse effect of inactivity on obesity.176

Small effects should not be disregarded, since theymay have considerable importance at the populationlevel, or lead to larger cumulative effects over longperiods of time.

Table 7 Relationship between post-walking physical activity and subsequent fatness

Study

Age at

outcome Description of activity Outcome Relationship

Raitakari et al20 18 ± 24y Leisure time activity (consistentlyactive vs consistently sedentary)

Subscapular skinfold Negative

BMI NoneTwisk et al172a 29y Total activityÐMETs=week Sum skinfolds representing

� 20% body fat males,� 30% females

Negative

Twisk et al173a 27y As above Sum four skinfolds Negativevan Lenthe et al174a 27y As above Subscapular skinfold None=positive (males)b

Negative=none femalesb

Beunen et al 175 17y Non-compulsory sport (active vsinactive)

Triceps, subscapular, suprailiac,calf skinfolds (separately)

None

Dietz and Gortmaker176 12 ± 17y TV watching hours=day Obesity based on triceps skinfold Positivec

Goran et al 44 8.3 y Total energy expenditure Rate of change of fat massadjusted for fat-free mass

Positive (total EE)

Resting energy expenditure None (resting EE)Klesges et al 45 6.4 y Increase in leisure activity Change in BMI Negative (leisure)

Structured activity None (structured)Aerobic activity Negative (aerobic)

Ku et al 170 8 y Total activity % body fat (UWW) None (girls)Negative (boys)

Maffeis et al 46 12.6 y Extra-curricular� vigorous play,min=day

Obesity based on relative BMI None

TV viewing min=dayMoore et al 167 6.5 y Total activity Caltrac monitor

counts=hourIncrease in triceps skinfold None=negativeb

Increase in subscapular skinfold NoneIncrease in BMI None

Parizkova177 15y Total activity % body lean (UWW) NegativeRobinson et al 178 14.4 y TV watching hours=day Change in BMI None

Change in triceps skinfold NoneShapiro et al 179 9 y Total activity Sum of four skinfolds Negative

Childhood to adulthood studies in bold type. MET�metabolic equivalent, EE�energy expenditure, UWW�underwater weighing.aThree papers from same study, various analyses reported. bRelationship dependent on method of analysis. cThis relationship ispositive between inactivity and fatness, therefore negative between activity and fatness.


S18

Impressions

The evidence from the studies identi®ed for thepresent review was inconsistent, but suggestiveof a protective effect of activity against developingobesity.Cross-sectional data from the UK and longitudinal

data from The Netherlands show that during adoles-cence, the time that children spend on exercisedecreases considerably.180 ± 184 The two childhood toadulthood studies identi®ed for the current reviewincluded this period, and both observed a protectiveeffect of activity on fatness measured by skinfoldthicknesses.20,172 It is possible that with the moremarked decline in activity in late adolescence, activeand inactive subjects are more easily identi®ed, andrelationships between activity and fatness becomeclearer. Some studies suggest that physical activitytends to track, and since tracking seems to becomestronger with age from teenage years through earlyadulthood,20 activity in adolescence may have impli-cations for later life. Although not examining trackingper se, the only long-term prospective data in the UKrelating to physical activity, indicate that an aboveaverage ability in school games was positively asso-ciated with activity at age 36.185

Intervention studies suggest that, whatever its rolein the genesis of obesity, increases in activity are animportant component of programmes designed totreat obesity and maintain weight. Indeed, althoughactivity may not be as effective as modi®cation ofdiet for actually losing body weight, it seems to playa crucial role in weight maintenance. There areconvincing data to support this in adults,186 butdata are lacking in children. The role of exercisein treatment of obesity in children has beenaddressed by ®ve randomized controlled trials ofat least 1 y in duration.10,11 Exercise showed bene-®cial effects when added to a diet programme187 ordietary and behavioural management techniques,188

or as part of a multi-component programme.189

Aerobic or lifestyle exercise was more effectivethan callisthenics (control group)190 in promotingweight loss and, interestingly, reinforcing adviceon decreasing time spent in sedentary activity wasmore effective than reinforcing increased physicalactivities or a combination of the two.191 As in thestudies identi®ed for the current review, outcomemeasures and exercise programmes or advice variedacross studies, yet even so, small but consistentbene®ts were observed.A few randomized and non-randomized controlled

trials of exercise in non-obese children have alsoshown small positive effects of exercise. In adoles-cents, a ®ve day a week exercise programme pre-vented fat increases in the thigh over 5 weeks. In pre-school children, a 30 week programme of additionalexercise 3 days a week showed a greater reduction inthe overall prevalence of obesity, although preventionof BMI gain was shown only in girls.192 After 2 y inanother study, children exposed to the exercise arm in

a school based trial tended to have lower levels ofbody fat.193

On the whole, intervention studies in children havenot so far provided more insight into the type ofexercise most important for prevention of obesity. Inthe one study comparing types of exercise, lifestyleexercise had the greatest success in weight loss,followed by aerobic exercise. The control groupperformed callisthenic exercise and gained weightover the year.190

Although the relationships between activity andobesity are of primary concern for this review, lookingone step further to health outcomes, there are dataexamining the relationships between ®tness, obesityand mortality, albeit in adults. Cardiorespiratory ®t-ness has been found to be a more important predictorof all-cause mortality than BMI in a large cohort ofmiddle-aged men.194,195 After adjusting for ®tness, thepositive relationship between BMI and mortality dis-appeared. Thus it would appear that if ®tness isincreased, even if fatness does not decrease, some ofthe risk associated with being fat is reduced. Thissuggests that levels of activity suf®cient to improvecardiorespiratory ®tness should be promoted regard-less of whether obesity is reduced.The observational longitudinal studies included in

this review are diverse in methods of measuringphysical activity and body fatness, and generally ofsmall sample size. There is a lack of studies spanningthe childhood into adulthood period. Despite theseproblems, the two studies with follow-up data inadulthood,20,172 and in particular the large study ofDietz et al176 suggest that activity may protect against,or inactivity promote, the development of obesity.Infant activity, which is likely to be a differentvariable from childhood activity, was not seen toin¯uence fatness in later childhood, although thisconclusion is based on few studies.

Dietary factors

Introduction

Energy intake falls on the opposite side of the energybalance equation to physical activity and other com-ponents of energy expenditure. If the balance tips infavour of energy intake and is sustained, body fat willaccumulate and obesity may develop. However, whilethe regulation of energy intake is considered crucial inthe maintenance of body weight and adiposity, therelationships may be more complicated than is sug-gested by a simple model of energy balance. There isincreasing discussion of the importance of dietaryfactors other than calorie intake, including infantfeeding method, dietary composition, particularlypercentage energy from fat, and energy density ofthe diet.


S19

Paradoxically, National Food Survey data196 indi-cate that average total energy intakes have fallenabout 20% since 1970, while the prevalence of obesityhas doubled. Yet metabolic studies suggest thatenergy cannot be considered a single currency, butmust be recognized in terms of its constituent macro-nutrients: fat, carbohydrate, protein and alcohol.197

Each macronutrient follows a particular metabolicpathway, and the ease with which it may be convertedto body fat for storage differs. Moreover, carbohydrateand protein stores seem to be closely regulated byadjusting oxidation to intake, whereas fat is almostexclusively used or stored in response to ¯uctuationsin energy balance.198 De novo fat synthesis appears tobe a very limited pathway for the deposition of bodyfat in humans199 and therefore constitutes only a smallproportion of body fat.200 Dietary fat is preferentiallystored, and therefore implicated in the major propor-tion of body fat. If energy intake is in excess, thesource of excess energy (fat or carbohydrate) maybecome important.198 Since general physical activitylevels appear to be falling, it is dif®cult to determineexactly what is happening to energy balance.Arguments about the potential in¯uence of infant

feeding on later obesity have centred around whetherthe infant's ability to regulate its intake, and=or themother's encouragement to feed, differs between breast-feeding and feeding of various formulas. Formula(bottle) feeding is frequently cited as a contributor toinfant obesity, since formula-fed infants have beenshown to be larger than those who are breast-fed.201

Length of breast feeding and age of introduction of solidfoods might also in¯uence the infants regulation ofenergy intake, since it has been suggested that theintroduction of solid foods leads to excess energyintake by increasing the energy density of the diet.201

Methods

All studies identi®ed from the results of the searchstrategy which investigated the in¯uence of dietary

intake or composition measured in childhood (< 18 y)on fatness at least 1 y later, were included in thissection of the review.

Results

The studies identi®ed fell clearly into two maingroups: those investigating infant feeding, and thoselooking at dietary intake during childhood. In each ofthese groups, only one study followed participantsfrom childhood into adulthood.

Infant feeding

Limitations of the data. Seventeen papers from 13studies looked at the relationship between feeding ininfancy (®rst year of life), and fatness later in child-hood (12 studies),47,73,74,94,96,97,130,202 ± 210 or in adult-hood (one study,211 Table 8 and Appendix I, TableA10). Two papers from a study by Agras et al,73,94

two papers from the DARLING study,203,204 and threepapers from a study by Kramer et al 74,96,97 wereidenti®ed. The only study directly relating infantfeeding to the risk of adult obesity211 was a smallsubset of a larger study. Of the 12 studies whichexamined the relationship between mode of infantfeeding, duration of breast feeding or energy intakein infancy and obesity in childhood, none had out-comes measured later than age 7. Seven studies madesome mention of socio-economic status (SES), ®ve ofwhich incorporated SES into multivariate ana-lyses,73,74,211 related it to fatness at follow-up,130 orto risk factor at baseline.202 In one study SES was notincluded in analyses with infant feeding47 and inanother study it was unclear whether SES was incor-porated into the analysis.210

Findings. Six of 13 studies investigated breast vsformula feeding, on fatness in childhood (®ve

Table 8 Relationships between infant feeding and subsequent fatness

Study Age at outcomeBreast vs formula

feeding Duration breast feedingAge at solid foodintroduction Energy intake

Agras et al 73,94a 6 y Positive=nonea None NoneBirkbeck et al 202 7 y NoneDewey et al 203=Heinig et al 204c 2 y Breast=formula=noneb NoneDine et al 205 5 y NoneKramer et al 74,96,97 2 y Negative NegativeMarmot et al211 32 y Breast (men)

None (women)O'Callaghan et al 47 5 y NonePoskitt206 5 y None NoneSveger207 4 y NoneVobecky et al 208 3 y NoneWells et al 209 2 ± 3.5 y Positive NoneWilson et al 210 7.2 y None NegativeZive et al 130 4 y None None

aRelationship dependent on outcome variable, positive for BMI, null for triceps skinfold. bRelationship dependent on outcome variable:weight-for-length, triceps and subscapular skinfolds, estimated % body fat greater in breast-fed children at 2 y. Biceps skinfold greaterin formula-fed children at 2 y. Flank and quadriceps and sum of ®ve skinfolds not signi®cantly different in breast and formula fedchildren at 2 y. cMore than one paper from same study, only more recent data used. Childhood to adulthood study in bold type.


S20

studies)202 ± 205,208,210 or adulthood (one study).211

Marmot et al found a small and marginally signi®cantassociation between breast feeding and greater fatnessat 32 y in men, and a non-signi®cant association in thesame direction in women.211 Of the studies withinchildhood, four found no in¯uence on fatness at 2 ±7 y.202,205,208,210 Dewey et al reported that, dependingon the outcome variable examined, formula feeding,breast feeding or neither was associated with greaterfatness at 2 y.203

Six studies looked at the effect of duration of breastfeeding on later fatness, three studies found no rela-tionship,47,130,206 one study suggested a negative rela-tionship (longer duration of breast feeding associatedwith lower risk of fatness),74 and one study suggesteda positive relationship (longer duration of breastfeeding associated with greater fatness).209 Dependingon the outcome variable analysed, Agras et al reportedeither no relationship, or an association betweenlonger duration of breast feeding and greater fatnessat 6 y.94

Five studies addressed the age of introduction ofsolid food; three studies observed no relationship withfatness at 1.5 ± 6 y94,202,210 and two studies suggestedthat earlier introduction of solids was associated witha lower risk of fatness at 1 or 7 y.74,210 Three studiesexamined the in¯uence of energy intake in infancy,but none found a relationship with fatness at 4 ±6 y.94,206,207

Summary. Overall, no consistent pattern emerged ofa relationship between mode of infant feeding, dura-tion of breast feeding, age of introduction of solids, orenergy intake in infancy and later risk of obesity. Itshould be emphasised that this conclusion is based ontwelve studies measuring fatness to 7 y, with only onestudy following through to adulthood. Few studies

considered potential confounders, particularly thewell-documented relationship between social classand breast feeding.212

Childhood dietary intake

Limitations of the data. Twelve papers from eightstudies investigated the in¯uence of childhood dietaryintake on later fatness (Table 9 and Appendix I, TableA10). Only one study, the Amsterdam Growth andHealth Study172 ± 174,213 reported data on diet in child-hood and adult fatness, in a sample of 182 individuals.The remaining seven studies reported on associationsbetween various aspects of childhood diet at 1 ± 8 y ofage, and fatness during later childhood.45,46,179,214 ± 218

These differences in age of measurement of bothexplanatory and outcome variables complicates inter-pretation of the data. All studies were relatively small,some particularly so.216,217 The majority of studiesestimated dietary intake using a dietary historymethod, which although it has the advantage of alonger measurement period, is generally considered afairly crude method of measuring intake.219 Possibleconfounders of the relationship between dietary intakeand fatness include parental fatness, socio-economicstatus, and energy expenditure or physical activity,which few studies took into account.

Findings. (i) Childhood to adulthood: The Amster-dam Growth and Health Study172 ± 174,213 was the onlystudy we found which estimated the in¯uence ofchildhood diet on adult fatness. Even within thisstudy, the in¯uence of macronutrient intake on fatnesswas inconsistent. This study included measurementsof diet and outcome at six time points between age 13and 27 y.Energy intake: Two of the four papers addressed

energy intake.174,213 Using data from all six time-

Table 9 Relationships between childhood dietary factors and subsequent fatness

StudyAge at

outcome Energy intake % energy from fat% energy fromcarbohydrate

% energy fromprotein

Cholesterolintake

Twisk et al172a 29y Negative=noneb None NonePost et al213a 27y Negative None None None NoneTwisk et al173a 27y None None Positivevan Lenthe et al174a 27y None (males) None (males

or females)None (males) None (males

or females)Negative (females) Positive

(females)Deheeger et al 214c 8 y None=positiveb

Rolland-Cachera et al 215c 8 y None None None PositiveGrif®ths et al 216 15y None (boys)

Negative (girls)Klesges et al 45 6.4 y None Positive NoneMaffeis et al 46 12y None None None NoneNicklas et al 217 7 y None None NoneShapiro et al 179 9 y NegativeShea et al 218 6 y None None

aFour papers from same study, various analyses reported.172 ± 174,213 bRelationship dependent on predictor, absolute or change in valueand time-point=interval considered. cTwo papers from same study, various analyses reported.214,215 Childhood to adulthood study inbold type.


S21

points in the study (from age 13 to 27 y), a negativerelationship was noted between change in energyintake and change in fatness.213 When males andfemales were separated, this relationship was con®nedto females, but did not persist when other lifestylevariables including physical activity were accountedfor.174

Fat: all four papers investigated percentage energyfrom fat, carbohydrate and protein, three using differ-ent analytical methods.172,173,213 No relationship wasobserved between change in percentage energy fromsaturated or unsaturated fat,213 or total fat,174 andchange in fatness over the study period, nor werethose consuming a higher proportion of energy fromfat more likely to be in the high risk group of fatnessat age 27.173 At the most recent follow-up, however,percentage energy from fat at 13 ± 16 y was negativelyrelated to fatness at 29 y.172 This relationship was notobserved for other time periods, (fat intake at 13 ± 21 yor 13 ± 29 y), nor did it persist after adjusting for otherlifestyle variables.Carbohydrate: three papers reported no relationship

between percentage energy from carbohydrate andlater fatness,172,173 or between change in carbohydrateand change in fatness.213 However, when males andfemales were separated, a positive relationshipsbetween carbohydrate intake and fatness was seen infemales,174 which did not persist after including otherlifestyle variables in analysis.Protein: three of the four papers found no relation-

ship between percentage energy from protein and laterfatness,172 or between change in protein and change infatness,174,213 but the fourth paper found that thosewith a high percentage energy from protein in the dietor those changing to such a diet were signi®cantlymore likely to be in the high risk group of fatness at27 y.173

(ii) Within childhood: Seven studies estimated diet-ary intake at 1 ± 8 y and investigated associations withfatness at 5 ± 15 y.Energy intake: ®ve studies considered energy

intake, two found no relationship with fatness at 8 ±12 y.45,46 One study found no relationship betweenabsolute energy intake, but a positive relationshipbetween change in energy intake between 4 and 6 y,and fatness at 8 y.214,215 One study found a negativerelationship with fatness at 9 y,179 and another found anegative relationship between energy intake and fat-ness at 15 y in girls but not boys.216

Fat: ®ve studies looked at percentage of energyfrom fat; four found no in¯uence on fatness at 5 ±12 y,46,215,217,218 and one study found a positive asso-ciation (P� 0.052) at 6 y.45 One study looked atcholesterol but found no relationship with fatness at5 ± 7 y.218

Carbohydrate: percentage energy from carbohy-drate was examined in four studies, none found arelationship with fatness at 2 ± 12 y.45,46,215,217

Protein: three studies investigated the effect ofpercentage energy from protein; two studies found

no in¯uence on fatness at 4 or 7 y,217 or 12 y,46 andone found a positive association with fatness at 8 y.215

Summary. The discussion is hampered by the smallsample sizes, paucity of data spanning childhood toadulthood, and dif®culty in measuring childhood diet-ary intake. Interpretation of data from studies of therelationship between diet at one point in childhoodand later fatness is also complicated by the differenttime points at which explanatory variables weremeasured. Nonetheless, the studies identi®ed hereprovide little evidence to suggest a relationshipbetween any aspect of childhood diet and fatness inadulthood. However, childhood dietary intake is verydif®cult to measure, making it likely that relationshipsmight be obscured. Few studies adjusted for potentialconfounders; three studies included parental fat-ness45,46,215 as covariates in analyses, two of whichalso included activity,45,46 the third study includedSES.215 These studies showed a positive relationshipbetween percentage energy from protein and fatnessafter adjustment for parental fatness or SES,215 anegative relationship between percentage energyfrom fat and fatness after adjusting for parental fatnessand activity,45 and no relationships between dietarycomposition and fatness after adjusting for parentalfatness and activity.46 Physical activity=®tness wasalso considered in some of the analyses of theAmsterdam Growth and Health Study172,174,213 and,after adjustment, the negative relationship betweenpercentage energy from fat and fatness172 disappearedand the positive relationship between percentageenergy from protein and fatness remained in one setof analyses,173 but was not observed in another.174

Impressions

During the past 2 ± 3 decades there has been consider-able debate about the role of infant feeding, particu-larly breast vs bottle feeding, in the promotion offatness both in infancy and possibly later in life. Overthe same period, there have been changes in thecomposition of infant formulas, and parental attitudesto feeding.220 As knowledge of breast milk composi-tion has increased, modern infant formulas have beenadapted to resemble breast milk more closely. It hasbeen demonstrated that infants fed old-style formulasconsumed signi®cantly more energy than those onmodern, adapted formulas or breast milk,221 andtherefore the relationships between infant feedingand fatness might have changed over time. Therecent UK estimated average requirements forenergy for formula fed infants,222 are based on com-bined evidence from energy expenditure data, andenergy intake data from infants receiving modernformulas, and are lower than earlier FAO=WHO=UNU guidelines.223 In addition to considera-tions of energy intake, formula feeding has beensuggested to promote infant obesity because themother tends to insist the child ®nishes the bottle, or


S22

provides over-strength formula. Other proposed con-tributors to infant obesity include early weaning, orintroduction of solid foods, which are generallyrelated to each other, but might not be of equalimportance in their in¯uence on later fatness.The evidence from longitudinal studies identi®ed

for the current review do not suggest that infantfeeding method has an in¯uence on fatness in child-hood. The only study to follow through to adulthoodsuggested that breast-feeding might be related to adultfatness,211 but there are dif®culties interpreting thesedata. This study is based on a subset of the 1946 UKbirth cohort, and reported that men who were breastfed had a triceps skinfold of 4mm higher, correspond-ing to 1.7% body fat,224 than men who were bottle fed.Sample sizes were not large, and the difference ofmarginal signi®cance (P� 0.06). The authors recog-nized that this relationship might be confounded bysocial class, but too few subjects were classi®ed asmanual to examine whether the difference existed inthe lower social classes, although it remained presentin those in the higher, non-manual social classes(itself a heterogeneous group). The relationshipsbetween infant feeding method and later fatness maynot remain constant over time, so that the relationshipmight differ for childhood fatness and for adultfatness. Studies with outcome measures at over 7 yof age will be needed to determine if this is the case.Similarly, no consistent evidence was identi®ed for

an in¯uence of duration of breast-feeding on child-hood fatness, and almost equal numbers of studiessuggested either no association, or a negative relation-ship between the age solid foods are introduced andlater childhood fatness. None of the three studiesinvestigating infant energy intake reported an associa-tion with childhood fatness. No study has followedparticipants for long enough to examine the contribu-tions of these factors to fatness in adulthood.Although there is a general acceptance that diet

plays a role in obesity development, the relativeimportance of various dietary factors is more contro-versial. Despite the biological plausibility that higherenergy intakes lead to increased energy storage andbody fat, many dietary surveys in Western countrieshave observed inverse relationships between energyintake and BMI. These studies, however, may easilybe confounded by other factors, particularly energyexpenditure. In the current review we found noevidence that absolute energy intake is positivelyrelated to later fatness, and only one study thatsuggested a change in energy intake was positivelyrelated to later fatness.214 One of the two studiesreporting a negative association between energyintake and later fatness attempted to investigatewhether increased energy expenditure explained therelationship seen in all subjects.213 Using ®tness as aproxy marker,213 the association was only partlyaccounted for. If, in the same study, males andfemales were separated174 and other lifestyle variables

including physical activity entered into multivariateanalyses, the relationship did not persist.Beyond energy intake itself, the hypotheses for

macronutrient composition in¯uencing fatness haveconcentrated around carbohydrate and fat, the majorsources of energy for the body. The levels of fat in thediet have received particular attention. Since the1970s absolute fat consumption has decreased, butbecause energy intakes have also fallen, fat as apercentage of energy has decreased very little. Arecent meta-analysis has suggested that at the begin-ning of the century, total fat intake was about 25% ofdietary energy, increasing to 33% by the late 1930sand peaking at 40% in the late 1970s before starting tofall.225 The National Food Survey ®gures are slightlyhigher, showing a peak of 42.6% in 1980, falling to39.1% in 1997.1,226 There is still some way to go tomeet the 1991 COMA recommendations,222 which arethat population averages for total fat intake should bereduced to 35% of total food energy, and saturated fatto 10% of energy. In the 1987 OPCS Nutritional andDietary Survey, only 12% of men and 15% of womenachieved the 35% limit of fat as percentage of foodenergy.227 Young children however, are much closerto the recommended levels, which are consideredappropriate for the whole population of children bythe age of 5 y.228 Between infancy and the age of 5 y,energy derived from fat is expected to fall from 50%,as supplied by breast milk or infant formula, to 35%.Recent data collected in 1992 ± 1993 suggest thatchildren aged 11

2± 41

2y, derive, on average, about

36% of energy from fat.228 Available data for olderchildren suggest that they may not be so close to therecommendations, but it should be noted that thesedata were collected in 1983. Although the averageconsumption of fat was about 38% of energy in 10 ±11 and 14 ± 15 y olds, three-quarters of children tookmore than 35% of their energy as fat.229

The relationship between dietary fat and obesity hasbeen recently reviewed.230,231 Ecological studies,which investigate the relationship between percentageenergy from dietary fat and relative weight acrosspopulations, have produced both positive and negativeassociations. Most cross-sectional studies withinpopulations have observed positive relationshipsbetween fatness in the population and fat intake,230

but prospective studies are less consistent.231 How-ever, the validity of all these epidemiological studieshas been criticized. Ecological, cross-sectional andprospective studies may all be subject to seriousconfounding by population or individual differencesin physical activity, smoking, socio-economic factorsand genetic susceptibility. Data from men and womenare not always separated, yet relationships may differbetween sexes. Furthermore, the precision and repro-ducibility of methods employed to measure energy orfat intake and relative weight vary greatly. Under-reporting of energy and fat intake by obese individualsis well recognized, and limits the possibility of study-


S23

ing the relationship in question in epidemiologicalstudies.Randomized controlled trials that investigated the

effect of a calorically unrestricted low fat diet onweight change, generally showed a modest weightreduction in the low-fat treatment group.230 How-ever, follow-up data suggest it is less clear whetherthe weight loss achieved in these relatively short-term studies could be maintained longer term. In arecent review of treatment of obesity, little successof only dietary therapy is seen, but sample sizes arevery small.10,11 Most of the studies discussed so far,however, do not attempt to disentangle the effect offat intake and energy intake. Because diets high infat are energy dense and highly palatable, a dietrelatively high in fat usually leads to increasedenergy intake. Conversely, decreasing fat content ofthe diet therefore usually simultaneously decreasesenergy intake. The role of energy density in theregulation of energy intake has been recentlyreviewed.197 Experimental studies show that if per-centage energy from fat is increased, and energydensity (kJ=100 g) allowed to increase in parallel,energy intake increases. If, however, energy densityis held constant, energy intake does not increase,despite the increasing fat content of the diet. Thiswould suggest that energy density of the diet has animportant role to play in regulating energy intake. Inpractice, with unmanipulated foods, a diet with lowenergy density is likely to also be low in fat.Despite the promotion of low fat foods and avoca-

tion of low fat diets, we identi®ed only one study ofsix that reported a positive relationship betweenpercentage energy from fat in childhood and greatersubsequent body fat. The appropriateness of recom-mending low fat diets for children and from what ageremains unclear.14,15 Low fat diets in childhood havebeen associated with de®cits in fat soluble vitaminsand poorer growth in height in some children.15 Therealso remains a danger that the recommendation of lowfat diets in childhood may lead to some parentsputting very young children onto extremely restricteddiets.Some studies have demonstrated that fatness in the

population decreases with increasing carbohydrate orsugar consumption, and that fat : sugar ratio predictsobesity better than fat alone.232 This can be explainedby the reciprocal relationship generally found betweenintakes of fat and simple sugars.233 None of thestudies addressing carbohydrate intake in the presentreview demonstrated any in¯uence on the develop-ment of fatness, in child- or adulthood.The role of protein intake in the development of

fatness has received little attention. In the currentreview, however, two of four studies observed apositive relationship between percentage energyfrom protein and later fatness.173,215 Rolland-Cacheraet al 215 found that a higher protein intake at 2 y wasassociated with an earlier adiposity rebound (seeMaturation section), which was associated with

increased adiposity at 8 y. The authors suggest thiseffect might result from protein intake leading to anincrease in IGF-1, triggering early adipocyte multi-plication. Participants in the Amsterdam Growth andHealth Study were older, and percentage energy fromprotein between ages 13 and 27 was seen to predictfatness at 27 y, for which the authors do not suggest anexplanation.

Overall, despite the prominent role that diet ispresumed to play in promoting adiposity, the currentreview found no consistent evidence that infant orchildhood diet is related to later fatness. This mayseem surprising in view of the belief that there islikely to be at least some tracking in food preferencesand eating habits in addition to any postulated phy-siological effects. The fact that little evidence wasidenti®ed linking childhood diet and adult fatness maypartly re¯ect the paucity of studies covering the wholeperiod. In addition, the studies in this section weregenerally small, the largest number of subjects atfollow-up was just over 400, increasing the possibilityof missing a relatively small but important associa-tion. Of particular importance in this area is thedif®culty in accurately estimating all elements ofdietary intake. This measurement error leads torandom misclassi®cation of exposure status andhence tends to bias studies against ®nding a relation-ship between dietary intake and later outcome. It hasbeen demonstrated, in a dataset of over 3000 subjects,that measurement error associated with dietary datamay not have a linear effect on regression coef®cientsand con®dence intervals, and may be marked whenmeasurement error variance accounts for more than10 ± 15% of the total variance.234

Interestingly, when relatively homogeneous groupswith markedly differing eating habits are compared,clear differences in fatness are also observed. Numer-ous studies (although not all), have observed lowerrelative body weights in vegetarians,235 ± 237 vegan,macrobiotic238 and Seventh Day Adventist239 popula-tion groups, compared with control groups or nationaldata. These ®ndings have been noted in both malesand females and across a large age range. Somestudies reported higher intakes of ®bre,236,240,241 orcarbohydrate242,243 or lower intakes of fat,235,240,244 orenergy,240,245 which might partly explain the differ-ences in body fatness, but few have simultaneouslyinvestigated levels of physical activity. Other factorsare also associated with vegetarian and other alter-native diets, such as lower incidence of smoking,lower alcohol consumption and generally increasedhealth consciousness. Because of the dif®culty ofmeasuring diet and adiposity, it is possible that thein¯uence of diet on fatness only becomes apparentwhen comparing more distinct, homogenous groups. Itis also likely that other health-related lifestyle habitsassociated with alternative diets, will contribute to therelationship between diet and fatness. The diet ± fat-ness relationship may itself be weak, but becomeevident when acting in combination with some other


S24

protective lifestyle factors, as observed in groupsfollowing alternative diets.

Behavioural and psychologicalfactors

Introduction

There is a diverse body of literature linking beha-vioural and psychological factors to obesity, much ofwhich concerns data from cross-sectional studies. Thelist of behavioural and psychological factors studiedin relation to obesity or weight loss is varied, andincludes, in children, self-esteem, temperament, per-sonality traits, feeding behaviour, nutritional knowl-edge, and maternal awareness of offspring eatingbehaviour. These factors have all been proposed byone mechanism or another, to in¯uence energy bal-ance. Temperament, which develops from birth, mayin¯uence not only subsequent feeding and activitybehaviour of the child,246 but also parental behaviour;for example, dif®cult children may be fed more oftento pacify them. Moreover, they may be fed sugary orenergy rich foods in these circumstances.246,247 Lowself-esteem has long been regarded as a factor whichcontributes to the development of fatness, and is thecentral element of the psychosomatic theory of obe-sity,248 in which food is associated with comfort.Unfortunately, the evidence for this hypothesis islargely from cross-sectional and case history data,with prospective observational studies almost non-existent.249 The relationship between self-esteem andobesity in children and adolescents has been quitewidely addressed, as summarized in a recent compre-hensive review.249 In contrast, personality has beensuggested to affect obesity by a little investigatedbiochemical mechanism.250 The way in which indivi-duals cope with stress can be described as an `active'reaction, involving help seeking and problem solvingbehaviour, or a `passive' reaction, characterized byavoidance of stressors, also considered a defeat reac-tion. A defeat reaction has been hypothesized to leadto hyperactivity of the hypothalamic ± pituitary ± adre-nal axis, causing increased cortisol secretion, whichpromotes fat accumulation.Additional behavioural and psychological factors

have been examined predominantly in studies ofadults, although they may have some relevance tochildren. Greater individual self-control, the abilityand propensity to engage in self-reinforcement,appears to enhance success at weight reduction.Eating behaviour patterns, in particular binge eating,are considered likely to in¯uence the development offatness. Binge eating disorder is de®ned as having apattern of regular and sustained binge eating, but incontrast to bulimia nervosa, neither starvation orcompensatory behaviours are present. In the UnitedStates, binge eating is a common and serious factor

among a subset of the overweight population,251 andhas been found to be more common in younger andheavier subjects, and signi®cantly related to dietaryrestraint. Increased restraint, the ability to eat less thandesired in resistance to physiological or psychologicalpressures to eat, has been observed to make anindividual more likely to overeat in response to anemotional challenge.252 Lastly, some individuals seemhave the ability to take corrective action when theirweight increases above some self-imposed level, abehaviour which may be acquired or learned. Whetherthis cognitive threshold may be modi®ed by training,however, or whether it is instinctive and cannot bealtered, is unknown.

Methodology

No attempt was made to de®ne behavioural or psy-chological factors, since predictors of fatness were notsearched for speci®cally in the search strategy. Thissection is a compilation of all studies identi®ed whichincluded a measure of behaviour or a psychologicalfactor during childhood (< 18 y) and a measure offatness at least 1 y later. Physical activity and inactiv-ity, which could be classi®ed as behaviour, is dealtwith in a separate section.

Results

Limitations of the data. Thirteen papers from 10studies were identi®ed investigating an aspect ofbehaviour or a psychological factor on the develop-ment of fatness47,73,74,94,96,131,209,246,253 ± 257 (Tables 10and 11, and Appendix I, Table A11); two papers fromthe study of Agras et al,73,94 two papers from the studyof Kramer et al,74,96 and two papers from the study ofWells et al.209,246 Two studies followed participantsfrom childhood to adulthood131,253 and the remainderfell within childhood, as summarized in the Table 10.Details of the behaviour or psychological factorsmeasured are given in Table 11. Some studies testedmany variables for relationships with fatness, inrelatively small sample sizes, increasing the likeli-hood of ®nding signi®cant relationships due tochance. Further problems associated with research inthis ®eld include the lack of standardized methods tomeasure psychological and behavioural variables,which are dif®cult to quantify. Use of global ratherthan speci®c measures of predictors such as self-esteem, personality or anxiety may mean that criticalvariables are missed. For example, general anxiety orself-esteem may not predict fatness, but that speci®cto body weight might.252

Findings. (i) Childhood to adulthood: one studyspanning the childhood to adulthood period assessedpersonality traits,253 and the other investigated vari-ables relating to sweet eating habits and mother'sknowledge of these habits in her offspring.131 Of the


S25

six personality traits investigated in the AmsterdamGrowth and Health Study (inadequacy, social inade-quacy, dominance, rigidity, achievement motivationand debilitating anxiety), only inadequacy at 13 ± 16 ywas related to increased fatness in adulthood.253 In thestudy of sweet eating,131 only mother's lack of knowl-edge had a signi®cant in¯uence on increasing thelikelihood of their offspring being overweight inadulthood, after adjustment for childhood BMI,gender and social factors. The other variables,mother's acceptance of child's sweet eating habits,amount of sweet money, child's frequency of eatingsweets,mother's acceptanceof sugary food, andamountof pocket money did not signi®cantly affect offspring'sobesity after adjustment for covariates.(ii) Within childhood: four studies within the child-

hood period were concerned with feeding behaviour,maternal attitudes to feeding behaviour, sleeplessnessand temperament of the infant.47,73,74,246 Agras etal 73,94 found that a vigorous feeding style, character-ized by high-pressure sucking, more rapid sucking,longer bursts of sucking and shorter interburstinterval, was related to increased adiposity up to 3 y,but not at 6 y. This vigorous feeding style was,however, also related to energy intake at the time ofmeasurement, 2 ± 4 weeks of age (see also dietarysection). In another study,47 infant feeding problems(unde®ned) had a protective effect on moderate, butnot on severe obesity at 5 y. Kramer et al 74,96 found

that neither infant temperament, maternal preconcep-tion of ideal body habitus or maternal feeding attitudewas related to fatness at 2 y, whereas Wells et al 246

found that infant distress to limitations (see Table 11for de®nition), was positively related and soothability(de®nition, Table 11) negatively related to fatness at2 ± 3.5 y.

The four remaining studies concerned childrenrather than infants,254 ± 257 and addressed the in¯uenceof temperament, self-esteem, family functioning, andattitudes towards and knowledge of nutrition, onfatness 1 ± 5 y later. Carey et al 254 found that, ofnine categories of temperament, decreased adaptabil-ity, increased intensity and increased withdrawal wererelated to increased weight-for-height. No in¯uence ofthe other six categories, activity, distractibility, mood,persistence, rhythmicity=predictability, or sensorythreshold was seen. Klesges et al 256 assessed fourtypes of self-esteem in children: cognitive ability,physical ability, peer relations, general self-esteemand also family function (reported by parents). Aftera year, higher physical esteem and more positivefamily situations as viewed by the father, were relatedto greater increases in body fat in girls, smallerincreases in boys. But these relationships becameweaker with length of follow-up. At year 3, onlyfamily functioning was related to BMI (not tricepsskinfold), with opposite effects in boys and girls.French et al 255 found physical appearance and

Table 10 Behavioural and psychological factors investigated in relation to the development of fatness

Study

Age at predictor and

outcomemeasurements Outcomemeasure Risk factor

(i) Childhood to adulthood

van Lenthe et al 253 13 y (14, 15, 16, 21 y) 27 y Sum biceps, triceps, subscapularand suprailiac skinfolds

Personality traits

Lissau et al 131 9 ± 10y Obesity and overweight basedon BMI

Variables relating to sweet eating habitsand mother's knowledge of such habits

20 ± 21y(ii) Within childhood

Agras et al 94 2 and 4 weeks Triceps skinfold, log BMI Sucking behaviour, active feeding time,sucking pressure, interburst interval,no. feeds=day

2yAgras et al 73 2 and 4 weeks As above As above

3 y, 6 yCarey et al 254 3 ± 7y Weight-for-height Nine categories of temperament.

