Food and Chemical Toxicology 54 (2013) 35–42

Contents lists available at SciVerse ScienceDirect

Food and Chemical Toxicology

journal homepage: www.elsevier .com/locate / foodchemtox

Benefit–risk assessment of plant sterols in margarine: A QALIBRA case study

Jeljer Hoekstra a,⇑, Heidi P. Fransen a, Jan C.H. van Eijkeren a, Janneke Verkaik-Kloosterman a,Nynke de Jong a, Helen Owen b, Marc Kennedy b, Hans Verhagen a, Andy Hart b

a National Institute of for Public Health and the Environment (RIVM), P.O. Box 1, 3720 BA Bilthoven, The Netherlandsb The Food and Environment Research Agency (FERA), Sand Hutton, York YO41 1LZ, United Kingdom

a r t i c l e i n f o a b s t r a c t

Article history:Available online 5 September 2012

Keywords:PhytosterolsBenefit–risk assessment

0278-6915/$ - see front matter � 2012 Elsevier Ltd. Ahttp://dx.doi.org/10.1016/j.fct.2012.08.054

⇑ Corresponding author. Tel.: +31 30 274 2204; faxE-mail address: jeljer.hoekstra@rivm.nl (J. Hoekstr

This paper presents the benefit–risk assessment of adding plant sterols to margarine as an illustration ofthe QALIBRA method and software. With the QALIBRA tool health effects, risks as well as benefits areexpressed in a common metric (DALY) which allows quantitative balancing of benefits and risks of foodintake. The QALIBRA software can handle uncertainties in a probabilistic simulation. This simple casestudy illustrates the data need and assumptions that go into a quantitative benefit–risk assessment.The assessment shows that the benefits of plant sterols added to margarine outweigh the risks, if any.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

The QALIBRA tool (Hart et al., this issue; Hoekstra et al., this is-sue) is used to perform a quantitative benefit–risk assessment ofplant sterol fortified margarines. The plant sterol case is used todemonstrate the data demand and the assumptions that are neces-sary for a quantitative benefit–risk assessment in general and morespecifically when the QALIBRA tool is used.

The market launch of functional foods forces the balancing ofbenefits and risks of dietary factors to be an important publichealth topic. As such health claims are a very hot topic in the EU,in particular following the consensus report of the Passclaim pro-ject (Aggett et al., 2005) and the publication of EU Regulation1924/2006 (Verhagen et al., 2010). In most cases it is unclearhow much benefit can be achieved as the market introductionsof these foods are based on a safety assurance only. This couldmean that foods that introduce a small risk but a much larger ben-efit are kept off the market. Indeed and in contrast to the vastattention for premarket scientific substantiation of health claims(Aggett et al. 2005; Verhagen et al., 2010), post market effective-ness monitoring of beneficial effects of functional foods is hardlygetting any attention (De Jong et al., 2007, 2008). Therefore it isof interest to know to what extent a specific food may provide pop-ulation health benefits and to what extent concomitant healthrisks are introduced, if any. Within the EU-funded QALIBRA projecta tool is developed that can precisely do that. Benefits and risks arequantified and expressed in a common metric (DALY). The benefit–risk assessment of plant sterols is used to illustrate the QALIBRA

ll rights reserved.

: +31 30 274 4466a).

method and tool and shows the data that is needed and theassumptions that are made to perform the assessment.

In the Netherlands, plant sterol-fortified margarines are allowedon the market since 2000. Thus, one may assume there are onlynegligible risks. It is claimed there are benefits in reduction of cho-lesterol with the assumption that that will decrease heart diseases.Indeed, many reviews support this (e.g., Law, 2000; Katan et al.,2003;Demonty et al., 2009; AbuMweis et al., 2008; Musa-Velosoet al., 2011). Despite the vast literature on lowering blood choles-terol levels there is to date no study investigating the ultimateproof of efficacy: lowering the incidence of coronary heart disease.Within this context, we use the QALIBRA method to quantify thebenefits and risks of plant sterol fortified margarines.

The recommended daily intake of phytosterol-fortified marga-rines is equivalent to 2–3 g of phytosterols and is regarded as theoptimal dose to reduce the LDL cholesterol levels effectively byabout 9–14% (Katan et al. 2003; Law 2000; EFSA, 2008; EFSA,2009; Demonty et al., 2009; AbuMweis et al., 2008; Musa-Velosoet al., 2011; Talati et al., 2010). As no further LDL cholesterol reduc-tions are achieved with intakes above 3 g/d, the former ScientificCommittee on Food concluded to avoid intakes above 3 g of phy-tosterols per day (SCF 2000, 2003a,b). Recently, EFSA, 2012a,b)has again evaluated the efficacy of plant sterols and plant stanolsat dose levels around 3 grams per day. EFSA concluded that plantstanol esters at a daily intake of 3 g (range 2.7–3.3 g) lower LDL-cholesterol by 10–13% and that plant sterols and stanol esters ata daily intake of 3 g (range 2.6–3.4 g) also lower LDL-cholesterolby 10–13%. Mensink et al. (2010) and Gylling et al. (2010) find fur-ther decreasing LDL-cholesterol levels with more than 3 g/d of sta-nol intake. The cholesterol lowering effect is established within afew weeks, and has been shown to remain stable for at least 1 year(Law, 2000; Katan et al., 2003; Demonty et al., 2009; AbuMweis

0

0.2

0.4

0.6

0.8

1

qual

ity

of li

fe

Healthy

Disease

Disability weight (w)

Years Lived with Disease

(YLD

Life Expectancy (LE)

DALY = wYLD +YLL

Years of Life Lost

(YLL)

36 J. Hoekstra et al. / Food and Chemical Toxicology 54 (2013) 35–42

et al., 2008; Musa-Veloso et al., 2011). Although no trials have di-rectly tested the effects of phytosterols on coronary heart disease(CHD) incidence, data from drug trials indicate that a reductionin LDL cholesterol level of about 10% could be expected to reducethe population incidence of ischemic heart disease by about 12–20% (Katan et al. 2003). (Law et al. (1994)) suggested that a reduc-tion in serum total or LDL cholesterol of 0.6 mmol/L (about 10% fortotal and 15% for LDL cholesterol) would reduce mortality fromischemic heart disease by 54% at age 40 to 19% at age 80. Baigentet al. (2005) refer to CHD risk reductions independent of age, vary-ing between 12% and 30% with a 10% to 1.0 mmol/L reduction inLDL level, respectively. Currently, few signals of (un)expected ad-verse health effects of phytosterols have been reported. The mainconcern is that phytosterols decrease serum a- and b-carotene lev-els (SCF, 2002).

In addition, there are several other reported potentially adverseside effects with phytosterols from natural sources but not withthose from plant sterol-enriched margarines and these have notbeen considered relevant in the safety evaluation. Some data havebeen presented that higher plasma levels of plant sterols are asso-ciated with risk of coronary heart disease, suggesting that plantsterols may be atherogenic (Sudhop et al. 2002), but EFSA con-cluded that this did not change their conclusions (EFSA, 2005).There are also suggestions that the consumption of phytosterolsmay promote the development of stroke* (Ratnayake et al. 2000Naito et al., 2003) or lead to the formation of unwanted hormonesfrom b-sitosterol and have an impact on estrogenic activity� (Maliniand Vanithakumari 1993; Mellanen et al. 1996). Several studies haveindicated an additive effect of phytosterols on LDL cholesterol lower-ing when combined with a statin. However, the use of statins hasalso been associated with an increase of cholesterol-standardizedserum plant sterol concentration, which may so induce extra ath-erogenicity (Miettinen et al. 2000). On a behavioural level, one mightthink of interference of phytosterol fortified foods in a patient’s drugtherapy compliance. A patient may think that because he consumesphytosterols, taking his drugs is not so important anymore whichmight induce a potential downside as has been investigated by Eus-sen et al. (2010a,b, 2011). On the beneficial side, in addition to theircholesterol lowering properties, consumption of phytosterols mayalso possess anti-cancer, anti-inflammatory, anti-atherogenicity,and anti-oxidant activities (see for an overview Berger et al. 2004)but evidence for this is still speculative.

