The Choices Programme is an internationally applicable nutrient profiling system with nutrition criteria for trans fatty acids (TFA), saturated fatty acids, sodium, added sugar and for some product groups energy and fibre. These criteria determine whether foods are eligible to carry a “healthier option” stamp. In this paper a nutrient intake modelling method is described to evaluate these nutritional criteria by investigating the potential effect on nutrient intakes.
Data were combined from the 2003 Dutch food consumption survey in young adults (aged 19–30) and the Dutch food composition table into the Monte Carlo Risk Assessment model. Three scenarios were calculated: the “actual intakes” (scenario 1) were compared to scenario 2, where all foods that did not comply were replaced by similar foods that did comply with the Choices criteria. Scenario 3 was the same as scenario 2 adjusted for the difference in energy density between the original and replacement food. Additional scenarios were calculated where snacks were not or partially replaced and stratified analyses for gender, age, Body Mass Index (BMI) and education.
Calculated intake distributions showed that median energy intake was reduced by 16% by replacing normally consumed foods with Choices compliant foods. Intakes of nutrients with a maximal intake limit were also reduced (ranging from −23% for sodium and −62% for TFA). Effects on intakes of beneficial nutrients varied from an unintentional reduction in fat soluble vitamin intakes (−15 to −28%) to an increase of 28% for fibre and 17% calcium. Stratified analyses in this homogeneous study population showed only small differences across gender, age, BMI and education.
Citation: Roodenburg AJC, van Ballegooijen AJ, Dötsch-Klerk M, van der Voet H, Seidell JC (2013) Modelling of Usual Nutrient Intakes: Potential Impact of the Choices Programme on Nutrient Intakes in Young Dutch Adults. PLoS ONE 8(8): e72378. https://doi.org/10.1371/journal.pone.0072378
Editor: François Blachier, National Institute of Agronomic Research, France
Received: March 31, 2013; Accepted: July 9, 2013; Published: August 28, 2013
Copyright: © 2013 Roodenburg et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This study was supported by a grant of The Netherlands Organisation for Scientific Research (NWO). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: AR was employed and AJvB was an intern at Unilever R&D, The Netherlands at the time of the study. MDK is still employed at Unilever R&D. There are no patents, products in development or marketed products to declare. This does not alter the authors′ adherence to all the PLOS ONE policies on sharing data and materials.
Overconsumption of energy dense, nutrient poor diets is one of the largest problems in modern society, resulting in an increasing prevalence of chronic, non-communicable diseases in many countries . In the Dutch population , , but also in other populations, adherence to dietary guidelines is low .
Following a diet consistent with dietary recommendations  may reduce the risk of chronic diseases. Therefore in 2004, WHO launched the Global Strategy on Diet, Physical Activity and Health  in which one of the recommendations to the private sector was to limit the levels of TFA, SAFA, salt and free sugars in existing products in order to contribute to reducing the burden of chronic diseases .
This discrepancy between dietary recommendations and actual intakes forms the basis of various initiatives in defining targets for food reformulation , . A nutrient profiling system, which is a systematic method for categorising foods according to their nutritional quality, is often used as basis for food reformulation. Over the years, different systems have been launched, each with another approach and purpose –. It is difficult to compare these systems as there is no gold standard for comparison, however, validation of nutrient profiling systems is of paramount importance .
The Choices Programme is an internationally applicable nutrient profiling system with criteria that determine whether foods are eligible to carry a “healthier option” stamp . The aims of the Choices Programme are to stimulate product reformulation and to help consumers by making healthier choices easier to identify. To develop the nutrient profiles for the Choices Programme, the generic criteria for energy and the key nutrients (TFA, SAFA, sodium, added sugar and fibre) were based on international nutrient intake recommendations for daily diets . There is increasing recognition and appreciation of the stamp by consumers  and since the launch of the initiative in 2006 it has driven food reformulation into a more healthy direction . The ultimate goal is to meet the recommendations for population intakes. It is hypothesized that if consumers choose food products that comply with these criteria, the calculated daily intake of the key nutrients should improve in the direction of the nutrient intake recommendations. We evaluated the potential impact of Choices on a broad range of nutrient intakes in a Dutch population of young adults aged 19–30 years using a Monte Carlo Risk Assessment (MCRA) model .
This paper builds on a short communication published earlier . Now we give a more in depth description of the methodology used; we expanded the analyses to a broader set of nutrients (including vitamins and minerals); we studied the effects of replacing snacks in more detail; we also evaluated whether effects differed between gender, BMI and level of education, to illustrate the possibilities of the intake modelling methodology used.
Background on the Modelling of Usual Nutrient Intakes
Usual nutrient intake is defined as the long-run average of daily nutrient intakes aggregated over all foods consumed. At the individual level, usual intake is generally unknown because only a few days are observed. However, the distribution of usual intakes in a population can be estimated from short-term measurements such as repeated 24-h recalls by statistical methods , .