8 ± 9yFrench et al 255 12 ±15y BMI Nine types of self-esteem

15 ±18yKlesges et al 256 3 ± 6y % ideal body fat based on triceps

skinfoldFour types of self-esteem

6± 9y BMI Family Relationship Index (FRI) assessingfamily function

Kramer et al 74,96 1 ± 3 days for IBH�MFA BMI, sum triceps� subscapular�suprailiac skinfolds

Infant temperament (ITQ)Maternal preconception of ideal infantbody habitus (IBH)

Maternal feeding attitudes (MFA)2 weeks for ITQ1y, 2 y

Morris et al 257 10 y D weight ± height index Attitudes towards nutrition11 y D skinfold index Knowledge of nutrition

O'Callaghan et al 47 6months Obesity based on BMI Feeding problems4 ± 6y Sleeplessness

Wells et al 209,246 12 weeks Sum triceps� subscapularskinfolds, % body fat

Infant temperament

2 ± 3.5 y


S26

Table 11 Details of behavioural and psychological factors investigated in relation to the development of fatness

Study Risk factor Description of risk factor

(i) Childhood to adulthood

van Lenthe et al 253 Personality traits:inadequacy Vague feelings of malfunctioning, anxiety,

vague physical=psychosomatic complaints, depressedmood

social inadequacy Neurotic shyness, uncomfortable feelings in socialsituations, avoidance of unfamiliar people or situations

dominance rigidity Need for regularity, having ®xed habits�principles,sense of duty, perseverance

achievement motivation Need to achieve, will to reach achievementsdebilitating anxiety Fear of failure, leading to lower achievements, especially

in unstructured tasksLissau et al 131 Variables relating to sweet eating

habits and mother's knowledge ofsuch habits

Mother's acceptance of child's sweet eating habitsAmount of sweet money, child's frequency of eating sweetsMother's knowledge of child's sweet eating habitsMother's acceptance of sugary foodAmount of pocket money

(ii) Within childhood

Agras et al 73,94 Sucking behaviour, active feedingtime, sucking pressure, interburstinterval, no. feeds=day

Carey et al 254 Nine categories of temperament: `Dif®culty of child'activity Amount of physical motion during sleep, eating, play,

dressing, bathing etc.rhythmicity (younger children) orpredictability (older children)

Regularity of physiologic functions such as hunger,sleep, elimination (younger children) or predictabilityand quality of organization e.g. doing things on time(older children)

approach=withdrawal Nature of initial responses to new stimuliÐpeople,places, foods, toys, procedures

adaptability Ease or dif®culty with which reactions to stimuli can bemodi®ed in a desired way

intensity Energy level of responses, regardless of quality or directionmood Amount of pleasant� friendly or unpleasant�unfriendly

behaviour in certain instancespersistence=attention span Length of time particular activities are pursued by child

with or without obstaclesdistractibility Effectiveness of extraneous environmental stimuli in

interfering with ongoing activitiessensory threshold Amount of stimulation, such as sounds or light, necessary

to evoke discernible responses in childFrench et al 255 Nine types of self-esteem: Harter Self-perception Pro®le for Adolescents

globalphysical appearanceathleticscholasticsocial acceptancejob competenceromantic appealbehavioural conductclose friendship

Klesges et al 256 Four types of self-esteem: Harter perceived Competence Scale for Childrencognitive ability Emphasizes academic abilitiesphysical ability Focuses on sport and outdoor gamespeer relations Assesses standing in peer groupgeneral self-esteem Focuses on self-perceptions of general all-round competence

Family Relationship Index (FRI)assessing family function

(a) cohesionÐdegree of helpfulness and support amongfamily members

(b) expressivenessÐextent to which family membersare encouraged to be open and express their feelings

(c) con¯ictÐextent to which family relations arecharacterized by open expression of anger and aggressionand general con¯ictual situations

Kramer et al 74,96 Infant temperament (ITQ): Modi®cation of questionnaire as used by Careymaternal preconception of ideal infantbody habitus (IBH)

maternal feeding attitudes (MFA)Morris et al 257 Attitudes towards nutrition

Knowledge of nutritionO'Callaghan et al 47 Feeding problems Unde®ned

Sleeplessness

Table11Continued


S27

social acceptance self-esteem were inversely relatedto change in BMI over 3 y in girls, but not boys.Morris et al 257 found that better knowledge of nutri-tion or nutritious meals were protective againstincreasing fatness, but that there was no in¯uence ofpositive attitudes to good nutrition, knowledge relat-ing to fattening puddings, or preferences of nutritiousmeals or fattening puddings.

Summary. Few studies have addressed the potentialin¯uence of behavioural or psychological factors onthe development of fatness, particularly in the longerterm. The studies identi®ed for the current reviewwere very diverse in the factors they investigated,making it dif®cult to draw unifying themes from thoseidenti®ed. There is potential for considerable overlapbetween behaviour and psychological factors andsocial, dietary and physical activity factors, and themechanisms by which behaviour might disturb energybalance appears largely untested.

Impressions

Many different constructs have been used to investi-gate the relationships between behavioural or psycho-logical variables and fatness, and include among thestudies presented here, personality traits, tempera-ment, self-esteem, sweet eating habits, infant feedingbehaviour, knowledge of nutrition and various mater-nal attitudes. Many of these were tested only in onestudy, providing little evidence to support or refuteprospective relationships with the development offatness. Factors such as child's knowledge of nutritionor mother's knowledge of sweet eating habits, couldalmost be regarded as socio-economic factors, ormight at least be highly correlated with other socialfactors. Some studies have taken socio-economicfactors into account,131 others have not.257

Temperament and self-esteem were investigated inmore than one study, and although assessment toolswere not identical, at least some common elementswere identi®ed. Kramer et al 74 and Wells et al 246

investigated temperament in infancy, using differentmodi®cations of the questionnaire used by Carey,254

who looked at temperament in childhood. Wellset al 246 found that greater irritability in infancy, asindicated by distress to limitations, was related toincreased fatness, and less irritability, as indicatedby the child being more easily soothable, was relatedto lower fatness. Soothability in this study is related tointensity, adaptability, sensory threshold and mood, asused in the study of Carey et al and described byRothbart.258 The ®ndings in the two studies broadlyagree, with the study of Carey et al,254 ®nding thatincreased `dif®culty' of the child, as indicated bydecreased adaptability or increased intensity, wasrelated to greater subsequent fatness. Conversely, ina larger but shorter study, Kramer et al 74 found norelationships between temperament and later fatness.In addition to being less fat, Wells et al 246 found thatmore easily soothable children were more `awake andactive' at follow-up, suggesting that infant tempera-ment may also relate to subsequent physical activitybehaviour. There was no relationship between distressto limitations and activity, but more irritable children,who were fatter, watched less television, an observa-tion which does not easily reconcile with the greateractivity of more soothable (less irritable) children.Measurement of activity however, was fairly crude.Neither soothability nor distress to limitations ininfancy were related to energy intake at follow-up,as measured by dietary history. Neither Carey et al 254

nor Kramer et al 74 investigated the possibility thattemperament might also be related to energy intake orexpenditure. A recent study of overfeeding in adultshas reported that the 10-fold differences in fat storagewere largely accounted for by changes in non-exerciseactivity, such as the activities of daily living, ®dget-ing, spontaneous muscle contraction and maintainingposture when not recumbent.259 Those who increasedtheir non-exercise activity the most gained less fat,suggesting that activity such as ®dgeting, which isvery dif®cult to measure in epidemiological studies,may be an important factor in resisting weight gain.Whether non-exercise activity is related to tempera-ment requires con®rmation.

Two studies looked at relationships between self-esteem and subsequent fatness, one in children, the

Table 11 Continued

Study Risk factor Description of risk factor

Wells et al 209,246 Infant temperament: `Irritability of child'time spent fussy Unsettled, irritable, restless, or fractious. may be vocalizing,

but not continuously cryingtime spent crying Periods of prolonged distressed vocalization

frequency of distress to limitations Measure of a child's fussing, crying, or showing distresswhile (a) waiting for food, (b) refusing food, (c) beingin a con®ning place or position, (d) being dressedor undressed, or (e) being prevented access to anobject toward which the child was directing his=her

attentionfrequency of soothability Child's reduction of fussing, crying or distress when

soothing techniques used by parent=caretaker of child(related to intensity, adaptability, sensory threshold and

mood as used by Carey258)


S28

other in adolescents. Each used a self-esteem scalespeci®c to the age of the subjects, but both scales weredesigned by Harter, and overlapped in the areas ofself-esteem assessed. In children, lower self-esteem intheir physical ability was related to greater subsequentfatness, a relationship which weakened with length offollow-up. However, in adolescents, lower self-esteemin their physical appearance or social acceptance wasrelated to greater subsequent fatness, whereas athletic(or physical ability) self-esteem did not relate to laterfatness. However, this may not be surprising; ifphysical appearance and social acceptance is moreimportant to adolescents than their physical or sport-ing ability, the self-esteem relationships might beexpected to change over time. Although self-esteemand fatness has been quite widely investigated, sincethe vast majority of the studies are cross-sectional,249

these issues have yet to be addressed. Existingresearch into self-esteem and obesity also showscontradictory ®ndings which deserve further attention.The review of French et al 249 suggested that self-esteem was often lower in obese adolescents, yet onelarge study has shown that self-esteem in overweightand non-overweight 16 ± 24 y olds was similar,and that their weight status did not predict theirself-esteem 6 y later.8 This apparent contradictionmay be due to differences in measurement methods,or because the relationships are complex, and maydiffer according to the weight status, social andcultural background of the population. The samereview found that weight loss treatment programmesincreased self-esteem, although the increase was notpredictive of weight loss.249 Which aspects of self-esteem are related to obesity, and whether they protectagainst obesity or are more important for weight loss,remains unclear.Perhaps the best current evidence that behavioural

and psychological factors may be important, is pro-vided by intervention studies which aim to treatobesity, incorporating a behavioural component.Behaviour modi®cation, usually related to manage-ment of eating behaviour and=or physical activity, hasshown success in increasing weight loss in those withmild to moderate obesity, but whether behaviourchange and weight loss are maintained is question-able.260 Recently, data from randomized controlledtrials has been reviewed.10,11 Studies were few,sample sizes small and interventions varied, yetsome aspects of behaviour appeared to aid weightloss. In adults, cue avoidance (avoidance of situationsthat provide the temptation to overeat), daily weightcharting, behavioural therapy by correspondence, andextension of the intervention period for behaviouraltherapy all showed promise as a component of aweight-loss programme. In children, there is someevidence that either treating obese parents and chil-dren together,190 or involving parents without directlytargeting them for weight loss, may enhance weightloss or prevent obesity progressing in children.261,262

The `Shapedown' programme, a combination of cog-

nitive, behavioural, affective and interactional techni-ques to encourage `small sustainable modi®cations indiet, exercise, relationships, lifestyle, communicationsand attitudes', has been successful in treating obesityin adolescents.263 The manner in which childhoodbehaviour is in¯uenced may be crucial; reinforcingdecreased sedentary behaviour alone has been foundto be more effective for weight loss than reinforcingincreased physical activity or a combination of thetwo.191

Further studies support the hypothesis that the wayin which behaviour is modi®ed matters. Both posi-tively reinforcing children for being less sedentary orpunishing them for being sedentary seems to be moreeffective in increasing activity than restricting accessto sedentary activities.191 Making sedentary activitiescontingent on being physically active also results indecreased time spent on sedentary activities.264 Thesechanges in activity patterns may contribute to usefulstrategies to enhance weight loss or prevent obesity.Eating behaviour may also be modi®ed. Children havebeen shown to learn to dislike foods they are encour-aged to eat to obtain rewards (e.g. eat your vegetablesand you can watch TV) and prefer foods eaten inpositive social contexts as rewards (e.g. clean up yourtoys and you can have some cookies).265 How thedevelopment of food preferences relates to later fat-ness remains unknown.Finally, there are interesting ®ndings from beha-

vioural intervention studies where the interventionwas not aimed to in¯uence obesity itself, yet aneffect on obesity was found. Children whose motherstook part in practitioner ±mother interviews duringthe ®rst 5 y of their child's life, were less obese at 27 ±29 y.266 The interviews aimed to enhance the mother'sself-worth, foster gentle physical interaction withchild, and adopt a positive attitude to modifyingchild's behaviour, with the intention of improvingthe mother ± child relationship. In the Know YourBody Evaluation Project, a health education pro-gramme to reduce coronary heart disease, increasedbaseline self-esteem was related to greater decreasesin ponderosity in those who were obese at baseline butnot at follow-up.267

Although longitudinal, observational evidence forthe contributions of behaviour and psychologicalfactors to the development of fatness is limited, therange of data provided by other types of study suggestthat further longitudinal investigation may proveuseful. From existing literature, the picture appearscomplex, since behaviour and psychological factorsclearly overlap with social factors, dietary intake andphysical activity. How these factors interrelate isunclear, but it may be important if elements ofbehaviour or personality lead to disruptions ofenergy balance, and therefore predict susceptibilityto obesity. Studies of behavioural modi®cation forweight loss and treatment of obesity give clues tothese interrelationships, yet because treatment strate-gies are frequently multifactorial, they can be dif®cult


S29

to disentangle. Behavioural strategies that are success-ful in aiding weight loss may or may not be effectivein protecting against obesity at the population level,and caution may need to be exercised in comparingthe two.

Discussion

Although the links between adult obesity and adversehealth outcomes are established, those between childor adolescent obesity and morbidity5 and mortality6 inadulthood are less well documented. The health con-sequences of obesity in youth have been recentlyreviewed268 and evidence for cross-sectional relation-ships between childhood obesity and factors related tomorbidity in adults summarized. Increased bloodlipids, glucose intolerance, hypertension, andincreases in liver enzymes associated with fattyliver, have all been observed to be more common inobese children or adolescents. Since these factors tendto cluster and track with age, it seems likely that anincreased risk for morbidity will carry forward intoadulthood. However, as pointed out by Dietz,268 it isunclear whether the risk of morbidity associated withobesity varies with age of onset, severity, duration orthe factors responsible for the onset. Childhood obe-sity is also associated with adverse social conse-quences,7,8 and as the prevalence of obesity inchildhood increases the potential associated morbidityand psychosocial consequences in adulthood may be acause for concern.

Limitations of the evidence

Few studies have addressed long term predictors ofobesity, and therefore the current review identi®edfew datasets with measurements of risk factors inchildhood, and outcome measures in adulthood. Partof the problem appears to be that existing data are notfully utilized. In particular, there are several largecardiovascular disease, blood pressure and cancerstudies with few or no published longitudinal analysesof predictors of obesity that we identi®ed. These datacould be usefully analysed (or analysed more exten-sively) since sample sizes are generally large, andheight and weight data are usually available, so thatfatness can be gauged in terms of body mass index(BMI). Obesity, hypertension, cardiovascular diseaseand cancer share common lifestyle factors that arerelated to all four conditions, and so in many cases itis likely that data for risk factors such as thoseconsidered in the current review will be available,even if they have been measured relatively crudely.The lack of available analyses from these studies islikely to be due, ®rst to the fact that obesity was notthe main interest of the study, and second to publica-tion bias. Publication bias has been de®ned as atendency towards preparation, submission and pub-

lication of research ®ndings based on the nature anddirection of the results. It has been demonstrated thatpositive and signi®cant results are more likely to bepublished, and published more rapidly than negative,null or non-signi®cant ®ndings.269 Therefore thoseinvolved in relevant longitudinal studies, but whereobesity is not the main focus or outcome measure,may be less inclined to prepare and publish non-signi®cant ®ndings. In addition, selective reportingof outcomes and risk factors may occur so that onlythose demonstrating positive or signi®cant results arepublished. The result of publication bias is to increasethe observed size of effect.270 For some of the riskfactors in the current review, for example diet andactivity, there were few studies and effects varied, butit might be more of a problem in studies concerningbirth weight, maturation or social class, for which dataare more routinely collected and therefore readilyavailable.

Because risk factors need to be measured prospec-tively, longitudinal studies provide the strongest datafor evidence of relationships between risk factors anddevelopment of obesity. Unfortunately, inherent pro-blems in de®ning and measuring risk factors andoutcomes may tend to obscure real relationships,although this problem extends to most other studydesigns (as well as longitudinal studies). The de®ni-tion of obesity remains a matter of debate, although itis anticipated that the recent work of the WHO expertcommittee,271 and the International Task Force onObesity,272 in establishing a consensus of de®nitionof obesity in adults and children and adolescents,respectively, will lead to greater consistency of report-ing measures of obesity. The appropriateness of mea-surement methodologies to assess lifestyle factorssuch as diet and physical activity, and for behaviouralcharacteristics is equally debated. This lack of con-sensus was re¯ected in the studies identi®ed for thisreview, in which many different de®nitions and meth-ods of measurement were employed for both riskfactors and outcome measures, complicating thesynthesis and interpretation of the ®ndings. Weacknowledge that measurement standardization ofsome risk factors, notably physical activity and dietaryintake would be very dif®cult to achieve, and in somecases, such as for a social group, possibly not desir-able, since it will be country speci®c. However, itwould be particularly useful if raw data were pre-sented, and used in analyses, rather than using groupsde®ned by arbitrary cut-offs, to facilitate comparisonsacross studies.

Summary of ®ndings

Parental obesity has been consistently shown toincrease the risk of fatness in offspring, althoughfew studies have looked at this relationship overlonger periods of childhood and into adulthood. Thecontribution of genes and inherited lifestyle factors tothe parent ± child fatness association remains largely


S30

unknown, and may well vary between societiesexposed to different environment in¯uences. Therelationship between low SES in childhood andincreased fatness in adulthood is remarkably consis-tent in studies of large sample sizes. In relation tofatness in childhood, however, the association withSES is less consistent. Women who change socialclass (social mobility) show the prevalence of obesityof the class they join, an association which is notpresent in men. The in¯uence of other social factors,family size, number of parents at home, child-care andothers, have been little researched. There is goodevidence from large and reasonably long-term studiesfor an apparently clear relationship for increasedfatness with higher birth weight, but in studieswhich attempted to address potential confounding,the relationship was less consistent. The relationshipbetween earlier maturation and greater subsequentfatness was investigated in predominantly smaller,but also a few large studies. Again, this relationshipwas very consistent, but could easily be confoundedby other factors which were generally not accountedfor. Studies investigating the relationships betweendiet or activity and fatness were generally based onsmall samples, and included diverse methods of riskfactor measurement. There was almost no evidencefor an in¯uence of activity in infancy on later fatness,and inconsistent but suggestive evidence for a protec-tive effect of activity in childhood on later fatness. Noclear evidence for an effect of infant feeding on laterfatness emerged, but there were gaps in the timing ofoutcome measure, with only one study with a mea-surement after 7 y. A few, diverse studies investigatedassociations between behaviour or psychological fac-tors and fatness, but mechanisms by which energybalance might be in¯uenced were rarely addressed.In summary, although the associations may be

subject to confounding, a number of childhood factorswere predictive of later obesity; parental fatness, lowSES, higher birth weight, earlier=more rapid matura-tion and inactivity.

Interrelationships between risk factors

There are several issues that concern the interrelation-ships between risk factors: confounding, effect modi-®ers and cumulative effects. Many of the suggested riskfactors are highly correlated, ormay operate as proximaland distal aetiological factors on the same causal path-ways. Controlling for confounding in individual studies,and attempts to disentangle relationships when combin-ing studies is complicated by problems in preciselyde®ning many of these risk factors.Among interrelationships we identi®ed, higher

socio-economic status (SES) is associated with greaterbirth weight, and in most studies, earlier maturation,although recent data from developed countries showan association with later maturation.136,137 Mothers inhigher social groups are more likely to breast feed,and breast feed for longer.212 Those of higher SES

have been reported to eat a healthier diet, and to havean increased level of physical activity, as demon-strated by data from the 1946 and 1958 British birthcohorts.93,273 Data from the USA show similar pat-terns; the prevalence of regular, and regular andvigorous activity has been reported to be greateramong Whites, and to increase with level of educa-tion.274 In children, SES has been reported to beunrelated to energy intake, but fat consumption tobe lower and protein and carbohydrate intake higher,in children of higher social groups.275 The recentHealth Survey for England data also demonstratesassociations between higher social class and morehealthy eating habits, in children but not in youngadults (aged 16 to 24 y).63 The proportion of childrenaged 2 ± 15 y consuming fruit and vegetables morethan once every day, decreased from the higher tolower social classes or income, while the proportion ofchildren consuming sweet foods, soft drinks andcrisps more than once a day, or chips at least 5 daysa week, increased. There was no relationship betweenoverall activity and social class but participation indifferent activity types varied with social class orincome group. In younger children (aged 2 ± 10 y),participation in sports and exercise (structured activ-ities) decreased from higher to lower social classes orincome group. A similar trend was seen in girls aged11 ± 15 y, but not in boys. Participation in active playor walking did not vary across social classes, but theproportion of children involved in active play orwalking on 5 days or more in the previous weekwas higher among the manual social classes. Amonggirls, levels of participation in housework or garden-ing were higher in the manual social classes, a trendthat was not seen in boys. The Health Survey forEngland also showed that behavioural, emotional orrelationship dif®culties were signi®cantly related toSES, with the lowest proportion of children withdif®culties in social class I, increasing to the highestproportion in social class V. Since in adults, womenfrom higher social groups tend to be leaner,69 SESmay also be linked to inheritance of obesity, becauseit is associated with maternal fatness.Nutrition and physical activity have both been

linked to timing of maturation. Childhood malnutri-tion has been associated with retarded maturation, asmeasured by delayed bone age and age atmenarche,276,277 whereas earlier maturation has beenclosely correlated with improved nutrition.135

Menarcheal age of athletes who begin trainingbefore menarche is delayed, each year of premenarch-eal training delaying menarche by about 5 months.138

Similar relationships have been observed in cohortstudies: late maturing boys and girls have been foundto be more active than early maturers,147 andincreased sports activity to be associated with latermenarche.139 Energy and protein intakes have beensuggested to be higher in late maturers after adjustingfor body weight,147 and energy-adjusted fat intake tobe associated with earlier menarche.139


S31

Increased energy expenditure or overall level ofphysical activity is necessarily accompanied by anincrease in energy intake if body weight is to bemaintained, and therefore each has the potential toconfound the relationship between the other andfatness. Unfortunately, both risk factors are verydif®cult to measure, and only three studies identi®edin the current review investigated (or reported ana-lyses of) diet and activity simultaneously.45,46,173,174

Beyond the relationship between energy intake andexpenditure, the macronutrient composition of energyintake might vary with energy expenditure=activitylevel. Children with a high activity level have beenobserved to consume more energy than those with alow activity level, but to consume a higher percentageof energy as carbohydrate and a lower percentage asfat.278 Increased inactivity, particularly televisionwatching, might also be related to increased dietaryintake, either during viewing, or in response to foodadvertising, a subject addressed in a recent review ofeating behaviour in children.265 In the USA themajority,265 and in the UK approximately half 279 theadvertisements during children's television are forfood products. Of these, 80% in the USA are forfoods with low nutritional value, including breakfastcereals high in simple sugars and snack foods high insugar, fat and salt.265 Similarly in the UK, 60% offood advertisements were for breakfast cereals, snackfoods and confectionery.279 There are few data linkingtelevision advertising and dietary intake, but childrenexposed to advertisements have been reported toselect more sugared foods,280 and in Australia, teen-agers who watch television more extensively havebeen suggested to eat healthy foods less often andunhealthy foods more often.281 Inactive pastimes suchas watching television may also displace periods ofmore active play or exercise from children's lives,further reducing overall activity. Increased televisionviewing has been associated with decreased level ofphysical activity, albeit weakly.178 A small studysuggests that, while watching television, children'smetabolic rate is lower than that during rest, whichmight contribute to the relationships between televi-sion watching and fatness.Certain aspects of behaviour might also be related

to physical activity and thereby fatness. Childrenreported by teachers to have high levels of energy inadolescence, and who were judged to be good atgames and were extrovert in personality, reportedtaking more exercise as adults, at 36 y of age.185

Behavioural and dietary factors are also likely to berelated in infancy. For example, mothers may use foodas a pacifying technique with less soothable or moreirritable infants, and wean infants to reduce infantdistress.209,246 Thus dietary factors could contribute tothe relationships between soothability, or weaning,and later fatness.Some relationships between risk factors and fatness

seem paradoxical. For example, babies with higherbirth weights, and babies born to parents of lower SES

subsequently become fatter. However, babies ofhigher birth weight are those from higher socialgroups and would therefore be expected to beleaner, rather than fatter, in later life. As previouslydiscussed (Intra-uterine growth section), the relation-ship between birth weight and later fatness may itselfbe confounded by other factors such as maternal bodysize or fatness.

Perhaps one of the most important potential con-founders of the relationship between early life riskfactors and later obesity, is parental fatness. Parentsprovide genetic predispositions, but also the environ-ment in which these predispositions are expressed,and therefore the inheritance of parental fatness isinevitably a combination of inherited genes andshared environment. The shared environment islikely to include behaviours that children have learntfrom their parents. Statistical adjustment for parentalfatness in analyses therefore not only takes account ofparental fatness, but also parental lifestyle factorswhich might be associated with their child's as wellas their own fatness. For example, fatter parents mayhave lower physical activity levels, which could inturn be related to offspring activity, which itself mightbe a predictor of offspring fatness. The contributionsof parental genes and shared common environmentare almost impossible to disentangle, but for thisreason we would strongly advocate that study resultsshould be presented both with and without adjustmentfor parental fatness.

Confounding is clearly a problem in attempting todetermine the important risk factors for obesity.Acknowledgement of the many interrelationshipsbetween risk factors and attempts to address them inanalyses, however, is important.

Factors associated with obesity might act either asconfounding factors, as described above, or as effectmodi®ers, that is, when one particular factor modi®esthe effect of another. The concept of genetic suscept-ibility is really only useful in the context of othercontributors to obesity, and gene ± environment inter-actions have been reported in relation to obesity inadults. Data from the Finnish Twin Cohort Study havesuggested that, in men, although not in women,genetic effects on change in BMI were dependent onphysical activity level.282 In Swedish women, thosewith a family history of obesity had a much higherrisk of weight gain on a high fat diet.283 Since the genepool has remained relatively constant while obesityhas been increasing, it is likely that changes in theenvironment and lifestyle are allowing expression ofgenes that were previously not expressed. If, forexample, an individual has genes predisposing toobesity, under extreme conditions of food limitationor enforced activity, the individual is unlikely tobecome obese. Today's environment in developingcountries, with sophisticated transport systems, seden-tary jobs and abundance of low cost, relatively high-fat foods, may be considered to highly predispose toobesity. Such factors may well modify the risk for


S32

those children with pre-existing risks, such as havingan overweight parent(s).Finally, cumulative effects on the development of

obesity are likely, both over time for speci®c risks orat any particular time over a range of risk factors. Therelationship between each risk factor and obesity isusually weak, but small effects should not be disre-garded, since they may amount to larger cumulativeeffects over time. So far, it remains unknown how riskfactors in childhood exert their effects on adulthood.There might be a physiological effect of, for example,physical activity in childhood on fatness in adulthood,or the effect might be due to tracking behaviour, withphysical activity in childhood in¯uencing later fatnessvia later activity patterns. In addition, `healthy' and`unhealthy' lifestyle factors tend to cluster together, sothat those who exercise are more likely to eat a dietcloser to current recommendations and are less likelyto smoke.273 Moreover, `healthy' lifestyle factors tendto be associated with higher social groups.93,273

Clearly, the relative contribution of each risk factorto the development of obesity is very dif®cult toassess, and because the effects of individual riskfactors are weak, an effect may manifest only incombination with other factors.

Population and individual level in¯uences

In¯uences on obesity may occur at either or both thepopulation level, determining the position of the dis-tribution as a whole, and at the individual level,affecting the position of an individual within thepopulation distribution. Because the gene pool isrelatively stable, genetic make-up may have a strongereffect on an individual's position in the distribution,than on the position of the distribution itself. At least indeveloped countries, socio-economic conditions havechanged greatly, and there has been an overallimprovement in living standards, which is likely tohave contributed to the upward shift in adipositydistribution of the population. Socio-economic group,however, is a relative measure that determines theindividual's position, rather than the population posi-tion. Lifestyle factors such as diet and activity, are alsolikely to exert an in¯uence at both population andindividual levels. Overall decreases in physical activ-ity, or increases in fat in the diet, might shift thepopulation adiposity distribution upwards, but changein an individual's diet and exercise habits could alsochange their position within the distribution. Themulti-level in¯uences of diet and activity mightexplain why small effects of each on obesity areobserved, but that larger effects are seen on comparisonof groups at the extremes of a risk factor (for example,very active compared with inactive). Possible popula-tion changes in behaviour are beyond the scope of thisreview, but behavioural factors might well in¯uencewhere an individual is located in the distribution.From the data currently available, it is dif®cult to

draw conclusions for simple policy implementation.

There is an added need for caution in childhood, toensure normal processes of growth and maturation arenot disrupted.

Research priorities

The major research gap identi®ed by the currentreview is the lack of long-term follow-up data span-ning the childhood to adulthood period. This gapcould be ®lled to some extent, ®rstly by follow-upof groups on whom good quality baseline data havealready been collected, and secondly, by furtherexploitation of existing datasets. For example, theBritish National Diet and Nutrition Survey of Chil-dren aged 11

2± 41

2y228 and the Nutritional Surveillance

Survey of the Diets of British Schoolchildren229 havecollected extensive dietary data on over 1500 and3000 children, respectively. In addition, the HealthSurvey for England data from children might prove avaluable source of baseline data.63 In the currentreview, parental fatness, low SES, higher birthweight, earlier maturation and inactivity are shownto be associated with later fatness. Further work mightexamine the sensitivity and speci®city of using suchmeasures or combinations to predict risk as a basis forfuture intervention strategies. Furthermore, given thefailure of many previous studies to account for con-founding variables, further insight into the causalchains of risk for obesity could be gained. Issuesthat remain unresolved include identi®cation of themechanism by which SES in early life in¯uencesobesity in adulthood and whether the relationshipsbetween birth weight and maturation and later obesitypersist after accounting for confounding factors. Evi-dence for relationships between dietary factors andactivity and later fatness are largely con®ned to smalldatasets, and there is little information as to whetherthese relationships are due to a direct effect, or totracking in dietary or activity behaviour. The inter-relationships between psychological factors and beha-viours that could in¯uence energy balance, andtherefore fatness, are largely unexplored. Given thatthe present day Western environment may predisposeto obesity, an additional neglected area of research isthe identi®cation of factors which predict the main-tenance of an appropriate and healthy relative weight,which may or may not be the opposite of predictors ofobesity. Acknowledging the inherent problems inmeasurement of risk factors and outcomes, an addi-tional approach to addressing these issues may be touse large samples for whom more basic measures ofrisk factors have been collected. This might allow thein¯uences of tracking of lifestyle behaviours and howgroups particular clusters of risk factors predict obe-sity to be investigated.The relationships between some of the risk factors

considered in the current review and secular trend datawould seem compatible, for example, higher birthweight is associated with greater fatness and bothbirth weight and fatness have increased over recent


S33

years. Similarly, maturation has been occurring ear-lier, and earlier maturation is related to fatness. Thesecular changes and evidence from longitudinal stu-dies for dietary factors and physical activity are lessconsistent, yet secular data may inform further stu-dies. If for example, the population distribution ofactivity has shifted so that the majority is inactive andat risk of obesity, and the range of activity level isnarrow, relationships between activity and obesitymight be obscured. There is therefore more agreementbetween secular trend data and ®ndings from popula-tion studies for risk factors which are more easilymeasured, and measured with more precision, but lessfor risk factors which are more dif®cult to measure.The extent to which lifestyle factors track over thelifecourse is debatable, but could be highly relevant tothe development of obesity. Cumulative effects ofdifferent risk factors also deserve attention, sinceweak effects may amplify over time. The challengeto future research will be to reconcile current know-ledge on individual and population data in order todiscern the important, and modi®able risk factors, orclusters of factors, on the causal pathways in thedevelopment of obesity.

Acknowledgements

This work was undertaken by TJ Parsons, C Power, SLogan and CD Summerbell, who received fundingunder the joint Department of Health=MedicalResearch Council Nutrition Research Initiative; theviews expressed in this publication are those of theauthors and not necessarily those of the sponsors. Theauthors would also like to acknowledge the time andinput from collaborators and members of the steeringgroup (Appendix III), DH of®cials and peer reviewerswho acted for DH, and additional experts consulted.We would be grateful for any comments on thereview, or information on studies we may have over-looked.

References1 Charlton J, Murphy M (eds). The health of adult Britain

1841 ± 1994, Vol. I, Decennial supplement no. 12. TheStationary Of®ce: London 1997.

2 Peckham CS, Stark O, Simonite V, Wolff OH. Prevalence ofobesity in British children born in 1946 and 1958. Br Med J1983; 286: 1237 ± 1242.

3 Hughes JM, Li L, Chinn S, Rona RJ. Trends in growth inEngland and Scotland, 1972 to 1994. Arch Dis Child 1997;76: 182 ± 189.

4 James WPT. The epidemiology of obesity. In: Chadwick DJ,Cardew G (eds). The origins and consequences of obesity. JWiley: Chichester, 1996, pp 1 ± 16.

5 Power C, Lake JK, Cole TJ. Measurement and long-termhealth risks of child and adolescent fatness. Int J Obes 1997;21: 507 ± 526.

6 Must A, Jacques PF, Dallal GE, Bajema CJ, Dietz WH.Long-term morbidity and mortality of overweight adoles-cents. A follow-up of the Harvard Growth Study of 1922 to1935. New Engl J Med 1992; 327: 1350 ± 1355.

7 Stunkard A, Burt V. Obesity and the body image. II. Age atonset of disturbances in the body image. Am J Psychiat 1967;123: 1443 ± 1447.

8 Gortmaker SL, Must A, Perrin JM, Sobol AM, Dietz WH.Social and economic consequences of overweight in adoles-cence and young adulthood. New Engl J Med 1993; 329:1008 ± 1012.

9 Rissanen AM. The economic and psychosocial consequencesof obesity. Ciba Found Symp 1996; 201: 194 ± 201.

10 Glenny AM, O'Meara S, Melville A, Sheldon TA, Wilson C.Systematic review of interventions in the treatment andprevention of obesity. CRD 10, NHS Centre for Reviewsand Dissemination: York, 1997.

11 Glenny AM, O'Meara S, Melville A, Sheldon TA, Wilson C.The treatment and prevention of obesity: a systematic reviewof the literature. Int J Obes 1997; 21: 715 ± 737; and websitehttp:==www.york.ac.uk=inst=crd=obesity.htm.

12 Douketis JD, Feightner JW, Attia J, Feldman WF. Periodichealth examination, 1999 update: 1. Detection, prevention andtreatment of obesity. Can Med Assoc J 1999; 160: 513 ± 525.

13 Epstein LH, Valoski AM, Kalarchian MA, McCurley J. Dochildren lose and maintain weight easier than adults: acomparison of child and parent weight changes from sixmonths to ten years. Obes Res 1995; 3: 411 ± 417.

14 Nutrition and your health: Dietary guidelines for Amercians,4th edn, Home and Garden Bulletin no. 232. US Deptart-ments of Agriculture and Health and Human Services:Washington DC, 1996.

15 Nicklas TA, Webber LS, Koschak M, Berenson GS. Nutrientadequacy of low fat intakes for children: The Bogalusa heartstudy. Pediatrics 1992; 89: 221 ± 228.

16 Serdula MK, Ivery D, Coates RJ, Freedman DS, WilliamsonDF, Byers T. Do obese children become obese adults? Areview of the literature. Prevent Med 1993; 22: 167 ± 177.

17 Power C, Lake JK, Cole TJ. Body mass index and heightfrom childhood to adulthood in the 1958 British born cohort.Am J Clin Nutr 1997; 66: 1094 ± 1101.

18 Dietz WH. Critical periods in childhood for the developmentof obesity. Am J Clin Nutr 1994; 59: 955 ± 959.

19 Kemper HC, van Mechelen W, Post GB, Snel J, Twisk JW,van Lenthe FJ, Welten DC. The Amsterdam Growth andHealth Longitudinal Study. The past (1976 ± 1996) and future(1997 ± ?). Int J Sports Med 1997; 18 (Suppl 3): S140 ± 150.

20 Raitakari OT, Porkka KV, Taimela S, Telama R, Rasanen L,Viikari JS. Effects of persistent physical activity and inac-tivity on coronary risk factors in children and young adults.The Cardiovascular Risk in Young Finns Study. Am JEpidemiol 1994; 140: 195 ± 205.

