Estimation of micronutrient intake distributions: development of methods to support food and nutrition policy making

Introduction
Adequate and safe micronutrient intake is important. Both insufficient and excessive intakes should be prevented as these can be associated with negative health effects. Therefore, the population intake distribution will ideally lay between insufficient and excessive intakes. For the development and evaluation of nutrition and food policy a good estimation of dietary micronutrient intake is of great importance.

Aim
Three challenges were addressed to improve the estimation of population micronutrient intake distributions: 1) how to estimate current habitual micronutrient intake when (detailed) data are lacking or data from different sources should be combined, 2) how to predict future intakes in order to support policy making, and 3) how to estimate a maximum safe fortification level per food item. The aim of this PhD-thesis is to further develop and apply statistical models which can cope with these challenges.

Methods & Results
Data from the Dutch National Food Consumption Surveys (DNFCSs) were used to develop and apply statistical models which can cope with the defined challenges. In addition, data from the Dutch food composition database (NEVO) and the Dutch dietary supplement database (NES) were used.

Three main methodological improvements have been made. First, the combination of a deterministic approach with probabilistic approaches to be able to take into account uncertainty and variability were needed. This method was applied to estimate habitual iodine and salt intake distributions. From DNFCSs no detailed information was available on the discretionary use of (iodized) salt and no up to date information was available on the use of iodized salt in industrially processed foods. Estimates of the proportion of the population discretionarily using (iodized) salt and the proportion of industrially processed foods applying iodized salt were obtained from other data sources. The model accurately estimates habitual iodine and salt intake distributions when compared with studies measuring urinary iodine and sodium excretion. Additionally a framework was developed to simulate the habitual intake distribution for potential scenarios of future fortification strategies. Within this framework, deterministic and probabilistic approaches were combined when uncertainty or variability had to be taken into account. This framework was illustrated by the estimation of habitual folate-equivalent intake for different scenarios of mandatory or voluntary fortification with folic acid. Further this framework was applied to estimate the habitual iodine intake for several potential changes in the Dutch iodine policy and also for several scenarios of salt reduction strategies.

A second methodological improvement was the development of a new statistical model to estimate habitual total micronutrient intake aggregated from food and dietary supplements. Within this 3-part model, habitual intake is estimated separately for a) intake from food for non-users of dietary supplements, b) intake from food for users of dietary supplements, and c) intake from dietary supplements for users only. Habitual total intake for the whole population was obtained by combination of the three separate habitual intake distributions (‘first shrink then add’). This 3-part model was illustrated by vitamin D intake for young children. With a more simple ‘first add then shrink’ approach the estimation of habitual total vitamin D intake distribution may give inconsistent results for the distribution of intake from foods and dietary supplements combined as compared to the intake from food only. In addition, this more simple approach may not be able to cope with multi modal distributions. With the newly developed model this inconsistency problem was solved and the multi-modal shape of the distribution as observed in the ‘raw’ data was preserved.

Third, a model calculating the maximum safe fortification level per 100 kcal of a food was developed for the Dutch situation. By considering the tolerable upper intake level and reasonable high micronutrient intakes from food and dietary supplements, the ‘free space’ for voluntary fortification was calculated. This amount was divided over the amount of energy intake that can and may be fortified. The model was applied to derive safe maximum fortification levels for vitamin A, D, and folic acid. Based on these results the risk manager decided to legally allow voluntary fortification with vitamin D and folic acid up to a maximum level of 4.5 and 100 μg/100 kcal respectively.

Conclusion
The methodological improvements have resulted in higher accuracy for estimations of habitual intake distributions, which are essential for nutritional and food policy making. Furthermore, scenario analyses provide (under specific conditions) quantitative insight into proposed changes or areas such as maximum safe fortification levels. Several results and methods described are currently being used in research to assist Dutch and European food and nutrition policy making, which shows these methodologies are of immediate value to the practice of policy development and support.