In the current survey, the available consumption data were two 24-h dietary recalls for 750 persons . First, multiplying daily food consumption with nutrient levels per food gives the daily intakes per food on the two days per person for which consumption data are available. Second, aggregating these intakes over all foods gives the total nutrient intake per person per day. A simple estimate of usual intake is to calculate mean intake over the available two days. However, these so-called observed individual means (OIMs) measure the actual usual intake with appreciable random error, and as a consequence the distribution of OIMs is too broad: both the low-end tail and the high-end tail of the OIM distribution over-estimate the frequency of true low and high intakes (see Figure 1, green curve, mean two days).
A simple estimate of usual intake is to calculate mean intake over the available two days. However, this measure still includes appreciable random error, resulting in a too broad distribution: Both the low-end tail and the high-end tail of the “Mean for two days” distribution over-estimate the frequency of true low and high intakes.
For nutrients, estimating usual intake is done as follows: first transform the data to a scale where the distribution is approximately normal. This is needed because the error-correction model is based on this assumption. A simple transformation often used for this purpose is the Box-Cox transformation . The second step is fitting a variance components model to the transformed intakes to distinguish between between-person and within-person variation: each intake for a person-day is modelled as , where for individual i the term ui represents the usual intake and eij is the intake on day j minus the usual intake. By fitting the model, the variance of ui (Varu) (between-person variance) and the variance of eij (Vare) (within-person variance) are estimated. The total variance is the sum of the two variance components: . For the purpose of assessing the sufficiency of nutrient intakes day-to-day variations (within person) are irrelevant because health status is generally reflecting long-term rather than short-term intake. Therefore at this stage, to represent the usual intake distribution, the within-person variation is no longer needed. We can continue with only the between-person variation around the mean. Finally, the estimated distribution of ui is back-transformed to the original scale. An indicator to characterise the difference between the distribution of single-day intakes and the estimated distribution of usual intakes is the so-called “shrinkage factor” which is defined as the ratio of standard deviations for between-person and total variation at the transformed scale ().
In this study, we applied the BetaBinomial-Normal (BBN) model that has been developed to quantify the risk of exposure to chemicals from the diet , , . This model can be applied for nutrients as well as foods, where many of the intakes -at day level- are often zero. In which case, in addition to the normal distribution fitted to the transformed intake amounts, the BetaBinomial part of the model estimates the frequency of consumption of non-daily used foods. However, the BetaBinomial part is not needed for nutrients that are consumed daily as in the current study.
Not only intakes vary between persons, but also nutrient requirements. If assumptions are satisfied, the Average Nutrient Requirement (ANR) cut-point method ,  provides a way to estimate the prevalence of inadequate nutrient intake in a population: by calculating the proportion of estimated usual intake distribution below the ANR. Note that the Average Nutrient Requirement (ANR) and not the level recommended for most individuals (Recommended Daily Allowance, RDA) should be used as a point of comparison, as is fully explained in reference .
Assumptions are that intakes are accurately measured and corrected for within-person error, that intakes and requirements are uncorrelated, that the requirements distribution is symmetrical, that the actual prevalence is not very high or very low, and that the variability of requirements in the population can be assumed to be much smaller than the variability in usual intakes.
The population nutrient intake goals of FAO/WHO are also meant to represent average rather than individual values in a healthy population , therefore the cut-point method using these limits is appropriate to estimate the proportion of inadequate (too high) intakes of saturated fatty acids, transfatty acids, sodium and added (free) sugar.
For a selection of nutrients (carbohydrates, protein, fat, SAFA, TFA, sodium, total sugar, fibre, polyunsaturated fatty acids (PUFA), monounsaturated fatty acids (MUFA), calcium, potassium, iron, folic acid, vitamin A, B1, B2 B6, B12, C, D, E) and energy, three scenarios were calculated and compared: In scenario 1 data from the Dutch National Food Consumption Survey 2003  were used to estimate the usual intake distribution (see Background). In this paper we identify this basic estimate of the usual intake distribution as the “Actual intake”. This was compared to scenario 2, which was the same as scenario 1, except that all foods that did not comply with the Choices criteria were replaced –where possible– by similar foods that did comply with the Choices criteria. Scenario 3 was the same as scenario 2, but corrected for the difference in energy density between the original and the replacement food. Two additional scenarios (4 and 5) were calculated for a selection of nutrients: In scenarios 4 and 5, snacks were either not replaced or partially replaced, to estimate the contribution of the different types of snacks to the nutrient intakes.
We applied the BetaBinomial-Normal (BBN) model that was developed to quantify the risk of exposure to chemicals from the diet , , . The BBN model is available in the Monte Carlo Risk Assessment (MCRA) program . Inputs for this model were food consumption data  and food composition data  as described below. We estimated optimal Box-Cox transformations for achieving approximate normality, and checked by visual inspection if a normal approximation was appropriate .