21 Dovey SM, Reeder AI, Chalmers DJ. Continuity and changein sporting and leisure time physical activities during adoles-cence. Br J Sports Med 1998; 32: 53 ± 57.

22 Prentice AM. ObesityÐ the inevitable penalty of civilisa-tion? Br Med Bull 1997; 53: 229 ± 237.

23 Jensen MD. Research techniques for body compositionassessment. J Am Diet Assoc 1992; 92: 454 ± 460.

24 Overweight and obese patients: principles of managementwith particular reference to the use of drugs. The RoyalCollege of Physicians of London: London, 1997.

25 Clinical guidelines on the identi®cation, evaluation, andtreatment of overweight and obesity in adults: executivesummary. Expert Panel on the Identi®cation, Evaluation,and Treatment of Overweight in Adults. Am J Clin Nutr1998; 68: 899 ± 917; and website http:==www.nhlbi.nih.-gov=nhlbi=cardio=obes=prof=guidelns=ob_home.htm.

26 Hortobagyi T, Israel RG, O'Brien KF. Sensitivity and speci-®city of the Quetelet index to assess obesity in men andwomen. Eur J Clin Nutr 1994; 48: 369 ± 375.

27 Lissner L, Odell PM, D'Agostino RB, Stokes J III, KregerBE, Belanger AJ, Brownell KD. Variability of body weightand health outcomes in the Framingham population. NewEngl J Med 1991; 324: 1839 ± 1844.


S34

28 Marshall JD, Hazlett CB, Spady DW, Conger PR, QuinneyHA. Validity of convenient indicators of obesity. Hum Biol1991; 63: 137 ± 153.

29 Barlow SE, Dietz WH. Obesity evaluation and treatment:Expert Committee recommendations. The Maternal andChild Health Bureau, Health Resources and Services Admin-istration and the Department of Health and Human Services.Pediatrics 1998; 102: E29; and website http:==www.pediatrics.org=cgi=content=full=102=3=e29.

30 Cole TJ, Freeman JV, Preece MA. Body mass indexreference curves for the UK, 1990. Arch Dis Child 1995;73: 25 ± 29.

31 Cole TJ. Changing prevalence of childhood obesity in theWestern world. (Abstract). Int J Obes 1998; 22 (Suppl 4): S1.

32 Program Abstracts, NAASO Annual Meeting. Obes Res1997; 5 (Suppl 1).

33 Eighth International Congress on Obesity Abstracts. Int JObes 1998; 22 (Suppl 3).

34 Critical appraisal of studies. Cochrane Collaboration Hand-book [updated September1997], Section 6. In: Mulrow CD,Oxman AD (eds). The Cochrane Library [database on diskand CD ROM]. The Cochrane Collaboration=Update Soft-ware: Oxford, 1997, Issue 4.

35 Roede M, Stoelinga G, van de Kar M, van 't Hof M. TheNijmegen growth study. Anthropometrical data on healthyDutch children aged 5 and one half, 8 and one half and 10and one half years (in Dutch). Ned Tijdschr Geneeskd 1976;120: 133 ± 140.

36 Kubat K, Kourim J, Renc Z. Relationship between the breastfeeding with the body composition and the number of acutediseases in a group of children followed up to 6 years of life(in Czech). Cesk Pediatr 1974; 29: 606 ± 608.

37 Pastorini M, Silli U. Longitudinal study of body constitutionin children aged 2 to 10 years (in Italian). Riv Clin Pediatr1969; 82: 155 ± 160.

38 Locard E, Mamelle N, Munoz F, Miginiac M, Billette A, ReyS. Life style of children and obesity in a population of 5-year-old children (in French). Rev Epidemiol Sante Publique1992; 40: 460 ± 466.

39 Wada J, Ueda K. Correlation between changes in obesityfrom adolescence to young adulthood and family obesityÐthe results of cross sectional and longitudinal studies(in Japanese). Nippon Koshu Eisei Zasshi 1990; 37:837 ± 842.

40 Maes HHM, Neale MC, Eaves LJ. Genetic and environmen-tal factors in relative body weight and human adiposity.Behav Genet 1997; 27: 325 ± 351.

41 Kaplowitz HJ, Wild KA, Mueller WH, Decker M, TannerJM. Serial and parent ± child changes in components of bodyfat distribution and fatness in children from the LondonLongitudinal Growth Study, ages two to eighteen years.Hum Biol 1988; 60: 739 ± 758.

42 Lake JK, Power C, Cole TJ. Child to adult body mass indexin the 1958 British birth cohort: Associations with parentalobesity. Arch Dis Child 1997; 77: 376 ± 381.

43 Whitaker RC, Wright JA, Pepe MS, Seidel KD, Dietz WH.Predicting obesity in young adulthood from childhood andparental obesity. New Engl J Med 1997; 337: 869 ± 873.

44 Goran MI, Shewchuk R, Gower BA, Nagy TR, CarpenterWH, Johnson RK. Longitudinal changes in fatness in whitechildren: no effect of childhood energy expenditure. Am JClin Nutr 1998; 67: 309 ± 316.

45 Klesges RC, Klesges LM, Eck LH, Shelton ML. A long-itudinal analysis of accelerated weight gain in preschoolchildren. Pediatrics 1995; 95: 126 ± 130.

46 Maffeis C, Talamini G, Tato L. In¯uence of diet, physicalactivity and parents' obesity on children's adiposity: a four-year longitudinal study. Int J Obes 1998; 22: 758 ± 764.

47 O'Callaghan MJ, Williams GM, Andersen MJ, Bor W,Najman JM. Prediction of obesity in children at 5 years:a cohort study. J Paediatr Child Health 1997; 33:311 ± 316.

48 Rosenbaum PA, Elston RC, Srinivasan SR, Webber LS,Berenson GS. Cardiovascular risk factors from birth to 7years of age: the Bogalusa Heart Study. Predictive valueof parental measures in determining cardiovascular riskfactor variables in early life. Pediatrics 1987; 80 (Suppl):807 ± 816.

49 Ayatollahi SM. Obesity in school children and their parentsin southern Iran. Int J Obes 1992; 16: 845 ± 850.

50 Burns TL, Moll PP, Lauer RM. The relation betweenponderosity and coronary risk factors in children and theirrelatives. The Muscatine Ponderosity Family Study. Am JEpidemiol 1989; 129: 973 ± 987.

51 Moll PP, Burns TL, Lauer RM. The genetic and environmentalsources of body mass index variability: the Muscatine Ponder-osity Family Study. Am J Hum Gen 1991; 49: 1243 ± 1255.

52 Garn SM, Bailey SM, Solomon MA, Hopkins PJ. Effect ofremaining family members on fatness prediction. Am J ClinNutr 1981; 34: 148 ± 153.

53 Grinker JA, Garn SM, Rosenberg KR. Effects of parentalweight and fatness on relative weight and fatness of youngchildren. Comparison of cross-sectional and prospectiveclinical data. Ann NY Acad Sci 1989; 575: 521 ± 524.

54 Guillaume M, Lapidus L, Beckers F, Lambert A, BjorntorpP. Familial trends of obesity through three generations: theBelgian ± Luxembourg child study. Int J Obes 1995; 19(Suppl 3): S5 ± 9.

55 Hashimoto N, Kawasaki T, Kikuchi T, Takahashi H,Uchiyama. In¯uence of parental obesity on the physicalconstitution of preschool children in Japan. Acta PaediatrJpn 1995; 37: 150 ± 153.

56 Khoury P, Morrison JA, Laskarzewski PM, Glueck CJ.Parent ± offspring and sibling body mass index associationsduring and after sharing of common household environ-ments: the Princeton School District Family Study. Metabo-lism 1983; 32: 82 ± 89.

57 Maffeis C, Micciolo R, Must A, Zaffanello M, Pinelli L.Parental and perinatal factors associated with childhoodobesity in north-east Italy. Int J Obes 1994; 18: 301 ± 305.

58 Rice T, Bouchard C, Perusse L, Rao DC. Familial clusteringof multiple measures of adiposity and fat distribution in theQuebec Family Study: a trivariate analysis of percent bodyfat, body mass index, and trunk-to-extremity skinfold ratio.Int J Obes 1995; 19: 902 ± 908.

59 Tienboon P, Rutishauser IH, Wahlqvist ML. A family studyof coronary risk factors in Geelong. Aus J Public Health1992; 16: 20 ± 25.

60 Vuille JC, Mellbin T. Obesity in 10-year-olds: an epidemio-logic study. Pediatrics 1979; 64: 564 ± 572.

61 Wilkinson PW, Parkin JM, Pearlson J, Philips PR, Sykes P.Obesity in childhood: a community study in Newcastle uponTyne. Lancet 1977; 1: 350 ± 352.

62 Weinberg R, Webber LS, Berenson GS. Hereditary andenvironmental in¯uences on cardiovascular risk factors forchildren: the Bogalusa Heart Study. Am J Epidemiol 1982;116: 385 ± 393.

63 Prescott-Clarke P, Primatesta P (eds) Health survey forEngland: the health of young people '95 ± '97. The Station-ary Of®ce: London, 1998.

64 Allison DB, Heshka S, Neale MC, Heyms®eld SB. Raceeffects in the genetics of adolescents' Body Mass Index. Int JObes 1994; 18: 363 ± 368.

65 Brook CG, Huntley RM, Slack J. In¯uence of heredity andenvironment in determination of skinfold thickness in chil-dren. Br Med J 1975; 2: 719 ± 721.

66 Sorensen TI, Holst C, Stunkard AJ. Childhood body massindexÐgenetic and familial environmental in¯uencesassessed in a longitudinal adoption study. Int J Obes 1992;16: 705 ± 714.

67 Vogler GP, Sorensen TI, Stunkard AJ, Srinivasan MR, RaoDC. In¯uences of genes and shared family environment onadult body mass index assessed in an adoption study by acomprehensive path model. Int J Obes 1995; 19: 40 ± 45.


S35

68 Stunkard AJ, Sorensen TI, Hanis C, Teasdale TW, Chakra-borty R, Schull WJ, Schulsinger F. An adoption study ofhuman obesity. New Engl J Med 1986; 314: 193 ± 198.

69 Sobal J, Stunkard AJ. Socioeconomic status and obesity: areview of the literature. Psychol Bull 1989; 105: 260 ± 275.

70 White A, Nicolaas G, Foster K, Browne F, Carey S (eds)Health survey for England 1991. The Stationary Of®ce:London, 1993.

71 Prescott-Clarke P, Primatesta P (eds) Health survey forEngland '96. The Stationary Of®ce: London, 1998.

72 Duran-Tauleria E, Rona RJ, Chinn S. Factors associated withweight for height and skinfold thickness in British children.J Epidemiol Community Health 1995; 49: 466 ± 473.

73 Agras WS, Kraemer HC, Berkowitz RI, Hammer LD. In¯u-ence of early feeding style on adiposity at 6 years of age.J Pediatr 1990; 116: 805 ± 809.

74 Kramer MS, Barr RG, Pless IB, Bolisjoyly C, McVey-WhiteL, Leduc DG. Determinants of weight and adiposity in earlychildhood. Can J Public Health 1986; 77 (Suppl 1): 98 ± 103.

75 Arnesen E, Forsdahl A. The Tromso heart study: coronaryrisk factors and their association with living conditionsduring childhood. J Epidemiol Community Health 1985;39: 210 ± 214.

76 Blane D, Hart CL, Smith GD, Gillis CR, Hole DJ, HawthorneVM. Association of cardiovascular disease risk factors withsocioeconomic position during childhood and during adult-hood. Br Med J 1996; 313: 1434 ± 1438.

77 Braddon FE, Rodgers B, Wadsworth ME, Davies JM. Onsetof obesity in a 36 year birth cohort study. Br Med J 1986;293: 299 ± 303.

78 Hardy R, Wadsworth M, Kuh D. Association between child-hood obesity and social class and adult obesity across the lifecourse in a British national cohort. (Abstract). J EpidemiolCommunity Health 1998; 52: 693.

79 Charney E, Goodman HC, McBride M, Lyon B, Pratt R.Childhood antecedents of adult obesity. Do chubby infantsbecome obese adults? New Engl J Med 1976; 295: 6 ± 9.

80 Garn SM, Sullivan TV, Hawthorne VM. The juvenile-onset,adolescent-onset and adult-onset obese. Int J Obes 1991; 15:105 ± 110.

81 Goldblatt PB, Moore ME, Stunkard AJ. Social factors inobesity. JAMA 1965; 192: 97 ± 101.

82 Lissau-Lund-Sorensen I, Sorensen TI. Prospective study ofthe in¯uence of social factors in childhood on risk of over-weight in young adulthood. Int J Obes 1992; 16: 169 ± 175.

83 Lissau I, Sorensen TI. School dif®culties in childhood andrisk of overweight and obesity in young adulthood: a ten yearprospective population study. Int J Obes 1993; 17: 169 ± 175.

84 Lissau I, Sorensen TI. Parental neglect during childhood andincreased risk of obesity in young adulthood. Lancet 1994;343: 324 ± 327.

85 Rasmussen F, Johansson M. The relation of weight, lengthand ponderal index at birth to body mass index and over-weight among 18-year-old males in Sweden. Eur J Epidemiol1998; 14: 373 ± 380.

86 Ravelli GP, Belmont L. Obesity in nineteen-year-old men:family size and birth order associations. Am J Epidemiol1979; 109: 66 ± 70.

87 Teasdale TW, Sorensen TI, Stunkard AJ. Genetic andearly environmental components in sociodemographic in¯u-ences on adult body fatness. Br Med J 1990; 300: 1615 ±1618.

88 Wannamethee SG, Whincup PH, Shaper G, Walker M.In¯uence of fathers' social class on cardiovascular diseasein middle-aged men. Lancet 1996; 348: 1259 ± 1263.

89 Garn SM, Hopkins PJ, Ryan AS. Differential fatness gainof low income boys and girls. Am J Clin Nutr 1981; 34:1465 ± 1468.

90 Lasker GW, Mascie-Taylor CG. Effects of social classdifferences and social mobility on growth in height, weightand body mass index in a British cohort. Ann Hum Biol 1989;16: 1 ± 8.

91 Power C, Moynihan C. Social class and changes in weight-for-height between childhood and early adulthood. Int J Obes1988; 12: 445 ± 453.

92 Power C, Hertzman C, Matthews S, Manor O. Social differ-ences in health: life-cycle effects between ages 23 and 33 inthe 1958 British birth cohort. Am J Public Health 1997; 87:1499 ± 1503.

93 Power C, Matthews S. Origins of health inequalities in anational population sample. Lancet 1997; 350: 1584 ± 1589.

94 Agras WS, Kraemer HC, Berkowitz RI, Korner AF, HammerLD. Does a vigorous feeding style in¯uence early develop-ment of adiposity? J Pediatr 1987; 110: 799 ± 804.

95 Berkowitz RI, Agras WS, Korner AF, Kraemer HC, ZeanahCH. Physical activity and adiposity: a longitudinal studyfrom birth to childhood. J Pediatr 1985; 106: 734 ± 738.

96 Kramer MS, Barr RG, Leduc DG, Boisjoly C, McVey-WhiteL, Pless IB. Determinants of weight and adiposity in the ®rstyear of life. J Pediatr 1985; 106: 10 ± 14.

97 Kramer MS, Barr RG, Leduc DG, Boisjoly C, Pless IB.Infant determinants of childhood weight and adiposity.J Pediatr 1985; 107: 104 ± 107.

98 Lindgren G. Height, weight and menarche in Swedish urbanschool children in relation to socio-economic and regionalfactors. Ann Hum Biol 1976; 3: 501 ± 528.

99 Seidman DS, Laor A, Gale R, Stevenson DK, Danon YL. Alongitudinal study of birth weight and being overweight inlate adolescence. Am J Dis Child 1991; 145: 782 ± 785.

100 de Spiegelaere M, Dramaix M, Hennart P. The in¯uence ofsocioeconomic status on the incidence and evolution ofobesity during early adolescence. Int J Obes 1998; 22:268 ± 274.

101 Weststrate JA, Van Klaveren H, Deurenberg P. Changes inskinfold thicknesses and body mass index in 171 children,initially 1 to 5 years of age: a 51

2-year follow-up study. Int J

Obes 1986; 10: 313 ± 321.102 Zack PM, Harlan WR, Leaverton PE, Cornoni-Huntley J. A

longitudinal study of body fatness in childhood and adoles-cence. J Pediatr 1979; 95: 126 ± 130.

103 Boulton TJ, Cockington RA, Hamilton-Craig I, MagareyAM, Mazumdar J. A pro®le of heart disease risk factorsand their relation to parents' education, fathers' occupationand family history of heart disease in 843 South Australianfamilies: the Adelaide Children's WHO Collaborative Study.J Paediatr Child Health 1995; 31: 200 ± 206.

104 Hunter SM, Frerichs RR, Webber LS, Berenson GS. Socialstatus and cardiovascular disease risk factor variables inchildren: the Bogalusa Heart Study. J Chronic Dis 1979;32: 441 ± 449.

105 Kimm SY, Obarzanek E, Barton BA, Aston CE, Similo SL,Morrison JA, Sabry ZI, Schreiber GB, McMahon RP. Race,socioeconomic status, and obesity in 9- to 10-year-old girls:the NHLBI Growth and Health Study. Ann Epidemiol 1996;6: 266 ± 275.

106 Patterson ML, Stern S, Crawford PB, McMahon RP, SimiloSL, Schreiber, GB, Morrison JA, Waclawiw MA. Socio-demographic factors and obesity in preadolescent black andwhite girls: NHLBI's Growth and Health Study. J Natl MedAssoc 1997; 89: 594 ± 600.

107 Khoury PR, Morrison JA, Laskarzewski P, Kelly K, MelliesMJ, King P, Larsen R, Glueck CJ. Relationships of educationand occupation to coronary heart disease risk factors inschoolchildren and adults: The Princeton School DistrictStudy. Am J Epidemiol 1981; 113: 378 ± 395.

108 Persson LA. Dietary habits and health risks in Swedishchildren. Hum Nutr Clin Nutr 1984; 38: 287 ± 297.

109 Greenlund KJ, Liu K, Dyer AR, Kiefe CI, Burke GL, YunisC. Body mass index in young adults: associations withparental body size and education in the CARDIA Study.Am J Public Health 1996; 86: 480 ± 485.

110 Webber LS, Wattigney WA, Srinivasan SR, Berenson GS.Obesity studies in Bogalusa. Am J Med Sci 1995; 310 (Suppl1): S53 ± 61.


S36

111 Sargent JD, Blanch¯ower DG. Obesity and stature in adoles-cence and earnings in young adulthood. Analysis of aBritish birth cohort. Arch Pediatr Adolesc Med 1994; 148:681 ± 687.

112 Feldman W, Feldman E, Goodman JT. Culture versus biol-ogy: children's attitudes toward thinness and fatness. Pedia-trics 1988; 81: 190 ± 194.

113 Dornbusch SM, Carlsmith JM, Duncan PD, Gross RT, MartinJA, Ritter, PL, Siegel-Gorelick B. Sexual maturation, socialclass, and the desire to be thin among adolescent females.J Dev Behav Pediatr 1984; 5: 308 ± 314.

114 Rosenbaum M. The changing body image of the adolescentgirl. In: Sugar M (ed). Female adolescent development.Brunner=Mazel: New York, 1979, pp 235 ± 252.

115 Drever, F, Whitehead, M. (eds) Health Inequalities. Decen-nial supplement, Series DS no. 15. The Stationary Of®ce:London, 1997.

116 Sobal J. Group dieting, the stigma of obesity, and over-weight adolescents: contributions of Natalie Allon tothe sociology of obesity. Marriage Family Rev 1984; 7:9 ± 20.

117 Chamberlain R, Chamberlain G, Howlett B, Claireaux A.British Births 1970, Vol 1: The First Week of Life. Heine-mann Medical: London, 1975.

118 Nafstad P, Jaakkola JJ, Hagen JA, Pedersen BS, Qvigstad E,Botten G, Kongerud J. Weight gain during the ®rst year oflife in relation to maternal smoking and breast feedingin Norway. J Epidemiol Community Health 1997; 51:261 ± 265.

119 Metzger BE, Silverman BL, Freinkel N, Dooley SL, OgataES, Green OC. Amniotic ¯uid insulin concentrationas a predictor of obesity. Arch Dis Child 1990; 65:1050 ± 1052.

120 Ravelli GP, Stein ZA, Susser MW. Obesity in young menafter famine exposure in utero and early infancy. New Engl JMed 1976; 295: 349 ± 353.

121 Alberman E. Are our babies becoming bigger? J R Soc Med1991; 84: 257 ± 260.

122 Allison DB, Paultre F, Heyms®eld SB, Pi Sunyer FX. Is theintra-uterine period really a critical period for the develop-ment of adiposity? Int J Obes 1995; 19: 397 ± 402.

123 Curhan GC, Chertow GM, Willett WC, Spiegelman D,Colditz GA, Manson, JE, Speizer FE, Stampfer MJ. Birthweight and adult hypertension and obesity in women. Cir-culation 1996; 94: 1310 ± 1315.

124 Hulman S, Kushner H, Katz S, Falkner B. Can cardiovascularrisk be predicted by newborn, childhood, and adolescentbody size? An examination of longitudinal data in urbanAfrican Americans. J Pediatr 1998; 132: 90 ± 97.

125 Sorensen HT, Sabroe S, Rothman KJ, Gillman M, Fischer P,Sorensen TI. Relation between weight and length at birth andbody mass index in young adulthood: cohort study. Br Med J1997; 315: 1137.

126 Barker M, Robinson S, Osmond C, Barker DJ. Birth weightand body fat distribution in adolescent girls. Arch Dis Child1997; 77: 381 ± 383.

127 Fomon SJ, Rogers RR, Ziegler EE, Nelson SE, Thomas LN.Indices of fatness and serum cholesterol at age eight years inrelation to feeding and growth during early infancy. PediatrRes 1984; 18: 1233 ± 1238.

128 Harland PSE, Watson MJ, Ashworth L. The effect of meta-bolic programming on atherosclerosis and obesity risk fac-tors in UK adolescents living in poor socioeconomic areas.Ann NY Acad Sci 1997; 817: 361 ± 364.

129 Taylor SJ, Whincup PH, Cook DG, Papacosta O, WalkerM. Size at birth and blood pressure: cross sectional studyin 8 ± 11 year old children. Br Med J 1997; 314: 475 ±480.

130 Zive MM, McKay H, Frank-Spohrer GC, Broyles SL, NelsonJA, Nader PR. Infant-feeding practices and adiposity in 4-y-old Anglo- and Mexican-Americans. Am J Clin Nutr 1992;55: 1104 ± 1108.

131 Lissau I, Breum L, Sorensen TI. Maternal attitude to sweeteating habits and risk of overweight in offspring: a ten-yearprospective population study. Int J Obes 1993; 17: 125 ± 129.

132 Vik T, Jacobsen G, Vatten L, Bakketeig LS. Pre- and post-natal growth in children of women who smoked in preg-nancy. Early Hum Dev 1996; 45: 245 ± 255.

133 Bayley N. Size and body build of adolescents in relation torate of skeletal maturity. Child Dev 1943; 14: 47 ± 90. .

134 Bayley N. Skeletal maturing in adolescence as basis fordetermining percentage of completed growth. Child Dev1943; 14: 1 ± 46.

135 Eveleth PB, Tanner JM. Worldwide Variation in HumanGrowth, Cambridge University Press: Cambridge. 1991.

136 Brundtland GH, Liestol K, Walloe L. Height, weight andmenarcheal age of Oslo schoolchildren during the last 60years. Ann Hum Biol 1980; 7: 307 ± 322.

137 Jenicek M, Demirjian A. Age at menarche in French Cana-dian urban girls. Ann Hum Biol 1974; 1: 339 ± 346.

138 Frisch RE. The right weight: body fat, menarche and fertility.Proc Nutr Soc 1994; 53: 113 ± 129.

139 Merzenich H, Boeing H, Wahrendorf J. Dietary fat and sportsactivity as determinants for age at menarche. Am J Epidemiol1993; 138: 217 ± 224.

140 Freeman JV, Power C, Rodgers B. Weight-for-height indicesof adiposity: relationships with height in childhood and earlyadult life. Int J Epidemiol 1995; 24: 970 ± 976.

141 Stark O, Peckham CS, Moynihan C. Weight and age atmenarche. Arch Dis Child 1989; 64: 383 ± 387.

142 Gasser T, Kneip A, Ziegler P, Molinari L, Prader A, LargoRH. Development and outcome of indices of obesity innormal children. Ann Hum Biol 1994; 21: 275 ± 286.

143 Gasser T, Ziegler P, Largo RH, Molinari L, Prader A. Alongitudinal study of lean and fat areas at the arm. Ann HumBiol 1994; 21: 303 ± 314.

144 Rolland-Cachera MF, Deheeger M, Guilloud-Bataille M,Avons P, Patois, E, Sempe M. Tracking the developmentof adiposity from one month of age to adulthood. Ann HumBiol 1987; 14: 219 ± 229.

145 Rolland-Cachera MF, Deheeger M, Bellisle F, Sempe M,Guilloud-Bataille, M, Patois E. Adiposity rebound in chil-dren: a simple indicator for predicting obesity. Am J ClinNutr 1984; 39: 129 ± 135.

146 Van Lenthe FJ, Kemper CG, Van Mechelen W. Rapidmaturation in adolescence results in greater obesity in adult-hood: the Amsterdam Growth and Health Study. Am J ClinNutr 1996; 64: 18 ± 24.

147 Post GB, Kemper HC. Nutrient intake and biological matura-tion during adolescence. The Amsterdam growth and healthlongitudinal study. Eur J Clin Nutr 1993; 47: 400 ± 408.

148 Guo SS, Chumlea WC, Roche AF, Siervogel RM. Age- andmaturity-related changes in body composition during adoles-cence into adulthood: the Fels Longitudinal Study. Int J Obes1997; 21: 1167 ± 1175.

149 Siervogel RM, Roche AF, Guo SM, Mukherjee D, ChumleaWC. Patterns of change in weight=stature2 from 2 to 18years: ®ndings from long-term serial data for children inthe Fels longitudinal growth study. Int J Obes 1991; 15:479 ± 485.

150 Garn SM, LaVelle M, Rosenberg KR, Hawthorne VM.Maturational timing as a factor in female fatness and obesity.Am J Clin Nutr 1986; 43: 879 ± 883.

151 Wellens R, Malina RM, Roche AF, Chumlea WC, Guo S,Siervogel RM. Body size and fatness in young dultsin relation to age at menarche. Am J Hum Biol 1992; 4:783 ± 787.

152 Knishkowy BN, Palti H, Adler B, Go®n R. A follow-upstudy of adiposity and growth of Jerusalem schoolchildrenfrom age 6 to 14 years. J Adolesc Health Care 1989; 10:192 ± 199.

153 Hediger ML, Scholl TO, Schall JI, Cronk CE. One-yearchanges in weight and fatness in girls during late adoles-cence. Pediatrics 1995; 96: 253 ± 258.


S37

154 Burke GL, Savage PJ, Manolio TA, Sprafka JM, Wagen-knecht LE, Sidney S, Perkins LL, Liu K, Jacobs DR Jr.Correlates of obesity in young black and white women: theCARDIA Study. Am J Public Health 1992; 82: 1621 ± 1625.

155 Sherman B, Wallace R, Bean J, Schlabaugh L. Relationshipof body weight to menarcheal and menopausal age: implica-tions for breast cancer risk. J Clin Endocrinol Metab 1981;52: 488 ± 493.

156 Prokopec M, Bellisle F. Adiposity in Czech children fol-lowed from 1 month of age to adulthood: analysis ofindividual BMI patterns. Ann Hum Biol 1993; 20: 517 ± 525.

157 St. George IM, Williams S, Silva PA. Body size and themenarche: the Dunedin Study. J Adolesc Health 1994; 15:573 ± 576.

158 Beunen G, Malina RM, Lefevre J, Claessens AL, Renson R,Simons J, Maes H, Vanreusel B, Lysens R. Size, fatness andrelative fat distribution of males of contrasting maturitystatus during adolescence and as adults. Int J Obes 1994;18: 670 ± 678.

159 Miller FJ, Billewicz WZ, Thomson AM. Growth from birthto adult life of 442 Newcastle upon Tyne children. Br JPrevent Soc Med 1972; 26: 224 ± 230.

160 Garn SM, Haskell JA. Fat thickness and developmental statusin childhood and adolescence. A M A J Dis Children 1960;99: 746 ± 751.

161 Ferro-Luzzi A, Martino L. Obesity and physical activity. In:Chadwick DJ, Cardew G (eds). The origins and conse-quences of obesity. J Wiley: Chichester, 1996, pp 207 ± 227.

162 Allied Dunbar Health and Fitness Survey. A report onactivity patterns and ®tness levels. Sports Council andHealth Education Authority: London, 1992.

163 DiGuiseppi C, Roberts I, Li L. In¯uence of changing travelpatterns on child death rates from injury: trend analysis. BrMed J 1997; 314: 710 ± 713. [Published erratum appears inBr Med J 1997 314:1385.]

164 Pullinger, J. (ed) Social trends 28. The Stationary Of®ce:London, 1998.

165 Young and active? Draft policy framework for young peopleand health enhancing physical activity. Health EducationAuthority: London, 1997.

166 Wareham NJ, Rennie KL. The assessment of physical activ-ity in individuals and populations: why try to be more preciseabout how physical activity is assessed? Int J Obes 1998; 22:S30 ± S38.

167 Moore LL, Nguyen US, Rothman KJ, Cupples LA,Ellison RC. Preschool physical activity level and changein body fatness in young children. The FraminghamChildren's Study. Am J Epidemiol 1995; 142: 982 ±988.

168 Davies PSW, Day JME, Lucas A. Energy expenditure inearly infancy and later body fatness. Int J Obes 1991; 15:727 ± 731.

169 Davies PSW, Connolly C, Day JME. Energy expenditure ininfancy and later body composition. (Abstract.) Int J Obes1993; 17: 35.

170 Ku LC, Shapiro LR, Crawford PB, Huenemann RL. Bodycomposition and physical activity in 8-year-old children. AmJ Clin Nutr 1981; 34: 2770 ± 2775.

171 Wells JC, Stanley M, Laidlaw AS, Day JM, Davies PS. Therelationship between components of infant energy expendi-ture and childhood body fatness. Int J Obes 1996; 20:848 ± 853.

172 Twisk JW, van Mechelen W, Kemper HC, Post GB. Therelation between ``long-term exposure'' to lifestyle duringyouth and young adulthood and risk factors for cardio-vascular disease at adult age. J Adolesc Health 1997; 20:309 ± 319.

173 Twisk JWR, Kemper HCG, van Mechelen W, Post GB.Which lifestyle parameters discriminate high- from low-risk participants for coronary heart disease risk factors.Longitudinal analysis covering adolescence and young adult-hood. J Cardiovasc Risk 1997; 4: 393 ± 400.

174 van Lenthe FJ, van Mechelen W, Kemper HCG, Post GB.Behavioral variables and development of a central pattern ofbody fat from adolescence into adulthood in normal-weightwhites: The Amsterdam Growth and Health Study. Am J ClinNutr 1998; 67: 846 ± 852.

175 Beunen GP, Malina RM, Renson R, Simons J, Ostyn M,Lefevre J. Physical activity and growth, maturation andperformance: a longitudinal study. Med Sci Sport Exerc1992; 24: 576 ± 585.

176 Dietz WH Jr, Gortmaker SL. Do we fatten our children at thetelevision set? Obesity and television viewing in children andadolescents. Pediatrics 1985; 75: 807 ± 812.

177 Parizkova J. Longitudinal study of the development of bodycomposition and body build in boys of various physicalactivity. Hum Biol 1968; 40: 212 ± 225.

178 Robinson TN, Hammer LD, Killen JD, Kraemer HC, WilsonDM, Hayward C, Taylor CB. Does television viewingincrease obesity and reduce physical activity? Cross-sectional and longitudinal analyses among adolescent girls.Pediatrics 1993; 91: 273 ± 280.

179 Shapiro LR, Crawford PB, Clark MJ, Pearson DL, Raz J,Huenemann RL. Obesity prognosis: a longitudinal study ofchildren from the age of 6 months to 9 years. Am J PublicHealth 1984; 74: 968 ± 972.

180 Kemper HC, Post GB, Twisk JW, van Mechelen W. Lifestyleand obesity in adolescence and young adulthood: resultsfrom the Amsterdam Growth And Health LongitudinalStudy (AGAHLS). Int J Obes 1999; 23 (Suppl 3): S34 ± S40.

181 Armstrong N, Balding J, Gentle P, Kirby B. Patterns ofphysical activity among 11 to 16 year old British children. BrMed J 1990; 301: 203 ± 205.

182 Riddoch C, Savage JM, Murphy N, Cran GW, BorehamC. Long term health implications of ®tness andphysical activity patterns. Arch Dis Child 1991; 66: 1426 ±1433.

183 Heartbeat Wales. Technical Report Exercise for Health,Report no. 17. Health Promotion Authority for Wales:Cardiff, 1989.

184 School-aged sport in Scotland, Report no. 6. Scottish SportsCouncil: Edinburgh, 1989.

185 Kuh DJL, Wadsworth MEJ. Physical health status at 36 yearsin a British national birth cohort. Soc Sci Med 1993; 37:905 ± 916.

186 Pavlou KN, Krey S, Steffee WP. Exercise as an adjunct toweight loss and maintenance in moderately obese subjects.Am J Clin Nutr 1989; 49: 1115 ± 1123.

187 Epstein LH, Wing RR, Koeske R, Valoski A. Effects of dietplus exercise on weight change in parents and children.J Consult Clin Psychol 1984; 52: 429 ± 437.

188 Epstein LH, Wing RR, Penner BC, Kress MJ. Effect of dietand controlled exercise on weight loss in obese children.J Pediatr 1985; 107: 358 ± 361.

189 Mellbin T, Vuille JC. The relative importance of rapidweight gain in infancy as a precursor of childhood obesity.Pediatr Adolesc Endocr 1983; 1: 78 ± 83.

190 Epstein LH, Valoski A, Wing RR, McCurley J. Ten-yearoutcomes of behavioral family-based treatment for childhoodobesity. Health Psychol 1994; 13: 373 ± 383.

191 Epstein LH, Valoski AM,Vara LS,McCurley J,Wisniewski L,Kalarchian, MA, Klein KR, Shrager LR. Effects of decreasingsedentary behavior and increasing activity on weight change inobese children. Health Psychol 1995; 14: 109 ± 115.

192 Mo-Suwan L, Pongprapai S, Junjana C, Puetpaiboon A.Effects of a controlled trial of a school-based exerciseprogram on the obesity indexes of preschool children. AmJ Clin Nutr 1998; 68: 1006 ± 1011.

193 Sallis JF, McKenzie TL, Alcaraz JE, Kolody B, Hovell ME,Nader PR. Project SPARK. Effects of physical education onadiposity in children. Ann NY Acad Sci 1993; 699: 127 ± 136.

194 Barlow CE, Kohl HW, Gibbons LW, Blair SN. Physical®tness, mortality and obesity. Int J Obes 1995; 19 (Suppl 4):S41 ± S44.


S38

195 Blair SN, Kohl HW III, Paffenbarger RS Jr, Clark DG,Cooper KH, Gibbons LW. Physical ®tness and all-causemortality. A prospective study of healthy men and women.JAMA 1989; 262: 2395 ± 2401.

196 Household food consumption and expenditure 1990: with astudy of trends over the period 1940 ± 1990. Annual report ofthe National Food Survey Committee=National Food SurveyCommittee. Ministry of Agriculture, Fisheries and Food.HMSO: London, 1991.

197 Poppitt SD, Prentice AM. Energy density and its role in thecontrol of food intake: evidence from metabolic and com-munity studies. Appetite 1996; 26: 153 ± 174.

198 Prentice AM. Manipulation of dietary fat and energy densityand subsequent effects on substrate ¯ux and food intake. AmJ Clin Nutr 1998; 67: 535S ± 541S.

199 Acheson KJ, Flatt JP, Jequier E. Glycogen synthesis versuslipogenesis after a 500 gram carbohydrate meal in man.Metabolism 1982; 31: 1234 ± 1240.

200 Schwarz J, Neese R, Dare D, Turner S, Hellerstein M. Effectof carbohydrate overfeeding and underfeeding on de novolipogenesis and insulin levels. (Abstract.) Proc Nutr Soc1995; 54(1): 47A.

201 Present-day practice in infant feeding. Report of a workingparty on the panel on child nutrition. Report on Health andSocial Subjects no. 9. The Stationary Of®ce: London, 1974.

202 Birkbeck JA, Buck®eld PM, Silva PA. Lack of long-termeffect of the method of infant feeding on growth. Hum NutrClin Nutr 1985; 39C: 39 ± 44.