It is recognised that the balance between risks and benefitsassociated with a functional food, food component or total dietshould form the ultimate basis for future nutrition policy on func-tional foods (EFSA, 2006; Tijhuis et al., 2012a; Tijhuis et al., 2012b).The purpose of integrated benefit–risk calculations is to expressthe beneficial health outcomes of functional food consumption inrelation to any adverse health outcomes. The ultimate result ofthe benefit–risk analyses can be used in benefit–risk management.Based on the outcomes, policy makers may want to adapt the label,stimulate the consumption of a certain food or diet, may ask foradditional studies or may even withdraw a product from the mar-ket. Within the EU-FP6 project QALIBRA, a quantitative benefit–risk assessment has been made of the population health effects ifan active policy is pursued regarding fortification of margarineswith phytosterols. This assessment is made to test and demon-strate the QALIBRA method and software tool.

In this assessment we answer the following specific benefit–riskquestion: Is there a public health gain if a policy is pursued regard-ing fortification of 100% of margarines with plant sterols comparedwith no fortification.

⁄ Animal experiment.� Animal and in vitro experiment.

2. Methods

A computer model simulation of the scenarios is performed to compute thehealth effects associated with different levels of plant sterol intake in the Dutchpopulation. A quantitative benefit–risk assessment involves a reference scenarioand (at least one) alternative scenario (Hoekstra et al. 2010). The health effects inthe alternative scenarios where 100% of margarine is fortified with 7.5 g of plantsterols per 100 g of margarine are compared with the reference scenario in whichmargarines are not fortified. The fortified margarine that is used most often inthe Netherlands is fortified with 7.5 g of plant sterols per 100 g of margarine. Thehealth effects are expressed in a common metric, the DALY (Disability Adjusted LifeYears) (Murray, 1994). The metric incorporates morbidity and mortality. DALYs aremeasured as the sum of the number of years lived with the disease adjusted with aweight that represents the severity of the disease, and the number of years lost dueto earlier death as a result of the disease. Fig. 1 schematically shows how DALYs arecalculated for an individual.

To compute the health effects we need the following:

� Intake distributions that describe the intake of plant sterols in the population ofinterest in the fortification and in the reference scenario.� Identification of relevant health effects that are associated with plant sterol

intake.� Characterization of the effects that result from plant sterol intake. This is dose–

response functions that describe the relation between plant sterol intake andeach relevant health effect.� Disease statistics, such as severity weights and survival data to convert the sim-

ulated incidences in DALYs.

More specifically, we use the QALIBRA tool to compute the difference in the an-nual average amount of DALYS between the reference, no fortification and the 100%fortification scenario. With this method the DALYs are counted if they result fromincidences occurring in 1 year. The difference between the scenarios can be inter-preted as the long-term annual effect of fortification.

2.1. QALIBRA method

The benefit–risk assessment method and software tool that is used is describedin detail in Hart et al. (this issue). The overall framework is similar to the PIF ap-proach (Barendregt and Veerman, 2010), (Murray et al., 2003) and also similar tothe quantitative benefit–risk analysis described in (Hoekstra et al., 2010; Hoekstraet al., 2008), but in more generalised form (facilitating its application to a widerange of benefit–risk questions) and adding the capability to quantify uncertaintiesat each stage of the process. In Hoekstra et al. (2010) a tiered approach is suggested.In the first tier an assessment is made of the risk and benefits separately. In theplant sterol case we could conclude that there is negligible risk (otherwise theauthorities would not allow it on the market) and a possible benefit. Hence, it isnot necessary to use sophisticated and data demanding simulation tools such asQALIBRA to assess whether there are health gains in using plant sterol fortified mar-garines. However, we use this case as a simple illustration of the QALIBRA tool.

The QALIBRA tool (www.qalibra.eu) quantifies and weighs benefits and risks byexpressing them in a common metric, the DALY. The DALYs are calculated for thecurrent year. Incidences that occur in the current year due to the intervention resultin years lived with the disease and years of live lost as a result of the disease.Assuming that we consider a long-term intervention with a demographically stablepopulation, the outcome can be regarded as the annual health gain or health lossexpressed in DALYs/year when the intervention or policy is implemented.

0 20 40 60 80 100

age

Fig. 1. Schematic representation of the DALY calculation.

1.00%

J. Hoekstra et al. / Food and Chemical Toxicology 54 (2013) 35–42 37

The calculation of a DALY is based on two components: the number of yearssomeone suffers from a disease and the number of years someone loses if dying ear-lier because of the disease. The calculation of DALYs takes account of two possibil-ities, either someone develops the disease or does not. If someone does, threealternative outcomes are possible: an individual may

� recover,� die early as a result of the disease,� survive with the disease until the normal life expectancy i.e., suffer chronically.

The probability that someone develops the disease depends on exposure. It isdenoted by peffect. peffect is a dose–response relationship that associates the intakeof a nutrient or contaminant with the disease. Often peffect depends on age andsex and possibly other individual characteristics.

The expected total DALY loss of an individual due to a single disease occurringin the current year can then be estimated as:

DALY ¼ peffect ½precYLDrecwþ pdieðYLDdiewþ LE� CA� YLDdieÞ þ ð1� pdie � precÞ� ðLE� CAÞw� ð1Þ

where peffect probability of onset of the disease in the current year (1/year); prec prob-ability of recovery from the effect (0–1); pdie probability this effect causes death (0–1); YLDrec duration of disease for those who recover (year); YLDdie duration of disease(years lived with disease) for those who die of it (year); CA current age of individualin year of disease onset (year); LE normal life expectancy� (year) and w disabilityweight for disease (0–1)

Between brackets the unit of each parameter is given. Each term (line) in Eq. (1)describes one of the three possibilities, recovery, premature death, or chronic suf-fering. Each term is composed of a probability that it will happen, the years livedwith the disease and the years lost due to early death. Eq. (1) refers to an individualwith a certain age, CA, a certain exposure and possibly other properties such as sexand weight etc. on which the variables in Eq. (1) may depend. Parameters in Eq. (1)can be functions of intake and other individual characteristics. At least one of theparameters should depend on intake otherwise there is no need to simulate differ-ent intake scenarios. In general peffect depends on exposure. Note that LE and peffect

will generally differ between age and sex groups and that other parameters may doalso (e.g. the severity of a disease might depend on exposure and the age at which itoccurs).

The incidence of a disease in the current year is estimated by summing peffect

over all individuals in the population.

inc: ¼X

peffect ð2Þ

In order to obtain an appropriate estimate for the average annual DALY loss of awhole population, the calculation in Eq. (1) needs to be repeated and summed foreach disease and for each characteristic individual in the population. Characteristicindividuals are constructed in proportion to the age structure, the intake distribu-tion (potentially per age and sex group) and other properties of the population.Our model population consists of 10,000 model individuals, representative for theDutch population in terms of age and sex.

In a benefit–risk assessment the focus is on evaluating the net health impact ofa dietary change or intervention, in essence a change in intake. This is representedas the change from a reference scenario to an alternative scenario. Thus, the nethealth impact of an intervention in DALYs is calculated as:

DDALY ¼X

DALYalt �X

DALY ref ð3Þ

whereP

DALY ref ; sum over all individuals and all health effects of DALY losses forthe reference scenario;

PDALYalt , sum over all individuals and all health effects of

DALY losses for the alternative scenarioThe following paragraphs describe the scenarios i.e., population and intake and

every variable from Eq. (1) which is necessary to simulate the health effects of thescenarios.

0.00%

0.25%

0.50%

0.75%

0 20 40 60 80 100

freq

uen

cy

men

women

2.2. Exposure assessment: scenarios of plant sterol intake in the Netherlands

The scenarios describe the public health effects in the Dutch population when0% (reference scenario) or 100% (alternative scenario) of margarines is fortified with7.5 g of plant sterols/100 g margarine, which the fortification level that is used inthe margarine that has the largest market share in the Dutch market.