Estimating the usual intake distribution allows to calculate not only the mean nutrient intake, but also the percentage of the population at risk of not complying with nutrient requirements. In addition, the contribution of foods to nutrient intakes was given.
The usual nutrient intake distributions were compared with intake limits to estimate the percentage of the population meeting the nutrient requirements using the “cut-point method” (see Background)). For the intake limits of beneficial nutrients we used the Average Nutrient Requirements (ANRs), which is the international harmonized term as proposed by the United Nations University , also known as Estimated Average Requirements (EARs) . For a selection of nutrients for which intakes need to be limited, maximal intake limits were used as defined by the FAO/WHO (population nutrient intake goals)  and by the Health Council for the Netherlands (for sodium) .
Food Consumption Data
Food consumption data were based on the Dutch National Food Consumption Survey 2003 . This survey was conducted among 750 Dutch young adults (aged 19–30) by trained dieticians using two independent computerized 24-h dietary recalls.
Food Composition Data and Food Replacement Scenarios
Food composition data from the Dutch food composition database (NEVO, 2006)  were evaluated against the Choices criteria . For scenario 1, this food composition database was used as published. For scenarios 2 and 3, all foods that were reported to be consumed by the participants in the survey and that did not comply with the Choices criteria were replaced by similar foods that did comply with the Choices criteria. In a few cases it was not possible to find a replacement food. In these cases, the food was either not replaced (e.g. egg) or occasionally replaced by a better alternative, which still did not comply with the criteria (e.g. chocolate was replaced by sugar free chocolate). In total around 430 foods out of the more than 1600 items in Dutch Food composition database  were replaced.
A small number of new foods with the Choices stamp were added to the database and used as replacement foods. Information on the composition of these products was taken from the label. In addition, about 350 foods and their compositions were added to the food composition table. These foods were consumed less frequently in the 2003 survey, and therefore not taken up in the published food composition table. For these foods the sodium, PUFA and MUFA content needed to be estimated based on similar food products of the food composition database. This approach enabled us to estimate the potential (maximum) shift in intakes, while staying as close as possible to the eating habits.
Added sugar is one of the key nutrients with criteria for logo eligibility. However, in the Dutch food composition database, there are only data available on total sugar. For the evaluation of foods against the Choices criteria, estimates for added sugar were used based on other food composition databases , . The calculations of the outcome variables was based on the original food composition data  and are therefore given for total sugar only. In Table 1 examples of replacements are given.
For scenario 3, an energy adjustment was applied: Scenario 2 showed a reduction in energy intake with consumption of a Choices compliant diet. It was hypothesised that consumers may compensate for this decrease in energy intake by eating more of a food with a lower energy density (kcal/g). Therefore, when a food (e.g. full fat 48+ cheese: 384 kcal/100 g) was replaced by a food with a lower energy density (in this case: reduced fat 30+ cheese: 279 kcal/100 g), a multiplication factor was applied (in this case: 384/279 = 1.38) so that the total amount of consumed energy was the same as the amount of energy delivered by the food that was replaced. This was done on a product by product basis with exceptions for low calorie soft drinks and some dairy and meat products, because correction would lead to unrealistic amounts being consumed. For these foods, an upper limit was set at 300% for meat and dairy products and 100% for soft drinks.
For scenario 2 all non-complying snacks were replaced by healthier alternatives. There were however, very few snacks in the food composition table  that were eligible for a healthier choice stamp. To replace all snacks, including apple pies, chocolate bars and other energy dense foods by the few healthier alternatives such as rice wafers and muesli bars seemed unrealistic. Therefore scenario 4 and 5, were calculated, in which snacks were either not replaced or partially replaced. In the partially replaced snack scenario, snacks were only replaced when the healthier alternative was reasonably realistic for e.g. a biscuit or a cookie. High energy apple pies with whipped cream or chocolate bars were not replaced (Table 1). This resulted for this scenario, in the replacement of approximately 40% of the snacks.
Stratified analyses were performed to explore if the scenarios would have a different effect for subgroups. Intake distributions were stratified for gender (men/women) age in years (<25/≥25), BMI kg/m2 (<25/≥25) and education (low/middle/high) and compared to the results of the actual scenario.
Changes in Nutrient Intakes
Figure 2a shows the percentage change in median nutrient intakes compared to the “actual” intake scenario for energy, fat, SAFA, TFA, sodium, total sugar, fibre and also for protein, total carbohydrates, monounsaturated fatty acids (MUFA) and polyunsaturated fatty acids (PUFA). A reduction was seen for energy intake (−16%), as well as for nutrients with a maximal intake limit (between −23% for sodium and −62% for TFA); but also for total carbohydrates (−16%), MUFA (−31%) and PUFA (−8%). When the data were adjusted for energy intake, these reductions were still present, but reversed for PUFA (+2%).