203 Dewey KG, Heinig MJ, Nommsen LA, Peerson JM, Lonner-dal B. Breast-fed infants are leaner than formula-fed infantsat 1 y of age: the DARLING study. Am J Clin Nutr 1993; 57:140 ± 145.

204 Heinig MJ, Nommsen LA, Peerson JM, Lonnerdal B, DeweyKG. Intake and growth of breast-fed and formula-fed infantsin relation to the timing of introduction of complementaryfoods: the DARLING study. Davis Area Research on Lacta-tion, Infant Nutrition and Growth. Acta Paediatr 1993; 82:999 ± 1006.

205 Dine MS, Gartside PS, Glueck CJ, Rheines L, Greene G,Khoury P. Where do the heaviest children come from? Aprospective study of white children from birth to 5 years ofage. Pediatrics 1979; 63: 1 ± 7.

206 Poskitt EM. Overfeeding and overweight in infancy and theirrelation to body size in early childhood. Nutr Metabolism1977; 21 (Suppl 1): 54 ± 55.

207 Sveger T. Does overnutrition or obesity during the ®rst yearaffect weight at age four? Acta Paediatr 1978; 67: 465 ± 467.

208 Vobecky JS, Vobecky J, Shapcott D, Demers PP. Nutrientintake patterns and nutritional status with regard to relativeweight in early infancy. Am J Clin Nutr 1983; 38: 730 ± 738.

209 Wells JC, Stanley M, Laidlaw AS, Day JM, Davies PS.Energy intake in early infancy and childhood fatness. Int JObes 1998; 22: 387 ± 392.

210 Wilson AC, Forsyth JS, Greene SA, Irvine L, Hau C, HowiePW. Relation of infant diet to childhood health: seven yearfollow up of cohort of children in Dundee infant feedingstudy. Br Med J 1998; 316: 21 ± 25.

211 Marmot MG, Page CM, Atkins E, Douglas JW. Effect ofbreast-feeding on plasma cholesterol and weight in youngadults. J Epidemiol Community Health 1980; 34: 164 ± 167.

212 Foster K, Lader D, Cheesbrough S. (eds) Infant feeding 1995.The Stationary Of®ce: London, 1995.

213 Post GB, Kemper HC, Twisk J, van Mechelen W. Theassociation between dietary patterns and cardio vasculardisease risk indicators in healthy youngsters: results covering®fteen years of longitudinal development. Eur J Clin Nutr1997; 51: 387 ± 393.

214 Deheeger M, Akrout M, Bellisle F, Rossignol C, Rolland-Cachera MF. Individual patterns of food intake developmentin children: a 10 months to 8 years of age follow-up studyof nutrition and growth. Physiol Behav 1996; 59:403-407.

215 Roland-Cachera MF, Deheeger M, Akrout M, Bellisle F.In¯uence of macronutrients on adiposity development: afollow up study of nutrition and growth from 10 months to8 years of age. Int J Obes 1995; 19: 573 ± 578.

216 Grif®ths M, Payne PR, Stunkard AJ, Rivers JP, Cox M.Metabolic rate and physical development in children at riskof obesity. Lancet 1990; 336: 76 ± 78.

217 Nicklas TA, Farris RP, Smoak CG, Frank GC, Srinivasan SR,Webber LS, Berenson GS. Dietary factors relate to cardio-vascular risk factors in early life. Bogalusa Heart Study.Arteriosclerosis 1988; 8: 193 ± 199.

218 Shea S, Basch CE, Stein AD, Contento IR, Irigoyen M,Zybert P. Is there a relationship between dietary fat andstature or growth in children three to ®ve years of age?Pediatrics 1993; 92: 579 ± 586.

219 Bingham SA. The dietary assessment of individuals; meth-ods, accuracy, new techniques and recommendations. NutrAbstr Rev (Ser A) 1987; 57: 705 ± 742.

220 Taitz LS. Infantile overnutrition among arti®cially fed infantsin the Shef®eld region. Br Med J 1971; 1: 315 ± 316.

221 Whitehead RG, Paul AA. Diet and growth in healthy infants.Hong Kong J Paediatr 1988; 5: 1 ± 20.

222 Dietary reference values for food energy and nutrients for theUnited Kingdom. Report of the Panel on Dietary ReferenceValues of the Committee on Medical Aspects of Food Policy.Report on Health and Social Subjects no. 41: London, 1991.

223 Energy and protein requirements. Report of a joint FAO=WHO=UNU Expert Consultation. World Health OrganizationTechnical Report Series no. 724. World Health Organization:Geneva, 1985.

224 Durnin JV, Womersley J. Body fat assessed from total bodydensity and its estimation from skinfold thickness: measure-ments on 481 men and women aged from 16 to 72 years. Br JNutr 1974; 32: 77 ± 97.

225 Stephen AM, Sieber GM. Trends in individual fat consump-tion in the UK 1900 ± 1985. Br J Nutr 1994; 71: 775 ± 788.

226 National food survey 1997: annual report on food expendi-ture, consumption and nutrient intakes=National FoodSurvey Committee. Ministry of Agriculture, Fisheries andFood. The Stationary Of®ce: London, 1998.

227 Gregory J, Foster K, Tyler H, Wiseman M (eds). The dietaryand nutritional survey of British adults. HMSO: London,1990.

228 Gregory J, Collins DL, Davies PSW, Hughes JM, Clarke PC(eds). National diet and nutrition survey: children aged 11

2to

412years. HMSO: London, 1995.

229 The diets of British schoolchildren. Sub-committee on Nutri-tional Surveillance. Committee on Medical Aspects of FoodPolicy. Report on Health and Social Subjects no. 36: London,1989.

230 Lissner L, Heitmann BL. Dietary fat and obesity: evidencefrom epidemiology. Eur J Clin Nutr 1995; 49: 79 ± 90.

231 Seidell JC. Dietary fat and obesity: an epidemiologic per-spective. Am J Clin Nutr 1998; 67: 546S ± 550S.

232 Bolton Smith C, Woodward M. Dietary composition and fatto sugar ratios in relation to obesity. Int J Obes 1994; 18:820 ± 828.

233 Gibney MJ. Dietary guidelines: a critical appraisal. J HumNutr Diet 1990; 3: 245 ± 254.

234 Paeratakul S, Popkin BM, Kohlmeier L, Hertz-Picciotto I,Guo X, Edwards LJ. Measurement error in dietary data:implications for the epidemiologic study of the diet ± diseaserelationship. Eur J Clin Nutr 1998; 52: 722 ± 727.

235 Woo J, Ho SC, Mak YT, Law LK, Cheung A. Nutritionalstatus of elderly patients during recovery from chest infectionand the role of nutritional supplementation assessed by aprospective randomized single-blind trial. Age Ageing 1994;23: 40 ± 48.

236 Appleby PN, Thorogood M, Mann JI, Key TJ. Low bodymass index in non-meat eaters: the possible roles ofanimal fat, dietary ®bre and alcohol. Int J Obes 1998; 22:454 ± 460.


S39

237 Barr SI, Prior JC, Janelle KC, Lentle BC. Spinal bonemineral density in premenopausal vegetarian and nonvege-tarian women: cross-sectional and prospective comparisons.J Am Diet Assoc 1998; 98: 760 ± 765.

238 Parsons TJ, van Dusseldorp M, van der Vliet M,van de Werken K, Schaafsma G, van Staveren WA.Reduced bone mass in Dutch adolescents fed a macrobioticdiet in early life. J Bone Miner Res 1997; 12: 1486 ±1494.

239 Sabate J, Lindsted KD, Harris RD, Johnston PK. Anthropo-metric parameters of schoolchildren with different life-styles.Am J Dis Child 1990; 144: 1159 ± 1163.

240 Dagnelie PC, van Dusseldorp M, van Staveren WA, HautvastJGAJ. Effects of macrobiotic diets on linear growth in infantsand children until 10 years of age. Eur J Clin Nutr 1994; 48:S103 ± S112.

241 Sanders TAB, Manning J. The growth and development ofvegan children. J Hum Nutr Diet 1992; 5: 11 ± 21.

242 Janelle KC, Barr SI. Nutrient intakes and eating behaviorscores of vegetarian and nonvegetarian women. J Am DietAssoc 1995; 95: 180 ± 186.

243 Slattery ML, Jacobs DR Jr, Hilner JE, Caan BJ, van Horn L,Bragg C, Manolio TA, Kushi LH, Liu KA. Meat consump-tion and its associations with other diet and health factors inyoung adults: the CARDIA study. Am J Clin Nutr 1991; 54:930 ± 935. [Published erratum appears in Am J Clin Nutr1992; 55(1): iv.]

244 Shultz TD, Leklem JE. Dietary status of Seventh-day Adven-tists and nonvegetarians. J Am Diet Assoc 1983; 83: 27 ± 33.

245 Darke SJ, Disselduff MM, Try GP. A nutrition survey ofchildren from one-parent families in Newcastle upon Tyne in1970. Br J Nutr 1980; 44: 237 ± 241.

246 Wells JC, Stanley M, Laidlaw AS, Day JM, Stafford M,Davies PS. Investigation of the relationship between infanttemperament and later body composition. Int J Obes 1997;21: 400 ± 406.

247 Barr RG, Quek VS, Cousineau D, Oberlander TF, Brian JA,Young SN. Effects of intra-oral sucrose on crying, mouthingand hand-mouth contact in newborn and six-week-oldinfants. Dev Med Child Neurol 1994; 36: 608 ± 618.

248 Kaplan HI, Kaplan HS. The psychosomatic concept ofobesity. J Nerv Ment Dis 1957; 125: 181 ± 201.

249 French SA, Story M, Perry CL. Self-esteem and obesity inchildren and adolescents: a literature review. Obes Res 1995;3: 479 ± 490.

250 Bjorntorp P. Visceral fat accumulation: the missing linkbetween psychosocial factors and cardiovascular disease?J Intern Med 1991; 230: 195 ± 201.

251 Marcus MD. Binge eating and obesity. In: Brownell KD,Fairburn CG (eds). Eating disorders and obesity. A compre-hensive handbook. The Guilford Press: New York, 1995.

252 Brownell KD. Behavioral, psychological, and environmentalpredictors of obesity and success at weight reduction. Int JObes 1984; 8: 543 ± 550.

253 van Lenthe FJ, Snel J, Twisk JWR, van Mechelen W,Kemper HCG. Coping, personality and the development ofa central pattern of body fat from youth into young adult-hood: the Amsterdam Growth and Health Study. Int J Obes1998; 22: 861 ± 868.

254 Carey WB, Hegvik RL, McDevitt SC. Temperamentalfactors associated with rapid weight gain and obesityin middle childhood. J Dev Behav Pediatr 1988; 9: 194 ±198.

255 French SA, Perry CL, Leon GR, Fulkerson JA. Self-esteemand change in body mass index over 3 years in a cohort ofadolescents. Obes Res 1996; 4: 27 ± 33.

256 Klesges RC, Haddock CK, Stein RJ, Klesges LM,Eck LH, Hanson CL. Relationship between psychosocialfunctioning and body fat in preschool children: a longitudinalinvestigation. J Consult Clin Psychol 1992; 60: 793 ±796.

257 Morris RW, Rona RJ, Murray M, Green V, Shah D. Thesmoking and dietary behaviour of Lambeth schoolchildren II.The relationship between knowledge, attitudes and beha-viour. Public Health 1984; 98: 225 ± 232.

258 Rothbart MK. Measurement of temperament in infancy.Child Dev 1981; 52: 569 ± 578.

259 Levine JA, Eberhardt NL, Jensen MD. Role of nonexerciseactivity thermogenesis in resistance to fat agin in humans.Science 1999; 283: 212 ± 214.

260 Stunkard AJ, Penick SB. Behavior modi®cation in the treat-ment of obesity. The problem of maintaining weight loss.Arch Gen Psychiat 1979; 36: 801 ± 806.

261 Brownell KD, Kelman JH, Stunkard AJ. Treatment of obesechildren with and without their mothers: changes in weightand blood pressure. Pediatrics 1983; 71: 515 ± 523.

262 Epstein LH, Wing RR, Penner BC, Kress MJ. Effects offamily based behavioural treatment on obese 5 to 8 year oldchildren. Behav Ther 1985; 16: 205 ± 212.

263 Mellin LM, Slinkard LA, Irwin CE Jr. Adolescent obesityintervention: validation of the SHAPEDOWN program. J AmDiet Assoc 1987; 87: 333 ± 338.

264 Saelens BE, Epstein LH. Behavioral engineering of activitychoice in obese children. Int J Obes 1998; 22: 275 ± 277.

265 Birch LL, Fisher JO. Development of eating behaviorsamong children and adolescents. Pediatrics 1998; 101(Suppl): 539 ± 549.

266 Cullen KJ, Cullen AM. Long-term follow-up of the Bussel-ton six-year controlled trial of prevention of children'sbehavior disorders. J Pediatr 1996; 129: 136 ± 139.

267 O'Brien RW, Smith SA, Bush PJ, Peleg E. Obesity, self-esteem, and health locus of control in black youths duringtransition to adolescence. Am J Health Promotion 1990; 5:133 ± 139.

268 Dietz WH. Health consequences of obesity in youth: child-hood predictors of adult disease. Pediatrics 1998; 101(Suppl): 518 ± 525.

269 Dickersin K. How important is publication bias? A synthesisof available data. AIDS Educ Prevent 1997; 9: 15 ± 21.

270 Egger M, Davey Smith G, Schneider M, Minder C. Bias inmeta-analysis detected by a simple, graphical test. Br Med J1997; 315: 629 ± 634.

271 WHO Expert Committee. Physical status: the use and inter-pretation of anthropometry. WHO Technical Report Seriesno. 854. WHO: Geneva, 1995.

272 Dietz WH, Robinson TN. Use of the body mass index (BMI)as a measure of overweight in children and adolescents.J Pediatr 1998; 132: 191 ± 193.

273 Wadsworth ME, Kuh DJ. Childhood in¯uences on adulthealth: a review of recent work from the British 1946national birth cohort study, the MRC National Survey ofHealth and Development. Paediatr Perinat Epidemiol 1997;11: 2 ± 20.

274 DiPietro L. Physical activity, body weight and adiposity: anepidemiologic perspective. Exerc Sport Sci Rev 1995; 23:275 ± 303.

275 Guillaume M, Lapidus L, Lambert A. Obesity and nutritionin children. The Belgian Luxembourg Child Study IV. Eur JClin Nutr 1998; 52: 323 ± 328.

276 McCance RA. Overnutrition and undernutrition. II. Effects.Lancet 1953; i: 740 ± 745.

277 Frisch RE. Fatness and fertility. Sci Am 1988; 258: 88 ± 95.278 Deheeger M, Rolland Cachera MF, Fontvieille AM. Physical

activity and body composition in 10 year old French chil-dren: linkages with nutritional intake? Int J Obes 1997; 21:372 ± 379.

279 Lewis MK, Hill AJ. Food advertising on British children'stelevision: a content and experimental study with nine-yearolds. Int J Obes 1998; 22: 206 ± 214.

280 Goldberg ME, Gorn GJ, Gibson W. TV messages for snackand breakfast foods; do they in¯uence children's prefer-ences? J Consumer Res 1978; 5: 73 ± 81.


S40

281 Woodward DR, Cumming FT, Ball PJ, Williams HM,Hornsby H, Boom JA. Does television affect teenagers'food choices? J Hum Nutr Diet 1997; 10: 229 ± 235.

282 Heitmann BL, Kaprio J, Harris JR, Rissanen A, Korkeila M,Koskenvuo M. Are genetic determinants of weight gainmodi®ed by leisure-time physical activity? A prospectivestudy of Finnish twins. Am J Clin Nutr 1997; 66: 672 ± 678.

283 Heitmann BL, Lissner L, Sorensen TI, Bengtsson C. Dietaryfat intake and weight gain in women genetically predisposedfor obesity. Am J Clin Nutr 1995; 61: 1213 ± 1217.

Appendix I. Tables of includedstudies

Abbreviations used in tablesD change inANOVA analysis of varianceBMI body mass index (weight=height2) kg=m2

CI con®dence interval, 95% unless stated otherwiseDZ dizygoticf femalef-up follow-uphgt heightm malemax maximumMETs metabolic equivalentsmin minimummo. monthMZ monozygoticOR odds ratioref. referenceRR relative riskTBW total body watervs versuswgt weightwks weeksy year

The number superscripted after the ®rst author of eachpaper included in the following tables refers to thereference list in this review. The reference numbersgiven as part of the description of the papers includedin the following tables refer to the reference list inthose papers.


S41

Table

A1

GeneticÐ

longitudinalstudies

Author=journal=

year=country

Agesatoutcome,

measurements=no.ofsubjects

Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Childhoodto

adulthood

Kaplowitzetal,41

1988,UK

LondonLongitudinalGroup

Every

3mo.in

1sty;1.5y;

annuallyuntil20y?,

n�224,�50%

female

Prospectivelongtitudinal.

Mothers

selectedin

approxim

ately

random

sample

inwestLondon

duringpregnancy

Meanoflogbiceps,

triceps,subscapular,

suprailiacskinfolds

Parentalmeanoflogbiceps,


suprailiacskinfolds.Allmothers

and57fathers

measured

Mother±childcorrelations

(<0.1

to�0.4)greater

thanfather±child

correlations(<0.1

to�0.25).Mother±child

correlationstendedto

increasewithtime,with

theexceptionofa

decreasein

late

adole-

scence.Father±child

correlationstendedto

remain

stable

or

decreaseovertime

Lakeetal,421997,UK

UK

1958birth

cohort

0,7y,11,16,23and33y;

Duration�33y;n�17,378;

baseline,n�11,407at33y

Prospectivelongitudinal.

Participants

members

of1958birth

cohort,all

births3±9March1958.

Data

from

immigrants

subsequentlyincluded

BMI

ParentalobesitybasedonBMI,

self-reportedwhenchild11y.

Parents

classi®edasunderw

eight,

norm

alweight,overw

eight,obese.

Obese:BMI>85th

percentile.

Ingeneral,child's

BMIat

allagesincreasedwith

increasingparental

obesity.Formalesand

females,oddsratioof

beingobeseat33y

increasedwithincreasing

parentalfatness.OR�8.4

forboth

parents

obese

comparedto

both

norm

alweight.Tracking

ofobesity(correlation

betw

eenBMIat7yand

at33y)wasstrongerif

both

parents

obesethan

ifboth

norm

alweight.

Males:both

parents

obese,r�0.46,both

norm

alr�0.25;Females:

both

parents

obese,

r�0.54,both

norm

al

r�0.32

TableA1Continued

TablesA1±A4:GeneticÐ

inheritanceofphenotypepapers


S42

Table

A1

Continued

GeneticÐ

longitudinalstudies

Author=journal=

year=country

Agesatoutcome,


Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Whitakeretal,43

1997,USA

1±2y;3±5y;6±9y;10±14y;

15±17y;21±29y;duration

�24.5y(m

ean).

Targetn�1333,n�854,

64%

female

Retrospectivelongitudinal.

Participants

long-term

members

ofGroup

HealthCooperative

inW

ashingtonState.

Includedifborn

1965±1970,with

measuredweightat

�21y,heightat

�18y(m

ales)orat

�16y(females).

Data

extracted

from

medicalrecords.

94%

non-H

ispanic

Whites

ObesitybasedonBMI

expressedasZ-score.

ChildhoodÐ

obese:BMI

�85th

percentile;very

obese:BMI>95th

percentile

(basedon

NHANESI�

IIdata).

Adulthood(21±29y)

obesity:menBMI

�27.8;womenBMI

�27.3

Parentalobesity:menBMI�27.8;

womenBMI�27.3

Basedonsingle

adultheightand

multiple

weightmeasurements

usedto

estimate

parentBMIat

mid-pointofeachchildage

interval.794pairsparents

(some

with

>1child)measured.747

mother'srecords,699father's

records

Calculationofoddsratios

showedriskofobesity

atallagessigni®cantly

greaterifeitherparent

obese.OR

forobese

mother2.8±3.6,for

obesefather2.4±2.9.

Prevalenceofobesitymuch

greaterin

childrenwith

�1obeseparentatall

ages.Multivariate

logistic

regressionusinggener-

alizedestimatingequa-

tions,withBMIweighted

fortimeintervaland

incorporatingadjustm

ent

forcorrelationsbetw

een

siblingdata

withchild-

hoodandparental

obesityincludedas

variablesofinterest:

before

3y,parental

obesity,childhood

obesitynotapredictor;

over3y,both

parental

obesityandchildhood

obesitypredictors,>9y

parentobesityless

importantthanchild

obesity.Parentalobesity

especiallyim

portantup

to6y


S43

Table

A2

GeneticÐ

cross-sectionalandsiblingstudies

Author=journal=year=country

Agesatoutcome

measurements=

no.ofsubjects

Participants

Outcomemeasure

Parent=siblingfatness

measure

Mainfinding

Health

Survey

forEngland63

1995±1997,UK

2±15y,

n�6114

(father±

childassociation);n�8163

(mother±childassociation)

Nationallyrepresentative

sample

selectedbyrandom

samplingtechnique

BMI

BMI

Proportionofchildrenin

top

BMIquintile

increaseswith

mother'sorfather's

increasingBMIin

boysand

girls

Ayatollahi,491992,Iran

6±12y,n�1207

(642m,565f)

Participants

representative

sample

ofschoolchildrenin

Shiraz(seerefs.7and8)

Obesityindex:

weight=height2.5

Obesityindex:

weight=height1

Correlation:father'sobesity

relatedto

child's

obesity,

r�0.15,P<0.00001.

Mother'sobesityrelatedto

child's

obesity,r�0.17,

P<0.00001.Father'sobesity

relatedto

son's

obesity,

r�0.199,P<0.001.Mother's

obesityrelatedto

daughter's

obesity,r�0.195,P<0.001

Burns

etal,501989,USA

MuscatineFamilyStudy

�15.5y,n�284families

(fourgroupsindividuals

selected,n�70random,

n�72lean,n�70weight

gain,n�72heavy)

ParticipantsÐ

seeref.11

Relativeweight

(weight=medianweightfor

appropriate

ageandsex

group�100)grouped:lean,

random,leangain,heavy

gain,heavy

BMI

ANOVA:afteradjustingfor

age,BMIofmothers,fathers,

siblingsandaunts=uncles


fatnessgroupofoffspring

(lean,random,weightgain,

heavy).

Sim

ilar

results

for

skinfolds,data

notshown

Moll

etal,511991,USA

MuscatineStudy

12±24y,numbers

approxim

ately

asabove

Asabove

Asabove

Asabove

From

abstract,%

variationin

BMIexplainedby:35%

explainedbysingle

recessive

locus;42%

bypolygenic

loci;

23%

adjustedvariationNOT

explainedbygeneticfactors;

spousesÐ

12%

byshared

environment;siblingsÐ

10%

bysharedenvironment

Garn

etal,521981,USA

Tecumseh

n�11,015,individuals

infourmemberfamilies

ParticipantsÐ

seerefs.5±7

Tricepsskinfold

Parentandsiblingtriceps

skinfold

Probabilityofindividual

beingobeserises

dramaticallywhenallthree

remainingfamilymembers

are

obese,comparedto

all

lean

TableA2Continued


S44

Table

A2

Continued

GeneticÐ



Agesatoutcome


Participants

Outcomemeasure


measure

Mainfinding

Goranetal,441998,USA

5.2y,annualto

8.3y,duration3y,n�75

baseline,n�53at8.3y

Participants

recruitedby

newspaperadvertisements

andword

ofmouth

Rate

ofchangein

fatmass

adjustedforfat-freemass

(fatandfatfreemass

measuredbyDXA)

Parentalobesity:BMI>

85th

percentile

forageandsex

ANOVA;rate

ofchangein

fat

mass

adjusted

for

fat-free

mass

related

toparental

obesity,P<0.001.Rate

of

changein

fatmassadjusted

forfat-freemasshigherin

childrenwithtw

oobese

parents

thanoneornoobese

parents

inboysandgirls

Grinkeretal,531989,USA

TenState

Study

1±12y

ParticipantsÐ

nodetail

given

Tricepsskinfold

Obesityofmothers

based

ontricepsskinfold:lean,

medium

orobese

Childrenofobesemothers

hadgreatertricepsskinfolds

from

3yonwards

Grinkeretal,531989,USA

MichiganStudy

0±4y

ParticipantsÐ

seeref.6

Weight-for-height

Obesityofmothers

based

onweight=height.Mothers

selectedforobesityor

norm

alweightcriteria

Infants

ofobesemothers

were

ofhigher

weight=heightthanthoseof

norm

alweightmothers

at1±

18mo.butnotthereafter

Guillaumeetal,541995,

Belgium

6±13y,70%

eligible

children

participated.n�1028(527m,

501f)

Participants

selectedat

random

from

schoolclasses.

PopulationW

allonianorigin,

little

admixture

by

immigration.Rural

BMI

Tricepsskinfold

ParentalBMIbasedon

reportedweightandheight

Children's

BMIwasrelatedto

parents

BMI(statistical

methodunclear,no

summary

statisticsgiven).In

boys,afteradjustingforage,

tricepsskinfold

relatedto

BMIquintile

offathers,

P�0.038

Hashim

oto

etal,551995,

Japan

1±6y,n�3187(1716m,

1471f)

Volunteers

for`Preventionof

ChildhoodObesityin

Niigata

City'project,from

municipal

schools

Obesity:>15%

overw

eight

forage,heightandsex

usingidealweights

from

Welfare

MinistryofJapan,

1990

ParentalBMIÐ

reported.

Obesity:BMI>95th

percentile.BMI>27.4

fathers;BMI>26.0

mothers

UsingChi-squaredtest,

incidenceofobesityin

childrenwithobesefathers

greaterthanincidencein

thosewithnon-obese

fathers,11.5%

vs6.2%,

P<0.01.Incidenceofobesity

inchildrenwithobese

mothers

greaterthanwith

non-obesemothers,14.5%

vs6.2%,P<0.01.Ifsexes

separated,nosigni®cant

differencein

boysorgirls

obesevsnon-obesefathers.

Boysandgirls

withobese

mothers

havehigher

incidenceobesity.Overall

obesityincidencefortw

oparents

obese�30.8%,for

oneparentobese�22.7%,

twoparents

non-obese5.9%

TableA2Continued


S45

Table

A2

Continued

GeneticÐ



Agesatoutcome


Participants

Outcomemeasure


measure

Mainfinding

Khoury

etal,561983,USA

PrincetonSchoolDistrict

Study

Twogroups,<20yand

�20y,totaln�177

ParticipantsÐ

random

recall

subsetoflargerstudy

BMI

BMI

Father±son,mother±son,

father±daughter,mother±

daughterBMIcorrelationsall

signi®cantforthose

<20y

notforthose

�20y.Sibling

correlationsalsosigni®cant

Klesgesetal,451995,USA

4.4,5.4,6.4y,

duration2y,n�203

baseline,n�146at6.4y

Participants

recruited

throughpaediatricians,

day-care

centres,churches.

Obeseover-sampled

DBMI

Numberofparents

overw

eight(BMI>

75th

percentile

forage)

Multiple

regressionanalysis

(dependentvariable�DBMI),

includingadjustm

entfor

baselineBMI,gender,age,

lifestyle

factors

(dietand

activity).Afteradjustm

ent,

boyswithboth

parents

overw

eightshowedaverage

increasein

BMIof0.67kg=m

2

comparedwithadecreaseof

0.31kg=m

2in

boyswithboth

parents

norm

alweight.In

girls,thosewithboth

parents

norm

alweightoroverw

eight,

BMIdecreasedby

0.23kg=m

2,butincreasedby

0.4kg=m

2in

thosewhose

fatherwasoverw

eight

Maffeis

etal,571994,Italy

4,8,10,12y,n�1523total,

n�1363this

paper(676m,

687f)

Retrospective.Participants

arandom

sample

from

four

schoolclassesin

six

areas

ofnorth-eastItaly.This

paperexcludeschildrenwith

low

gestationalageand

includessubjects

forwhom

complete

data

wasavailable

Obesity:BMI>95th

percentile

forageand

gender,usingFrenchtables

asreference(seeref.13)

ParentalBMIfrom

reported

weightandheight

Oddsratioscalculatedfor

likelihoodofobesity,parental

BMIcategoriesunspeci®ed.

Afteradjustingforage,

increasingpaternaland

maternalBMIincreased

likelihoodofobesity,in

boys

OR�1.14,P<0.001and

OR�1.11,P<0.001,andin

girls

OR�1.19,P<0.001and

OR�1.16,P<0.001

respectively

TableA2Continued


S46

Table

A2

Continued

GeneticÐ



Agesatoutcome


Participants

Outcomemeasure


measure

Mainfinding

Maffeis

etal,461998,Italy

8.6y;12.6y;duration4y,

n�298baseline,n�112at

12.6y

Participants

recruited

throughschools

BMI;obesity:relativeBMI

>120%

(Rolland-Cachera

BMItablesasreference)

ParentalBMI(unclear

whethermeasuredor

reported)

Threedifferenttypesof

multivariate

analysis,with

age,gender,dietary

variables,parentalBMI,TV

viewingtimeandvigorous

activityincludedas

covariates.1,Mother'sand

father'sBMIsigni®cantly

positively

relatedto

child's

relativeBMIat12y,b�2.5,

P<0.0001,b�2.1,P<0.0001

respectively.2,In

logistic

regressionmodelbasedon

whetherchildincreasedor

decreasedBMIoverstudy

period,parentalBMInot

relatedto

changein

child

BMI.3,Ifchangein

childBMI

expressedrelativeto

average

changein

BMI,no

relationship

with

parental

BMI.

O'Callaghanetal,471997,

Australia

MUSP(M

aterUniversity

StudyofPregnancy)

Baseline0y;follow-up5y;

n�7357baseline,n�4062

at5y(2133f,1929m)

Mothers

(n�8556)enrolled

atpregnancyhospitalvisit

(ref.20).Atbirth

n�7357,not

allofthesechildrenfollowed

updueto

insuf®cientfunds

ObesityÐ

severe

obesity:

BMI>94th

percentile;

moderate

obesity:BMI85±

94th

percentile

within

cohort

Maternalpre-pregnantBMI

basedonreportedweight,

measuredheight.Paternal

BMIbasedonmaternal

reportsofheightandweight

Univariate

analyses:RR,95%

CIs,Chi-squaredtestfor

signi®cance.RRfor

moderate

andsevere

obesity

at5yincreaseswithmaternal

orpaternalobesity.

Multivariate

analyses

includingmaternalBMI,

maternaleducation,

maternalincomeand

paternalBMI,adjustedORs

calculatedwith95%

CIs.OR

formoderate

andsevere

obesityat5yincreaseswith

maternalorpaternalobesity.

Riceetal,581995,Canada

QuebecfamilyStudy

15y,n�903,parents

n�727

Familiesrecruitedvia

media

1978±1981to

studygenetic

effects

onseveral

physiologicaland

biochemicaltraits.

BMI;%

bodyfatby

underw

aterweighing

BMI;%

bodyfatby

underw

aterweighing

Cross-traitcorrelations

suggestingfamilial

determ

inants

signi®cantonly

forBMI.

TableA2Continued


S47

Table

A2

Continued

GeneticÐ



Agesatoutcome


Participants

Outcomemeasure


measure

Mainfinding

Rosenbaum

etal,481987,

USA

BogalusaHeart

Study

0y;1,2,3,4y;7y;duration

7y,n�440baseline,

n�250at7y,analyseson

n�69±180

Alllivebirthsto

mothers

residingin

ward

ofBogalusa,

Louisianafrom

1=1=74to

30=6=75eligible.Of447,

sevenrefused.Population

mixed,BlackandW

hite

BMI;subscapularskinfold

ParentalBMI;parental

subscapularskinfold,both

measuredwhenchildaged

2y.192mothers,99fathers

measured

Univariate

regression

analyses.Mother'sand

father'sBMIpositively

relatedto

child's

BMIatall

ages,P<0.05.Regression

coef®cientforfather±child

BMIincreasedfrom

0.15at

6mo.to

0.29at7y,andfor

mother±childBMIfrom

0.11

at6mo.to

0.19at7y.

Father±childsubscapular

skinfold

coef®cient

decreasedfrom

0.18at6mo.

to0.09at7y,andmother±

childcoef®cientincreased

from

70.1

at6mo.to

0.16

at7y

Tienboonetal,591992,

Australia

14.9y,n�126families

Schoolchildrenin

Geelong

district.Allfamilieswithchild

born

Jul±

Dec1972,atschool

in1987invited.1215families

invited,low

responserate

justi®edadequate

forwithin

familycorrelations.

BMI;skinfolds:biceps,


suprailiac,thigh

ParentalBMI;skinfolds(as

forchildren);obesity:BMI

>30

Obesityin

oneortw

oparents

associatedwithincreaseof

oneunitBMIin

adolescents

(notsigni®cant).Correlations:

fathers

vschildren,r�0.21,

P<0.05;mothers

vschildren

r�0.04,ns

Vuille

andMellbin,601979,

Sweden

10y,n�345

Participants

selectedfrom

largeroriginalsampleÐ

seeref.18

Relativeweight;sum

triceps�subscapular

skinfolds

ParentalBMIcalculatedfrom

reportedweightandheights,

atorafterbirth

ofchild

Mother'sBMIpositively

relatedto

BMIin

boysand

girls,andsum

skinfoldsin

girls

(contributedatleast

0.015to

R2in

multivariate

regressionmodel).No

contributionofparent

fatnessto

sum

skinfoldsin

boys

WadaandUeda391990,

Japan

14±15y,19±20y,

duration5y,baseline

n�522;at19±20y

n�400(198m,212f)

Noinform

ationon

participantrecruitment

BMI

ParentBMI,basedon

reportedweightandheight,

seemingly

atbaseline

BMIat19±20ysigni®cantly

correlatedto

parentalBMI

inmalesandfemales,

mother±childslightlyhigher

thanfather±child

correlations.At19±20yin

malesandfemales,BMI

signi®cantlyhigherifoneor

both

parents

obesethanif

neitherparentobese,

P<0.05.Prevalenceof

obesity,at19±20y

increasedwithnumberof

parents

obese.Multiple

regression:afteradjusting

TableA2Continued


S48

Table

A2

Continued

GeneticÐ



Agesatoutcome


Participants

Outcomemeasure


measure

Mainfinding

WadaandUeda39

(continued)

forBMIat14±15y,in

boys,

mother'sandfather'sBMI

positivleyrelatedto

BMIat

19±20y,b�0.12,P�0.04,

b�0.06,P�0.04

respectively.In

girls,

mother'sandfather'sBMI

positively

relatedto

BMIat

19±20y,b�0.07,P�0.04,

b�0.07,P�0.04,respectively

Weinberg

etal,62

1982,USA

BogalusaHeart

Study

5±17y,n�1856

(1740fullsiblings,

116half-siblings)

Participants

of

BogalusaHeart

study

screenedduringthe

1973±4and1976±7

schoolyears

±seerefs

1,2,13.

Tricepsskinfold

Fullsiblingsvs

halfsiblings

Intraclasscorrelation

coef®cients

andheritability

analysis.Intraclass

correlationcoef®cients

for

tricepsskinfold

greaterin

full

siblingsthanhalfsiblings.

Heritabilityestimate

not

signi®cant,sharedcommon

environmentstatistic

signi®cantin

Whites,not

Blacks.

Wilkinsonetal,61

1977,UK

NewcastleStudy

ofChildDevelopment

(NSCD)

0,6,12mo.,5,10,

12y;at12-y

subgroupof

n�60obese,n�60

controls

studied;

n�8000in

NSCD

at10y

Casesandcontrols

identi®ed

at10yfrom

alongtitudinal

cohort

study.Casesand

controls

born

inoneofthe

threeyears

ofrecruitment

(1961)providedmore

detailedinform

ation

(n�120)

Obesecases(w

gtfor

hgt>97th

percentile

using

Scott's

tables)orcontrols

(25±75th

percentile)and

matchedforage,sexand

school

Parents

tricepsskinfold

(of

subgroupofchildren

n�120)measuredwhen

child12y.Obesity±

mother:triceps>20mm;

father:triceps>10mm

Mothers

andfathers

ofobese

childrenhadsigni®cantly

greatertricepsskinfolds.39%

siblingsofobesegrouphad

weight-for-height>90th

percentile,only

10%

siblings

incontrolgroupw2�23.2,

P>0.001

TableA2Continued


S49

Table

A3

GeneticÐ

twin

studies


Agesatoutcome,


Participants

Outcomemeasure

Riskfactor

Mainfinding

Allisonetal,641994,USA

12±18y;n�238pairs;39

pairsMZmales,31pairsDZ

males,69pairsMZfemales,

51pairsDZfemales,48pairs

DZm=f

ParticipantsÐ

noinform

ation

onrecruitment,seeref.24

BMI

Heritability,MZvsDZ

Correlationandvariance-

covariancematricesforBox-

Coxtransform

edBMIs

calculated.MZcorrelations

muchhigherthanDZ

correlationsandmuchhigher

thanoppositesexDZ

correlationssuggesting

substantialheritability.