We assume that every individual in the population will consume the sameamount of margarine in each scenario. This is the amount estimated from the DutchNational Food Consumption Survey (DNFCS) intake database depending on age andsex. So, we assume that subjects will not alter their margarine consumption whenmargarines are fortified.

� Note: life expectancy is sometimes defined as the number of years of liferemaining, but here it is defined as the expected age at death.

2.2.1. PopulationThe scenarios were simulated with a population of 10,000 individuals repre-

senting the Dutch population. The Dutch population age distribution for both sexesin 2007 was obtained from Statistics Netherlands (Website CBS Statline http://stat-line.cbs.nl/statweb/). From this distribution (see Fig. 2) 10,000 individuals were ran-domly drawn. When random sampling a subject from the population, thecumulative frequency is employed to determine the subjects age. Whether the sub-ject would be a male or female is determined by the age dependency of the prob-ability of a subject being a female. The sampled 10,000 model individualsrepresent the Dutch population i.e., the distribution of age and sex in the modelpopulation is the same as in the Dutch population.

2.2.2. Margarine intakeData of the population wide Dutch National Food Consumption Survey pub-

lished in 2003 (DNFCS) were used to simulate several scenarios of plant sterol in-take. This DNFCS was conducted in 1997–1998. The respondents (N = 6250, aged1–97 years) recorded their food intake via dietary records on two consecutive days(Hulshof et al., 2003). In the period this DNFCS was conducted, plant sterol fortifiedfoods were not yet introduced on the market. The Dutch food composition tabledoes not contain information about phytosterols naturally occurring in foods. How-ever, it is expected that plant sterol intake from fortified margarines surpasses theintake from natural sources. For example, Hearty et al. (2008) report a mean back-ground exposure for phytosterols of 287 mg/d. An average slice of bread containsabout 6 g of margarine, corresponding to 450 mg of plant sterols using the currentfortification level of 7.5/100 g. Thus, plant sterol intake through fortified margarineconsumption by far outweighs plant sterol intakes from natural sources. Therefore,background intake was ignored in the scenarios.

From the DNFCS for each subject the mean (low-fat) margarine intake over2 days was calculated. The (low-fat) margarine intakes during hot meals were ex-cluded. Subjects with a mean intake of zero grams over the 2 days were considerednon-consumers of the product group.

From the DNFCS data the fraction of non-consumers is established and an intakedistribution for margarine consumers is established, depending on age and sex. Thedistributions are illustrated in the graph in Fig. 3. The two clusters that can be iden-tified in the graph are ‘men’ and ‘women’. The graph shows that men tend to eatsomewhat more margarine than women.

2.2.3. Plant sterol intakeIn the policy scenarios it was assumed that every (low-fat) margarine was for-

tified with plant sterols and that only these fortified products were used by themargarine consumers. For the reference scenario we assumed that no margarineswere fortified with plant sterols. For (low-fat) margarine the intake from hot mealswas not taken into account, as plant sterol fortified (low-fat) margarines are in-tended to be used as bread spread only. In the scenarios (low-fat) margarine wassimulated to be plant sterol fortified with a level of 7.5 g plant sterols per 100 gmargarine.

From the intake distributions of the policy scenario mentione above, age- andsex-specific margarine intakes were drawn and multiplied by the level of plant ster-ols in fortified (low-fat) margarine (i.e. 7.5/100 g) to calculate the total plant sterolintake from fortified foods.

2.3. Benefit identification and characterization

Katan et al. (2003) present summary estimates from randomised placebo-con-trolled trials of the relative reductions in LDL cholesterol (%) to plant sterol/stanoldose (g/d). More recently other reviews confirmed this relationship between LDLcholesterol decrease and plant sterol intake (EFSA, 2008; EFSA, 2009; AbuMweiset al., 2008; Demonty et al., 2009; Talati et al., 2010). Musa-Veloso et al. (2011) con-

age

Fig. 2. The age distribution of men and women in the Dutch population.

0%

25%

50%

75%

100%

0 20 40 60 80 100

margarine intake gram/day

per

cen

tiles

men 20-29 yr

women 20-29 yr

men 40-49 yr

women 40-49 yr

men 60-69 yr

women 60-69 yr

Fig. 3. A cumulative distribution of margarine intake in the Dutch populationstratified for age and sex.

Table 1Estimated parameter values for the relative LDL serum level decrease (%) versus steroldose (g/d) model Eq. (4). The estimates of Demonty et al. (2009) are shown betweenbrackets.

Rmax (%) d1/2 (mmol/L)

2.5 percentile 10.1 (9.99) 0.803 (0.62Mean 11.9 (12.68) 0.803 (1.12)97.5 percentile 13.7 (15.38) 0.803 (1.63)

Table 2Optimized parameter values for the LDL serum level (mmol/L) versus age (year)model Eq. (5).

Percentile C (mmol/L) M (mmol/L) Age1/2a (year)

Male 50 3.17 4.15 13.15 2.39 3.06 13.1

95 4.64 5.28 13.1Female 50 2.91 9.46 13.1

5 2.18 7.09 13.195 4.03 11.3 13.1

a Fixed to value found for calibration on (Gotto et al., 1984) data.

38 J. Hoekstra et al. / Food and Chemical Toxicology 54 (2013) 35–42

cluded in the meta-analysis on relatively high levels of cholesterol lowering at highdose levels of phytosterols, albeit that their methodological robustness has alreadybeen questioned (Demonty et al., 2011). Law et al. (1994, 2003) present an agedependent relationship between absolute LDL cholesterol reduction (mmol/L) andrelative (%) decrease in IHD (Ischemic Heart Disease) events. We assume that a de-crease in cholesterol due to plant sterol intake will result in decreased incidence inIHD, although this has not yet been demonstrated in a clinical trial.

Based on these two dose–response relationships, we developed a combined(external) dose–response relationship between intake of plant sterols (g/d) and de-crease in the probability of IHD events, that is dependent on the LDL cholesterol le-vel at the start of the intervention. The combination of this dose–responserelationship together with age dependent relations between serum LDL cholesterollevel and IHD incidence allows for estimating absolute IHD probability in depen-dence on plant sterol dose and age (peffect,IHD).

The relationship between plant sterol intake and reduction of IHD incidence(peffect,IHD) combines several calculation steps:

(1) plant sterol intake (g/d) ? relative LDL cholesterol reduction,(2) relative LDL cholesterol reduction (%) ? absolute LDL cholesterol reduction

(using baseline LDL cholesterol data),(3) absolute LDL cholesterol reduction ? relative reduction of the probability

of an IHD event,(4) relative IHD probability ? absolute IHD probability (using baseline IHD

incidence),(5) IHD incidence ? DALY.

2.3.1. Plant sterol intake versus relative LDL cholesterol reductionData presented in Katan et al. (2003) Fig. 2, summarizing estimates from ran-

domised placebo-controlled trials of the percentage reductions in LDL related todose were modelled following Eq. (4).

% Reduction LDLðdÞ ¼ Rmax � 1� exp � lnð2Þ � dd1=2

� �� �ð4Þ

where Rmax is the maximally attained reduction (%), d is sterol dose (g/d) and d1/2 thehalf maximum LDL reduction dose. The corresponding confidence interval bound-aries were estimated by fitting Eq. (1) through the 2.5 and 97.5 percentiles.

While fitting Eq. (1) to the data in Katan et al. (2003) Fig. 2, the last data at about4 g/d was omitted (in accordance with the summary in Katan et al. (2003) Table 3)because of its great uncertainty and thus avoiding a dose–response relationshipthat might be too optimistic. When fitting the 2.5 and 97.5 percentiles, the valuefor the half dose d1/2 was fixed at is value found when fitting the mean. The esti-mated model parameter values for the mean, 2.5 and 97.5 percentile are tabulatedin Table 1.

According to the dose–response relationship (Eq. (1)), the predicted relative LDLcholesterol lowering effect of the recommended daily dose of phytosterols (2 g)would be 9%, similar to the estimate of Demonty et al. (2009)). We also are awarethat there are suggestions for even more lowering of cholesterol at higher doses ofphytosterols, albeit that there is discussion on this (Musa-Veloso et al. 2011; De-monty et al., 2011).