Results are expressed as percentage change in median intakes for macronutrients (A) and micronutrients (B) if the Dutch population (aged 19–30) would eat ‘only’ foods that comply with Choices (with and without adjustment for differences in energy density) as compared to the “actual” intake. Assuming a normal distribution, median values approximate average intake levels. SAFA: saturated fatty acids, TFA: trans fatty acids, MUFA: mono unsaturated fatty acids, PUFA: poly unsaturated fatty acids.
Figure 2b shows the percentage change in median nutrient intakes for micronutrients compared to the “actual” intake scenario: calcium, iron, folate (natural), and the vitamins A (as retinol equivalents), B1, B2, B6, B12 C, D, E. Increases were seen for fibre (28%), calcium (17%), iron (14%) and folate (5%) intake when the Choices scenario was applied. Increases were shown as well for vitamin B1 (8%) and B6 (7%). All these changes were even larger when the data were adjusted for energy. A decline was observed for potassium (−12%) and vitamin B12 (−5%) although the effects were neutralised or even reversed when adjusted for energy (−2%) and (24%) respectively.
For the fat soluble vitamins negative changes were observed (between −28% for vitamin A, −15% for vitamin D and −25% for vitamin E). The decline was proportional compared to the decrease in total fat intake (−29%) and became smaller after adjustment for energy intake.
Comparison with Dietary Recommendations
The usual intake distributions were compared to nutrient requirements (Population intake goals and ANRs, Table 2 and 3). For this comparison the BBN model was applied: When usual intakes were estimated, variation is reduced for many nutrients, when compared with intakes on a person-day basis calculated directly from the survey data (see Background). This so-called “shrinkage factor” for the nutrients (in the “actual” scenario) ranged from 0.45 to 0.71. To illustrate this, for example, without this correction, the percentage of the population with insufficient vitamin B12 intakes in the “actual” scenario (shrinkage factor 0.53) would have been estimated erroneously as 24.3% instead of 5.9% (Table 3).
With the “actual” scenario, intake limits for the key nutrients SAFA, TFA, sodium and total sugar were not reached by 88, 51, 81 and 95% of the population respectively. After replacing non-complying products with products complying with Choices criteria, the proportion of the population not meeting recommended intake limits reduced to 36, 0.1, 40, and 71% respectively. After adjustments for energy, the percentage that did not comply with the criteria for SAFA, TFA, sodium and total sugar increased (Table 2).
The results for beneficial nutrients varied. After replacement with products complying with the Choices criteria, the percentage of the population at risk of intakes lower than the Average Nutrient Requirements (ANRs) decreased for iron, most of the B-vitamins and vitamin C. However, the opposite was shown for the fat soluble vitamins (A, E). The percentages of the population at risk of lower intakes than the ANRs for vitamin A and vitamin E increased to 34% and 91%, respectively (Table 3). For vitamin D, calcium, fibre and potassium no ANR was established.
Table 4 shows the intake distributions for energy SAFA, TFA, sodium, and total sugar for the five different scenarios, including the different snack-scenarios. When evaluating the median values (Figure 3), the snacks scenarios indicated that not replacing snacks with Choices compliant alternatives resulted in substantially higher intakes for sugar (12%) and SAFA (8%) and especially TFA (68%), indicating snacks being important sources of TFA and also, but to a lesser extent, sugar and SAFA. For energy and sodium, these effects were much smaller (1 and 4%, respectively). The intake distributions became narrower for the Choices scenario and the snack scenarios, as can be seen by lower inter-quartile ranges (IQRs). This indicates less variability and shorter tails of the distributions (Table 4, Figure 4). This is also illustrated for TFA in Figure 4 where the distributions are shown for all five scenarios.
Results are expressed as percentage difference of median nutrient intakes for the different scenarios as compared to the Choices scenario: If the Dutch population (aged 19–30) would eat ‘only’ foods that comply with Choices (set at = 0), with adjustment for differences in energy density (Choices energy adjusted), with partial or with no replacements of snacks (Snacks, partially or not replaced). Assuming a normal distribution, median values approximate average intake levels. SAFA: saturated fatty acids, TFA: trans fatty acids.
“Actual” TFA intake is compared to the Choices scenario: if the Dutch population (aged 19–30) would eat ‘only’ foods that comply with Choices; with and without adjustment for differences in energy density (Choices and Choices, energy adjusted); with partial or with no replacements of snacks (Snacks, partially or not replaced). Maximal intake limit for TFA is 2.2 g/d (1 en%, 2000 kcal).