Differencetendsto

begreater

infemales,suggesting

greaterheritabilityin

females.Nomeaningful

differencesin

heritabilityof

BMIbetw

eenBlacksand

Whites.Degreeto

which

geneticandenvironmental

factors

in¯uenceBMIvaries

byracebutnotgender.

Brooketal,651975,UK

3±15y;n�222pairs;38pairs

MZmales,67pairsDZmales,

40pairsMZfemales,44pairs

DZfemales

Participants

alltw

ins

registeredatGOSH

aged

3±15ylivingin

greater

London.Additional

volunteers

byappeal

afterTVprogramme

ontw

ins

Tricepsskinfold,

subscapularskinfold

Heritability,MZvsDZtw

ins

Correlationcoef®cients:

tricepsÐ

femalesMZ>DZ,

geneticfactors

important;

malesMZ�DZ,genetic

factors

notsoim

portant;

subscapularÐ

malesand

femalesMZ>DZ,genetic

factors

important.Heritability

estimates:boysÐ

<10y,

environmentim

portantfor

triceps,geneticfor

subscapular;

>10y,genetic

factors

importantfortriceps

andsubscapular;girlsÐ

<10y,environment

importantfor

triceps�

subscapular;

>10y

geneticfactors

importantfor



S50

Table

A4

GeneticÐ

adoptionstudies


Agesatoutcome


Participants

Outcomemeasure

Riskfactor

Mainfinding

Vogleretal,671995,

Denmark

Danishadoptionstudy

42y;n�660adoptees;225

bio

fathers,330bio

mothers,

202adoptfathers,309adopt

mothers,177bio

siblings,

321of568maternalhalfsibs

included,252of491paternal

halfsibsincluded

Alladoptionsin

Denmark

1924±1947in

whichneither

parentis

abiologicalrelative

oftheirchild.Inform

ationon

3691of5455adoptions

BMIbasedonself-reported

weightandheight

BMIofbiologicaland

adoptiveparents

when

participantatschoolÐ

reportedeitherbyparentor

offspring

Complexstatistical

analysisÐ

referto

appendix

ofpaper.Concludethat

familialresemblancein

BMI

isentirely

dueto

genetic

relationships.Sharedfamily

environmenthasno

in¯uence

Sorensenetal,661992,

Denmark

Danishadoptionstudy

7±13y,n�269

Asabove,subgroupselected

BMIbasedonweights

and

heights

from

schoolrecords

BMIofbiologicaland

adoptiveparents

when

participantatschoolÐ

reportedeitherbyparentor

offspring

Correlationsbetw

een

adopteesandadoptive

parents

weakerthanthose

withbiologicalparents.

Correlationsbetw

een

adoptiveparents

andfull

siblingsgreaterthanfor

adoptees

Stunkard

etal,681986

Danishadoptionstudy

42y,n�60thin,n�61

median,n�61

overw

eight,n�57obese

Asabove,subgroupselected

torepresentrangeofBMI

BMIbasedonself-reported

weightandheight

BMIofbiologicaland

adoptiveparentsÐ

assumed

current,notexplicit.

Overw

eight:BMI>25;

obeseBMI>30

Proportionofoverw

eight

plusobesebiological

mothers

andfathers


weightclassof

offspring,P�0.001,P�0.04,

respectively.Proportionof

overw

eightplusobese

adoptivemothers

no

relationship

withfatnessof

offspring(P

�0.98);

proportionadoptivefathers

decreasedwithincreasing

weightclassofoffspring,

P�0.04.Proportion

ofonly

obesebiologicaland

adoptivemothers

andfathers


fatnessofoffspring,only

signi®cantforbiological

mothers,P�0.03


S51

Table

A5

SocialÐ

longitudinalstudies


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcomemeasure

Riskfactor

Mainfinding

(i)Childhoodto

adulthood

ArnesenandForsdahl,75

1985,

Norw

ay

TromsoHeart

Study

20±54y;n�14,652

(7405m,7247f)

Retrospective.Population

notselected,municipal

birth

cohort

(referto

ref.

3),

69%

ofthoseinvited

BMI

Economic

conditionsin

familyduringchildhood,

reportedretrospectively.

Fourgroups:very

dif®cult,

dif®cult,good,very

good.

Usingequality

ofmeans,

lineartrendanddifference

betw

eenextremegroups,no

relationship

foundbetw

een

povertyin

childhoodandBMI

inadulthoodin

menor

women

Blaneetal,761996,UK

(Scotland)

35±64y,duration30±

60y;n�5645allmen


from

27work

places

selectedto

representall

levels

andtypesof

employment.Men

<35y

excluded

BMI

Father'smain

occupation,

socialmobility

Usingmultiple

regression

analyses,adjustingforage

andownsocialclass,BMI

signi®cantlyrelatedto

father'ssocialclass,

b�0.2,P<0.0001.Thosein

lowersocialclasshigherBMI.

Norelationship

betw

een

socialmobilityandBMI

Braddonetal,771986

1946UKbirth

cohort

0,7,11,14,20,26,36y;

duration36y;n�5362

baseline,n�3249at36y

Prospective.Participants

inbirth

cohort,allbirthsin

1week,March1946.Strati®ed

sample

only

includedin

this

paper.Pregnantwomen

excluded

Overw

eightÐ

men:

BMI�

25.0±29.9;women:

BMI�

24.3±29.1;obesityÐ

men:BMI>29.9;women:

BMI>29.1

SESbasedon`socialclass

andeducation'ofparents.

Fourgroups:upperand

lowernon-m

anualand

manual.Socialmobility:

changebetw

eenSESin

childhood(asabove)and

SESin

adulthood

(unspeci®ed)

At36yboth

menandwomen

from

non-m

anualclass

homeswere

signi®cantly

more

likely

tohavenorm

alor

underw

eightBMI(m

en

w2�30,P<0.001,women

w2�52,P<0.001)than

others,andthosefrom

manualhomesmore

likely

tobeoverw

eightorobese.In

multiple

regression

modelsÐ

men:after

adjustingforrelativeweight

at11y,SESin

childhoodand

educationwere

signi®cantly

relatedto

BMIat36yasin

aboveanalyses,P<0.05;

women:afteradjustingfor

relativeweightat11y,parity

andmaritalstatus,SESin

childhoodandeducation

were


BMIat36yasin

above

analyses,P<0.05.Social

mobility:womenwhomoved

upfrom

manualto

non-

manualSESshowed

signi®cantlylower

prevalenceofobesitythan

thosewhoremainedin

manualSES,4.7%

vs11.2%.

Noeffectin

men

TableA5Continued

TablesA5andA6:Socialfactorspapers


S52

Table

A5

Continued

SocialÐ

longitudinalstudies


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Hardyetal,781998,UK

1946UKbirth

cohort

0,7,11,14,20,26,36,43y;




inbirth

cohort,allbirthsin

1week,March1946

Obesity:BMI�30kg=m

2Childhoodsocialclass

Usinggeneralizedestimating

equations,childrenfrom

manualsocialclasshave

greaterriskofobesityat43y

thanthosefrom

non-m

anual,

OR�1.56(1.12,2.18)

Charneyetal,791976,

USA

20±30y;duration20±

30y,n�366


born

1945±1955,birth

weight>2270g,withheight

andweightrecordedseveral

timesin

childhood.90%

of

thosecontactedparticipated

Overw

eight:

>10%

above

medianweightforheight

andage.Obese:>20%

weightforheightandage

(NationalHealthSurvey

1960±1962usedas

referencedata)

Socialclassbasedon

educationandoccupation.

Ordinalpositionofchildin

family

Nocorrelationbetw

een

ordinalpositionofchildand

overw

eight.Prevalenceof

overw

eightwas18%

insocial

class1,17%

insocialclass2,

20%

insocialclass3,and

31%

inclasses4and5

Garn

etal,891981,USA

TecumsehCommunity

HealthStudy

8y,26y;duration

18y,n�1117(553m,564f);

drop-outnosnotgiven,

sim

ilarforhighandlow

incomegroups

Prospective.Subjects

were

thosemeasuredin

rounds1

and4ofTecumseh

CommunityHealthStudy

Subscapularskinfold,

Dsubscapularskinfold

Incomepercapitaseemsto

bemeasuredatbaseline

At26y(inGarn's

paper

table

1andtextdisagree

aboutlength

ofstudy),low

incomegrouphadgreater

skinfold

thanmedium

and

highincomegroupsin

boys

andgirls.In

girls,medium

groupwasalsogreaterthan

highincomegroup.No

mentionofstatisticaltest.

Changein

skinfold

thickness

wasgreatestin

low

income

groupandlowestin

high

incomegroupusingdata

from

allsubjects,signi®cant

usingt-tests

andtestfor

trend(levelnotspeci®ed)

TableA5Continued


S53

Table

A5

Continued

SocialÐ

longitudinalstudies


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Garn

etal,801991,

USA

TecumsehCommunity

HealthStudy

0±34y,5±39y,duration

5y;n�8648at5±39y,

drop-out<5%


of

TecumsehCommunity

HealthSurvey

Obesitybasedonsummed

tricepsandsubscapular

skinfoldsnorm

alisedZ-

scores.Obesity�Z-score

�1

Norm

alizedZ-score

ofyears

ofparentaleducationlikely

tobereportedatbaseline

althoughnotexplicitly

stated

Parents

ofchildrenwho

becameobeseby5±9ywere

slightlyaboveaverage

educationallevel,Z-score

�

0.11,notsigni®cantly

differentfrom

zero,but

signi®cantlydifferentfrom

othergroups.Parents

of

childrenwhobecameobese

by10±14ywere

slightly

below

averageeducational

level,Z�

70.15,not

signi®cantlydifferentfrom

zero.Parents

ofthosewho

becameobeseby15±19y

and20±39ywere

oflower

thanaverageeducationlevel,

Z�

70.21andZ�

70.16,

respectively,signi®cantly

differentfrom

zero

Goldblattetal,811965,

USA

MidtownManhattan

Study

20±59y;n�1,660(690m,

970f)

Unclearwhetherprospective

orretrospectiveÐ

see

refs.1±33.ParticipantsÐ

see

refs.1±3

Obesitybasedonself-

reportedheightandweight

SESbasedonfather's

occupationandeducation

whenparticipant8y.Social

mobility:changein

SES

betw

een8yandadulthood

basedonoccupation,

education,income,rent

SESoforigin

inversely

relatedto

obesityin

women

w2�66.5,P<0.001andin

menalthoughrelationship

muchweaker,statisticsnot

given.More

obesityin

downward

sociallymobile

thansociallystatic,andmore

obesityin

sociallystaticthan

inupward

sociallymobile;in

women

w2�20.5,

P<0.001,

inmennotsigni®cant

TableA5Continued


S54

Table

A5

Continued

SocialÐ

longitudinalstudies


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcomemeasure

Riskfactor

Mainfinding

LaskerandMascie-Taylor,90

1989,UK

1958UKbirth

cohort

(NationalChilddevelopment

study,NCDS)

0,7,11y;16y;duration

betw

eeneachfollow-up

usedin

analyses,f-

upn�18,559


inbirth

cohort,98%

births3±9

March1958.Data

from

ethnic

minoritiesexcludedforthis

paper

BMI,DBMI

SESbasedonfather's

occupation,reportedat

baselineandfollow-ups.

Socialmobility:changein

SESbetw

eenfollow-ups

Noin¯uenceofSES(assume

atage7y)onchangein

BMI

7±11y.Signi®canteffectof

SES.Thosein

lowestSES

grouphadgreatestincrease

inBMIat11±16y.DSES

(socialmobility)0(assumed)

Ð7ynotrelated

toBMIat7y.D

SES7±11y

notrelatedto

BMIat11y,D

SES11±16ynotrelatedto

BMIat16y.Nomentionof

statisticaltests

or

signi®cancelevels

PowerandMoynihan,911988,

UK

1958UKbirth

cohort

7y;23y;duration

16y;n�12,274(6133m,

6141f)


inbirth

cohort,98%

births3±9

March1958

Standard

deviationscore

(SDS)forBMIfrom

reported

heightandweightat23y

Father'soccupation,

reportedatbaseline

Atage23,signi®cantly

greaterpercentageofmen

andwomenfrom

manual

classeshadstandard

deviationscore

0.5±1.5

and

>1.5

comparedwithnon-

manualclasses,22vs16%

and8.2

vs4.6%

formen,18

vs14%

and9.2

vs4.2%

for

women.Betw

eenage7and

23,more

from

manual

classesmaintainedSDSof

>1.5

compared

tonon-

manualclasses,31vs18%

formen,39vs22%

for

women,P<0.01.More

from

manualclassesincreased

SDSto

0.5±1.5

orto

>1.5

TableA5Continued


S55

Table

A5

Continued

SocialÐ

longitudinalstudies


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Poweretal,921997,UK

1958UKbirth

cohort

0,7,11,16,23y;33y;

duration33y;n�17,414

baseline,n�8,459at33y

Asabove

Overw

eightandobese

basedonreportedBMI.

Overw

eightÐ

men:BMI�

25.0±30.0;

women:BMI�

23.8±28.6.

ObesityÐ

men:BMI>30.0;

women:BMI>28.6

Father'soccupationreported

atbaseline.Fourgroups

Resultspresentedasodds

ratios,representing

proportionalincreasein

odds

from

topto

bottom

ofsocial

hierarchy.In

menand

women,OR

ofbeing

overw

eightorobeseat23y

or33ysigni®cantlygreater

forlowestcomparedto

highestsocialgroup,

P<0.05.At33y,ORfor

overw

eightwas1.60in

men,

1.54in

womenandfor

obesity,2.19in

menand1.99

inwomen.Socialgroup

differencesin

overw

eightand

obesitylessenedbetw

een23

and33ybutnotsigni®cantly

PowerandMatthews,93

1997,UK

1958UKbirth

cohort

0,7,11,16,23y;33y;

duration33y;n�17,414

baseline,n�12,264at

7y;n�10,121at33y

Asabove

BMI(m

easuredat7y,

reportedat33y)

Father'soccupation

reported

atbaseline.Fourgroups

Inboth

malesandfemales,

nodifferencein

BMIacross

socialgroupsat7ybut

signi®cantincreasingBMI

withlowersocialclassat

33y,P<0.01,MantelHaenzel

w2testfortrend

Lissau-Lund-Sorensenand

Sorensen,821992,

Denmark


10y;n�1258baseline,

n�522at20±21y


arandom

sample,25%

of3rd

graders

inCopenhagen

Obesity:BMI>95th

percentile;overw

eight:BMI

>90th

percentile

from

reportedheightandweight,

malesandfemales

separately

Maternalandpaternal

education,occupationof

householder,childhood

rearingareas(goodvspoor),

reportedatbaseline

Bivariate

regression;those

withmotheroflow

education

more

likely

tobeoverw

eight

thanthosewithmotherof

middle

leveleducation,

OR�2.4,P<0.05.Those

rearedin

poorareamore

likely

tobeoverw

eightthan

thosein

goodarea,

OR�2.2,P<0.001.No

relationshipsforfather's

educationorhouseholder

occupation.Aftercontrolling

forBMIin

childhoodand

gender,only

effectofrearing

arearemained(O

R�2.4,

P<0.001).Resultssim

ilarif

thosealreadyoverw

eightin

childhoodexcluded

TableA5Continued


S56

Table

A5

Continued

SocialÐ

longitudinalstudies


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcomemeasure

Riskfactor

Mainfinding

LissauandSorensen,831993,

Denmark



n�756at20±21y

Asabove

Asabove

Scholarlydif®culties,

learningdif®culties,

scholasticpro®ciency,

specialeducation,reported

byform

teacheratbaseline

Usinglogisticregression

analysis

andadjustingfor

childhoodBMI,genderand

confounderscore

toaccount

forparents

education,

householder'soccupation

andresidentialarea,obesity

more

likely

iflearning

dif®cultiespresentin

childhood,OR�4.2,

P�0.0003,scholastic

pro®ciencybelow

average,

OR�2.8,P�0.006,received

specialeducation,OR�2.7,

P�0.007,orhadscholarly

dif®cultiesin

childhood,

OR�1.5,P�0.006.Odds

ratiosforbeingoverw

eight,

values1.5±2.0

andless

signi®cant.Resultssim

ilarif

separate

adjustm

entmade

forsocialfactors

ratherthan

usingconfounderscore.If

childrenalreadyoverw

eight

orobesein

childhood

excluded,resultssim

ilarbut

lesssigni®cant

LissauandSorensen,841994,

Denmark



n�756at20±21y

Asabove

Asabove

Familyfactors,no.of

siblings,parentalsupport,

child's

hygiene,atbaseline

Usinglogisticregression

analysis

andadjustingfor

childhoodBMI,age,gender

andconfounderscore

(as

above),familystructure

had

nosigni®canteffectonadult

obesity.Childrenreceiving

noparentalsupport

had

higherriskofadultobesity,

OR�7.1,P<0.0001than

thosewithharm

onius

support.Thosereceiving

overprotectivesupport

had

non-signi®cantlyincreased

riskofobesity,OR�2.3,

P�0.2.Dirty=neglected

childrenhadmuchgreater

riskofobesitythanaveragely

groomedchildren,OR�9.8,

P<0.0001.Risksfor

overw

eightasforobesity

TableA5Continued


S57

Table

A5

Continued

SocialÐ

longitudinalstudies


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Rasmussenand

Johansson,851998,

Sweden

18.2y,n�75,638

Retrospective.Single

male

birthsidenti®edfrom

birth

registry1973±1976.Linked

tofollow-updata

from

military

serviceregistry.78%

ofidenti®edbirthswere

term

births,andheightandweight

data

wasavailable

at18y.

Analysesrelevantto

this

review

completedon1975±

1976birthsonly,36%

of

identi®edbirths

Overw

eight:25�BMI<

30.

Severe

overw

eight:BMI�30

Mother'sage,mother's

education,mother'sparity,

mother'smaritalstatus,

livingarea

Multiple

logisticregression

analyses,oddsratiosand

95%

CIs

calculated.

Independentvariables

included:birth

weight,

mother'sage,mother's

education,mother'sparity,

mother'smaritalstatus,

livingarea.Comparedwith

livingin

alargecity,

likelihoodofbeing

overw

eightsigni®cantly

greaterlivingin

amedium-

sizedtown,countryareaor

ruralenvironment,P<0.05.

Comparedwithmothers

aged25±30y,mothers

<20ymore

likely

tohave

overw

eight(O

R�1.20,CI

1.10±1.32)offspringat18y.

Comparedwithmothers

of

higheducationlevel,mothers

oflow

educationlevelmore

likely

tohaveoffspring

overw

eightat18y,OR�1.72,

CI1.62±1.82.Likelihoodof

obesityincreasedwith

�4

childrenin

familycompared

withtw

ochildren,OR�1.22,

CI1.10±1.35,andincreasedif

only

oneparentin

family

comparedwithtw

o,

OR�1.11,CI1.06±1.17.

Resultsforsevere

overw

eightwere

sim

ilar

TableA5Continued


S58

Table

A5

Continued

SocialÐ

longitudinalstudies


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcomemeasure

Riskfactor

Mainfinding

RavelliandBelm

ont,861979,

TheNetherlands

19y,n�283,028,allmale

Retrospective.Dutchmen

examinedformilitary

induction,born

betw

een

1=1=44and31=12=47

Obesity:�120%

weight-for-

height.W

HO

data

usedas

reference

Father'soccupation(m

anual

ornon-m

anual),birth

order,

familysize

Overallprevalenceofobesity

higherin

manualclasses

(2.09%)thannon-m

anual

(1.54%).Obesitydeclined

withfamilysize,from

3.2%

inonly

childrento

1.5%

inchildrenwhowere

1of5,

w2�218,P<0.0005fortrend.

Prevalenceofobesityin

single

childrensigni®cantly

more

thanchildrenwhowere

1of2,3,4or5,w2�221,

P<0.0005.Only

inmanual

socialgroup,in

familiesof

fourchildren,obesityamong

®rst-born

signi®cantlyless

thanamonglast-born.

Prevalenceofobesitynot

consistentlyrelatedto

birth

order

Teasdale

etal,87

1990,Denmark

Copenhagenadoptions

3mo.(m

edian),42y,

duration

�42y,n�2015

(1155m,860f)

Prospective.Copenhagen

registerofnon-familial

adoptions,totaln�5455

LogBMIfrom

reported

weightandheight

Biologicalandadoptive

father'soccupationattime

ofadoption.Geographical

regionofadoption

Pearsoncorrelations:logBMI

negatively

relatedto

biologicalfather'ssocialclass

(r�

70.05,P<0.05),

adoptivefather'ssocialclass

(r�

70.09,P<0.05).Log

BMIto

geographicalregion

r�0.08,P<0.05,those

adoptedinto

provincial

regionhadhigherBMIthan

thoseinto

urban.Multiple

regression:logBMI�

age�

sex�geographical

region�

bio

father's

SES�adoptivefather's

SES�bio

father'sSES*age,

allSESvariablessigni®cant.

logBMI�

age�sex�

geographicalregion�bio

father'sSES�adoptive

father'sSES�bio

father's

SES*age�adoptee's

SES,all

SESvariablessigni®cant.In

both

models,relationship

betw

eenbiologicalfather's

SESandlogBMIis

inverse

andincreaseswithage

TableA5Continued


S59

Table

A5

Continued

SocialÐ

longitudinalstudies


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Wannametheeetal,881996,

UK

BritishRegionalHeart

Study

53±72y,n�5643

Retrospective.Menselected

from

age±sexregisters

of

generalpractices,onefor

eachof24townsin

theUK.

6528oforiginal7735men

aliveandcontacted,5934

replied,5643(73%

oforiginal

cohort)withcomplete

inform

ation

Obesityunde®nedin

paper

butcutoffusedwasÐ

obesity:BMI�80th

percentile

oforiginal7735

menaged40±59y,

equivalentto

�28kg=m

2

(personalcommunication)

SESbasedonfather's

longestheld

occupation,

manualornon-m

anual

t-Test:%

obesein

groupof

non-m

anualfathervsmanual

father.Greaterprevalenceof

obesityin

manualfather

group(20.1%)vsnon-m

anual

fathergroup(14.1%),

P<0.0001.Afteradjustm

ent

forageandownadultsocial

class,19.7%

obesein

manual

fathergroup,14.5%

innon-

manualgroup,P<0.0001

(ii)W

ithin

childhood

Agrasetal,731990,USA

2,4wks,12mo.;20mo.;2y,

3y;6

y;n�99baseline,

n�54at6y

Prospective.Children

recruitedfrom

hospitalin

®rst

weekoflife

Tricepsskinfold,logBMI

Age,maritalstatus,

educationandoccupationof

parents

Univariate

correlations:at6y,

educationofparents

negatively

relatedto

log

BMI,r�

70.31,P<0.05.

Multivariate

regression:after

adjustm

entforBMIatbirth,

severaldietary

factors

and

parentBMI:at6y,parent

educationsigni®cantly

negatively

relatedto

logBMI,

partialR2�0.13,P�0.02.No

associationbetw

eenparent

educationandchildBMIat

anyage

<6yorchildtriceps

skinfold

atanyage.No

analysesforotherriskfactors

Agrasetal,941987,

USA

2,4wks,12mo.;20mo.;2y;

duration2y;n�99

baseline,n�79at2y

Asabove


Parenteducation

Multivariate

regression:after

adjustm

entfortriceps

skinfold

at2wks,several

dietary

factors

andparent

triceps:parenteducation

negatively

relatedto

triceps

skinfold

at1y,P<0.02.No

in¯uenceat2yoronBMIat1

or2y

Berkowitzetal,951985,

USA

1±3days,6.5y;duration

�6.5y;n�112

baseline,n�52(25m,27f)

at6.5y

Prospective.Recruitment

detailsnotgiven,seeref.9

LogBMI.Tricepsskinfold

SESbasedonfather's

occupationandeducationas

reportedbymother

Nosigni®cantcorrelation

betw

eenSESandlogBMIor

tricepsskinfold

at6.5y

TableA5Continued


S60

Table

A5

Continued

SocialÐ

longitudinalstudies


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Krameretal,741986,

Canada

Baseline1±3days,further

measurements

at2,4,6wks,

2,3,4mo.Follow-upat1,2y.

Duration2y,n�462



from

largeuniversity

teachinghospitalserving

sociallydiverse

population.n�553eligible,

462enrolled

BMI,sum

triceps�

subscapular�suprailiac

skinfolds

SESbasedonmother's

educationandheadof

household's

occupation

Multiple

regressionmodels,

stepwiseprocedure,

constructedforBMIand

skinfoldsat1and2y.

Independentvariables

included:sex,birth

weight,

parentalrelativeweights,

socialclass,andbehaviour

anddietary

variables.No

in¯uenceofsocialclasson

BMIorskinfoldsat1or2y

Krameretal,961985,Canada

Asabove

Asabove

Asabove

Asabove

Asabove


Asabove

Asabove

Asabove

Asabove

Asabove

Lindgren,981976,

Sweden

SLU

Project

9y;semi-annuallyuntil17yf,

18ym;duration8±9y;


at17±18y


asample

ofrandomly

selected

children(seeLjung,1974)

Weight-for-height

Father'soccupationfamily

income,combinationofthe

abovetw

o

UsingANOVA,in

girls,those

from

manualworkers

tended

tohavehigherweight-for-

heightthanhighersocial

groupsatallages,only

signi®cantatage11±15y.

Sim

ilarrelationshipsin

boys

butweakerandnot

signi®cantatanyage.

Comparingextremegroups

(father'soccupationand

familyincomecombined)

greaterweight-for-heightin

boysoflowersocialgroup

notsigni®cant,butsigni®cant

ingirls

atages11,13and15,

P<0.05.Norelationship

betw

eenno.ofparents

or

mothers

workingoutof

home=in

homeandweight-

for-heightatanyage

TableA5Continued


S61

Table

A5

Continued

SocialÐ

longitudinalstudies


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Seidmanetal,991991,

Israel.

Jerusalem

PerinatalStudy

Baseline0y,f-up17y,

duration17y;baselinen�?;

f-upn�33,413

Prospective.Allbirthsfrom

threemajorwardsin

Jerusalem

1964±1971.At

17ythosenotin

defence

forcesnotincluded(61.7%

femalesmissing)

Overw

eight:BMI�90th

percentile

for17yoldsin

defenceforces,24.6

form

�f.

Severe

overw

eight:BMI

�97th

percentile,27.8

form,

27.7

forf

Paternaleducation(years

of

schooling)seemsto

be

assessedin

childhood=at

birth.Municipaltaxlevel

(areaofresidence)

Chisquare

test.Males:%

severe

overw

eightless

inlowestpaternaleducation

level,P<0.01.Females:%

overw

eightandsevere

overw

eightgreaterin

lowest

paternaleducationlevel,

P<0.001,P<0.01

respectively.%

overw

eight

less

inhighesttaxlevel.

Multiple

regressionalsoused

butnotclearwhether

paternaleducation,included

inmodel,wassigni®cant

covariate

deSpiegelaere

etal,1001998,

Belgium

12.3y14.8y,duration

2.4y,n�2607


from

six

schoolmedical

centresin

Brussels

region.

Medical®leschosenin

®ling

orderto

obtain

�500®les

from

eachof®vesocial

groups

ObesitybasedonBMI.

Moderate

obesity:BMI120±

140%

forageandsex;severe

obesity:BMI>140%

forage

andsex(Rolland-Cachera's

data

usedasreference)

SESbasedonparent's

professionandactivitystatus

(inoroutofwork).Unclear

whenmeasured,presumably

whenenteredschoolsince

from

medicalrecords

Chisquare

test:in

girls,at12

and15ysigni®cantinverse

relationship

betw

eenSES

andobesityprevalence(P

valuenotquoted).In

boys,no

relationship.Usingalldata

(girls

andboys),obesity

prevalencesigni®cantly

decreased,7

4%,in

interm

ediate

socialgroup

andsigni®cantlyincreased,

3.3%

inlowestsocialgroup.

RR(95%

CI)ofobesityin

lowestvshighestSESgroup

at12yin

girls

1.8

(1.3±2.6)

andin

boys1.0

(0.7±1.5),and

at15yin

girls

2.1

(1.5±3.1)

andin

boys1.4

(1.0±2.0).RR

(95%

CI)ofsevere

obesityin

lowestvshighestSESgroup

at12yin

girls

5.1

(2.0±13.1)

andin

boys1.6

(0.7±3.9)at

15yin

girls

6.7

(2.4±18.7)

andin

boys1.6

(0.7±3.4).

Changein

obesity:greater

percentageofchildrenfrom

low

SESbecameobese

(8.6%)thanfrom

inter-

mediate

(3.5%)orhigh

SES(4.5%),w2test,P<0.01.

RR(95%

CI)ofbecoming

obese,lowestvshighestSES

group:g

irls2.1(0.9±4.6);boys

1.8

(0.9±3.8);all1.9

(1.1±3.3);

RRofbecomingnorm

al

BMI,lowestSESgroupvs

highestgroup:girls

0.5

(0.3±

1.0);boys0.4

(0.2±1.1);all

0.5

(0.3±0.9) TableA5Continued


S62

Table

A5

Continued

SocialÐ

longitudinalstudies


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Weststrate

etal,1011986,

TheNetherlands


5.5y,n�171(104m,67f)

Prospective.ParticipantsÐ

noinform

ationon

recruitmentÐ

seeref.23

Changein

tertileofsum

triceps,biceps,subscapular

andsuprailiacskinfolds

SESbasedonmean

educationallevelofboth

parents,threelevels,

reportedatbaselineandf-up

(authorcommunication),

unclearwhichused,little

difference

Usingw2test,signi®cantly

more

childrenofparents

of

lowesteducationlevel

increasedfatness(tertileof

skinfold

thickness)than

childrenofparents

ofhighest

educationlevel,19=45vs

13=70,w2�7.62

Wilkinsonetal,611977,

UK

NewcastleStudyofChild

Development(N

SCD)

0,6,12mo.,5,10,12y,

duration12y,n�8000in

NSCD

at10y,n�262(161

obese,161controls).

Subgroupofn�60

obese,n�60controls

Casesandcontrols

identi®ed

at10yfrom

alongitudinal

cohort

study.Casesand

controls

born

inoneofthe

threeyears

ofrecruitment

(1961)providedmore

detailedinform

ation

Obesecases(w

gtforhgt

>97th

percentile

using

Scott's

tables)orcontrols

(25±75th

percentile)and

matchedforage,sexand

school

Allchild

ren:socialclass(at

birth?);subgroupn�120;age

ofmotheratbirth,care

of

childandstate

ofhome

reportedbyhealthvisitors

at

0±3y

Agreaterpercentageof

obesechildren(n

�161)

camefrom

socialclassesIV

andVthanexpected

comparedwithNCDS

population(n

�8000),

w2�11.4,P<0.001.In

subgroup(60obese,60

controls),more

obese

childrenhadmotherover

35yatbirth

(w2notgiven)

comparedwithcontrols.Care

ofchildrenandstate

oftheir

homesduringthe®rst3yof

life

sim

ilarin

obese

comparedwithcontrol

children

Zacketal,1021979,

USA

Data

source:NHANESII

andIII

6±11y,12±17y;duration

2±5y,n�2177


from

US

HealthExamination

Surveys,cycle

II(N

HANESII)

andcycle

III(N

HANESIII),

probabilitysamplesofnon-

institutionalizedpopulation

Sum


skinfolds

Familyincomeatbaseline

Usingmultiple

regression,

familyincomeaddedslightly,

butsigni®cantly,to

the

predictionofsubsequent

bodyfatnessin

whitemales

(n�339),in

thepresenceof

skinfoldsandageatbaseline.

Directionofassociationnot

given.Noin¯uencein

other

sex=racegroups


S63

Table

A6

SocialÐ

cross-sectionalstudies


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Boultonetal,1031995,

Australia

AdelaideChildrenW

HO

8.6y,n�856(448m,408f);

mothers

n�838,

fathers

n�640

Participants

from

cohort

of

2000childrenborn

inparticularhospital,1975±

1976

Sum

biceps,triceps,

subscapularandsupra-iliac

skinfolds(SSF).BMI

Mother'seducation,father's

education,father's

occupation

UsingANOVA,both

mother's

andfather'seducation

negatively

relatedto

child's

BMIandSSF,P<0.0001,

P<0.01respectively.Father's

occupationnegatively

related

tochild's

BMIandSSF,

P<0.01,P<0.0001

respectively

Gortmakeretal,81993,

USA

NLSY(N

ationalLongitudinal

SurveyOfLabourMarket

Experience,Youth

cohort)

16±24y;n�10,039

(4901m,5138f)

Nationalprobabilitysample

whichoversampledBlacks,

Hispanicsandpoornon-

Hispanic

Whites

Overw

eightbasedonBMI

from

reportedheightand

weight.Overw

eight:BMI

>95th

percentile

forageand

sexfrom

NHANESIdata

Mother'seducation,father's

education.Household

income

Inwomen,overw

eight

associatedwithlower

maternalandpaternal

educationlevel,and

household

income,P<0.01

(t-test?

unclear).In

men,no

association

Greenlundetal,1091996,

USA

CARDIA

18±30y;n�4006

Participants

randomly

recruited.Samplingdesigned

toincludeequalnumbers

by

age,sex,ethnicity,education

level

BMI

Father'seducation,mother's

education

UsingANCOVA,adjustingfor

fatherandmother'sbodysize

andparticipant'sbaseline

ageandeducation,BMI

decreasedwithincreasing

father'seducationlevelin

Blackmen,P�0.03,and

Whitewomen,P�0.0001,

notin

Blackwomenand

Whitemen,P>0.05

Hunteretal,1041979,

USA

BogalusaHeart

Study

5±14y,n�2837with

parentaldata,of

totaln�3524

Participants

93%

ofall

children5±14ylivingin

Bogalusa,Louisianaduring

1973±1974schoolyear,

i.e.n�3524

Weight±heightindex:

weight=height2.77;triceps

skinfold

Headofhousehold's

education,sevencategories

(n�2837)

Age-andsex-adjusted

meanscomparedacross

parentaleducationcategories

forBlackandW

hites

separately.Linearand

quadraticmodels

®tted.

Blacks:quadraticmodel

signi®cant,indicating

childrenofparents

with

highestandlowesteducation

levels

hadhighestskinfolds,

P<0.01.W

hites:no

differencein

skinfoldsover

educationcategories.Sim

ilar

resultsforweight±height

index

TableA6Continued


S64

Table

A6

Continued

SocialÐ



Agesatmeasurements=

no.ofsubjects

Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Khoury

etal,1071981,

USA

6±19y;n�893

(459m,434f)

Participants

arandom

15%

sample

from

astudyofsix

communities,biracial

BMI.

Tricepsskinfold

Educationofheadof

household.Occupationof

headofhousehold

Headofhousehold's

educationnegatively

correlatedwith

BMIr�

70.17,P<0.001and

tricepsskinfold,r�

70.12,

P<0.05in

girls

andwith

BMI,r�

70.13,P<0.01but

nottricepsin

boys.Headof

household's

occupation

negatively

correlatedwith

BMI,r�

70.12,P<0.05,but

notskinfold

inboysandno

relationshipsin

girls.Noneof

theserelationships

signi®cantafteradjustingfor

ageandrace

Kim

metal,1051996,

USA

NGHSofNHLBI

10y;n�2357girls

(1158

White,1199Black)

Schools

intw

ostates

selectedforapproxim

ately

equalproportionsofBlack

andW

hitechildrenandleast

racialdisparity

inincome.

Girls

only,sampledto

obtain

evendistributionin

SES.74±

81%

eligible

girls

enrolled

Obesity:BMI>75th

percentile

(20.5kg=m

2)from

combineddata

allsubjects

Parentaleducation.Total

household

income.No.

parents

athome

UsingBartholomew's

testfor

monotonic

trend:W

hite

girlsÐ

prevalenceofobesity

lowerathigherhousehold

income,P<0.001fortrend;

prevalenceofobesitylower

athigherparentaleducation,

P<0.001fortrend;black

girlsÐ

norelationship

betw

eenprevalenceof

obesityandhousehold

income,parentaleducation

orno.parents.Univariate

regressionanalysesgave

sameresultsandwhitegirls

withtw

oparents

signi®cantly

lesslikely

tobeobese,

P<0.001.Multiple

regression:W

hitegirlsÐ

lowereducationandone

parentfamilysigni®cant

predictors,incomenot

signi®cant.BlackgirlsÐ

neithereducation,incomeor

no.parents

signi®cant

TableA6Continued


S65

Table

A6

Continued

SocialÐ



Agesatmeasurements=

no.ofsubjects

Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Pattersonetal,1061997,

USA

NGHS

ofNHLBI

9±10y,n�2357girls

(1158

White,1199Black)

Asabove

Obesity:BMI�85th

percentile,ageandsex-

speci®cdata

from

NHANES

II.BMI�19.6

at9y;BMI

�20.9

at10y

Household

income.Max

parenteducation.Parent

employmentstatusoverprior

12mo.No.parents=

guardians.Agefemale

guardian.No.siblings

UsingANOVA,W

hitegirls

from

household

withlower

incomesigni®cantlymore

likely

tobeobese,Blackgirls

norelationship.Whiteswith

parents

oflowereducation

more

likely

tobeobese,

Blacksnorelationship.