2.3.2. Converting relative LDL cholesterol reduction to absolute LDL cholesterolreduction using baseline LDL cholesterol

Relative LDL cholesterol values can easily be converted in absolute values whenbaseline absolute levels are known. Data on LDL cholesterol levels in dependence onage and sex were obtained from (Gotto et al., 1984; Verschuren et al., 2008). Data in(Gotto et al., 1984) suggested a constant level for ages till about 20 years and a stea-dy increase thereafter. So, the following model was applied:

LDLðAge;=; sÞ ¼CðsÞ Age < 20

CðsÞ þ ðMðsÞ � CðsÞÞ � 1� exp � logð2Þ Age�20Age1=2 ðsÞ

� �� �20 < Age

(

ð5Þ

where LDL is the baseline LDL cholesterol level, M is an asymptotic maximum LDL-cholesterol level (that need not be representative of LDL levels encountered in prac-tise), Age1/2 the half age at which LDL level is halfway in between C and M and s cor-responds to sex category, male or female.

The model in Eq. (5) was fitted to the data both for men and women in (Gottoet al., 1984) concerning the median (50th percentile) and the 5th and 95th percen-tiles of the observed distribution. The half age value, estimated for the median val-ues data was fixed when estimating the 5th and 95th percentile data. From themodel and data comparison it was decided that the model assumption is adequate.To profit from more recent data in Doetinchem (Verschuren et al., 2008) the modelwas recalibrated. This data on a Dutch sub-population may be more informativeconcerning current Dutch LDL-cholesterol serum levels, however was only availablefor ages between 30 and 80 years in age classes of 10 years. It was found that themedian levels of this data compared to those from USA 1984 are about 10% higherfor men and 15% for women. Because of paucity of data it was decided to fix the halfage value to that found for the former data. The model parameters C(s) and M(s)were corrected in accordance with the new data. Furthermore, because the datafor the class 70–79 years was based on sampling only a relative small number ofsubjects (14 men and 27 women) these data were left out of recalibration. The finalparameter values are tabulated in Table 2.

The relative LDL serum level decrease (%) can be combined with baseline LDLserum level to obtain absolute LDL serum level decrease:

DLDLðd; age;sexÞ ¼ Rmax

1001� exp � lnð2Þ d

d1=2

� �� �LDLðage;sexÞ ð6Þ

where DLDL represents the absolute LDL cholesterol decrease (mmol/L).

2.3.3. Absolute LDL cholesterol reduction ? relative reduction of the probability of anIHD event

Data presented in Law et al. (2003, Table 7) relates expected decrease in IHDevents (%) to a specified decrease in LDL-cholesterol serum level (mmol/L) basedon 10 large cohort studies in a meta analysis. These were fitted to Eq. (7) in a similarmanner to Eq. (4):

% IHD decreaseðD;AgeÞ ¼ 100 � 1� exp � lnð2Þ � DD1=2ðAgeÞ

� �� �ð7Þ

where the age dependent decrease D1=2 (half IHD decrease LDL-cholesterol serum le-vel) is the LDL-cholesterol decrease when IHD decrease is 50%.

To account for age dependency, we assumed a linear relationship between halfcholesterol decrease and age for ages between 50 and 70 years:

D1=2ðAgeÞ ¼ a � Ageþ b ð8Þ

In Law et al. (2003), data for the three ages (50, 60 and 70 years) are tabulated,representing age classes of 45–54, 55–64 and 65–74 years, respectively. These datawere simultaneously fitted to the model as defined by the Eqs. (7) and (8). Esti-mated parameter values are

Table 3Estimated parameter values in Eq. (10) depending on sex.

A T c

Male 29.6 63.8 15.9Female 21.0 67.1 13.4

J. Hoekstra et al. / Food and Chemical Toxicology 54 (2013) 35–42 39

a ¼ 0:0501ðmmol=L=yearÞandb ¼ �1:66ðmmol=LÞ ð9Þ

The resulting model calculations are compared to those of Law et al. (2003).From the comparison it is evident that the modelling approach in Law et al.(2003) is equivalent to our approach in Eq. (7). Incorporation of Eq. (8) into themodel however, allows for IHD decrease estimates for any age between 45 and75 years. In order to estimate variability, it was observed that for the age class55–64 years the inverse variance weighted mean for 27% of eight investigationscould be estimated by median polish. Therefore, the lower and upper confidencelimits of the eight investigations were median polished to obtain a confidence inter-val of 20–33%. As this range of 13% represents the 95% confidence interval of width4r the coefficient of variation was estimated to be 100 � (13/4)/27) = 12%.

In Law et al. (1994), Table 2 also one data for estimated IHD reduction of 54% atage 40 (range between 35 and 44 years) is presented inclusive a 95% confidenceinterval of 45–62%. Model calculated decrease is 70%, which shows that modelbased estimates are too optimistic. Therefore, when estimating relative IHD reduc-tion, for all ages below 45 year, the age of 45 year will be substituted. This is a con-servative approach. Moreover, the majority of IHD incidence decrease derives fromages above 45 year and therefore the resulting underestimation is expected to beirrelevant.

2.3.4. Relative IHD probability ? Absolute IHD probability (using baseline IHD risk)Data regarding absolute baseline IHD incidence in the Dutch population in 2003

were obtained from the Nationaal Kompas Volksgezondheid [www.rivm.nl/nation-aalkompas]. These data concern the two sexes and several age classes. IHD inci-dence is given per 1000 subjects of a given sex and age class. The data could bemodelled using a hyperbolic tangent shaped function:

IHDðage;sÞ ¼ AðsÞ2

1þ tanhage� TðsÞ

cðsÞ

� �� �� 1þ tanh

�TðsÞcðsÞ

� �� �� �ð10Þ

where, A is the asymptotic maximum, T is the age at half maximum and c determinesthe rate of increase and s is the corresponding sex. IHD in Eq. (10) should be inter-preted as the baseline IHD incidence (1/1000) of individuals of sex s and age Age.

The model Eq. (10) was fitted to the data for men and for women. Estimated val-ues are presented in Table 3. The data and model are compared in Fig. 4.

The absolute decrease in the probability of an IHD event is obtained by multi-plication of the expressions in the Eqs. (4) and (10). As a result, the absolute prob-ability of an IHD event depending on age, sex and intake of plant sterols is given byEq. (8)

peffect;IHDðd; Age;sÞ ¼ % IHDðd; Age; sÞ � IHDðAge;sÞ ð11Þ

2.3.5. IHD incidence ? DALYDALYs are computed according to Eq. (1). It combines mortality expressed as

the number of years lost (YLL) and morbidity expressed as the number of years livedwith the disease (YLD) weighted by a factor (w) that expresses the severity of thedisease.

0 20 40 60 80 100age (year)

0

10

20

30

IHD

inci

denc

e / 1

000(

of a

ge a

nd s

ex)

Fig. 4. IHD incidence per 1000 in the Dutch population in 2003 for men (⁄) andwomen (s). Comparison of modelling approach (lines) and data (symbols).

We assume that an individual will either live with the disease for the rest of hislife, or he will die after 1 year. Therefore, the probability of recovery, prec and con-sequently YLDrec are equal to 0. According to Statistics Netherlands (www.stat-line.nl, 2005 the mortality of CHD-patients was 5.8% for men and 8.6% for womanin the first year after hospitalisation. Thus, YLDdie is set to 1 and the probabilityof death, pdic is 0.058 and 0.086 for men and women respectively. We assume mor-tality due to IHD does not depend on plant sterol intake. The disability weight, w, is0.29 (Stouthard et al., 1997).

2.4. Risk identification and characterization

Phytosterols are allowed on the European markets so we could conclude thereare no risks of concern, which is acknowledged in the respective EFSA and SCF opin-ions (EFSA, 2012a; EFSA, 2012b; SCF 2002, 2003a,b). However, there are humanstudies that show an adverse effect of plant sterol intake on serum levels of b-car-otene. Because we are introducing QALIBRA, a quantitative benefit risk tool, we liketo incorporate a risk in the assessment however small. Therefore, we choose toincorporate the effect on b-carotene levels.