In Table 5 the foods that contribute to the intake of the SAFA and vitamin A are given. For SAFA, dairy products (milk, cheese), meat and fats are major contributors in the “actual” scenario. For the Choices scenarios, major contributors to SAFA intake are dairy products and snacks.
For vitamin A, intakes were substantially lower with the Choices scenarios. Table 5 shows that when high-fat, liver-based meat products, were in the Choices scenarios replaced by meat sources with lower vitamin A content, this resulted in a larger contribution of carrots, and fats to overall vitamin A intake. Carrots are in general 10–20 times less rich in vitamin A (or carotene) as compared to liver based meat products . In the Choices scenarios, foods that contribute to the intake of vitamin E shift from mayonnaise as a main contributor towards nut-based products. In addition, fats remained an important source of vitamin E. In all scenarios, vitamin D came from fat, meat and fish products (data not shown).
The results of the analysis stratified for gender showed only marginal differences. In general, women showed slightly larger changes than men −65.0% vs. −60.0% for TFA, respectively, when the Choices criteria were applied, except for most vitamins were the effects in men were slightly larger (vitamin E, −20.9% women vs. −27.9% men). A close look at the BMI groups showed that the values of the usual intakes of the high BMI group were consequently lower than the low BMI group (energy 2137 vs. 2343 kcal). However, despite these differences in intakes, the Choices scenario and the Choices energy adjusted scenario had comparable effects for different BMI groups. Stratification for education and age resulted only in minor differences in impact between the groups (data not shown).
In this nutrient intake modelling study we used the MCRA model to investigate the potential impact of the Choices Programme on nutrient intakes. In general the nutrient intake distributions based on the 2003 Dutch food consumption survey shifted into the direction of the recommendations  when non-complying foods were replaced with Choices compliant products.
Using nutrient intake modelling, we can evaluate whether a nutrient profiling system is able to improve the nutritional quality of consumers’ diets. The calculation of intake distributions enabled us to calculate the fraction of the population that has a long-term average nutrient intake which is above or below nutrient requirements (FAO/WHO population nutrient intake goals or ANR). This approach allows a quantitative look into effects of specific reformulations, on intakes from specific food groups, such as snacks, possible differences in intakes related to population characteristics (such as BMI, gender and education). The model calculated the contribution of specific foods to nutrient intakes (Table 5) and has the possibility to include market share information, as done by Temme et al. .
The results for the nutrients saturated fatty acids (SAFA), trans fatty acids (TFA), sodium, sugar, energy and fibre, for which criteria are defined for logo eligibility in the Choices Programme, showed a shift in intake distributions in a beneficial direction when the population would consume a diet conform the Choices criteria. With regard to total sugar intakes, the results were less strong. While there was an improvement in the percentage of the population meeting the recommended intake limit for sugar, a substantial proportion of the population would still consume more than is recommended when non-compliant foods were replaced with Choices compliant foods. One reason for this could be that the Choices criteria for added sugar are not strict enough. However evidence for an optimal level of added sugar intake does not allow firm conclusions .
Regarding the macronutrients, protein intake increased probably due to the increased consumption of low-fat, protein-rich, animal based products. The replacement of non-complying foods with Choices compliant products decreased the intakes of carbohydrates and unsaturated fatty acids. Regarding the minerals and micronutrients, potassium intakes were reduced. In addition, the percentage of the population that had estimated intakes below the Average Nutrient Requirements (ANR) was higher for the fat-soluble vitamins (A and E). This unintentional effect could be related to the reduction in total fat intakes and the replacement of foods such as high-fat, liver-based meat products that are specifically high in, for example, vitamin A (Table 5). Also, high fat mayonnaise, which is non-compliant because of energy density, was a good source of vitamin E. For vitamin D, it was less clear which food replacement caused the reduction. Intakes of micronutrients like calcium, iron and the water soluble vitamins were higher with the Choices scenarios, resulting in a reduction of the percentage of the population at risk for inadequacies. It must be noted that the ANR represents the average daily nutrient intake level estimated to meet the requirement of half of the healthy individuals in a defined population , . De Lauzon et al. validated and justified the ANR as a cut-point for the prevalence of nutrient inadequacy at a population level . For nutrients for which no ANR was defined such as vitamin D, calcium, fibre and potassium, this cannot be applied. We compared the percentages of the population at risk for inadequacies in the “actual” scenario with those in the same age group in the latest Dutch survey . For iron, vitamin B1, B6, B12 and C, the numbers were comparable. We found a higher percentage of the population at risk for low intakes of folate, vitamin B2 and E, while for vitamin A this percentage was lower. Some of these differences (folate and vitamin E) were larger than expected based on anticipated differences based on weighing factors or effect of time (2003 vs. 2010) between two surveys of the same age group in the same population. Therefore these absolute percentages of the population at risk for inadequacies need to be interpreted with caution. Alternatively, the relatively low intakes of some of the beneficial nutrients might also be related to the fact that food composition data is incomplete for some of the micronutrients, leading to underestimated intakes and higher estimated risks for inadequacies.