Whiteswithneitherparent

employedmore

likely

tobe

obesethanthosewithoneor

twoparents

employed,

Blacksnorelationship.

Whiteswithoneparentmore

likely

tobeobesethanthose

withtw

oparents,Blacksno

relationship.Both

Blackand

Whitegirls

were

more

likely

tobeobeseif

motherolder,

less

likely

tobe

obese

with

more

siblings.Multiple

logistic

regression:mother's

ageandno.ofsiblings

signi®cantpredictors

forboth

BlackandW

hitegirls,max

parenteducationand

employmentsigni®cantfor

Whitesonly

Persson,1081984,

Sweden

4,8or13y,n�738,935

invited,901eligible

Random

selectionof

childrenfrom

three

speci®edgeographical

areas

Overw

eight:110±119.9%

of

standard

obesity:�120%

of

standard

weight-for-height,

from

Swedishreference

values,Karlberg,1976

SESbasedonmother's

education

Usinglogisticmultiple

regression,afteradjustingfor

agegroup,sexand

geographicalarea,SESwas

notasigni®cantpredictorfor

overw

eightbutwasfor

obesity.Theshorterthe

mother'seducationthemore

likely

thechildto

beobese


S66

Table

A7

Intrauterinegrowth

papers


Agesatoutcome


Design=participants

Outcomemeasure

Birth

weighta

Mainfinding

(i)Childhoodto

adulthood

Allisonetal,1221995,

USA

Minnesota

Twin

Registry

28±52y(m

ean40y),n�4020

twin

pairs;699pairsMZ

males,609pairsDZmales,

939pairsMZfemales,880

pairsDZfemales,893pairs

DZmale=female

Retrospective.Members

of

Minnesota

twin

Registry

(refs.22,23).Onepairtw

ins

excludeddueto

intra-pair

difference

inbirth

weight

>1.9kg

BMI

Birth

weight

Correlations:Birth

weight

positively

associatedwith

adultBMI,r�0.078,

P<0.0005(sim

ilarresults

whetherallindividuals

includedor1memberof

eachpair).Intra-pair

differencein

birth

weightnot

associatedwithintra-pair

differencein

adult

BMI,r�0.026,P�0.331.

Multivariate

analysesÐ

ordinary

leastsquares

regression:withdependent

variable

intra-pairdifference

inadultweight,independent

variable

ofinterestintra-pair

differencein

birth

weightand

covariatesintra-pair

differencein

adultheight,

age,sex,meanbirth

weight;

semi-partialr2

fordifference

inbirth

weightanddifference

inadultweight�0.03,

P�0.034,thusbirth

weight

hasnoin¯uenceonadult

relativeweight

Braddonetal,771986,

UK

1946UK

birth

cohort

0,7,11,14,20,26y;36y;


baseline;n�3249at36y


inbirth

cohort,allbirthsin

1week,March1946.Strati®ed

sample

only

includedin

this

paper.Pregnantwomen

excluded

Overw

eightÐ

men:

BMI�

25.0±29.9;women:

BMI�

24.3±29.1.ObesityÐ

men:BMI>29.9;women:

BMI>29.1

Birth

weight

Upto

14y,prevalenceof

overw

eight(relativeweight

111±130%)andobesity

(relativeweight>130%)was

greaterin

boysandgirls

of

greaterbirth

weight.Atage

36,overw

eightorobesemen

hadsigni®cantlyhigherbirth

weights,F�12.29,P<0.01.

Nodifferenceat36yin

women

TableA7Continued


S67

Table

A7

Continued

Intrauterinegrowth

papers


Agesatoutcome

measurements,=no.ofsubjects

Design=participants

Outcomemeasure

Birth

weighta

Mainfinding

Charneyetal,791976,

USA

20±30y;duration20±

30y;n�366


born

1945±1955,birth

weight>2270g,withheight

andweightrecordedseveral

timesin

childhood.90%

of

thosecontactedparticipated

Overw

eight:

>10%

above

medianweightforheightand

age.Obese:>20%

weightfor

heightandage(N

ational

HealthSurvey1960±1962

usedasreferencedata)

Birth

weightunclearwhether

from

recordsorreported

Higherbirth

weight

associatedwithgreater

prevalenceofoverw

eightin

adulthood.Forthosewith

birth

weight>75th

percentile,32%

were

overw

eightasadults,for

thosewithbirth

weight

�75th

percentile,20%

were

overw

eightasadults.

Statisticaltestnotstated,

P�0.04

Curhanetal,1231996,

USA

NursesHealthStudies

(NHS)I�

II

NHS

I,30±55y,n�52,824

inthis

paper

NHS

II,25±42y,n�84,839

inthis

paperBMI

Retrospective.NHSI:121,700

female

registerednursesin

1976.96%

White.NHS

II:

116,686female

registered

nursesin

1989.94%

White

BMI

Birth

weightreportedby

participants

Logisticregressionanalysis,

OR

and95%

CIcalculated.

Afteradjustingforage,ORof

beingin

highestquintile

vs

lowestquintile

ofBMIwas

increasedforwomenwith

birth

weightgreaterthan

referencegroup(7.1±8.5lb).

Forbirth

weight8.6±10.0lb,

OR�1.2,CI1.1±1.3

inNHSI

andOR�1.3,CI1.2±1.4

inNHSII.Forbirth

weight

>10.0lb,OR�1.6,CI1.4±1.9

inNHSIandOR�2.0,CI

1.6±2.4

inNHSII.ORalso

higherthanreferencegroup

forlow

birth

weight<5.0lb,

showndiagrammatically

only.In

NHS

I,subjects

also

dividedaccordingto

mother's®gure

atage50y,

thin,medium

orheavy.Only

inmedium

groupwas

increasedbirth

weight

associatedwithincreased

BMI

Hulm

anetal,1241998,

USA

28y,n�137(67m,70f)

(withcomplete

anthropometric

data

from

birth

±28y,a

sample

ofcohort)

Retrospective.African

Americans,offspringof

mothers

enrolledin

Philadelphia

Perinatal

CollaborativeProjectÐ

see

ref.15

Obesity:BMI>27.3

females;BMI>27.8

males

Birth

weight

Chisquare

testto

compare

groups:norelationship

betw

eenbirth

weightgroup

(<2500,2500±3200,3201±

3800and

>3800g)and

obesity,w2�0.88,P�0.83

TableA7Continued


S68

Table

A7

Continued

Intrauterinegrowth

papers


Agesatoutcome


Design=participants

Outcomemeasure

Birth

weighta

Mainfinding

Rasmussenand

Johansson,851998,

Sweden

18.2y,n�75,638

Retrospective.Single

male

birthsidenti®edfrom

birth

registry1973±1976.Linked

tofollow-updata

from

military

serviceregistry.78%

ofidenti®edbirthswere

term

birthsandheightandweight

data

wasavailable

at18y.

Analysesrelevantto

this

review

completedon1975±

1976birthsonly,36%

of

identi®edbirths

Overw

eight:25�BMI<

30;

severe

overw

eight:BMI�30

Birth

weight

Multiple

logisticregression

analyses,oddsratiosand

95%

CIs

calculated.After

adjustingforlivingareaand

mother'sage,educational

level,parity

andmarital

status:forgroupslowerthan

thereferencegroup,no

associationbetw

eenbirth

weightforgestationalage

andoverw

eightorsevere

overw

eight;forhigherthan

referencegroups,positive

associationsbetw

eenbirth

weightforgestationalage

andoverw

eightandsevere

overw

eight;foroverw

eight,

oddsratiosincreasedfrom

1.07to

1.67withbirth

weight

group,forsevere

overw

eight,

oddsratiosincreasedfrom

1.09to

1.66withincreasing

birth

weightgroup

Sorensenetal,1251997,

Denmark

18±26y,n�4305,allmales

Retrospective.Menin

5th

conscriptiondistrictof

Denmark,born

after1=1=73

andexaminedatdraftboard

1=8=93±31=7=94.505men

excludeddueto

asthma

BMI

Birth

weightobtainedfrom

birth

registry

Steadyincreasein

meanBMI

from

22.7

atbirth

weight

<2500gto

24.9

atbirth

weightof4500g,no

statisticaltestmentioned.

Multiple

regressionmodels,

adjustingforgestationalage,

birth

order,mother'smarital

status,ageandoccupation,

birth

weightassociatedwith

BMI,b�0.82,s.e.�

0.17

(ii)W

ithin

childhood

Barkeretal,1261997,

UK

14±16y,n�216,348

recruited,maternityrecords

obtainedfor216of242

singletongirls

born

locally

Retrospective.Girls

recruited

from

®veschools

inSouthamptonarea,chosento

givemix

ofsocialclasses.

83%

agreedto

participate

(n�348).This

paperincludes

singletongirls

born

locally

BMI.Subscapularand

tricepsskinfold

Birth

weight

Correlation:birth

weight

positively

associatedwith

BMI,signi®cance

borderlineP�0.08.After

adjustingforBMI,birth

weightnegatively

associated

withsubscapularskinfold.

Subscapularskinfold

increasedby7%

forevery

1kgdecreasein

birth

weight,

CI1±13%,P�0.02.No

associationbetw

eenbirth

weightandtricepsskinfold.

Theseassociations

independentofsocialclass

TableA7Continued


S69

Table

A7

Continued

Intrauterinegrowth

papers


Agesatoutcome


Design=participants

Outcomemeasure

Birth

weighta

Mainfinding

Fomonetal,1271984,

USA

8y,n�471atbaseline,

n�432at8y(244m,

188f)


were

childrenwhohad

completedapreviousstudy

(ref.16).Forthis

paper,

infants

fedform

ulas

containingbutterfator

differingappreciably

incompositionfrom

commercialform

ulas

excluded

BMI.Triceps,subscapular

skinfolds

Birth

weightmeasured

Resultsanalysedbyfeeding

andsexgroups:breastfed

males,n�73,form

ula

fed

malesn�171,breastfed

femalesn�83,form

ula

fed

femalesn�105.Using

correlationanalyses,birth

weightwaspositively

related

toBMIat8y,signi®cantonly

inform

ula

fedmales(r�0.16,

P<0.05).Associationnot

signi®cantin

othergroups.

Usingmultiple

regression

analysesto

adjustforweight

at112days,gain

inweight

5±112daysandBMIat112

days,birth

weightwas

positively

relatedto

BMIat

8yin

form

ula-fedmalesand

breastfedfemales,P<0.05.

Noanalysesreportedfor

skinfolds

Guillaumeetal,541995,

Belgium

6±13y,n�1028(527m,

501f)


selectedatrandom

from

schoolclasses.Population

Wallonianorigin,little

admixture

byim

migration.

Rural.70%

eligible

children

participated

BMI

Birth

weightreportedby

mother(a

sub-sample

checkedwithwrittenrecords,

nostatisticaldifference)

UsingANOVA,adjustingfor

age,birth

weightpositively

relatedto

BMIat6±13y,

P<0.0001

Harland,etal,1281997

UK

12±16y,n�?at

baseline,n�188at12±16y


students

ofastate

schoolin

sociallydeprivedareain

England

BMI(assumedmeasured)

Birth

weightÐ

notstated

how

obtained

Usingcorrelations;birth

weightnotrelatedto

BMI

Krameretal,74,96,971986,

and1985,Canada

1,2y;n�462at



from

largeuniversity


sociallydiverse

population.n�

553eligible,

462enrolled

BMI,sum

triceps�

subscapular�superiliac

skinfolds

Birth

weight

Multiple

regressionmodels,

stepwiseprocedure,


skinfoldsat12and24mo.

Birth

weightassociatedwith

BMIat12and24mo.

(b�0.23,P<0.0001forboth

relationships)in

model

includingdurationofbreast

feeding,genderandmaternal

score

of`idealinfantbody

habitus'.Birth

weight

associatedwithsum

skinfoldsat12mo.(b

�0.13,

P�0.02)and24mo.(b

�0.17,

P�0.002)in

modelincluding

ageatsolidfood

introduction,durationbreast

feeding,gender,maternal

relativeweightandmaternal

feedingattitude

TableA7Continued


S70

Table

A7

Continued

Intrauterinegrowth

papers


Agesatoutcomemeasurements=

no.ofsubjects

Design=participants

Outcomemeasure

Birth

weighta

Mainfinding

Maffeis

etal,571994,

Italy

4,8,10,12y;n�1523

total,n�1363this

paper

(676m,687f)

Retrospective.Random

sample

from

fourschool

classesin

six

areasofnorth-

eastItaly.This

paper

excludeschildrenwithlow

gestationalageandincludes

subjects

forwhom

complete

data

wasavailable

Obesity:BMI>95th

percentile

forageand

gender,usingFrenchtables

asreference(seeref.13)

Birth

weightreportedby

parents

Oddsratioscalculatedfor

likelihoodofobesity,birth

weightcategories

unspeci®ed.Afteradjusting

forage,higherbirth

weight

increasedlikelihoodof

obesity,in

boysOR�1.001,

P<0.002,andgirls

OR�1.001,P�0.004.Using

logisticregression,adjusting

forageandincludingboth

parents'BMIs,birth

weight

waspositively

relatedto

BMI

inboys(O

R�1.001,P�0.02)

andgirls

(effectvarieswith

mother'sBMI),P<0.05


Australia

MUSP(M

aterUniversity

StudyofPregnancy)

5y;n�7357at


(2133f,1929m)

Prospective.Mothers

(n�8556)enrolledat

pregnancyhospitalvisit

(ref.20).Atbirth

n�7357,

notallofthesechildren

followedupdueto

insuf®cientfunds

Obesity:severe

obesityÐ

BMI>94th

percentile;

moderate

obesityÐ

BMI

85±94th

percentile

within

cohort

Birth

weight

Univariate

analyses:RR,95%

CIs.Birth

weightin

85±94

percentile

and

�95th

percentile

positively

related

tosevere

obesity,RR�1.7,CI

1.2±2.9,RR�1.8,CI1.1±2.9

respectively.Birth

weight

�95th

percentile

positively

relatedto

moderate

obesity

RR�1.8,CI1.3±2.5.

Multivariate

analyses

includedbirth

weight,sizefor

gestationalage,gender,

feedingproblems,duration

ofbreastfeeding,

sleeplessness,parentalBMIs,

maternaleducation,levelof

income.AdjustedORs

calculatedwith95%

CIs.Birth

weight�95th

percentile

positively

relatedto

severe

obesityOR�1.8,CI1.1±2.9,

notmoderate

obesity.Birth

weight85±94percentile

positively

relatedto

moderate

obesityOR�2.0

CI

1.3±3.2,notto

severe

obesity

TableA7Continued


S71

Table

A7

Continued

Intrauterinegrowth

papers


Agesatoutcome


Design=participants

Outcomemeasure

Birth

weighta

Mainfinding

Prescott-Clarkeand

Primatesta,631998,UK

HealthSurveyforEngland

2±15y,n�13,563

(6846m,

6717f)

Retrospective.Nationally

representativesample

selectedbyrandom

sampling

technique

BMI

Birth

weightreported

Inmalesandfemales,with

increasingbirth

weightthe

proportionofchildrenin

the

topquintile

ofBMIincreased,

andtheproportionin

the

bottom

quintile

decreased

Seidmanetal,991991,

Israel.

Jerusalem

PerinatalStudy

0y,17y;duration17y;n�?

atbaseline,n�20,747at17y,

allmales

Prospective.Allbirthsfrom

threemajorwardsin

Jerusalem

1964±1971.At

17ythosenotin

defence

forcesnotincluded.61.7%

femalesmissing

Overw

eight:BMI�90th

percentile,24.6;severe

overw

eight:BMI�97th

percentile,27.8

Birth

weight

Whenprevalenceof

overw

eightandsevere

overw

eightexpressedby

birth

weight,strongand

linearrelationship

observed

atbirth

weights

>3000g(no

statisticaltestmentioned).

Usingmultiple

logistic

regression,including

parentaleducation,areaof

residence,ethnic

origin

and

birth

order,ORforbeing

overw

eightincreasedwith

increasingbirth

weight,

comparedto

birth

weightof

3000±3499gasreference

group.Modelforsevere

overw

eightincludedpaternal

educationandethnic

origin,

sim

ilarresultsobtained

Tayloretal,1291997,

UK

TenTownsStudy

8±11y,n�3010


from

astrati®edrandom

sample

of10schools

ineach

of10townsselectedonbasis

ofadultcardiovascular

mortality.3728participated,

3181withdata

onbirth

weight,3010singletonbirths

BMI

Birth

weightreportedby

parents

whenchildage

8±11y

Birth

weightpositively

correlatedto

BMIat8±

11y,r�0.09,P<0.0001

Ziveetal,1301992,

USA

SCAN

study

4y,n�270,�50%

m,

50%

fRetrospective.Recruitment

via

state-fundedpre-schools,

children's

centres.Low

SES

familiestargettedforstudy.

Anglo-andMexican

Americans.331respondents,

complete

data

for270

BMI,sum

triceps�

subscapularskinfold

Birth

weightreportedby

mother

Correlations:birth

weight

signi®cantlycorrelatedwith

BMI,r�0.28,P<0.001and

withsum

skinfoldsr�0.16,

P�0.01

aBirth

weightmeasuredunlessstatedotherw

ise.


S72

Table

A8

Maturationpapers


Agesatrisk

factorandoutcome


Design=participants

Outcomemeasure

Riskfactor

Mainfinding

(i)Childhoodto

adulthood

Beunenetal,1581994,

Belgium

LeuvenGrowth

Study

13y,annuallyuntil18y,30y,

duration17y;n�588at


(allmale)

Prospective.Subgroupof

boys(n

�588)from

nationally


of

Belgiansecondary

school

boys(n

�21,175)

BMI,triceps,subscapular,

suprailiac,medialcalf

skinfolds

Maturity

basedonPHVage.

Earlymaturers:PHVage

<13.37y;averagematurers:

13.85�PHVage�14.8y;late

maturers

PHVage>15.27y

One-w

ayANOVA

with

Duncanmultiple

rangetest

usedto

compare

groups.At

30y,nodifferencebetw

een

maturity

groupsin

BMI,

triceps,medialcalfskinfolds.

Subscapular

and

suprailiac

skinfoldssigni®cantlylargest

inearly>average>late

maturers,P<0.05,P�0.07,

respectively.Correlations

usedto

relate

PHVageand

bodydim

ensionsat30y.

Correlationcoef®cients

mostlylow,r<0.1.PHVage

relatedto

subscapular

(r�

70.34,P<0.05)and

medialcalf(r�

70.18,

P<0.05)skinfolds

Burkeetal,1541992,

USA

CARDIA

18±30y,n�2658

(allfemale)


randomly

recruited.

Samplingdesignedto

includeequalnumbers

by

age,sex,ethnicity,education

level

BMI

Ageatmenarche

Nostatisticaltestreported.In

both

BlackandW

hite

women,BMIincreasedwith

decreasingageofmenarche

Garn

etal,1501986,

USA

NationalCollaborative

PerinatalProject(N

CPP)

20±35y,n�16,868

(allfemale)


orretrospective(seerefs.6±

8).Participants

ofNational

CollaborativePerinatal

Project.Teenagers

insample

excluded

BMI

Ageatmenarche:early

<11y;interm

ediate

12±13y;

late

�14y

ANOVA

andt-tests

usedto

compare

groups.Early

maturers

showedgreater

BMIthanlatermaturers,

61.17vs56.96,P<0.001.

Differencepersistedif

subjects

groupedbyage,

20±25y,26±30y,31±35y,

butbecamelargerwith

increasingage.Difference

persistedifsample

restricted

onbasis

ofparity

or

education

Garn

etal,1501986,

USA

TecumsehCommunity

HealthStudy

20±35y,n�476

(allfemale)


orretrospective(seerefs.6±

8).Participants

ofTecumseh

roundII.Teenagers

insample

excluded

Triceps,subscapular,iliac,

abdominalskinfolds;obesity:

abdominalskinfold

>85th

percentile

Ageatmenarche:early

<11y;interm

ediate

12±13y;

late

�14y

ANOVA

andt-tests

usedto

compare

groups.All

skinfoldsgreaterin

early

maturers

thanin

later

maturers,P�0.001.Subjects

ofsim

ilarageandSES.RRof

obesityin

earlymaturing

group�1.7

TableA8Continued


S73

Table

A8

Continued

Maturationpapers


Agesatrisk

factorandoutcome


Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Gasseretal,1421994,

Switzerland

ZurichGrowth

Study

1,3,6,9,12,18,24mo.

annuallyuntil9y(f),10y(m

);biannually,thenannually

dependingon

hgtuntil

�18y;additionaladulthood

measurementformost

subjects

17±25y,

baselinen�321,follow-

upn�232(112f,120m)


noinform

ationgiven,see

ref.Gasser1990

BMI.`Fat'group:adultBMIin

toptertile;`lean'group:adult

BMIin

bottom

tertile.

Extremityskinfolds

(triceps�biceps).Trunk

skinfolds(subscapular�

suprailiac)

Pattern

ofvelocityofchange

inBMIthroughoutchildhood

Structuralaveragegrowth

curvesconstructed.

Relationship

ofmeanvelocity

BMIwithagefor`fat'and

`lean'groupsdisplayed

graphically,malesand

femalesseparately.`Fat'and

`lean'groupscompared.`Fat'

malesandfemalesshowed

largerpeaksandtroughsin

BMIvelocity.Difference

especiallyclearatpuberty

where

`fat'groupsshowed

largerpeak.In

general`fat'


higherBMIvelocityfrom

3y

into

adulthood.Comparing

meanincreasein

BMI

betw

een3and10y(3

and

12yboys)andbetw

een15.7y

andadult(16.6yboys),`lean'

groupsshowedmuch

smallerincreasesduring

thesepre-andpostpubertal

periods.Duringpuberty,`fat'

malesandfemales

demonstratedhigherpre-

andpostpubertalfatspurts,

asmeasuredbyextremity

andtrunkskinfolds

Gasseretal,1431994,

Switzerland

ZurichGrowth

Study

Asabove

Asabove

Asabove

Pattern

ofvelocityofchange

infatandleanareaofupper

arm

throughoutchildhood

Asabovebutgrowth

curves

constructedformeanvelocity

fatandleanareasofarm

.`Fat'and`lean'groups

compared.`Fat'malesand

femalesshowedhigher

velocityofleanareafrom

age

6to

puberty.In

males,

maxim

um

velocity

�6mo.

earlierin

`fat'males.`Fat'


largerpeaksandtroughsin

velocityoffatarea.Difference

especiallyclearatpubertyin

males

TableA8Continued


S74

Table

A8

Continued

Maturationpapers


Agesatrisk

factorandoutcome


Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Guoetal,1481997,USA

Fels

LongitudinalStudy

8y,23y;duration15y,

n�244(114m,130f)


of

Fels

study(seeref.17)with

�6bodycomposition

measurements

betw

een8

and23y.AllCaucasian

%bodyfatbasedon

underw

aterweighing

Rate

ofmaturation(skeletal

age7

chronologicalage

(SA7

CA)).MeanSA7

CA

overperiodfrom

8yto

maturity

obtained.Rapid

maturers:value

>1.

Interm

ediate:ÿ1<value<1.

Slow

maturers:value

<7

1

Usingmultiple

regression

analysesandadjustingfor

age,age2andage3,in

males,

slow

maturers

havelower

skinfoldsthanrapid

maturers,P<0.05.In

females,interm

ediate

maturers

havelower

skinfoldsthanrapid

maturers,P<0.05andalso

lowerthanslow

maturers,

notsigni®cant

Siervogeletal,1491991,USA

Fels

LongitudinalStudy

2y,semi-annuallyuntil18y;



(167m,163f)


of

Fels

study.AllCaucasianand

from

widerangeofSES

BMI

Ageatmaxim

um

BMI;age

atmaxim

um

velocityBMI;

ageatminim

um

BMI

Participants

analysedin

two

groupsdependingon

whetherthreeorfour

parameterpolynomial

growth

curveusedto

®tdata.

Forfourparameter

model,n�417.Ageat

minim

um

BMInegatively

correlatedto

BMIat18yin

boysr�

70.46,and

girls

r�

70.54,P<0.01.Age

atmaxim

um

BMIpositively

correlatedto

BMIat18yin

boysr�0.57,andgirls

r�0.64,P<0.01.Ageat

maxim

um

velocityBMI

unrelatedto

BMIat18y.For

threeparametermodel

n�56.Nocorrelations

betw

eenageatmaxBMI,min

BMIormaxvelocityBMIand

BMIat18y

Milleretal,1591972,UK

NewcastleThousand

FamiliesStudy

0y,3,5,9,13,14,15y;

22y;n�1142(m

�f)at

baseline;n�442(m

�f)at

22y(201m,241f)

Prospective.Originalsample

allbabiesborn

inNewcastle

inMayandJune1947.Only

data

from

femalesrelevantto

this

review

Weight-for-height(Kemsley

1962Britishdata

usedas

referencedata)

Ageatmenarche(unclear

whetherrecorded

prospectively

or

retrospectively)

Nostatisticaltestreported.

Ageofmenarchenegatively

relatedto

weight-for-height

at22y

TableA8Continued


S75

Table

A8

Continued

Maturationpapers


Agesatrisk

factorandoutcome


Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Poweretal,171997,UK

1958UK

birth

cohort

7,11y,33y,duration26y

(forthis

paper);n�

�17,000

atbaseline;n�11,407

at33y


all

births3±9March1958.

Immigrants

born

insame

weeksubsequentlyincluded

infollow-ups

BMI

Tannerstageat11y

(assessedbydoctor).At16y:

ageatmenarche(girls,

reported)stageofaxillary

hair(boys)

Testfortrendacross

maturationgroups.Basedon

pubertyassessedat11y:BMI

at33yincreasedwith

advancingstageofpubertyin

females,P�0.003,andmales

(relationship

notsigni®cant),

P�0.3.Basedonmaturation

assessedat16y:BMIat33y

increasedwithearlierageof

menarchein

females,

P�0.003,andadvancing

maturationin

males,

P�0.003

Freemanetal,1401995,UK

1958UK

birth

cohort

7,11y,23y;duration16y(for

this

paper);at7yn�6874m,

6422f;at11yn�6381m,

6118f;at23yn�6134m,

6145f

Asabove

Obesity:BMI�30

Maturation:earlymaturerÐ

tallat7ycomparedto

at16y;

late

maturerÐ

short

at7y

comparedto

at16y(both

at7

and16y,subjects

classi®ed

asshort,medium

andtall,

separate

cut-offsformales,

femalesat16y)

Nostatisticaltestmentioned.

At23y,prevalenceofobesity

higherin

earlymaturers

(tall

at7,short

at16y)greater

thanin

late

maturers

(shortat

7,tallat16y),4.3%

vs1.4%

inmales,9.5%

vs3.8%

infemales

Stark

etal,1411989,UK

1958UK

birth

cohort

7y,11y,16y,duration

9y,n�

�17,000baseline

(boys

and

girls);

n�3018

girls

withcomplete

data

for

this

paper

Asabove.Only

girls

includedin

this

paper

Relativeweight(observed

wgtas%

ofexpectedwgt

obtainedbyregressionof

logwgtonloghgt)

Ageatmenarche(reported)

Nomentionofstatisticaltest.

Withincreasingageof

menarche,percentageof

girls

withrelativeweight

101±120%

and

>120%

decreased,thosewith

relativeweight<100%

increased,atallages,i.e.7y,

11y,16y

ProkopecandBellisle,156

1993,CzechRepublic

Co-ordinatedbyCentre

Internationaldel'Entrance

(ClEP)

1,3,6,9,12mo.every

6mo.

until18y;duration18y,


at18y(80m,78f)


selectedatrandom

from

particulardistrict,children

born

onW

ednesdayincluded

BMI

Adiposityrebound

Notalldata

used,participants

withBMI>90th

percentile

(fat)andBMI<10th

percentile

(lean)at18y

compared.Males:fatn�9,

leann�10.Females:fat

n�10,leann�8.ANOVA

usedto

testforgroupmean

differences.Ageatadiposity

reboundsigni®cantlylower

inmalesfatat18y(4.6yvs

7.8y,P<0.02)andfemales

fatat18y(5.3yvs7.4y,

P<0.05)

TableA8Continued


S76

Table

A8

Continued

Maturationpapers


Agesatrisk

factorandoutcome


Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Rolland-Cachera

etal,144

1987,France

Co-ordinatedbyCentre

Internationaldel'Enfance

(CIEP)

�every

6mo.until16y;21y;

duration21y;n�164(85m,

79f)


noinform

ationgiven,see

below.This

paperincludes

subjects

measuredbeyond

16y

BMI,subscapularskinfold

Adiposityrebound:

<5.5y�advanced;

5.5±7y�average;

>7y�delayed

ANOVA

usedto

compare

maturity

groups.Boyswith

advancedadiposityrebound

showedhigherBMI(P

�0.01)

andsubscapularskinfold

(P�0.02)thanthosewith

delayedadiposityrebound.

Sim

ilarrelationships

observedin

girls,butless

signi®cant.Girls

with

advancedadiposityrebound

showedhigherBMI(P

�0.12)

andsubscapularskinfold

(P�0.08).Signi®cance

greaterifcomparisonmade

only

betw

eenadvancedand

delayedadiposityrebound

groups(forBMIP�0.04,

subscapularskinfold

P�0.03)

Rolland-Cachera

etal,145

1984,France

Asabove

�every

6mo.until16y,

16y;duration16y,n�?at

baseline,n�151at14y,

n�114at16y

(62m,52f)


noinform

ationgiven,see

refs.10,11,or12.This

paperincludessubjects

measureduntil14y

BMI

Adiposityrebound:

<5.5y�advanced;

5.5±7y�average;

>7y�delayed

Varianceanalysis:boysand

girls

withadvancedadiposity

reboundwere

more

adipose

at16ythanthosewith

averageadiposityrebound,

whowere

inturn

more

adiposeat16ythanthose

withdelayedadiposity

rebound(P

<0.01)

St.Georgeetal,1571994,

New

Zealand

Dunedin

Multidisciplinary

Healthand

Development

Study

3y,every

2yuntil15y,18y;

duration15y,n�1139

baseline(m

�f),n�415girls

withcomplete

data

inthis

paper


of

birth

cohortÐ

allbirthsin

Dunedin

over1y,1=4=72±

31=3=73

BMI

Ageatmenarche

Data

displayedgraphically,

nostatisticaltestmentioned.

BMIat18ywasgreatestin

thosewithmenarche

<12y,

lessin

thosewithmenarche

at12±13y,lessagain

inthosewithmenarcheat13±

14yandleastin

thosewith

menarche

>14y.This

was

trueatalltime-points

9±18y

inclusive

TableA8Continued


S77

Table

A8

Continued

Maturationpapers


Agesatrisk

factorandoutcome


Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Sherm

anetal,1551981,

USA

19±73y,n�1221


female

universitystudent

volunteers

anddaughters

of

this

cohort

BMIcurrentÐ

basedon

reportedcurrenthgtandwgt.

BMIat18yÐ

basedon

reportedcurrenthgtand

recalledwgtat18y

Ageofmenarche,recalled

whilestudentatuniversityfor

originalcohort,recorded

prospectively

fordaughters

Groupscomparedusing

Student'st-test.Basedon

BMIat18y,menarchein

the

highestquartileofBMIwas

12.4y,comparedto

13.0yin

thelowestBMIquartile,

P<0.001.Basedoncurrent

BMI,menarchein

thehighest

quartileofBMIwas12.5y,

comparedto

12.8yin

the

lowestBMIquartile,P<0.001

vanLentheetal,1461996,

TheNetherlands

AGHS

(Amsterdam

Growth

andHealthStudy)

13,14,15,16,21,27y;



(79m,98f)


complete

1stand2ndform

ofAmsterdam

school,no

refusals

toparticipate.SES

abovepopulationaverage

BMI,sum

ofskinfolds

(biceps,triceps,

subscapular,suprailiac)

Maturationasmeasuredby:

(i)skeletalageÐ

rapid

maturer:skeletalage

>3mo.

inadvanceofchronological

ageatallfourtime-points

13±16y;slow

maturer:

skeletalage

>3mo.behind

chronologicalageatallfour

time-points

13±16y;

(ii)PHVage(boysonly);

(iii)ageatmenarche(girls);

(ii)and(iii)early

maturer�bottom

tertile,late

maturer�top

tertile

Groupdifferencesanalysed

bysex-speci®cMANOVA

for

repeatedmeasures.Boys:

skinfoldsÐ

whethermaturity

basedonskeletalageorPHV,

decreasein

adolescence

occurredearlierin

rapid

maturers.BMIÐ

larger

increasein

slow

maturers

at

age16±21y.Ifmaturity

basedonskeletalage,BMI

andskinfoldsgreaterin

rapid

maturers

13±27ybut

differencenotsigni®cantif

maturity

basedonPHVage.

Girls:skinfoldsÐ

whether

maturity

basedonskeletal

ageorageatmenarche,

increaseat16±21ylargerin

slow

maturers.W

hether

maturity

basedonskeletal

ageorageatmenarche,both

skinfoldsandBMIgreaterin

rapid=earlymaturers

throughouttimeperiod

13±27y

TableA8Continued


S78

Table

A8

Continued

Maturationpapers


Agesatrisk

factorandoutcome


Design=participants

Outcomemeasure

Riskfactor

Mainfinding

PostandKemper,1471993,

TheNetherlands

AGHS(A

msterdam

Growth

andHealthStudy)

13,14,15,16,21y;duration

8y;baselinen�307at

21yn�200(93m,107f)

Asabove

Sum

ofskinfolds(biceps,


suprailiac)

Maturationasmeasuredby

skeletalage,seeabovefor

de®nition

Only

early(rapid)andlate

maturers

comparedwith

eachother.Groupdifferences

analysedusingANOVA.In

both

boysandgirls,early

maturers

demonstrated

higherskinfoldsthroughout

study,P�0.05

Wellensetal,1511992,

USA

1970cohort

19.2y,n�342

Retrospective.Participants:

cohortsofW

hitefemale

universitystudents

recruited

in1970

BMI,tricepsskinfold

Ageofmenarche

Differencesin

BMIand

skinfold

acrossfourgroups

ofmenarchealageanalysed

byANOVA.Asageof

menarcheincreased,BMI

andtricepsskinfold

decreasedalthoughnot

statisticallysigni®cant


USA

1987cohort

18.8y,n�109

Retrospective.Participants:

cohortsofW

hitefemale

universitystudents

recruited

in1987

BMI,tricepsskinfold

Ageofmenarche

Differencesin

BMIand

skinfold

acrossfourgroups


byANOVA.Asageof

menarcheincreased,BMI

andtricepsskinfold

decreasedalthoughnot

statisticallysigni®cant


USA

Fels

study

18.3y,n�234


femalesin

Fels

study,sim

ilar

tocohortsdescribedabove

BMI,tricepsskinfold

Ageofmenarche

Differencesin

BMIand

skinfold

acrossfourgroups


byANOVA.Thosewith

menarche

<12yshowed

greaterBMIthanthosewith

menarche12±12.9y,13±

13.9yor>13.9y,P<0.01.

Thosewithmenarche

<12y

showedgreatertriceps

skinfold

thanthosewith

menarche

>13.9y,P<0.05

TableA8Continued


S79

Table

A8

Continued

Maturationpapers


Agesatrisk

factorandoutcome


Design=participants

Outcomemeasure

Riskfactor

Mainfinding

(ii)W

ithin

childhood

Hedigeretal,1531995,

USA


1y,n�851atbaseline,

n�668at15±19y

(allfemale)


female

volunteers

from

two

highschools,grades9±11.

Threeroundsofrecruitment.

70±93%

oftotalvolunteered.