LDL cholesterol is an important transport vehicle for fat-soluble vitamins. If theabsorption of LDL cholesterol is lowered, for example due to the use of plant sterolfortified products, the concentration of these fat-soluble vitamins, like carotenoids,may also be affected. b-Carotene is one of the carotenoids that might be affected,albeit that the effect is controversial (Polagruto et al., 2006; Berger et al., 2004).

In some plant sterol trials, decreases of b-carotene concentrations of 8–19%were found when plant sterol fortified products were consumed (Law, 2000; Katanet al., 2003). Other studies do not find effects on carotenoid levels. (Berger et al.,2004) have summarized these results. Some studies have shown that if the plantsterol fortified products are consumed in a healthy diet, the carotenoid levels willstay within normal levels, within the seasonal variation (Ntanios and Duchateau,2002; Noakes et al., 2002).

Because the studies reporting a decrease of b-carotene level are diverse with re-spect to quantifying units and are not quantitatively relating sterol dose and serumb-carotene level decrease, a quantitative sterol dose–serum b-carotene level de-crease response relation cannot be obtained.

The health effects of the possible variation in blood carotenoid levels are largelyunknown. It is assumed that people with higher levels of b-carotene have a lowerrisk for cancer and CVD (SCF, 2002), but more studies are needed to look furtherinto this problem (Katan et al., 2003). As b-carotene is also an important sourceof vitamin A, (carotenoids can contribute to more than 40% of the vitamin A supply(Ref. SCF, 2002)). For the sole purpose of this benefit–risk study, we have chosen tolook at the lowering of vitamin A levels due to lower b-carotene levels. As such, weassume that a b-carotene level decrease will also result in a lower vitamin A pro-duction, leading subjects from near vitamin A deficiency into vitamin A deficiency.

Several health effects are reported to be related to vitamin A deficiency, withxerophthalmia being the first visible clinical effect. This disease consists of severalstages, nightblindness being the first symptom that is observed. No data were foundin the literature that could be used to derive a quantitative vitamin A status–night-blindness response relation was found. Therefore, and because the conversion fromb-carotene to vitamin A is proportional to the individual’s needs, we chose to createan ‘‘on/off’’-scenario for sterol intake and nightblindness. The model assumes that ifa person is nearly vitamin A deficient, the use of plant sterol fortified products leadsinevitably to vitamin A deficiency in this person.

2.4.1. Sterol dose – nightblindness relation, peffect,NB

Lack of clear and univocal qualitative data, let alone sufficient quantitative data,prohibited the development of a quantitative dose–response for plant sterol dose–vitamin A status and for vitamin A status–nightblindness response. Therefore, fourworst-case assumptions were made to obtain a highly conservative but univocalpopulation sterol dose–nightblindness relation, peffect,NB. These assumptions are:

(1) that any amount of sterol intake (yes/no) will decrease b-carotene level;(2) that people that are supposed to possess a vitamin A status that is near defi-

ciency are actually near deficient;(3) that decrease of b-carotene level, considered as a provitamin A, will lead a

person from near vitamin A deficiency to deficient;(4) that a deficient vitamin A status has nightblindness as effect.

To summarize, our highly conservative model assumption is that people whohave a poor vitamin A status and consume plant sterols will suffer from nightblind-ness. We are aware that this is a gross overestimation of the effect of a reducedbeta-carotene level due to plant sterol consumption. Not every person who is nearlyvitamin A deficient will actually become deficient if their b-carotene levels decreasedue to sterol intake and will develop nightblindness as a result. We hypothesisethat the cholesterol lowering effect of plant sterol intake outweighs the beta-caro-tene lowering effect. So, if a grossly overestimated beta-carotene effect is still out-weighed by the cholesterol lowering effect than there is no need to be moreaccurate and our simple unrealistic assumptions suffice.

Table 5Results of the simulation of the reference scenario 0% fortification and the 100%fortification scenario (95% confidence intervals between brackets).

Ischeamic Heartdisease

Nightblindness

Change in DALYs �8.6 (�11.1, �6.5) 0.37Incidence per 1000, Reference

scenario4.8 0

Incidence per 1000, 100% fortificationscenario

3.7 (3.4, 4.0) 18.7

Mortality per 1000, Referencescenario

0.33 0

Mortality per 1000, 100% fortificationscenario

0.25 (0.24, 0.28) 0

Table 4Distribution of worst-case vitamin A deficiency in the Dutch population. Based onWaijers and Feskens, 2004, per age and sex class and as part of the total population.

Vit. A deficientindividuals inage and sexclass (%)

Part of thepopulation inage and sexclass (%)

Vit. A deficientindividuals inthe totalpopulation (%)

Children 1–3y 0.2 4.7 0.01Children 4–8y 1.5 6.2 0.09Children 9–13y 1.5 6.0 0.0914–18y Men 2.5 3.1 0.0814–18y Women 4.1 3.0 0.1219–50y Men 1.6 22.5 0.3619–50y Women 4.8 22.1 1.0651–65y Men 1.3 9.5 0.1251–65y Women 2.6 9.4 0.2465+ Men 0.6 5.7 0.0365+ Women 1.8 7.8 0.14

40 J. Hoekstra et al. / Food and Chemical Toxicology 54 (2013) 35–42

2.4.2. Vitamin A statusThe distribution of poor vitamin A status in the Dutch population, classified on

sex and age ranges, was obtained from Waijers and Feskens, 2004. This distributionis based on food consumption data and the assumption that people who consumefood with too low a (pro-) vitamin A (e.g. beta-carotene) content are likely (at theverge of being) vitamin A deficient. It is realised that this in fact is a worst case sit-uation for the sole purpose of this integrated benefit–risk assessment which ishighly unlikely to occur in the Netherlands. The Dutch population sex and age dis-tribution was obtained from the Dutch central bureau of statistics (www.stat-line.nl). Both distributions have been combined to obtain the vitamin A statusdistribution in the total population.

The population distribution of people who are supposed to be nearly deficient isshown in Table 4 (second column) as the percentage of subjects of the correspond-ing sex and age class. The total population age distribution is also shown in Fig. 2. Ineffect, peffect,NB for sterol intake >0 is shown in the last column of Table 4.

2.4.3. DALY computation of nightblindnessThe resulting change in incidence in nightblindness is converted to DALYs by

using a disability weight for minor visual disturbance (w = 0.02) because no weightfactor for nightblindness is available. Furthermore, sufficient intake of vitamin Awill cure the disease is and it is presumed that it lasts no longer than 1 year. So,YLDrec and prec are 1 and YLDdie and pdie are 0.

§ Note that DALYs measure health loss so a negative DALY number is a health gain

3. Results

The simulated beneficial effect of using plant sterol fortifiedmargarines is that an IHD event is prevented in the current year.The simulated adverse effect is nightblindness. Individuals can beaffected in two ways, positively and negatively. But many peoplewill not be affected at all, they did not experience an IHD eventin the 0% fortification-scenario and hence in the fortification sce-nario they did not prevent that event and neither did they becomenightblind. Others only experience an adverse effect, they becomenightblind or a beneficial effect, they prevent IHD. The QALIBRA

method calculates incidences and the annual average change inDALYs. If margarines are 100% fortified, the incidence in night-blindness under worst-case assumptions increases with about19/1000 and the incidence in IHD decreases by about 1/1000.The corresponding IHD mortality decreases with slightly less than0.1/1000. When we combining the adverse and beneficial healtheffects in one metric, the difference between the reference andthe 100% fortification scenario is �8.2 DALYs with a 95% confidenceinterval of (�10.7,�6.1)§. Table 5 shows the results.

The results clearly indicate that the benefits of fortified marga-rine (decrease in the incidence of IHD) clearly outweigh any puta-tive risks (nightblindness). Therefore, the overall benefit–riskequation supports the use of plant sterols for blood cholesterollowering.