A way to overcome the unintentional negative side-effects on mineral and micronutrient intakes could be the inclusion of criteria for these nutrients in the Choices Programme as is done in other nutrient profiling systems , . This would, however, complicate implementation of such a nutrient profiling system, because it is necessary to collect or analyse these data, for the evaluation of foods for stamp eligibility. Recently, effects on product group-specific (micro)nutrient intakes of different nutrient profile systems have been estimated . The authors illustrated that a simultaneous decrease of estimated vitamin D and calcium intakes with a reduction in SAFA and sodium from dairy products, is affected by the type of nutrient profiling system and the inclusion of criteria for beneficial nutrients .
To compensate for the difference in energy density between the original and the replacement foods a multiplication factor was applied. The energy adjusted scenario can be seen as the “worst case scenario”, assuming consumers may compensate for the decrease in energy intake when switching to Choices compliant products. Nevertheless, results still showed a substantial shift in intake distributions into a more favourable direction. Stratified analysis for gender, BMI, education and age resulted only in marginal differences between the subgroups. This means that Choices had comparable effects for all subgroups in terms of potential impact on nutrient intakes.
In addition to the studies mentioned above , , also others reported on food replacement scenarios, mostly with a focus on SAFA only –. Schickenberg et al  replaced up to three products from three different product groups with low-SAFA alternatives, resulting in a mean reduction of 13.4 g SAFA. Lloyd Williams et al  replaced only one snack with a healthier alternative resulting in a smaller reduction of 4.4 g SAFA per day (−120 kcal). In the present study the median SAFA reduction for a total diet was larger (13.2 g, Table 3). However, we estimated a smaller contribution of snacks to median SAFA intake of 1.5 g (Table 4). An explanation for the different findings in these studies ,  can be that in our study the replacements were chosen to stay as close as possible to the normal situation. For example, to define a ‘realistic’ replacement, we replaced full fat cheese (Gouda 48+) by cheese 30+ instead of cheese 10+ as was done by Schickenberg et al . The potential effect of Choices on the general Dutch population (18–70 y) was studied by Vyth et al  who showed a reduction of average SAFA intake between 10 to 15 en% and for TFA 0.6 to 1.0 en%. These numbers related to a broader age-range of the Dutch population (18–70 y) are in the same range as our values (SAFA from 9 to 13 en% and TFA from 0.9 to 0.4 en%) for a population aged 19–30 y. In addition, results for SAFA, TFA and sodium intakes were also comparable as estimated in a study using a different methodology (Daily Menu Method) for measuring potential effects of the Choices Programme on nutrient intakes . Estimated intakes of (added) sugar were less reduced in the present study as compared to our earlier work, which was more subjective and lacked good added sugar data.
However, the present study may also have some limitations. Despite the fact that replacements were chosen on a product by product basis, dependent on the decisions of two nutritional experts, this may however still be susceptible to some subjectivity and bias. Also product acceptability is not taken into consideration. Consumers might prefer other alternatives. For sweet and salty snacks there were few Choices compliant replacements, thus the same replacement food has been used for a large number of snacks (Table 1), leading to unrealistic high consumption of a few snack products. Furthermore, snacks are usually eaten for indulgence; therefore it is unrealistic to assume that consumers will replace all snacks with one and the same healthier alternative.
Moreover, the use of generic food composition data such as NEVO  could be criticized because nutrient profiling systems are intended to be implemented on real food products. Many branded products are not included, meaning that several products are averaged to one nutrient composition, which may have influenced individual intakes. New products are not in the food composition table. Nutrient data may not be correctly measured or calculated. But, these generic food composition databases are the best open source of nutritional information of a wide range of foods and are used in surveys. Alternatives are food composition data from labels, especially useful for countries where labelling is obligatory. The role of food composition data in the benchmarking and evaluations of foods highlights the need for good up-to-date, quality data on nutrient composition of foods in the supermarket. This is a challenge for the fast changing global food supply, which is the reality in the current market.
Another limitation is the fact that survey data are known to be associated with underreporting of intake especially in subjects with a higher BMI . Therefore, results indicating the percentage of the population that complied with the nutrient requirements have to be interpreted with caution since it may be lower (for nutrients to limit) or higher (for beneficial nutrients). However, stratified analyses indicated that BMI was not likely a disturbing factor in the present study.
Despite mentioned limitations, this study offers opportunities for future research which may include quantifying the beneficial effects of these same modest changes on health outcomes for example chronic diseases and the disease burden. For example, Lloyd-Williams  showed that approximately 6000 cardiovascular deaths could be prevented annually in the UK by reducing cholesterol levels with 0.054 mmol/L and salt intake with 0.5 g on population level. Vyth et al. estimated slight positive potential effects on cholesterol levels .