Ethnicity:61%

White,16%

Black,17%

Puerto-Rican,6%

other

Dtricepsskinfold,D

subscapularskinfold

Gynaecologic

age

(chronologic

age7

ageatmenarche)

Multiple

linearregression

analysis

witheitherDBMI,

andDtricepsorD

subscapularskinfold

as

dependentvariable,and

whichappears

toincludeall

offollowingascovariates

(notentirely

clear):age,

gynaecologic

age,ethnicity,

late

maturation,cigarette

smoking,initialskinfold

status.Noin¯uenceof

gynaecologic

ageorlate

maturationonDtricepsor

subscapularskinfold

Knishkowyetal,1521989,

Israel

6y,14y,duration7.5y,

n�312fatbaseline,

n�288fat14y


boysandgirls

attw

oelementary

schools

inJerusalem.Only

data

from

girls

relevantto

review.39%

allmothers

Israeliborn,18%

Asian,26%

N.African,16%

European=American

BMI

Ageatmenarche

(reportedat

follow-up)

ANOVA

usedto

testgroup

differencesbetw

eengirls

menstruatingandnot

menstruatingatfollow-up

(14y).Menstruatinggirls

(n�215)hadhigherBMI,

20.1�2.6

thannon-

menstruatinggirls

18.3�2.1,

P<0.001.Menstruatinggirls

(n�215)hadhigherage,13y

10mo.�

3.9mo.thannon-

menstruatinggirls

13y

8.8mo.�

3.6mo.,P�0.01.

Differencesin

BMIpersisted

afteradjustingforageusing

ANCOVA,P<0.001

Adiposityrebound�lowestBMIrecordedbefore

secondrisein

adiposity,whichusuallyoccurs

at�6yonBMIandskinfold

charts.PHVage�ageatpeakheightvelocity.


S80

Table

A9

Physicalactivitypapers

Author=year=country

Agesatmeasurements=

no.ofsubjects

Design=participants

Outcome(m

easuredunless

statedotherw

ise)

Activity

risk

factor=method

ofmeasurement

Mainfinding

(i)Childhoodto

adulthood

Raitakarietal,201994,

Finland

CardiovascularRiskin

Young

FinnsStudy

12±18y,15±21y,18±24y,

duration6y,n�3,596,

baselinen�961at18±24y


randomly

chosenfrom

populationregister.No

inform

ationondropouts

BMI,subscapularskinfold

Leisure-tim

ephysicalactivity

assessedbyquestionnaire

andindexcalculated

accountingforintensity,

durationandmonthly

frequency

Studentt-testusedto

compare

subjects

whowere

active(n

�23f,44m)with

thosewhowere

sedentary

(n�55f,33m)atallthree

measurements.Subscapular

skinfold

waslowerin

active

subjects

inmales(9.9

vs12.1,

P�0.01)andfemales(10.5

vs

14.2,P�0.005).Nodifference

inBMIin

males(21.8

vs21.9)

orfemales(22.6

vs22.1)

(activevssedentary)

Twisketal,1721997,

TheNetherlands

Amsterdam

Growth

and

HealthStudy(A

GHS)

13,14,15,16,21,27,29y,

duration16y,n�233


(98f,83m)


complete

1stand2ndform

ofAmsterdam

school,no

refusals

toparticipate.SES


Sum

ofbiceps,triceps,

subscapular,suprailiac

skinfolds(SSF)

Totalactivitybystructured

interview

concerning

previous3mo.Durationand

intensitycombinedto

give

weightedscore

expressedin

MET's

Multiple

linearregression

analysesincludingvariables

basedonsigni®cancein

univariate

analyses.Daily

physicalactivityin

adolescenceandyoung

adulthoodwasnegatively

relatedto

SSF(b

�7

0.13,

P�0.05).Relationship

strongerifentire

longitudinal

periodconsidered(further

into

adulthood;b�

70.20,

P<0.01)

Twisketal,1731997,

The

Netherlands

AGHS

13,14,15,16,21,27y,

duration14y,n�233


(98f,83m)

Asabove

Asabove

Asabove

Statisticalmodel

incorporatingparameters

calculatedusinggeneralized

estimatingequations.After

adjustingforbiologicalage,

sex,sum

ofskinfolds,

VO

2max,andhighriskforsum

skinfolds(�20%

bodyfatfor

males,30%

forfemales),was

negatively

relatedto

daily

activity(O

R�0.81,P�0.01).

Participants

whoare

inactive=becomeinactiveare

more

likely

tobein

ormove

into

riskgroupforsum

of

skinfolds

TableA9Continued


S81

Table

A9

Continued


Author=year=country

Agesatmeasurements=

no.ofsubjects

Design=participants

Outcome(m

easuredunless

statedotherw

ise)

Activity

risk

factor=method

ofmeasurement

Mainfinding


TheNetherlands

AGHS

13,14,15,16,21,27y,

duration14y,n�233


(98f,84m)

Asabove


Asabove

`Generalestimating

equations'(longitudinal

regressionanalysis),

adjustingforsum

skinfolds

betw

een13and27y:

malesÐ

noin¯uenceof

activityonDsubscapular

skinfold;femalesÐ

activity

negatively

relatedto

D

subscapularskinfold,

b�

70.10,CI7

0.15,

70.05,P<0.001.Multiple

regressionanalysis

with

subscapularskinfold

at27y

asdependentvariable

and

initiallyincludingactivity,

diet,smokingandalcohol

intakeascovariates,non-

signi®cantcovariates

removedbybackwards

eliminationprocedure:

malesÐ

activitypositively

relatedto

subscapular

skinfold,b�0.15,SE�0.07,

P�0.03.FemalesÐ

no

in¯uence

(ii)Withinchildhood

Baselinemeasurementpre-1

yearofage(i.e.pre-w

alking)

Berkowitzetal,951985,

USA

1±3days,6.5y,duration

�6.5y,n�112baseline,

n�52at6.5y(25m,

27f)


recruitedfrom

cohort

of

anotherstudyÐ

seeref.9

LogBMI,tricepsskinfold

Neonatalnoncrying

activityÐ

frequencyand

intensitybyelectronic

activity

monitor

Neonatalnon-crying

amplitude(intensityof

movements)negatively

correlatedto

tricepsskinfold

at6.5y,r�

70.34,P<0.05

Daviesetal,1681991,

UK

12wks,2y,duration

2y,n�52baseline,

n�33at2y


recruitedfrom

large

maternityhospital

BMI,sum


skinfolds

Energyexpenditure

measuredbydoubly

labelled

water


betw

eenenergyexpenditure

at12wksandBMIor

skinfoldsat2y

Daviesetal,1691993,

UK

12wks,6.7y,duration

6y,n�?baseline,

n�24at6y

Asabove

Fatmassrelativeto

fat-free

mass

Energyexpenditure

measuredbydoubly

labelled

water


betw

eenbodyfatnessat6.7y

andenergyexpenditure

at

12wksexpressedrelativeto

fat-freemass,r�0.11,

t�0.52

TableA9Continued


S82

Table

A9

Continued


Author=year=country

Agesatmeasurements=

no.ofsubjects

Design=participants

Outcome(m

easuredunless

statedotherw

ise)

Activity

risk

factor=method

ofmeasurement

Mainfinding

Kuetal,1701981,

USA

NB.Alsofallsinto

`post-

walking'groupbelow

6mo.,1,2,3,4,8y,duration

8y,n�450

baseline,n�170

at8y(83m,87f)


forrecruitmentdetailssee

refs.13±17

%bodyfatat8yby

underw

aterweighing

Physicalactivitymeasuredat

6mo.,1,2,3,4ybyparents

by24hactivityrecord

for

every

5min.At8ychildren

recordedactivityassistedby

parents

Girls:nosigni®cant

correlationbetw

eenactivity

at6mo.and%

bodyfatat8y.

Boys:nosigni®cant

correlationbetw

eenactivity

at6mo.and%

bodyfatat8y


Australia

MUSP(M

aterUniversity

StudyofPregnancy)

Baseline0y,follow-up

5y,n�7357baseline,

n�4062at5y(2133

f,1929m)

Prospective.Mothers



(ref.20).Atbirth

n�7357,not

allofthesechildrenfollowed

updueto

insuf®cientfunds

Obesity:severe

obesityÐ

BMI>94th

percentile;

moderate

obesityÐ

BMI85±

94th

percentile

within

cohort

Frequencyof`overactivity'

reportedbymotherwhen

child6mo.Node®nition

givento

mother

Univariate

analyses:RR,95%

CIs.Norelationship

betw

een

activityat6mo.andobesity

at5y

Wellsetal,1711996,UK

12wks,2±3.5y,duration2±

3.5y,n�50baseline,n�24±

28at2±3.5y


recruitedfrom

large

maternityhospital

%fat(deuterium

dilution),

skinfolds,BMI

Components

ofinfantenergy

expenditure:sleeping

metabolicrate,totalenergy

expenditure,energyspenton

physicalactivity

Multiple

regressionanalyses:

afteradjustingforbody

weightorfat-freemass,no

relationshipsbetw

eentotal

orrestinginfantenergy

expenditure

andchildhood

fatness

Baselinemeasurementpost-

1yearofage(i.e.post-

walking)

Beunenetal,1751992,

Belgium

LeuvenGrowth

Study

12,13,14,15,1617y;

duration5y,n�?


Prospective.Subgroupof

boys(n

�588)from

nationally


of

Belgiansecondary

school

boys(n

�21,175)

Triceps,subscapular,

suprailiac,calfskinfolds

Active(>5h=wksport)vs

inactive(�1.5h=wksport)in

additionto

compulsory

physicaleductaionfor®rst

3yofstudyperiod

Generallinearmodelwith

repeatmeasures;no

differencein

anyofskinfolds

betw

eenactive(n

�32)and

inactive(n

�32)boys

Dietz

andGortmaker,1761985,

USA

(NHES)


6y,n�2,153


thosewhohadparticipatedin

both

NHESIandII

Tricepsskinfold.Obesity:

triceps�85th

percentile;

superobesity:triceps�95th

percentile

Hours

oftelevisionwatching

perdaybyquestionnaireto

parents

Weightedstepwise

regressions:positive

relationshipsbetw

een

televisionviewingatage

6±11yandobesity(b

�0.08,

P<0.07)andsuperobesity

(b�0.06,P<0.03)atage

12±17y,aftercontrollingfor

obesityatage6±11and

socio-economic

characteristics

TableA9Continued


S83

Table

A9

Continued


Author=year=country

Agesatmeasurements=

no.ofsubjects

Design=participants

Outcome(m

easuredunless

statedotherw

ise)

Activity

risk

factor=method

ofmeasurement

Mainfinding

Goranetal,441998,

USA

5.2y,annualto

8.3y,duration

3y,n�75baseline,n�53

at8.3y


recruitedbynewspaper

advertisements

and

word

ofmouth

Rate

ofchangein

fatmass

adjustedforfat-freemass(fat

andfatfreemassmeasured

byDXA)

TEE(totalenergy

expenditure)bydoubly

labelledwater,REE(resting

energyexpenditure)by

indirectcalorimetry

UsingANOVA,rate

ofchange

infatmassadjustedforfat-

freemasspositively

related

toTEE(r�0.28,P�0.025),

TEEadjustedforREE

(r�0.33,P�0.08)and

unrelatedto

REE(r�0.14,

P�0.24).Assume(not

explicitbutlikely)TEE,REE

are

from

baseline

measurements

Klesgesetal,451995,

USA

4.4y,5.4,6.4y,duration

2y,n�203baseline,

n�146at6.4y


recruitedthrough

paediatricians,day-care

centres,churches.Obese

over-sampled

DBMI

Leisure

activity,structured

activity,andaerobic

activity

allassessedbyboth

parents

(askedwhetherchildrendid

muchless,aboutthesame,

ormuchmore

thanother

children)

Multiple

regressionanalysis

includingadjustm

entfor


parentalfatnessand

interactionsbetw

eengender

andparentalfatness.In

modelincludingthese

covariatesandalsodietary

intake,over2ystudyperiod

higherbaselineaerobic

activityandincreasedleisure

activitywere

associatedwith

decreasein

BMI(b

�7

0.32,

P�0.03,andb�

70.04,

P�0.01respectively)

Kuetal,1701981,

USA

NB.Alsofallsinto

`pre-w

alking'group

above


8y,n�450

baseline,n�170

at8y(83m,87f)


forrecruitmentdetailssee

refs.13±17

%bodyfatat8yby

underw

aterweighing

Physicalactivitymeasuredat

6mo.,1,2,3,4ybyparents

by24hactivityrecord

for

every

5min.At8ychildren

recordedactivityassistedby

parents

Ingirls

nosigni®cant

correlationbetw

eenactivity

atanyageand%

bodyfatat

8y.In

boysnosigni®cant

correlationbetw

eenactivity

at6mo.,1y,2yor8yand%

bodyfatat8y,butsigni®cant

negativecorrelations

betw

eenactivityat3y,4y

and%

bodyfatat8y,and

betw

eenandsum

activity

6mo.±

8yand%

bodyfat

at8y

TableA9Continued


S84

Table

A9

Continued


Author=year=country

Agesatmeasurements=

no.ofsubjects

Design=participants

Outcome(m

easuredunless

statedotherw

ise)

Activity

risk

factor=method

ofmeasurement

Mainfinding

Maffeis

etal,461998,

Italy

8.6y,12.6y,duration

4y,n�298baseline,

n�112at12.6y


recruitedthroughprivate

and

publicschools.AllCaucasian

BMI,obesity:relativeBMI

>120%

(Rolland-Cachera


Extra-curricularphysical

activityandvigorousplay

(min=day)andtelevision

watchingmin=dayby

questionnairecompletedby

parents

withpaediatrician


multivariate

analysis,with

age,gender,dietary

variables,parentalBMI,

televisionviewingtimeand

vigorousactivityincludedas

covariates.Nosigni®cant

effects

oftelevisionviewing

timeorvigorousplay=activity

inanyanalysis

Moore

etal,1671995,

USA

(Framingham

OffspringStudy)

4y,5y,6y,duration

2.5y,n�106baseline,

n�97at6y


invitedonbasis

of

geographicalarea,good

health,andboth

parents

athome

BMI,triceps,subscapular

skinfolds

Physicalactivitymeasured

twiceeachyearbyCaltrac

accelerometerfor2�5days,

6mo.apart

Changein

tricepsskinfold

overstudyperiod:active

girls

�1.0mm,inactivegirls

�1.75mm

activeboys

ÿ0.75mm,inactiveboys

�0.25mm.Oddsratios

calculatedforrelativeriskof

increasinganthropometry

slopewithlow

levels

of

activity(below

median).For

inactivechildren,riskwas

greaterforincreasingtriceps

skinfold

OR�2.6,

subscapularskinfold

OR�1.3,orBMIOR�1.5,

non-signi®cant.Logistic

regressionanalysis:after

adjustingforage,hours

of

televisionwatchedperday,

energyintakeandBMIof

parents,loweractivity

increasesriskofincreasing

skinfold

slopeforleaner

subjects

(triceps�median)

OR�2.9,CI0.6±13.1,and

fattersubjects

(triceps>median)OR�5.8,

CI1.1±31.3

TableA9Continued


S85

Table

A9

Continued


Author=year=country

Agesatmeasurements=

no.ofsubjects

Design=participants

Outcome(m

easuredunless

statedotherw

ise)

Activity

risk

factor=method

ofmeasurement

Mainfinding

Parizkova,1771968,

CzechRepublic

11yannualuntil15y,

duration4y,n�146


Prospective.Nodetails

exceptn�146selectedfrom

largernumberofsubjects.

Sim

ilarSES,allattended

sim

ilarstate

schools

%bodyfatand%

bodylean

byunderw

aterweighing

Record

ofactivityin

sport

clubsandquestionnaires

concerningtimeandintensity

inunorganizedactivities.

Fourgroupsbasedon

consistentphysicalactivity

Statisticaltests

notstated,

probably

t-tests.Atbaseline

%leansim

ilaracrossactivity

groups.At15y,allthree

more

activegroupshigher%

leanthanleastactivegroup.

Overstudyperiod,%

lean

increasedsigni®cantlyin

mostactivegroupbutno

changein

leastactivegroup.

At15y,%

bodyfatlowerin

mostactivegroup

Robinson,1781993,

USA

12.4y,13.4y,14.4y;duration

2y,n�536baseline,n�279

at14.4y


the

randomly

allocatedcontrol

groupofall6th

and7th

grade

girls

from

fourschools

unless

withdrawn=refused

DBMI,Dtricepsskinfold

both

adjustedforsexualmaturity.

Obesity:BMI>85th

percentile

ortriceps>85th

percentile,adjustedfor

sexualmaturity

Televisionwatching

hours=dayandphysical

activitybyquestionnaireto

children

Spearm

an's

rankcorrelation

analyses:televisionviewing

wasnotrelatedto

sexual

maturity

adjustedDBMI

(r�0.03,P�0.62)ortriceps

skinfold

(r�0.03,P�0.54).

Multiple

linearregression:

afteradjustingforage,race,

parenteducation,parent

fatnessandbaselinephysical

activity,televisionviewing

wasnotrelatedto

sexual

maturity,adjustedDBMI

(b�0.05,P�0.82)orD

tricepsskinfold

(b�

70.2,

P�0.67).Resultswere

unchangedifanalysedwith

obesityasdependent

variable

inlogisticregression

Shapiroetal,1791984,

USA

6mo.,1,2,3,4,6y,9y;

duration8.5y,n�450


(83m,87f);analyses

onn�149

Prospective.Sample

selected

from

1193birth

records®led

in1969.Parents

whocould

belocatedat6mo.invitedto

participate

ifplanningto

bein

areafor3y

Sum

oftriceps,subscapular,

suprailiacandchestskinfolds

Physicalactivityassessedat

eachtimepointby1day

records,completedby

parents

at6mo.±

6yand

childrenat9y

Sum

ofactivityscore

over

studyperiodnegatively

relatedto

sum

ofskinfoldsat

9y,r�

70.18,P<0.05


S86

Table

A10

Dietary

papers


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcome(m

easured

unless

statedotherw

ise)

Activity

risk

factor=

methodofmeasurement

Mainfinding

Infantfeeding

Agrasetal,731990,

USA

2,4wks,12mo.,20mo.,

2y;duration2y,n�99

baseline,n�54at6y


recruitedfrom

hospitalin

®rstweekoflife


Suckingbehaviour

(measuredin

laboratory)Ð

caloricintakeduringsucking,

activefeedingtime,sucking

pressure,interburstinterval,

no.feeds=day,durationof

breastfeeding,introduction

solidfoods

Univariate

analyses:at3y,

solidfooddelayed

>5mo.,

breast-feeding

>5mo.,

suckingpressure

all

positively

relatedto

logBMI.

At6y,breast-feeding

>5mo.

positively

related,solidfood,

suckingpressure,no.

feedings=dayunrelatedto

log

BMI.Multivariate

analyses:

afteradjustm

entforseveral

factors

includingparent

educationandBMI:at3y,

suckingpressure

andsolid

foodsigni®cantlypositively

relatedto

logBMI.At6y,

breast-feeding

>5mo.

positively

relatedto

logBMI.

At6y,norelationships

betw

eenfeedingbehaviour

andtricepsskinfold

Agrasetal,941987,

USA

2,4wks,12mo.,20mo.,2,

3y,6y;duration6y,n�99

baseline,n�79at2y

Asabove

Asabove

Asabove

Inmultiple

regression

models

adjustingforseveral

factors,includingparent

educationandBMI,shorter

interburstintervalwas


tricepsskinfold

at1y

(P<0.02).Pressure

of

suckling(P

<0.006)andno.

offeedsperday(P

<0.01)

were

positively

associated

withtricepsskinfold

at2y.

Numberoffeedsperdaywas

positively

associatedwith

(log)BMIat1y.No


for(log)

BMIat2y.`Themore

adipose

infantat1and2ysucked

more

rapidly,athigher

pressure,withlongersuck

andburstdurationand

shorterburstinterval.'

Associatedwithhigher

caloricintake TableA10

Continued


S87

Table

A10

Continued

Dietary

papers


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcome(m

easured

unless

statedotherw

ise)

Activity

risk

factor=

methodofmeasurement

Mainfinding

Birkbecketal,2021985,

New

Zealand

Dunedin

Multidisciplinary

HealthandDevelopment

Study(D

MHDS)

0,3y,7y;duration

7y,n>2000baseline,

n�565at7y


derivedfrom

acohort

of

>2000consecutivebirthsat

Dunedin's

only

maternity

hospitalin

1972±1973

Tricepsskinfold,subscapular

skinfold,BMI

Exclusivebreastvsform

ula

feedingforatleast®rst

12wks

Studentt-testusedto

compare

breastfedand

form

ula

fedgroups.At7y,no

differencein

tricepsor

subscapularskinfold

orBMI

Deweyetal,2031993,

USA

Anthropometry:monthly

from

1±18mo.,

�21,24mo.

diet:every

3mo.from

3±

18mo.;duration2y,n�?

baseline,n�75at1y(41

breast-fed,34form

ula

fed)

Prospective.Strati®ed

matchingforbreastand

form

ula

groupsforseveral

factors

includingSES.

Recruitmentreferto

refs.15±17

Weight-for-length

Z-scores,

skinfold

thicknesses,

estimated%

bodyfatfrom

wgt,hgt,ageandsex

(predictionequation)

Breastorform

ula

feeding,

energyintakeby4day

weighedrecords,breast

milkincluded

ANOVA:weight-for-length

signi®cantlygreaterfor

form

ula-fedinfants

betw

een

7and24mo.Tricepsskinfold

signi®cantlygreaterin

form

ula-fedinfants

at9±

24mo.,subscapularskinfold

greaterat3±18mo.and

24mo.,¯ankandquadriceps

skinfold

greaterat9±16and

10±13mo.respectively.

Bicepsskinfold

signi®cantly

greaterin

form

ula-fedinfants

at13mo.,greaterin

breast-

fedinfants

at21and24mo.

Sum

skinfolds(®vesites)

greaterin

form

ula-fedinfants

at9±17mo.Estimated%

bodyfatsigni®cantlygreater

inform

ula-fedinfants

betw

een5and24mo.

Multiple

regressionmodel

includinggender,birth

weight,parentalbodysizeas

covariates.Norelationships

betw

eenenergyintake,

feedingmodeorintroduction

ofsolidfoods,andsum

skinfold

thicknessesat2y

Heinig

etal,2041993,

USA

Asabove,duration

18mo.,n�44breast-fed

Asabove

Weight-for-length

Z-scores

Tim

eofintroductionof

solidfoods

Within

breast-fedgroup,

weight-for-length

Z-scores

sim

ilarthroughout®rst

18mo.forinfants

who

receivedsolidfoodbefore

26wksandthosereceived

solidfoodat26�

weeks(not

enoughform

ula

infants

ineachgroupforanalysis)

TableA10

Continued


S88

Table

A10

Continued

Dietary

papers


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcome(m

easured

unless

statedotherw

ise)

Activity

risk

factor=

methodofmeasurement

Mainfinding

Dineetal,2051979,

USA

0,3,6,12mo.,2,3,4,5y,

duration�5y,n�620


withcomplete

data


throughpaediatric

practice.

Birth

weight�5.5lb.W

hite,

mid-uppersocialclass

Weight=height,

height=weight1=3,

weight=height2

Breastvsbottle

feeding

Student'st-testusedto

compare

breastandbottle

fedgroups.Nodifferencein

weight=height,ponderal

indexorBMIbetw

eenbreast

andbottle-fedinfants

over

®rst5yoflife

(P>0.1;breast-

fedn�70,bottle-fedn�

406)

Krameretal,741986,

Canada

1±3days,2,4,6wks,2,3,

4mo.,1,2y;duration2y,


at2y


from

largeuniversity


sociallydiverse


462enrolled

BMI,sum

triceps

�subscapular�suprailiac

skinfolds

Ageatintroductionofsolids,

durationofbreastfeeding

Multiple

regressionmodels,

stepwiseprocedure,

includingsex,birth

weight,

parentalrelativeweights,

socialclass,behaviourand

dietary

variablesas

covariates.Durationofbreast

feedingnegativepredictorof

BMIat1y(b

�7

0.21,

P<0.0001),2y(b

�7

0.11,

P�0.008),andofskinfoldsat

2y

(b�

70.16,

P�0.003).

Ageatintroductionofsolids

negativepredictorof

skinfoldsat1y(b

�7

0.14,

P�0.009)


Asabove

Asabove

Asabove

Asabove

Asabove


Asabove

Asabove

Asabove

Asabove

Asabove

Marm

otetal,2111980,UK

1946UK

birth

cohort

2mo.,2y,32y,duration

32y,n�238baseline,

n�172at32y

Prospective.Sample

selected

from

UK1946birth

cohort

on

basis

ofinfantfeedingand

geographicalarea

BMI,tricepsskinfold

Solely

breast-fedfor®rst

5mo.vsnotbreast-fedatall

for®rst5mo.(bottle-fed)

Statisticaltestnotstated,test

tocompare

groupsused.

Men:tricepsskinfold

greater

inbreast-fedgroup(15.0

vs

12.0mm,P�0.06),BMI

greaterin

breast-fedgroup,

nosummary

statisticsgiven.

Women:tricepsgreaterin

breast-fedgroup(24.1

vs

22.5mm)butnot

signi®cantly,Pvalue

notgiven

TableA10

Continued


S89

Table

A10

Continued

Dietary

papers


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcome(m

easured

unless

statedotherw

ise)

Activity

risk

factor=

methodofmeasurement

Mainfinding


Australia

MUSP(M

asterUniversity

StudyofPregnancy

0y,5y,duration5y,n�7357

baseline,n�4062at5y(2133

f,1929m)

Prospective.Mothers



(ref.20).Atbirth

n�7357,


followedupdueto

insuf®cientfunds

Obesity:severe

obesityÐ

BMI>94th

percentile;

moderate

obesityÐ

BMI85±

94th

percentile

within

cohort

Durationofbreastfeeding:

>6mo.;4±6mo.;7wks±

3mo.;3±6wks;<2wks;

notatall

Univariate

analyses:RR,95%

CIs,w2testforsigni®cance.

Durationofbreastfeedingno

in¯uenceonmoderate

obesity.RRforsevere

obesity

generallyhigherforshorter

durationofbreastfeeding.

Only

breastfeedingfor

7wks±3mo.showed

signi®cantlygreaterRRfor

severe

overw

eight

(comparedwithbreast

feeding

>6mo.),RR�1.6,CI

1.0±2.4,P<0.05.

Multivariate

analyses

includingbirth

weight,

gender,gestationalage,

feedingproblemsand

sleeplessness,adjustedORs

calculatedwith95%

CIs.No

in¯uenceofdurationof

breastfeedingonmoderate

orsevere

obesity

Poskitt,2061977,

UK

0±1y,5y;duration

5y,n�300

baseline,n�203

at5y


from

welfare

clinics,birth

weight�2.5kg

%expectedrelativeweight

(Tanner1996usedas

referencedata)

Energyintake,ageof

weaning

Nosigni®cantcorrelations

betw

eenageofweaningor

infancyenergyexpenditure

and%

expectedweightat5y

Sveger,2071978,

Sweden

0±1y,4y;

duration�4y,n�243

baseline,n�226f-up

Prospective.Forrecruitment

detailsreferto

refs.5,7

Relativeweight

Overfeedingduring®rstyear,

de®nedas�2s.d.mean

dailycalorieintakeofinfants

sameageandsex

26of243childrenoverfedat

�1offourmeasurements

during®rstyear.24ofthese

26were

followedupat4y,

nonewasobeseor

overw

eight.In

childrennot

overfedduring®rstyear,2%

of202followedupwere

obeseand5%

overw

eight.

Authors

concludethat

overfeedingis

notapredictor

foroverw

eightorobeseat4y

TableA10

Continued


S90

Table

A10

Continued

Dietary

papers


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcome(m

easured

unless

statedotherw

ise)

Activity

risk

factor=

methodofmeasurement

Mainfinding

Vobeckyetal,2081983,

Canada

1,2,3,4,5,6mo.thenevery

3mo.until3y;n�556

baseline,n�170selected

Prospective.Childrenwith

complete

data

from

larger

cohort

(seerefs.6±9)Mostly

highersocialclasses

Relativeweight(w

eightfor

heightandage),norm

aldata

from

largercohort

(refs.

6±10)

Breastfeeding

>2wks

vs bottle

feeding

Nodifferencein

relative

weightdistributionsbetw

een

breastandbottle

fedinfants

at1,2,or3y.At1,2,and3y±

males:relativeweightof

breastfed>bottle

fed;

females:relativeweightof

bottle

fed>breastfed;all

differencesnotsigni®cant

Wellsetal,2091998,

UK

12wks,2±3.5y;duration2±

3.5y;n�50

baseline,n�20

at

2±3.5y


recruitedfrom

large

maternityhospital

%fat(deuterium

dilution),

meanZ-score

oftricepsand

subscapularskinfolds

Weaningstatusat12wks,

milkvolumeintake(M

VI)

measuredbydeuterium

turnover,energyintake(EI)

Student'st-testusedto

compare

groups,those

weanedat12wkshadlower

skinfold

Zscore

at2±3.5y

thanthoseunweaned,1.20

vs7

0.54,P<0.05.Multiple

regressionanalyses:after

adjustingforinfantskinfolds,

weanedinfants

hadlower

skinfoldsthanunweaned,

b�

70.57,P�0.06.Results

sim

ilarifMVIorEIaddedto

model.MVIorEInot

signi®cantin

thesemodels.

Afteradjustingforinfant

skinfoldsandincluding

weaningstatusand

soothability,weaningnot

signi®cant,soothability

negatively

relatedto

skinfolds(seeBehaviour

section)

Wilsonetal,2101998,

UK

Dundeeinfant

feedingstudy

0y,regularlyuntil2y,7.2y;

duration7y,n�674


Prospective.Forrecruitment

detailsreferto

ref.3

BMI(n

�412),%

fatby

impedance(n

�397),

skinfoldsatfoursites

(n�405)

Exclusivebreast-feedingfor

15wksvspartialbreast-

feedingvsexclusivebottle

feeding.Solidsintroduced

before

15wksor15�

wks

Generallinearmodel:after

adjustm

entforsex,birth

weight,weightat®rstsolid

feed,%

bodyfatby

impedancesigni®cantly

greaterin

thosegivensolids

before

15wks(18.5%

vs

16.5%).Resultsforskinfolds

sim

ilar(P

<0.05)BMInot

signi®cantlydifferentin

solidsbefore

orafter15wks

TableA10

Continued


S91

Table

A10

Continued

Dietary

papers


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcome(m

easured

unless

statedotherw

ise)

Activity

risk

factor=

methodofmeasurement

Mainfinding

Ziveetal,1301992,USA

SCAN

study

4y,n�270,�50%

m,

50%

fRetrospective.Recruitment

via

state

fundedpre-schools,

children's

centres.Low

SES

familiestargettedforstudy.

Anglo-andMexican

Americans.331respondents,

complete

data

for270

BMI,sum

of


skinfold

Durationofbreast-feeding,

ageofintroductionto

form

ula,agewhencow's

milkorform

ula

fedmore

thanbreastmilk,agesolid

foodintroduced

Correlations:nosigni®cant

associationsbetw

een

nutritionvariablesandBMI

orskinfolds.Multiple

regression:in

modelwhich

appearedto

includeall

dietary

riskfactors,dietary

riskfactors

did

notcontribute

signi®cantlyto

BMIor

skinfolds

Childhooddietary

intake

(i)Childhoodto

adulthood

Twisketal,1721997,

TheNetherlands

Amsterdam

Growth

and

HealthStudy(A

GHS)

13,14,15,16,21,27,29y;

duration16y;n�233


(98f,83m)


complete

1stand2ndform

of

Amsterdam

school,no

refusals

toparticipate.SES


Sum

ofbiceps,triceps,


skinfolds

Dietary

intakebycross-check

dietary

history

forpast

month.Fat,protein,

carbohydrate

expressedas

%energyintake

Multiple

regressionanalysis:

afteradjustingforgender,%

fatintakeduringadolescence

(13±16y)wasnegatively

relatedto

skinfolds

b�

70.15,P�0.04.After

includingvariablesbasedon

signi®cancein

univariate

analyses,norelationships

betw

eendietandfatness

Postetal,2131997,

TheNetherlands

AGHS

13,14,15,16,21,27y;

duration14y,n�233


(98f,84m)

Asabove

Dsum

ofbiceps,triceps,


skinfolds

Methodasabove.Anim

al

protein,saturatedfat,poly-

unsaturatedfat,and

cholesterolexpressedas

%ofenergyintake

Longitudinalregression

analysis

(correctiononly

for

gender)

by`general

estimatingequations'

showednegativerelationship

betw

eenchangein

energy

intakeandDsum

ofskinfolds

(P�0.005)andno

relationship

betw

een%

energyofanim

alprotein,

saturatedfat,

polyunsaturatedfator

cholesterolandDsum

of

skinfolds

TableA10

Continued


S92

Table

A10

Continued

Dietary

papers


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcome(m

easured

unless

statedotherw

ise)

Activity

risk

factor=

methodofmeasurement

Mainfinding

Twisketal,1731997,The

Netherlands

AGHS

13,14,15,16,21,27y;

duration14y,n�233


(98f,83m)

Asabove

Sum

ofbiceps,triceps,


skinfolds

Asabove

Statisticalmodel

incorporatingparameters

calculatedusinggeneralized

estimatingequations.After

adjustingforbiologicalage,

sex,sum

ofskinfolds,

VO

2max,highriskforsum

skinfolds(�20%

bodyfatfor

males,30%

forfemales),was

positively

relatedto

protein

intake,OR�1.5,P<0.01.No

relationship

betw

eenriskof

skinfoldsandfator

carbohydrate

intake


TheNetherlands

AGHS

13,14,15,16,21,27y;

duration14y,n�233


(98f,84m)

Asabove


Methodasabove.Fat,

protein,carbohydrate

expressedas%

energy

intake

Using`generalestimating

equations'(longitudinal

regressionanalysis),

adjustingforsum

skinfolds

betw

een13and27y:in

males,noin¯uenceofdietary

intakeonDsubscapular

skinfold.In

females,energy

intakenegatively

relatedto

D

subscapularskinfold,

b�

70.25,CI7

0.31,

70.19,P<0.001,%

energy

from

carbohydrate

positively

related

toD

subscapular

skinfold,b�0.09,CI0.02,

0.16,P<0.001.Using

multiple

regressionanalysis

withsubscapularskinfold

at

27yasdependentvariable

andinitiallyincluding

activity,diet,smokingand

alcoholintakeascovariates

(non-signi®cantcovariates

removedbybackwards

eliminationprocedure)no

in¯uenceofdietary

factors

onskinfold

inmalesor

females

TableA10

Continued


S93

Table

A10

Continued

Dietary

papers


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcome(m

easured

unless

statedotherw

ise)

Activity

risk

factor=

methodofmeasurement

Mainfinding

(ii)W

ithin

childhood

Deheegeretal,2141996,

France

10mo.,2,4,6,8y,duration

7y;n�278baseline,

n�112at8y


examinedin

twopublic

healthcentresin

Paris.No

furtherinform

ationon

subjectgroup

BMI

Energyintakebydiethistory

ANOVA:increasein

energy

intakebetw

een4and6ywas

signi®cantlygreaterin

the

groupwhowere

fat(highest

tertileofBMI)at8y

comparedto

groupswho

were

medium

orlean(m

idandlowertertileofBMI)at8y

(P�0.01).Energyintakes

before

4yorafter6ynot

predictiveofBMItertileat8y

Rolland-Cachera

etal,215

1995,France

2y,8y;duration6y;

n�278baseline,

n�112at8y

Asabove

BMI,tricepsskinfold,

subscapularskinfold

Dietary

intakebydietary

history:energyintake,%

energyfrom

protein,fat,

carbohydrate

Energyintakeat2ypositively

correlatedwithBMIat

8yr�0.20,P�0.049,not

signi®cantlywithskinfolds.

Afteradjustingforenergy

intakeorBMIat2y,no

associationbetw

een%

energyfrom

fat,

carbohydrate

orprotein

and

BMIorskinfoldsat8y.After

adjustm

entforparentalBMI,

%energyfrom

protein

relatedto

BMI,r�0.22,

P�0.03,andsubscapular

skinfold

r�0.20,P�0.04,not

tricepsskinfold.No

relationshipsbetw

een%

energyfrom

fator

carbohydrate

andBMIor

skinfolds.Afteradjustm

ent

forparentalBMI,energy

intakeandBMIat2y,%

energyfrom

protein

related

toBMIr�0.28,P�0.008.

Afteradjustm

entforSES,

both

energyintakeand%

energyfrom

protein

were

positively

relatedto

BMI,

P<0.05

TableA10

Continued


S94

Table

A10

Continued

Dietary

papers


Agesatmeasurements,

no.ofsubjects

Design=participants

Outcome(m

easured

unless

statedotherw

ise)

Activity

risk

factor=

methodofmeasurement

Mainfinding

Grif®thsetal,216

Lancet1990

3±4y,�15y;duration

�11y;n�37baseline;

n�25at15y(15m,10f)


selectedfrom

obeseand

norm

alweightparents

BMI,BFMI(fatmass=hgt2)

Methodforenergyintakein

childhoodmeasurementnot

givenÐ

seerefs.3,4

Ingirls,energyintake=kg

bodyweightat3±4y

signi®cantlynegatively

correlatedto

BMIandBFMIat

15y(r�

70.73,P<0.0118,

r�

70.77,P<0.009

respectively).