4. Discussion

The discussion is divided in two parts. First we discuss resultsand assumptions of the plant sterol case study. Then we discussthe use of the QALIBRA tool.

4.1. On risks and benefits of plant sterol fortified margarine

Fortification of all margarines in the Netherlands would resultin a reduction of approximately eight disability adjusted life years,annually per 1000 inhabitants. Assuming a stable population, theincidence of ischemic heart disease would decrease by about20%.An average of eight healthy life years gained per 1000 peoplemay not seem a lot but it is each year as long as margarines are for-tified and represents a 20% decrease in the incidence of a major dis-ease changes the perspective. However, this effect is achieved onlyif every margarine that is used as bread spread is fortified with7.5 g of plant sterols per 100 g. In reality, fewer margarines areand are likely to be fortified so fewer individuals will prevent theirischaemic heart disease. The results represent an upper estimate ofthe net effect and show the potential gain of margarinefortification.

There are some caveats to this assessment. We have ignored po-tential adverse long term effects of plant sterols for which cur-rently there is no convincing evidence, albeit that none are known.

Furthermore, in the study of Law that we used to establish therelationship between LDL cholesterol reduction and IHD incidence(Eq. (7)) some subjects were using statins to lower their cholesterollevels. We have assumed that statins have no effect on IHD otherthan by lowering cholesterol. However, statins provide more pro-tection than the effect of cholesterol lowering. Therefore the effectof plant sterols on IHD is overstated in our Eq. (7) and thus in thefinal results.

We have also slightly underestimated the effect in DALYs. Whenan individual develops IHD, the attributed DALYs, notably the yearsof life lost are based on the general life expectancy not on the lifeexpectancy including a reduced plant sterol induced probability ofdying from IHD.

It should be noted that the adverse effect, becoming nightblind,is a gross overestimation of the negative effect and it is highly un-likely to occur in the Netherlands. However, because the effect isstill very small it does not influence the overall results.

It seems fair to conclude that the net health effect of phytosterolintake is positive, given that risks are hugely overestimated, bene-fits are probably overestimated, and the difference is large and infavour of the benefits.

J. Hoekstra et al. / Food and Chemical Toxicology 54 (2013) 35–42 41

4.2. On benefit–risk assessment and the QALIBRA tool

This case study supports the consideration of benefits and risksin one combined assessment in order to come to an overall judge-ment. Benefit–risk assessment is a highly interesting and develop-ing area in which many papers are now being produced. Theyoriginate from QALIBRA and other EU projects on benefit–riskassessment: BRAFO (Hoekstra et al., 2010; Watzl et al., 2011; Ver-hagen et al., 2011); BENERIS (this issue) and BEPRARIBEAN (Tijhuiset al., 2012a; Tijhuis et al., 2012b; Verhagen et al., 2012; Luteijnet al., 2012; Magnusson et al., 2012; Pohjola et al., 2012; Kalogeraset al., 2012; Ueland et al., 2012).

This case study illustrates how the QALIBRA tool is used in aquantitative assessment. At the same time, it shows the amountof data collection, data treatment and assumptions that are neededto make a quantitative benefit–risk assessment. For any quantita-tive assessment, the data need will be large and assumptions willhave to be made. A clear advantage of using the QALIBRA tool isthat the user is guided through the assessment and it is made clearbeforehand which data is needed.

The method of the QALIBRA tool is an extended version of themethod described in Hoekstra et al. (2008) incorporating recoveryand with the possibility to perform a probabilistic calculation (Hartet al. (this issue)). The tool is suited to calculate health effects ofrisks and benefits expressed in DALYs or QALYs. This is neededfor the last steps of tiered approach frameworks such as the BRAFOor EFSA approach (Hoekstra et al., 2010; EFSA, 2010 Fransen et al.,2010). For this specific case study a quantitative assessment maynot really be necessary, the same conclusions about plant sterolfortified margarines can and have been reached in a qualitativeassessment. This would be an earlier tier in a tiered approach. Plantsterol fortified margarines is relatively simple case that is used toclearly describe the functioning of the tool. Hoekstra et al., this is-sue apply the tool to a more complex case study, fish, with manybenefits and many risks.

Conflict of Interest

The authors declare that there are no conflicts of interest.

Acknowledgements

This work was supported by the European Commission throughthe QALIBRA Project (FOOD-CT-2006-022957): Quality of Life –Integrated Benefit and Risk Analysis. Web-based tool for assessingfood safety and health benefit. www.qalibra.eu. FERA’s contributionto the project was partly supported by the Food Standards Agency.

References

AbuMweis, S.S., Barake, R., Jones, P.J., 2008. Plant sterols/stanols as cholesterollowering agents. A meta-analysis of randomized controlled trials. Food Nutr.Res. 52.

Aggett, P.J., Antoine, J.M., Asp, N.G., Bellisle, F., Contor, L., Cummings, J.H., Howlett, J.,Muller, D.J., Persin, C., Pijls, L.T., Rechkemmer, G., Tuijtelaars, S., Verhagen, H.,2005. PASSCLAIM: consensus on criteria. Eur. J. Nutr. 44 (Suppl. 1), i5–30.

Baigent, C., Keech, A., Kearney, P.M., Blackwell, L., Buck, G., Pollicino, C., Kirby, A.,Sourjina, T., Peto, R., Collins, R., Simes, R., 2005. Efficacy and safety ofcholesterol-lowering treatment: prospective meta-analysis of data from90,056 participants in 14 randomised trials of statins. Lancet 366, 1267–1278.

Barendregt, J.J., Veerman, J.L., 2010. Categorical versus continuous risk factors andthe calculation of potential impact fractions. J. Epidemiol. Community Health64, 209–212.

Berger, A., Jones, P.J., Abumweis, S.S., 2004. Plant sterols: factors affecting theirefficacy and safety as functional food ingredients. Lipids Health Dis. 3 (1), 5.

De Jong, N., Klungel, O.H., Verhagen, H., Wolfs, M.C., Ocké, M.C., Leufkens, H.G., 2007.Functional foods: the case for closer evaluation. BMJ 334 (7602), 1037–1039.

De Jong, N., Ros, M., Ocké, M., Verhagen, H., 2008. A general postlaunch monitoringframework for functional foods tested with the phytosterol/-stanol case. TrendsFood Sci. Technol. 19 (10), 535.

Demonty, I., Ras, R.T., van der Knaap, H.C., Duchateau, G.S., Meijer, L., Zock, P.L.,Geleijnse, J.M., Trautwein, E.A., 2009. Continuous dose-response relationship ofthe LDL-cholesterol-lowering effect of phytosterol intake. J. Nutr. 139, 271–284.

Demonty, I., Ras, R.T., Trautwein, E.A., 2011. Comment on: ‘‘A comparison of theLDL-cholesterol lowering efficacy of plant stanols and plant sterols over acontinuous dose range: results of a meta-analysis of randomized, placebocontrolled trials’’ by Musa-Veloso K et al.. Prostaglandins Leucot. Essent. FattyAcids 85, 9–28.

EFSA (European Food Safety Authority), 2005. Opinion of the Scientific Panel onDietetic products, nutrition and allergies [NDA] related to two scientificpublications concerning aspects of serum levels of phytosterols. EFSA J. 211,1–6.

EFSA (European Food Safety Authority), 2006. Risk-Benefit Analysis of Foods,Methods and Approaches, EFSA scientific Colloquium Summary Report 6,Parma, Italy 2006.

EFSA(European Food Safety Authority), 2008. Plant sterols and blood cholesterol –Scientific substantiation of a health claim related to plant sterols and lower/reduced blood cholesterol and reduced risk of (coronary) heart disease pursuantto Article 14 of Regulation (EC) No. 1924/2006[1]. EFSA J. 781, 1–12.

EFSA (European Food Safety Authority), 2009. Scientific opinion of the panel ondietetic products nutrition and allergies on a request from the EuropeanCommission and a similar request from France in relation to the authorisationprocedure for health claims on plant sterols/stanols and lowering/reducingblood LDL-cholesterol pursuant to Article 14 of Regulation (EC) No. 1924/2006.EFSA J. 1175, 1–9.