In conclusion, data from this study in a Dutch young adult population shows the potential beneficial effects of Choices on nutrient intakes. There are, however concerns for some beneficial nutrients. Especially estimated fat-based beneficial nutrient intakes were unintentionally reduced with replacement of high-fat foods by alternatives with a better fat quality or lower energy content. It is recommended to study this effect in more detail. For the nutrients used in the Choices benchmarks, intakes shift substantially in a beneficial direction when people consume Choices compliant foods. By choosing healthier options in each product category, consumers could have substantially healthier diets that are more in line with the WHO recommendations. These changes could potentially have significant impact on public health in terms of chronic diseases. The results of the simulation study are promising, although field studies that monitor nutrient intakes are needed to confirm the feasibility and impact of the proposed strategy in real life.
We are greatly indebted to the many people that were involved in this study. The authors would like to thank Gerda Donkersgoed, Jacob van Klaveren, Anika de Mul, Olivia-Howell Davies, Robert Daamen, Liesbeth Temme and Renske Visser.
Conceived and designed the experiments: AJCR AJvB HvdV. Performed the experiments: AJCR AJvB HvdV. Analyzed the data: AJCR AJvB MD-K HvdV. Contributed reagents/materials/analysis tools: AJCR AJvB MD-K HvdV. Wrote the paper: AJCR AJvB MD-K HvdV JCS.
- 1. FAO/WHO (2003) Diet, nutrition and the prevention of chronic diseases. World Health Organ Tech Rep Ser 916. Geneva: World Health Organisation. 160p.
- 2. Ocke MC HK, Buurma-Rethans EJM, Rossum CTM van, Drijvers JJMM, Brants HAM, et al.. (2004) Food consumption survey, 2003: summary of procedures and evaluation. Available: http://www.rivm.nl/en/Documents_and_publications/Scientific/Reports/2004/oktober/National_food_consumption_survey_2003_Summary_of_procedures_and_evaluation. Accessed 2013 Jul 24.
- 3. van Rossum CTM, Fransen HP, Verkaik-Kloosterman J, Buurma-Rethans EJM, Ocke MC (2011) Dutch National Food Consumption Survey 2007–2010 : Diet of children and adults aged 7 to 69 years. Bilthoven: National Institute of public Health and the Environment. 143p.
- 4. Holt R (2003) The Food and Agriculture Organization/World Health Organization expert report on diet, nutrition and prevention of chronic diseases. Diabetes Obes Metab 5: 354–358.
- 5. Waxman A (2004) WHO global strategy on diet, physical activity and health. Food Nutr Bull 25: 292–302.
- 6. Scarborough P, Rayner M, Stockley L (2007) Developing nutrient profile models: a systematic approach. Public Health Nutr 10: 330–336.
- 7. Nijman CA, Zijp IM, Sierksma A, Roodenburg AJ, Leenen R, et al. (2007) A method to improve the nutritional quality of foods and beverages based on dietary recommendations. Eur J Clin Nutr 61: 461–471.
- 8. The National Heart Foundation of New Zealand (2002) Pick theTick. The Food Information Program. Guidelines for Tick Approval.Available: http://wwwheartfoundationorgnz/healthy-living/healthy-eating/heart-foundation-tick Accessed 2012 Jun 26.
- 9. Choices (2008) Choices International Foundation, Available: http://www.Choicesinternational.org/Accessed 2012 Jun 26.
- 10. Drewnowski A, Fulgoni VL, Young MK, Pitman S (2008) Nutrient-rich foods: applying nutrient navigation systems to improve public health. J Food Sci 73: H222–228.
- 11. Roodenburg AJ, Popkin BM, Seidell JC (2011) Development of international criteria for a front of package food labelling system: the International Choices Programme. Eur J Clin Nutr 65: 1190–1200.
- 12. Vyth EL, Steenhuis IH, Mallant SF, Mol ZL, Brug J, et al. (2009) A front-of-pack nutrition logo: a quantitative and qualitative process evaluation in the Netherlands. J Health Commun 14: 631–645.
- 13. Vyth EL, Steenhuis IH, Roodenburg AJ, Brug J, Seidell JC (2010) Front-of-pack nutrition label stimulates healthier product development: a quantitative analysis. Int J Behav Nutr Phys Act 7: 65.
- 14. de Boer WJ, van der Voet H (2011) A Web-Based Program for Monte Carlo Risk Assessment release 7.1. Available: https://mcra.rivm.nl. Accessed 2012 Jun 28.
- 15. Roodenburg AJC, Temme EHM, Howell Davies O, Seidell JC (2009) Potential impact of the Choices Programme on nutrient intakes in the Dutch population. Nutr Bull 34: 318–323.