No

association

inboys

Klesgesetal,451995,

USA

4.4,5.4y,6.4y,duration

2y,n�203baseline,

n�146at6.4y


recruitedthrough

paediatricians,day-care

centres,churches.Obese

oversampled

DBMI

%kcalfat,%

kcal

carbohydrate

measuredby

foodfrequencyquestionnaire

Multiple

regressionanalysis

includingadjustm

entfor


parentalfatnessand

interactionsbetw

eengender

andparentalfatness.In

modelincludingthese

covariatesandalsoactivity,

baseline%

kcalfrom

fat

positively

relatedto

2y

increasein

BMI(b

�0.03,

P�0.052).Changein

kcal

fatnegatively

relatedto

increasein

BMI(b

�7

0.04,

P�0.011),buttextdoesnot

agreewithtablesand

suggests

positive

relationship

Maffeis

etal,461998,

Italy

8.6y,12.6y;duration

4y,n�298baseline,

n�112at12.6y


recruitedthroughprivate

and

publicschools.AllCaucasian

BMI.Obesity:relativeBMI

>120%

(Rolland-Cachera


Energyintake,protein,

carbohydrate,fatintakeas%

energymeasuredbydietary

history

methodÐ

dietician

interviewedmothers

and

children,schoolmenus

reviewed


multivariate

analysis,with

age,gender,dietary

variables,parentalBMI,TV

viewingtimeandvigorous

activityincludedas

covariates.Nosigni®cant

effects

ofanydietary

variable

isanyanalysis

TableA10

Continued


S95

Table

A10

Continued

Dietary

papers


Agesatmeasurements=

no.ofsubjects

Design=participants

Outcome(m

easured

unless

statedotherw

ise)

Activity

risk

factor=

methodofmeasurement

Mainfinding

Nicklasetal,2171988,

USA


6.5y,n�440baseline,n�50

at7ywithcomplete

data


an18month

birth

cohort

from

semirural,biracial

population.n�50

participatedin

allsix

screenings

Subscapularskinfold.Rohrer

index(w

gt=hgt3)

%energyfrom

protein,

carbohydrate,fatmeasured

by24hrecall

Changein

totalprotein,fat,

starchandenergyfrom

6mo.

to4ysigni®cantlypositively

correlatedwithchangein

subscapularskinfold

(same

period),P<0.05.After

adjustm

entforenergyor

bodyweight,relationships

notsigni®cant.Using

statisticaltestto

compare

groups(unde®ned),those

withconsistentlyhighintakes

comparedwiththosewith

consistentlylowerintakesof

%energyfrom

anim

alfat,

totalfatandprotein

had

sim

ilarRohrerindexesat7y,

P>0.05

Shapiroetal,1791984,

USA

6mo.,1,2,3,4,6y,9y;

duration8.5y,n�450

baseline,n�149at9ywith

complete

data

Prospective.Sample

selected

from

1193birth

records®led

in1969.Parents

whocould

belocatedat6mo.were

invitedto

participate

ifgoing

tobein

areafor3y.Initial

participants

71%

White,19%

Black,10%

others

Sum

oftriceps,subscapular,

suprailiacandchestskinfolds

Energyintakeassessedat

eachtimepointby3day

dietary

records,completed

byparents

at6mo.±

6yand

childrenat9y

Sum

energyintakeover

studyperiodnegatively

correlatedto

sum

skinfoldsat

9y,r�

70.16,P<0.05

Sheaetal,2181993,

USA

3±4y,�severalfollow-up

measures,4±6y;duration

25mo.,n�238baseline,

n�215at5ywith

complete

data


recruitedthroughpaediatric

practice.Only

onechildper

familyeligible.

PredominantlyHispanic,

low-incomedistrict

Rate

ofchangein

BMI

(kg=m

2pery)

Dietary

intakebysemi-

quantitativequestionnaire

(averageofthreeduring®rst

year)and24hrecall(average

offourin

®rstyear)

ANOVA:nosigni®cant

(P<0.05)differencesin

rate

ofchangein

BMIacross

quintilesoftotalfatdensity,

saturatedfatdensityor

cholesterol(m

gper

1000kcal).Nodifferencein

rate

ofchangein

BMI

betw

eenchildrenconsuming

lessthan30%

orcalories

from

totalfatcomparedto

thoseconsuming30%

or

more


S96

Table

A11

Behaviouralandpsychologicalfactors

papers


Ageatpredictorandoutcome


Design=participants

Outcomemeasure

Riskfactor

Mainfinding

i)Childhoodto

adulthood


TheNetherlands

Amsterdam

Growth

and

HealthStudy(A

GHS)

13y(14,15,16,21y),27y;

duration14y;n�233


(83m,98f)


complete

1stand2ndform

ofAmsterdam

school,no

refusals

toparticipate.SES


Sum

ofbiceps,triceps,

subscapularandsuprailiac

skinfolds

Personality

traitsassessed

byquestionnaire(atallsix

time-points

duringstudy).

Six

traits:inadequacy,social

inadequacy,dominance,

rigidity,achievement

motivation,debilitating

anxiety

Generalizedestimating

equations(takinginto

accountdata

from

alltime-

points):inadequacyat13±

16yrelatedto

increasein

sum

skinfolds,b�0.08,95%

CI0.02,0.14,P<0.01.No

relationship

betw

een

inadequacyat13±21yor

13±27yoranyothertraitat

anyperiodandsum

skinfolds

Lissauetal,1311993,

Denmark


10y;n�1258atbaseline,

n�552at20±21y


arandom

sample,25%

of3rd

graders

inCopenhagen

BMI(reported)

overw

eight�90th

percentile

BMI,obesity�95th

percentile

Variablesrelatingto

sweet

eatinghabitsandmother's

knowledgeofsuchhabits

AftercontrollingforBMIin

childhoodandgender,odds

ratiosforbeingoverw

eightin

youngadulthoodwere

2.6

(P�0.02)formothers

acceptanceofsweeteating

habits,2.0

(P�0.04)for

providingmore

sweet

money,2.6

(P�0.04)for

mother'slackingknowledge

aboutoffspring's

sweet-

eatinghabits.After

controllingforsocialfactors,

only

mother'slacking

knowledgeremained

signi®cant,OR�4.5,

P�0.003

ii)W

ithin

childhood

Agrasetal,941987,

USA

2,4wks,2y;duration2y,

n�99baseline,n�79at2y


recruitedfrom

hospitalin

®rstweekoflife


Suckingbehaviour

(measuredin

laboratory)Ð

caloricintakeduringsucking,

activefeedingtime,sucking

pressure,interburstinterval,

no.feeds=day

Stepwisemultiple

regression

models

usingseveralfactors:

highersuckingpressure

(P<0.006)andfewerfeeds

perday(P

<0.01)were


tricepsskinfold

at2y.No


for(log)

BMIat2y.`Themore

adipose

infantat1and2ysucked

more

rapidly,athigher

pressure,withlongersuck

andburstdurationand

shorterburstinterval.'This

feedingpattern

was

associatedwithhighercaloric

intake(referto

dietary

section

ofreview)

TableA11

Continued


S97

Table

A11

Continued


papers




Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Agrasetal,731990,

USA

2and4wks,3y,6y;duration

6y;n�99baseline,n�54

at6y

Asabove

Asabove

Asabove

Correlations:at3y,sucking

pressure

relatedto

log

BMI,r�0.33,P�0.05,no.

feeds=daynotrelatedto

log

BMI.At6y,suckingpressure,

no.feeds=dayunrelatedto

logBMI.Stepwisemultiple

regressionmodels

using

severalfactors:at3y,sucking

pressure

relatedto

logBMI,

partialr2�0.26,P�0.002.At

6y,noaspectoffeeding

behaviourrelatedto

logBMI.

Nosigni®cantrelationships

betw

eenfeedingbehaviour

andtricepsskinfold

Careyetal,2541988,

USA


unspeci®ed(1

±5y);n�?

baseline,n�138at8±9y

(67m,71f)

Prospective.Private

practice

setting,predominantly

middle

class

Weight-for-heightpercentile

from

UnitedStatesPublic

HealthServicedata

Behaviourasassessedby

behaviourstyle

questionnaire(BSQ).BSQ

consists

of100itemswhich

condense

tonine

categories

oftemperament:activity,

adaptability,distractibility,

intensity,mood,persistence=

attentionspan,rhythmicity=

predictability,threshold,

withdrawal

Correlations:decreased

adaptabilityrelatedto

increasein

wgt-for-

hgt,r�0.18,P�0.02.

Increasedintensityrelatedto

increasein

wgt-for-

hgt,r�0.19,P�0.01.

Increasedwithdrawalrelated

toincreasein

wgt-for-

hgt,r�0.24,P�0.002

Frenchetal,2551996,

USA



n�1278at15±18y(656

f,622m)

Prospective.All7±9th

graders

insuburbanschool

districtinvitedto

participate.

Ofstudents

inyear1who

remainedin

districtfor3y,

83%

femalesand85%

males

measuredatfollow-up.No

mentionofstudents

who

movedoutofdistrict

BMI

Self-esteem:global,physical

appearance,athletic,

scholastic,socialacceptance,

jobcompetence,romantic

appeal,behaviouralconduct,

closefriendship

Partialcorrelations:girlsÐ

physicalappearanceself-

esteem

atbaselinenegatively

relatedto

BMIatfollow-

up,r�

70.11,P<0.05;

socialacceptanceself-

esteem

atbaselinenegatively

relatedto

BMIatfollow-up,

r�

70.13,P<0.05.BoysÐ

norelationshipsbetw

een

self-esteem

andBMI

TableA11

Continued


S98

Table

A11

Continued


papers




Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Klesgesetal,2561992,

USA


3y;n�203baseline

(assumedÐ

inform

ation

from

ref.45)n�132at

6±9y(77m,55f)

Prospective.Families

participatingin

largerstudy

ofcardiovascularriskfactors

inchildren.W

hite,upper-

middle

class.Obesechildren

oversampled

BMI,tricepsskinfold;obesity:

�120%

idealbodyfatbased

ontricepsskinfold

(norm

al

data

from

USDepartmentof

Health,Educationand

Welfare)

PerceivedCompetenceScale

forChildren(PCPS)assessing

fourtypesofself-esteem:

cognitiveability,physical

ability,peerrelations,general

self-esteem.Family

Relationship

Index(FRI)

assessingfamilyfunction:

cohesion,expressiveness,

con¯ict,reportedseparately

bymotherandfather

Multiple

regressionanalyses.

Afteradjustingforinitial

bodyfatandgenderat1y

follow-up,physicalself-

esteem

(b�

71.03),father's

FRI(b

�7

0.35),physical

self-esteem

*gender

(b�0.65)andfather'sFRI*

gender(b

�0.23)were


body

fat(tricepsskinfold),P<0.05.

Thushigherphysicalesteem

andmore

positivefamily

situationsasviewedbythe

father,were

relatedto

greater

increasesin

bodyfatin

girls,

smallerincreasesin

boys.At

year2,relationshipswere

weaker,only

physicalself-

esteem

wasrelatedto

gain

inbodyfat(b

�7

0.83,

P�0.036)andatyear3,none

ofthepsychosocialvariables

signi®cantlypredictedbody

fatbasedontricepsskinfold.

Atyear3,mother'sFRI*

gendersigni®cantlyrelated

toBMI,asfamilyfunctioning

increased,BMIincreasedin

girls,decreasedin

boys

Krameretal,74,961986

and1985,Canada

1±3daysforIBH�MFA;

2wksforITQ,1y,2y;

duration2y;n�462



recruitedfrom

large

universityteachinghospital

servingsociallydiverse


462enrolled

BMI,sum

oftriceps�

subscapular�

suprailiac

skinfolds

ITQÐ

infanttemperament

questionnaire;IBHÐ

maternalpreconceptionof

idealinfantbodyhabitus;

MFAÐ

maternalfeeding

attitudes,allassessedby

questionnaire

Multiple

regressionmodels,

stepwiseprocedure,


skinfoldsat1and2y.

Maternalpreconceptionof

infantbodyhabituswas

positively

associatedwith

BMIat12mo.(b

�0.10,

P<0.045)in

modelincluding

birth

weight,durationof

breastfeeding,andgender.

Noassociationat2y.No

associationsbetw

een

maternalfeedingattitudeor

infanttemperamentandBMI

orskinfoldsat1or2y

TableA11

Continued


S99

Table

A11

Continued


papers




Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Morris

etal,2571984,

UK

10y,11y;duration1y;

n�282atbaseline,

n�230at11y

Childrenrecruitedfrom

10

schools

inoneLondonarea.

Totaleligible

n�408

Dweight±heightindex

(basedonlog

transform

ation),Dskinfold

index(basedonlog

transform

ationofcombined


skinfolds)

Attitudestowardsnutrition,

knowledgeofnutrition

Betterknowledgeofnutrition

wasassociatedwithgreater

decreasein

skinfold

(7

0.6mm,P<0.05),in

boys,

notgirls.Identifying

nutritiousmeals

was


decreasein

weightforheight

index(7

0.3kg,P<0.05)in

girls

notboys.No

associationsbetw

een

positiveattitudeto

good

nutrition,preferringor

identifyingfattening

puddings,orpreferring

nutritiousmeals

andobesity

indexes


Australia

MUSP(M

aterUniversity

StudyofPregnancy)

0y,5y;duration5y;n�7357

atbaseline,n�4062at5y

(2133f,1929m)

Prospective.Mothers



(ref.20).Atbirth

n�7357,


followedupdueto

insuf®cientfunds

BMI,severe

obesity:BMI

>94th

percentile;moderate

obesity:BMI85±94th

percentile

within

cohort

Feedingproblems

(unde®ned),sleeplessness

Univariate

analyses:RR,95%

CIs.FeedingproblemsÐ

no

in¯uenceonmoderate

or

severe

obesity.

SleeplessnessÐ

alm

ost

signi®cantin¯uenceof

sleeplessnessfew

times=week(notmore

or

less)onmoderate

obesity

only,RR�1.2,CI0.9±1.7.

Multivariate

analyses

includingseveralfactors,

adjustedORscalculatedwith

95%

CIs.Feeding

problemsÐ

protectiveeffect

onsevere

obesityonly,

OR�0.3,CI0.1±0.8.

SleeplessnessÐ

signi®cant

in¯uenceofsleeplessness

few

times=week(notmore

or

less)onmoderate

obesity

only,OR�1.7,CI1.1±2.6

TableA11

Continued


S100

Table

A11

Continued


papers




Design=participants

Outcomemeasure

Riskfactor

Mainfinding

Wellsetal,2461997,

UK

12wks,2±3.5y;duration

2±3.5y;baselinen�50,

f-upn�25±30(depending

onoutcome)


recruitedfrom

large

maternityhospital

Sum

oftriceps�subscapular

skinfolds;%

bodyfatbyTBW

method

Infanttemperamentby

questionnaireto

mother:

fussiness;crying;distressto

limitations;soothability

Correlation:infantdistressto

limitationswaspositively

correlatedto

%bodyfatat2±

3.5y(r�0.4,P<0.05)butno

associationwithskinfolds.

Infantsoothabilitywas

negatively

correlatedto

sum

ofskinfolds(r�

70.66,

P<0.001),butnoassociation

with%

bodyfat.Multiple

regression:afteradjustm

ent

forinfant%

fatandchildhood

age,infantdistressto

limitationswaspositively

associatedwithchildhood%

bodyfat(b

�1.96,P�0.059,

borderlinesigni®cance).After

adjustm

entforinfantsum

of

skinfoldsand`awakeand

active'in

childhood,infant

soothabilitywasnegatively

associatedwithchildhood

sum

ofskinfolds(b

�7

1.65,

P�0.012)

Wellsetal,2091998,

UK

12wks,2±3.5y;duration

2±3.5y;n�50baseline,

n�20at2±3.5y

Asabove

%fat(deuterium

dilution);

meanZ-score

oftricepsand

subscapularskinfolds

Soothability

Usingmultiple

regression

analyses:afteradjustingfor

infantskinfoldsandincluding

weaningstatusand

soothability,weaningnot

signi®cant,soothability

negatively

relatedto

skinfoldsb�

70.41,P�0.01


S101

Appendix II

Search strategy: Medline (Ovid software interface),

1966 ±April 1998

1. exp Child=2. exp child nutrition=3. obes$.tw.4. adipos$.tw.5. ponderos$.tw.6. bodyfat$.tw.7. body fat$.tw.8. (fat adj3 mass).tw.9. fatness$.tw.10. subcutaneous fat.tw.11. body mass.tw.12. body masses.tw.13. bmi.tw.14. underweight.tw.15. overweight.tw.16. body size$.tw.17. body composition.tw.18. quetelet.tw.19. ponderal.tw.20. anthropomet$.tw.21. skinfold$.tw.22. skin fold$.tw.23. lean$.tw.24. thinness.tw.25. waist$.tw.26. exp obesity=27. exp skinfold thickness=28. exp body mass index=29. exp body weight=30. exp body weight changes=31. exp body composition=32. prospectiv$.tw.33. longitudinal$.tw.34. followup$.tw.35. (follow adj up$).tw.36. long term.tw.37. longterm.tw.38. track$.tw.39. serial$.tw.40. cohort$.tw.41. reexamin$.tw.42. re examin$.tw.43. reassess$.tw.44. re assess$.tw.45. restud$.tw.46. re stud$.tw.47. reevaluat$.tw.48. re evaluat$.tw.49. reinvestigat$.tw.50. re investigat$.tw.51. rereview$.tw.52. re review$.tw.53. remeasure$.tw.54. re measure$.tw.55. retrospectiv$.tw.

56. growth study.tw.57. case control$.tw.58. case?control$.tw.59. exp case control studies=60. exp cohort studies=61. epidemiologic methods=62. 1 or 263. or=3-3164. or=57-5965. or=26-3166. 64 and 6567. or=32-56,60-61,6668. 62 and 63 and 6769. Amsterdam Growth.tw,in.70. atherosclerosis risk in communities.tw.71. baltimore longitudinal study.tw.72. belgian luxembourg child.tw.73. birth cohort.tw.74. bogalusa heart study.tw.75. california child health.tw.76. cardia study.tw.77. cardiovascular health study.tw,in.78. (christchurch adj3 development).tw,in.79. darling.tw.80. dormont high.tw.81. dunedin multidisciplinary.tw,in.82. dunedin multi disciplinary.tw,in.83. fels.tw.84. framingham.tw.85. hhanes$.tw.86. kabi pharmacia.tw.87. leuven growth.tw.88. minneapolis children$ blood.tw.89. muscatine.tw.90. (national health adj3 examination adj3

survey$).tw,in.91. nhanes$.tw.92. nhlbi.tw.93. nurses health study.tw.94. princeton school district.tw.95. procam.tw.96. stanford five city.tw.97. tecumseh community.tw.98. Wroclaw growth study.tw,in.99. Wroclaw longitudinal twin study.tw,in.100. young finns.tw.101. youth health survey.tw,in.102. zurich growth.tw.103. or=69-102104. 103 and 63105. 68 or 104

Search strategy: Embase (Ovid software interface),

1980 ±May 1998

1. child=2. exp child=3. exp adolescent=4. exp newborn=5. exp adolescence=


S102

6. exp childhood=7. exp child nutrition=8. obes$.tw.9. adipos$.tw.10. ponderos$.tw.11. bodyfat$.tw.12. body fat$.tw.13. (fat adj3 mass).tw.14. fatness$.tw.15. subcutaneous fat.tw.16. body mass.tw.17. body masses.tw.18. bmi.tw.19. underweight.tw.20. overweight.tw.21. body size$.tw.22. body composition.tw.23. quetelet.tw.24. ponderal.tw.25. anthropomet$.tw.26. skinfold$.tw.27. skin fold$.tw.28. lean$.tw.29. thinness.tw.30. waist$.tw.31. exp obesity=32. exp skinfold thickness=33. exp body mass=34. exp body size=35. exp body weight=36. exp body composition=37. body fat=38. subcutaneous fat=39. prospectiv$.tw.40. longitudinal$.tw.41. followup$.tw.42. (follow adj up$).tw.43. long term.tw.44. longterm.tw.45. track$.tw.46. serial$.tw.47. cohort$.tw.48. reexamin$.tw.49. re examin$.tw.50. reassess.tw.51. re assess.tw.52. restud$.tw.53. re stud$.tw.54. reevaluat$.tw.55. re evaluat$.tw.56. reinvestigat$.tw.57. re investigat$.tw.58. rereview$.tw.59. re review$.tw.60. remeasure$.tw.61. re measure$.tw.62. retrospectiv$.tw.63. growth study.tw.64. exp cohort analysis=65. exp clinical study=

66. exp epidemiology=67. case control$.tw.68. case?control$.tw.69. or=1-770. or=8-3871. or=39-6872. 69 and 70 and 7173. Amsterdam Growth.tw,in.74. baltimore longitudinal study.tw.75. belgian luxembourg child.tw.76. birth cohort.tw.77. bogalusa heart study.tw.78. california child health.tw.79. cardia study.tw.80. (christchurch adj3 development).tw,in.81. darling.tw.82. dormont high.tw.83. dunedin multidisciplinary.tw,in.84. dunedin multi disciplinary.tw,in.85. fels.tw.86. framingham.tw.87. hhanes$.tw.88. kabi pharmacia.tw.89. leuven growth.tw.90. minneapolis children$ blood.tw.91. muscatine.tw.92. (national health adj3 examination adj3

survey$).tw,in.93. nhanes$.tw.94. nhlbi.tw.95. nurses health study.tw.96. princeton school district.tw.97. procam.tw.98. stanford five city.tw.99. tecumseh community.tw.100. young finns.tw.101. zurich growth.tw.102. atherosclerosis risk in communities.tw.103. cardiovascular health study.tw,in.104. youth health survey.tw,in.105. Wroclaw growth study.tw,in.106. Wroclaw longitudinal twin study.tw,in.107. or=73-106108. 107 and 70109. 72 or 108

Search strategy: CAB Abstracts (Ovid software

interface), 1973 ±May 1998

1. exp adolescents= or exp children= or expdaughters= or exp infants= or exp sons=

2. obes$.ab,do,ti.3. adipos$.ab,do,ti.4. ponderos$.ab,do,ti.5. bodyfat$.ab,do,ti.6. body fat$.ab,do,ti.7. (fat adj3 mass).ab,do,ti.8. fatness$.ab,do,ti.9. subcutaneous fat.ab,do,ti.10. body mass.ab,do,ti.


S103

11. body masses.ab,do,ti.12. bmi.ab,do,ti.13. underweight.ab,do,ti.14. overweight.ab,do,ti.15. body size$.ab,do,ti.16. body composition.ab,do,ti.17. quetelet.ab,do,ti.18. ponderal.ab,do,ti.19. anthropomet$.ab,do,ti.20. skinfold$.ab,do,ti.21. skin fold$.ab,do,ti.22. lean$.ab,do,ti.23. thinness.ab,do,ti.24. waist$.ab,do,ti.25. weight?for?height.ab,do,ti.26. weight?height.ab,do,ti.27. exp body fat= or exp obesity= or exp overeating=

or exp overfeeding= or exp overweight= or expthinness= or exp weight reduction=

28. exp body density= or exp body weight=29. exp body composition=30. prospectiv$.ab,do,ti.31. longitudinal$.ab,do,ti.32. followup$.ab,do,ti.33. (follow adj up$).ab,do,ti.34. long term.ab,do,ti.35. longterm.ab,do,ti.36. track$.ab,do,ti.37. serial$.ab,do,ti.38. cohort$.ab,do,ti.39. reexamin$.ab,do,ti.40. re examin$.ab,do,ti.41. reassess$.ab,do,ti.42. re assess$.ab,do,ti.43. restud$.ab,do,ti.44. re stud$.ab,do,ti.45. reevaluat$.ab,do,ti.46. re evaluat$.ab,do,ti.47. reinvestigat$.ab,do,ti.48. re investigat$.ab,do,ti.49. rereview$.ab,do,ti.50. re review$.ab,do,ti.51. remeasur$.ab,do,ti.52. re measur$.ab,do,ti.53. retrospectiv$.ab,do,ti.54. growth study.ab,ti,do.55. exp aetiology= or exp disease surveys= or exp

epidemiological surveys= or epidemiology=56. case control$.ab,do,ti.57. 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11 or

12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20or 21 or 22 or 23 or 24 or 26 or 27 or 28 or 29

58. 30 or 31 or 32 or 33 or 34 or 35 or 36 or 37 or 38or 39 or 40 or 41 or 42 or 43 or 44 or 45 or 46 or47 or 48 or 50 or 51 or 52 or 53 or 54 or 55 or 56

59. 1 and 57 and 58

Search strategy: PsycLIT (Silverplatter software

interface), 1966 ±March 1998

1. explode `̀ Children''2. `̀ Childhood-'' in DE3. `̀Adolescence-'' in DE4. explode `̀Adolescents''5. explode `̀ College-Students''6. explode `̀ Infants''7. `̀ Preadolescents-'' in DE8. explode `̀ Students''9. explode `̀ High-School-Students''10. explode `̀ College-Students''11. explode `̀ Preschool-Students''12. explode `̀ Elementary-School-Students''13. obes*14. adipos*15. ponderos*16. bodyfat*17. body fat*18. (fat near3 mass)19. fatness*20. subcutaneous fat21. body mass22. body masses23. bmi24. underweight25. under weight26. overweight27. over weight28. body size*29. body composition30. quetelet31. ponderal32. anthropomet*33. skinfold*34. skin fold*35. lean*36. thinness37. waist*38. weight?for?height39. weight?height40. explode `̀ Body-Weight''41. `̀ Weight-Control'' in DE42. prospectiv*43. longitudinal*44. followup*45. follow up*46. long term*47. longterm*48. track*49. serial*50. cohort*51. reexamin*52. re examin*53. reassess*54. re assess*55. restud*56. re stud*57. reevaluat*58. re evaluat*


S104

59. reinvestigat*60. re investigat*61. rereview*62. re review*63. remeasure*64. re measure*65. retrospectiv*66. growth study67. `̀ Causal-Analysis'' in DE68. explode `̀ Between-Groups-Design''69. explode `̀ Cohort-Analysis''70. explode `̀ Followup-Studies''71. explode `̀ Longitudinal-Studies''72. explode `̀ Repeated-Measures''73. `̀ Etiology-'' in DE74. case control*75. casecontrol*76. #1 or #2 or #3 or #4 or #5 or #6 or #7 or #8 or #9

or #10 or #11 or #1277. #13 or #14 or #15 or #16 or #17 or #18 or #19 or

#20 or #21 or #22 or#23 or #24 or #25 or #26 or#27 or #28 or #29 or #30 or #31 or #32 or #33 or#34 or #35 or #36 or #37 or #38 or #39 or #40 or#41

78. #42 or #43 or #44 or #45 or #46 or #47 or #48 or#49 or #50 or #51 or #52 or #53 or #54 or #55 or#56 or #57 or #58 or #59 or #60 or #61 or #62 or#63 or #64 or #65 or #66 or #67 or #68 or #69 or#70 or #71 or #72 or #73 or #74 or #75

79. #76 and #77 and #7880. amsterdam growth81. (atherosclerosis risk) near2 (communities)82. baltimore longitudinal study83. belgian luxembourg child84. birth cohort85. bogalusa heart study86. california child health study87. christchurch near3 development88. darling in ti,ab,su89. dormont high90. dunedin multi disciplinary91. dunedin multidisciplinary92. fels93. framingham94. hhanes95. kabi pharmacia96. leuven growth97. minneapolis children* blood98. muscatine99. (national health) near3 examination near3 survey100. nhanes101. nhlbi102. nurses health study103. princeton school district104. procam105. stanford five city106. tecumseh community107. Wrowclaw growth study108. Wroclaw longitudinal twin study109. young finns

110. youth health survey111. zurich growth112. #80 or #81 or #82 or #83 or #84 or #85 or #86 or

#87 or #88 or #89 or #90 or #91 or #92 or #93 or#94 or #95 or #96 or #97 or #98 or #99 or #100or #101 or #102 or #103 or #104 or #105 or#106 or #107 or #108 or #109 or #110 or #111or #112

113. #112 and #78114. #113 or #80

Search strategy: Sport Discus (Silverplatter software

interface), 1975 ±March 1998

1. explode `̀ CHILD''2. explode `̀ INFANT''3. explode `̀ADOLESCENT''4. OBES* in TI, AB5. ADIPOS* in TI,AB6. PONDEROS* in TI,AB7. BODYFAT* in TI,AB8. BODY FAT* IN TI,AB9. (FAT near3 MASS) in TI,AB10. FATNESS* in TI,AB11. SUBCUTANEOUS FAT IN TI,AB12. BODY MASS IN TI,AB13. BODY MASSES IN TI,AB14. BMI in TI,AB15. UNDERWEIGHT in TI,AB16. UNDERWEIGHT IN TI,AB17. OVERWEIGHT in TI,AB18. OVERWEIGHT IN TI,AB19. BODY SIZE* IN TI,AB20. BODY COMPOSITION IN TI,AB21. QUETELET in TI,AB22. PONDERAL in TI,AB23. ANTHROPOMET* in TI,AB24. SKINFOLD* in TI,AB25. SKIN FOLD* IN TI,AB26. LEAN* in TI,AB]27. THINNESS in TI,AB28. WAIST* in TI,AB29. `̀ OBESITY-'' IN DE30. `̀ SKINFOLD-THICKNESS'' IN DE31. explode `̀ BODY-WEIGHT''32. explode `̀ BODY-COMPOSITION''33. explode `̀ANTHROPOMETRY''34. PROSPECTIV* in TI,AB35. LONGITUDINAL* in TI,AB36. FOLLOWUP* in TI,AB37. (FOLLOW* near1 UP*) in TI,AB38. LONG TERM IN TI,AB39. LONGTERM in TI,AB40. TRACK* in TI,AB41. SERIAL* in TI,AB42. COHORT* in TI,AB43. RE?EXAMIN* in TI,AB44. RE?ASSESS in TI,AB45. RE?STUD in TI,AB46. RE?EVALUAT* in TI,AB


S105

47. RE?INVESTIGAT* in TI,AB48. RE?REVIEW* in TI,AB49. RE?MEASURE* in TI,AB50. RETROSPECTIV* in TI,AB51. GROWTH STUDY IN TI,AB52. CASE CONTROL* IN TI,AB53. CASE?CONTROL* in TI,AB54. explode `̀ CASE-STUDY''55. explode `̀ LONGITUDINAL-STUDY''56. explode `̀ PREVENTIVE-MEDICINE''57. `̀ EPIDEMIOLOGY-'' IN DE58. `̀ COMPARATIVE-STUDY'' IN DE59. explode `̀ PROSPECTIVE-STUDY''60. explode `̀ RETROSPECTIVE-STUDY''61. `̀ TREND-ANALYSIS'' IN DE62. #1 or #2 or #363. #4 or #5 or #6 or #7 or #8 or #9 or #10 or #11 or

#12 or #13 or #14 or #15 or #16 or #17 or #18 or#19 or #20 or #21 or #22 or #23 or #24 or #25 or#26 or #27 or #28 or #29 or #30 or #31 or #32 or#33

64. #34 or #35 or #36 or #37 or #38 or #39 or #40 or#41 or #42 or #43 or #44 or #45 or #46 or #47 or#48 or #49 or #50 or #51 or #52 or #53 or #54 or#55 or #56 or #57 or #58 or #59 or #60 or #61

65. #62 and #63 and #6466. AMSTERDAM GROWTH67. ATHROSCLEROSIS RISK NEAR2 COMMU-

NITIES68. BALTIMORE LONGITUDINAL69. BELGIAN LUXEMBOURG CHILD70. BIRTH COHORT71. BOGALUSA HEART STUDY72. CALIFORNIA CHILD HEALTH73. CARDIA STUDY74. CARDIOVASCULAR HEALTH STUDY75. CHRISTCHURCH near3 DEVELOPMENT76. DARLING STUDY77. DORMONT HIGH78. DUNEDIN MULTIDISCIPLINARY79. DUNEDIN MULTI DISCIPLINARY80. FELS81. FRAMINGHAM82. HHANES83. KABI PHARMACIA84. LEUVEN GROWTH85. MINNEAPOLIS CHILDREN* BLOOD86. MUSCATINE87. NATIONAL HEALTH NEAR3 EXAMINA-

TION NEAR3 SURVEY*88. NHANES89. NHLBI90. NURSES HEALTH STUDY91. PRINCETON SCHOOL DISTRICT92. PROCAM93. STANFORD FIVE CITY94. TECUMSEH95. WROCLAW GROWTH STUDY96. WROCLAW LONGITUDINAL TWIN STUDY97. YOUNG FINNS

98. YOUTH HEALTH SURVEY99. ZURICH GROWTH100. #66 or #67 or #68 or #69 or #70 or #71 or #72 or

#73 or #74 or #75 or #76 or #77 or #78 or #79 or#80 or #81 or #82 or #83 or #84 or #85 or #86 or#87 or #88 or #89 or #90 or #91 or #92 or #93 or#94 or #95 or #96 or #97 or #98 or #99

101. #100 and #63102. #65 or #101

Appendix III. Collaborators, steeringgroup and experts consulted

Collaborators

Professor Alan Lucas, MRC Childhood NutritionResearch Centre, Institute of Child Health, LondonWC1N 1EH, UK.Professor Shah Ebrahim, Department of Social Med-icine, Canynge Hall, Whiteladies Road, Bristol BS82PR, UK.Dr Julian Higgins, Department of Medical Statisticsand Evaluation, Imperial College School of Medicine,Hammersmith Hospital, Du Cane Road, London W120NN, UK.Professor Catherine Peckham, Epidemiology andPublic Health, Institute of Child Health, LondonWC1N 1EH, UK.

Steering group

Dr Andrew Prentice, MRC International NutritionGroup, London School of Hygiene & Tropical Med-icine, 49 ± 51 Bedford Square, London WC1B 3DP,UK.Dr Andrew Hill, Senior Lecturer in BehaviouralSciences, Division of Psychiatry and BehaviouralSciences, School of Medicine, The University ofLeeds, Leeds LS2 9LT, UK.Professor Jane Wardle, Health Behaviour Unit,Department of Epidemiology and Public Health, Uni-versity College, London WC1E 6BT, UK.Dr Ken Fox, Professor of Exercise and HealthSciences, Department of Exercise and HealthSciences, Priory House, Bristol University, BristolBS8 1TH, UK.

Additional experts contacted

Betsy Anagnostelis, Medical Library, Royal FreeHospital School of Medicine, Rowland Hill Street,London NW3 2PF, UK.Dr Laurel Edmunds, BHF Health Promotion ResearchGroup, Institute of Health Sciences, Oxford, OX37LF, UK.Dr Jonathan Wells, MRC Childhood NutritionResearch Centre, Institute of Child Health, LondonWC1N 1EH, UK.


S106

Dr Peter Whincup, Primary Care and PopulationSciences, Royal Free Hospital School of Medicine,London, UK.Dr Susan Jebb, MRC Resource Centre for HumanNutrition Research, Downhams Lane, CambridgeCB4 1XJ, UK.Lorna Rapoport, Health Behaviour Unit, Departmentof Epidemiology and Public Health, London WC1E6BT, UK.Sarah Lewington, Department of Gerontology, Rad-cliffe In®rmary, Oxford OX2 6HE, UK.Professor Tim Lang, Centre for Food Policy, ThamesValley University, Wolfson School of HealthSciences, London W5 2BS, UK.Nick Cavill, Health Education Authority, London, UK.Dr Doug Lamont, Department of Child Health, Uni-versity of Newcastle, The Royal Victoria In®rmary,Newcastle upon Tyne NE1 4LP, UK.Professor Ken Roberts, Department of Sociology,Social Policy and Social Work Studies, Universityof Liverpool, Liverpool L69 7ZA, UK.

Dr Gaston Beunen, Center for Physical DevelopmentResearch, Fac Physical Education, Physiotherapy,K.U. Leuven, Tervuursevest 101, B-3001 Leuven,Belgium.Professor Han Kemper, EMGO Institute, Faculty ofMedicine. Vrije Universiteit, Amsterdam, The Nether-lands.Professor William H Dietz, Division of Nutrition andPhysical Activity, Centre for Diseases Control andPrevention, Atlanta, GA 30341-3717, USA.Professor Michael I Goran, Division of Physiologyand Metabolism, Department of Nutrition Sciences,University of Alabama at Birmingham, Birmingham,AL 35294, USA.Dr Danielle R Reed, Behavioral Genetics CRB no.138, Center for Neurobiology and Behavior, Depart-ment of Psychiatry, University of Pennsylvania, 415Curie Blvd, Philadelphia, PA 19104, USA.Professor Redford B Williams, Department of Psy-chiatry and Behavioral Sciences, Duke University,Durham, NC 27706, USA.


S107

childhood predictors of adult obesity: a systematic review · lifestyle factors with a view to...

Documents