EFSA (European Food Safety Authority), 2010. EFSA Scientific Committee; guidanceon human health risk–benefit assessment of food. EFSA J. 8 (7), 1673. http://dx.doi.org/10.2093/j.efsa.2010.1673.

EFSA Panel on Dietetic Products, Nutrition and Allergies (NDA), 2012a. Scientificopinion on the substantiation of a health claim related to 3 g/day plant sterols/stanols and lowering blood LDL-cholesterol and reduced risk of (coronary) heartdisease pursuant to Article 19 of Regulation (EC) No. 1924/2006. EFSA J. 10 (5),2693. http://dx.doi.org/10.2903/j.efsa.2012.2693.

EFSA Panel on Dietetic Products, Nutrition and Allergies (NDA), 2012b. Scientificopinion on the substantiation of a health claim related to 3 g/day plant stanolsas plant stanol esters and lowering blood LDL-cholesterol and reduced risk of(coronary) heart disease pursuant to Article 14 of Regulation (EC) No. 1924/2006. EFSA J. 10 (5), 2692. http://dx.doi.org/10.2903/j.efsa.2012.2692.

Eussen, S.R., van der Elst, M.E., Klungel, O.H., Rompelberg, C.J., Garssen, J.,Oosterveld, M.H., de Boer, A., de Gier, J.J., Bouvy, M.L., 2010a. Apharmaceutical care program to improve adherence to statin therapy: arandomized controlled trial. Ann. Pharmacother. 44 (12), 1905–1913.

Eussen, S.R., de Jong, N., Rompelberg, C.J., Garssen, J., Verschuren, W.M., Klungel,O.H., 2010b. Effects of the use of phytosterol/-stanol-enriched margarines onadherence to statin therapy. Pharmacoepidemiol Drug Saf. 19 (12), 1225–1232.

Eussen, S.R., Bouvy, M.L., Rompelberg, C.J., van der Elst, M.E., Garssen, J., Oosterveld,M.H., de Boer, A., de Gier, J.J., van Kranen, H.J., Klungel, O.H., 2011. Influence ofthe use of functional foods enriched with phytosterols/-stanols on adherence tostatin therapy. Pharmacoepidemiol Drug Saf. 20 (8), 830–837.

Fransen, H., de Jong, N., Hendriksen, M., Mengelers, M., Castenmiller, J., Hoekstra, J.,van Leeuwen, R., Verhagen, H., 2010. A tiered approach for risk–benefitassessment of foods. Risk Anal. 30 (5), 808–816.

Gotto Jr, A.M., Bierman, E.L., Connor, W.E., Ford, C.H., Frantz Jr, I.D., Glueck, C.J.,Grundy, S.M., Little, J.A., 1984. Recommendations for treatment ofhyperlipidemia in adults. A joint statement of the Nutrition Committee andthe Council on Arteriocslerosis. Circulation 69, 1065A–1090A.

Gylling, H., Hallikainen, M., Nissinen, M.J., Miettinen, T.A., 2010. The effect of a veryhigh daily plant stanol ester intake on serum lipids, carotenoids, and fat-solublevitamins. Clin. Nutr. 29 (1), 112–118.

Hart, A., Hoekstra, J., Owen, H., Kennedy, M., Zeilmaker, M.J., de Jong, N.,Gunnlaugsdottir, H., this issue. Quantifying uncertainty and variability inhealth risk–benefit assessment using DALYs or QALYs. Food Chem. Toxicol.

Hearty, A.P., Duffy, E., Gibney, M.J., 2008. Intake estimates of naturally occurringphytosterols using deterministic and probabilistic methods in a representativeIrish population. Int. J. Food Sci. Nutr., 1–14.

Hoekstra, J., Verkaik-Kloosterman, J., Rompelberg, C., van Kranen, H., Zeilmaker, M.,Verhagen, H., de Jong, N., 2008. Integrated risk–benefit analyses: methoddevelopment with folic acid as example. Food Chem. Toxicol. 46, 893–909.

Hoekstra, J., Hart, A., Boobis, A., Claupein, E., Cockburn, A., Hunt, A., Knudsen, I.,Richardson, D., Schilter, B., Schutte, K., Torgerson, P.R., Verhagen, H., Watzl, B.,Chiodini, A., 2010. BRAFO tiered approach for benefit-risk assessment of foods.Food Chem. Toxicol. (Epub ahead of print)

Hoekstra, J., Hart, A., Owen, H., Zeilmaker, M., Bokkers, B., Thorgilsson, B.,Gunnlaugsdottir, H., this issue. Fish, contaminants and human health:quantifying and weighing benefits and risks. Food Chem. Toxicol.

Hulshof, K.F., Brussaard, J.H., Kruizinga, A.G., Telman, J., Lowik, M.R., 2003. Socio-economic status, dietary intake and 10 y trends: the Dutch National FoodConsumption Survey. Eur. J. Clin. Nutr. 57, 128–137.

Kalogeras, N., Odekerken, G., Gunnlaugsdottir, H., Magnusson, S.H., Holm, F., Leino,O., Luteijn, J.M., Pohjola, M.V., Tijhuis, M.J., Tuomisto, J.T., Ueland, O., White, B.C.,Verhagen, H., 2012. State of the art in benefit–risk analysis: marketing andfinance. Food Chem. Toxicol. 50, 56–66.

Katan, M.B., Grundy, S.M., Jones, P., Law, M., Miettinen, T., Paoletti, R., 2003. Efficacyand safety of plant stanols and sterols in the management of blood cholesterollevels. Mayo Clin. Proc. 78 (8), 965–978.

42 J. Hoekstra et al. / Food and Chemical Toxicology 54 (2013) 35–42

Law, M., 2000. Plant sterol and stanol margarines and health. BMJ 320 (7238), 861–864.

Law, M.R., Wald, N.J., Thompson, S.G., 1994. By how much and how quickly doesreduction in serum cholesterol concentration lower risk of ischaemic heartdisease? BMJ 308 (6925), 367–372.

Law, M.R., Wald, N.J., Rudnicka, A.R., 2003. Quantifying effect of statins on lowdensity lipoprotein cholesterol, ischaemic heart disease, and stroke: systematicreview and meta-analysis. BMJ 326 (7404), 1423.

Luteijn, J.M., White, B.C., Gunnlaugsdottir, H., Holm, F., Kalogeras, N., Leino, O.,Magnusson, S.H., Odekerken, G., Pohjola, M.V., Tijhuis, M.J., Tuomisto, J.T.,Ueland, O., McCarron, P.A., Verhagen, H., 2012. State of the art in benefit-riskanalysis: medicines. Food Chem.Toxicol. 50, 26–32.

Magnusson, S.H., Gunnlaugsdottir, H., van Loveren, H., Holm, F., Kalogeras, N., Leino,O., Luteijn, J.M., Odekerken, G., Pohjola, M.V., Tijhuis, M.J., Tuomisto, J.T., Ueland,O., White, B.C., Verhagen, H., 2012. State of the art in benefit-risk analysis: foodmicrobiology. Food Chem. Toxicol. 50, 33–39.

Malini, T., Vanithakumari, G., 1993. Effect of beta-sitosterol on uterinebiochemistry: a comparative study with estradiol and progesterone. Biochem.Mol. Biol. Int. 31 (4), 659–668.

Mellanen, P., Petanen, T., Lehtimaki, J., Makela, S., Bylund, G., Holmbom, B., Mannila,E., Oikari, A., Santti, R., 1996. Wood-derived estrogens: studies in vitro withbreast cancer cell lines and in vivo in trout. Toxicol. Appl. Pharmacol. 136, 381–388.

Mensink, R.P., de Jong, A., Lütjohann, D., Haenen, G.R., Plat, J., 2010. Plant stanolsdose-dependently decrease LDL-cholesterol concentrations, but notcholesterol-standardized fat-soluble antioxidant concentrations, at intakes upto 9 g/d. Am. J. Clin. Nutr. 92 (1), 24–33.