- 16. de Boer WJ, van der Voet H, Bokkers BGH, Bakker MI, Boon PE (2009) Comparison of two models for the estimation of usual intake addressing zero consumption and non-normality. Food Addit Contam Part A Chem Anal Control Expo Risk Assess 26: 1433–1449.
- 17. Dodd KW, Guenther PM, Freedman LS, Subar AF, Kipnis V, et al. (2006) Statistical methods for estimating usual intake of nutrients and foods: a review of the theory. J Am Diet Assoc 106: 1640–1650.
- 18. Box GEP, Cox DR (1964) An analysis of transformations. Journal ofthe Royal Statistical Society, Series B 26: 211–252.
- 19. van der Voet H, de Mul A, van Klaveren JD (2007) A probabilistic model for simultaneous exposure to multiple compounds from food and its use for risk-benefit assessment. Food Chem Toxicol 45: 1496–1506.
- 20. Nusser SM CA, Dodd KW, Fuller WA (1996) A semiparametric transformation approach to estimating usual daily intake distributions J Am Stat Assoc. 91: 1440–1449.
- 21. Beaton GH (1994) Criteria of an adequate diet. In: Shils ME, Olson JA, Shike M, editors. Modern Nutrition in Health and Disease, Philadelphia: Lea & Febiger Pp. 8th edition. p 1491–1505.
- 22. Institute of Medicine (IOM) (2000) Dietary Reference Intakes: Applications in Dietary Assessment. Washington, DC: National Academy Press.
- 23. King JC, Vorster HH, Tome DG (2007) Nutrient intake values (NIVs): a recommended terminology and framework for the derivation of values. Food Nutr Bull 28: S16–26.
- 24. (2006) NEVO-tabel Nederlands Voedingsstoffenbestand (Dutch Food Composition Table). The Hague: Stichting Nederlands Voedingsstoffenbestand.
- 25. Health Council of the Netherlands (2006) Guidelines for a healthy diet. Publication no. 2006/21. The Hague: Health Council of the Netherlands.
- 26. Schakel SF, Buzzard IM, Gebhardt SE (1997) Procedures for Estimating Nutrient Values for Food Composition Databases. Journal of Food Composition and Analysis 10: 102–114.
- 27. Agricultural Research Service USDA (2008) National Nutrient Database for Standard Reference. Release 21. Nutrient Data Laboratory.Available: http://www.ars.usda.gov/ba/bhnrc/ndl Accessed 2012 Jun 26.
- 28. Temme EH, van der Voet H, Roodenburg AJ, Bulder A, van Donkersgoed G, et al. (2011) Impact of foods with health logo on saturated fat, sodium and sugar intake of young Dutch adults. Public Health Nutr 14: 635–644.
- 29. de Lauzon B, Volatier JL, Martin A (2004) A Monte Carlo simulation to validate the EAR cut-point method for assessing the prevalence of nutrient inadequacy at the population level. Public Health Nutr 7: 893–900.
- 30. Darmon N, Vieux F, Maillot M, Volatier JL, Martin A (2009) Nutrient profiles discriminate between foods according to their contribution to nutritionally adequate diets: a validation study using linear programming and the SAIN,LIM system. Am J Clin Nutr 89: 1227–1236.
- 31. Trichterborn J, Drossard C, Kersting M, Harzer G, Kunz C (2012) The potential impact of nutrient profiles on dairy-related energy and nutrient intake in German children and adolescents. Eur J Clin Nutr 66: 466–473.
- 32. Schickenberg B, van Assema P, Brug J, Verkaik-Kloosterman J, Ocke MC, et al. (2009) Replacing foods high in saturated fat by low-saturated fat alternatives: a computer simulation of the potential effects on reduction of saturated fat consumption. Br J Nutr 102: 478–483.
- 33. Vyth EL, Hendriksen MA, Roodenburg AJ, Steenhuis IH, van Raaij JM, et al. (2012) Consuming a diet complying with front-of-pack label criteria may reduce cholesterol levels: a modeling study. Eur J Clin Nutr 66: 510–516.
- 34. Roodenburg AJ, Schlatmann A, Dotsch-Klerk M, Daamen R, Dong J, et al. (2011) Potential effects of nutrient profiles on nutrient intakes in the Netherlands, Greece, Spain, USA, Israel, China and South-Africa. PLoS One 6: e14721.
- 35. Lloyd-Williams F, Mwatsama M, Ireland R, Capewell S (2009) Small changes in snacking behaviour: the potential impact on CVD mortality. Public Health Nutr 12: 871–876.
- 36. Heerstrass DW, Ocke MC, Bueno-de-Mesquita HB, Peeters PH, Seidell JC (1998) Underreporting of energy, protein and potassium intake in relation to body mass index. Int J Epidemiol 27: 186–193.