Miettinen, T.A., Strandberg, T.E., Gylling, H., 2000. Noncholesterol sterols andcholesterol lowering by long-term simvastatin treatment in coronary patients:relation to basal serum cholestanol. Arterioscler. Thromb. Vasc. Biol. 20 (5),1340–1346.

Murray, C.J., 1994. Quantifying the burden of disease: the technical basis fordisability-adjusted life years. Bull. World Health Organ. 72, 429–445.

Murray, C.J., Ezzati, M., Lopez, A.D., Rodgers, A., Vander Hoorn, S., 2003. Comparativequantification of health risks conceptual framework and methodological issues.Popul. Health Metr. 1, 1pp.

Musa-Veloso, K., Poon, T.H., Elliot, J.A., Chung, C., 2011. A comparison of the LDL-cholesterol lowering efficacy of plant stanols and plant sterols over acontinuous dose range: results of a meta-analysis of randomized, placebo-controlled trials. Prostaglandins Leukot. Essent. Fatty Acids.

Naito, Y., Nagata, T., Takano, Y., Nagatsu, T., Ohara, N., 2003. Rapeseed oil ingestionand exacerbation of hypertension-related conditions in stroke pronespontaneously hypertensive rats. Toxicol. 187 (2-3), 205–216.

Noakes, M., Clifton, P., Ntanios, F., Shrapnel, W., Record, I., McInerney, J., 2002. Anincrease in dietary carotenoids when consuming plant sterols or stanols iseffective in maintaining plasma carotenoid concentrations. Am. J. Clin. Nutr. 75(1), 79–86.

Ntanios, F.Y., Duchateau, G.S., 2002. A healthy diet rich in carotenoids is effective inmaintaining normal blood carotenoid levels during the daily use of plant sterol-enriched spreads. Int. J. Vitam. Nutr. Res. 72 (1), 32–39.

Pohjola, M.V., Leino, O., Kollanus, V., Tuomisto, J.T., Gunnlaugsdottir, G., Holm, F.,Kalogeras, N., Luteijn, J.M., Magnusson, S.H., Odekerken, G., Tijhuis, M.J., Ueland,O., White, B.C., Verhagen, H., 2012. State of the art in benefit–risk analysis:environmental health. Food Chem. Toxicol. 50, 40–55.

Polagruto, J.A., Wang-Polagruto, J.F., Braun, M.M., Lee, L., Kwik-Uribe, C., Keen, C.L.,2006. Cocoa flavanol-enriched snack bars containing plant sterols effectivelylower total and low-density lipoprotein cholesterol levels. J. Am. Diet Assoc. 106(11), 1804–1813.

Ratnayake, W.M., L’Abbe, M.R., Mueller, R., Hayward, S., Plouffe, L., Hollywood, R.,Trick, K., 2000. Vegetable oils high in phytosterols make erythrocytes less

deformable and shorten the life span of stroke-prone spontaneouslyhypertensive rats. J. Nutr. 130, 1166–1178.

SCF, 2000. Opinion of the Scientific Committee on Food on a Request on the SafetyAssessment of the Use of Phytosterolesters in Yellow Fat Spread. EC, Brussels,Belgium (SCF/CS/NF/DOS/1 FINAL).

SCF, 2002. General view of the Scientific Committee on Food on the Long-TermEffects of the Intake of Elevated Levels of Plant Sterols from Multiple DietarySources, with Particular Attention to the Effects on Beta-Carotene. EC, Brussels,Belgium (SCF/CS/NF/DOS/20 ADD 1 Final).

SCF, 2003a. Opinion of the Scientific Committee on Food on applications for avariety of plant sterol-enriched foods. EC, Belgium (SCF/CS/NF/DOS/15 ADD2Final).

SCF, 2003b. Opinion of the Scientific Committee on Food on an application fromADM for approval of plantsterol-enriched foods. EC, Brussels, Belgium (SCF/CS/NF/DOS/23 ADD2 Final).

Stouthard, M.A.E., Essink-Bot, M.L., Bonsel, G.J., Barendregt, J.J., Kramer, P.G.N.,Gunning-Schepers, L.J., van der Maas, P.J., 1997. Disability Weights for Diseasesin The Netherlands. Department of Public Health Erasmus UniversityRotterdam.

Sudhop, T., Gottwald, B.M., von Bergmann, K., 2002. Serum plant sterols as apotential risk factor for coronary heart disease. Metabolism 51, 1519–1521.

Talati, R., Sobieraj, D.M., Makanji, S.S., Phung, O.J., Coleman, C.I., 2010. Thecomparative efficacy of plant sterols and stanols on serum lipids: asystematic review and meta-analysis. J. Am. Diet Assoc. 110 (5), 719–726.

Tijhuis, M.J., de Jong, N., Pohjola, M.V., Gunnlaugsdóttir, H., Hendriksen, M.,Hoekstra, J., Holm, F., Kalogeras, N., Leino, O., van Leeuwen, F.X., Luteijn, J.M.,Magnússon, S.H., Odekerken, G., Rompelberg, C., Tuomisto, J.T., Ueland, O.,White, B.C., Verhagen, H., 2012a. State of the art in benefit–risk analysis: foodand nutrition. Food Chem. Toxicol. 50, 5–25.

Tijhuis, M.J., Pohjola, M.V., Gunnlaugsdóttir, H., Kalogeras, N., Leino, O., Luteijn, J.M.,Magnússon, S.H., Odekerken-Schröder, G., Poto, M., Tuomisto, J.T., Ueland, O.,White, B.C., Holm, F., Verhagen, H., 2012b. Looking beyond borders: Integratingbest practices in benefit–risk analysis into the field of food and nutrition. FoodChem. Toxicol. 50, 77–93.

Ueland, O., Gunnlaugsdottir, H., Holm, F., Kalogeras, N., Leino, O., Luteijn, J.M.,Magnusson, S.H., Odekerken, G., Pohjola, M.V., Tijhuis, M.J., Tuomisto, J.T.,White, B.C., Verhagen, H., 2012. State of the art in benefit–risk analysis:consumer perception. Food Chem. Toxicol. 50, 67–76.

Verhagen, H., Vos, E., Francl, S., Heinonen, M., van Loveren, H., 2010. Status ofnutrition and health claims in Europe. Arch. Biochem. Biophys. 510, 6–15.

Verhagen, H., Andersen, R., Antoine, J.M., Finglas, P., Hoekstra, J., Kardinaal, A.,Nordmann, H., Pekcan, G., Pentieva, K., Sanders, T.A., van den Berg, H., vanKranen, H., Chiodini, A., 2011. Application of the BRAFO tiered approach forbenefit–risk assessment to case studies on dietary interventions. Food Chem.Toxicol. (Epub ahead of print).

Verhagen, H., Tijhuis, M.J., Gunnlaugsdottir, H., Kalogeras, N., Leino, O., Luteijn, J.M.,Magnusson, S.H., Odekerken, G., Pohjola, M.V., Tuomisto, J.T., Ueland, O., White,B.C., Holm, F., 2012. State of the art in benefit–risk analysis: introduction. FoodChem. Toxicol. 50, 2–4.

Verschuren, W.M., Blokstra, A., Picavet, H.S., Smit, H.A., 2008. Cohort profile: theDoetinchem cohort study. Int. J. Epidemiol. 37, 1236–1241.

Waijers, P.M.C.M., Feskens, E.J.M., Vitamine A voorziening van de Nederlandsebevolking, RIVM rapport 350010002/2004. http://www.rivm.nl/bibliotheek/rapporten/350010002.html. (in Dutch).

Watzl, B., Gelencsér, E., Hoekstra, J., Kulling, S., Lydeking-Olsen, E., Rowland, I.,Schilter, B., van Klaveren, J., Chiodini, A., 2011. Application of the BRAFO-tieredapproach for Benefit–risk assessment to case studies on natural foods. FoodChem. Toxicol. doi:10.1016/j.fct.2011.02.010 (Epub ahead of print).

Website CBS Statline http://statline.cbs.nl/statweb/.Website Nationaal Kompas http://www.nationaalkompas.nl.