Software tools for practical application of human nutrient requirements in food-based social science research

Nutrient requirements

The IOM DRIs contain nutrient requirements for micronutrients (vitamins and minerals), energy, and macronutrients (carbohydrates, protein, and fats) for 25 distinct population groups that differ by age, sex, and life stage (e.g. pregnancy and lactation)^3–5. The 2006 DRIs marked an important evolution in defining human nutrient requirements^{3 15}. In particular, the DRIs comprehensively included all requirement categories and all essential nutrients for the first time. Furthermore, the most important advancement of the 2006 DRIs was to establish an Estimated Average Requirement (EAR) appropriate for use in analyses of populations and groups. Using the EAR with survey data of individuals applies the median nutrient requirement to all individuals within the population (age, sex, and lifecycle group) for which it is defined. The UN organizations (and others) had described this as the appropriate requirement to use for populations and groups, but it had not been formally established in any prior publications^9,16. It was also the first time safe upper levels were established¹⁷.

Some researchers may question the use of the DRIs for studies globally, given that they were developed for use in North America. Recommendations have been developed by WHO and compiled in collaboration with FAO, UNICEF and UNU contained across multiple documents^9–14. These global recommendations, however, do not contain EARs, and were not as rigorously developed as the IOM DRIs¹⁸. The other ongoing international effort to set human nutrient requirements based on the most recent scientific evidence is carried out by the European Food Safety Authority (EFSA). There are current ongoing efforts to harmonize the nutrient requirements released by these and other scientific groups, detailed by Allen et al.¹⁸ who recently proposed an initial set of harmonized requirements integrating the IOM DRIs and EFSA dietary reference values for 26 macro- and micronutrients. We provide data files to utilize these harmonized values (Extended data, Supplement 6¹⁹) but detail the IOM DRIs values in the following sections, as our use case (the primary motivation for this software tools paper and what is contained in our source code) employs only the DRIs values.

Micronutrients. The DRIs contain four different types of requirements: the EAR (the daily amount necessary to meet the needs of 50% of the healthy population); the Average Intake (AI) (the amount estimated to be sufficient to meet the needs of nearly all individuals based on the average intake of healthy populations); the Recommended Daily Allowance (RDA) (the amount estimated to meet the needs of 97.5% of the healthy population, defined as two standard deviations above the EAR); and the tolerable upper intake level (UL) (the maximum daily intake which does not produce any signs of toxicity).

For each micronutrient per age-sex group, the IOM defines either 1) an RDA and EAR or 2) only an AI level. The RDA is a high level of intake that meets the needs of nearly all healthy people. Where an RDA can be set, beginning with the 2006 DRIs, the EAR is also established. The EAR, which defines the average nutrient requirement in a population, is the appropriate metric for social science researchers interested in questions about populations and groups^3,9,16. When there is insufficient evidence for an RDA and EAR, the IOM establishes an AI, which is the recommended average daily intake based on observed or experimentally determined approximations of nutrient intake in healthy individuals that are assumed to be adequate. AIs are also expected to meet the needs of most individuals. RDAs and AIs are used in clinical practice to ensure nutrient needs for any given individual are met, while EARs are used to estimate needs of populations and plan population-level interventions.

Though the EAR is the appropriate nutrient requirement to use when defining the needs or assessing the diets of populations and groups, setting an EAR requires a high degree of scientific evidence and therefore has not been established for all nutrients or all ages and sexes. Therefore, for analyses of populations and groups, we recommend including only those nutrients that have a defined EAR. Among the micronutrients with an EAR, we further recommend excluding three nutrients from food-based analyses: vitamin D, iodine, and molybdenum. Vitamin D should be excluded given high variation within populations in the quantity produced by the body and variation due to sun exposure and race/ethnicity⁴. Iodine is often fortified in salt and typically not measured in food composition tables²⁰. Molybdenum is similarly not contained in food composition tables, is present in many foods, and diseases of deficiency have never been observed in healthy humans³.

Some micronutrients can also produce harmful effects if too much is consumed. Tolerable Upper Intake Levels (ULs) are set at the level of intake that produces symptoms of toxicity. The documentation for each nutrient specifies whether the UL is possible to reach from food sources or only supplements. A new requirement termed the Chronic Disease Risk Reduction (CDRR) level was established for sodium in 2019⁵. The reason for establishing a new type of nutrient requirement was that the committee found sufficient evidence to establish upper bound limits on sodium intake to reduce chronic disease risk, but these are not the same as symptoms of toxicity, the criteria required to set a UL.

When defining healthy diets, such as in the least-cost diet application described below, we recommend including the UL for all nutrients for which it is defined and can be reached with food sources. Additional considerations exist for vitamin A, however: the UL for vitamin A pertains only to retinol, the preformed form of vitamin A found in animal source foods as well as some medications. The EAR for vitamin A includes both preformed and provitamin A carotenoids. Carotenoids (e.g. beta carotene) are found in many foods and are converted by the body to an active form only as needed, and therefore pose no risks of toxicity. Excessive intake of retinol, on the other hand, can cause severe consequences; for example if consumed in excess by pregnant women it causes birth defects³. For this reason, we treat vitamin A and retinol as separate nutrients, with the former having only an EAR and the latter having only a UL value⁴. Finally, although not formally defined as a UL, we recommend including the CDRR for sodium as an upper bound on sodium intake which takes the same form in practice as a UL when defining healthy diets. The 2019 DRIs defined the CDRR given the high degree of scientific evidence that higher consumption of sodium substantially increases the risk of chronic disease, though it does not produce symptoms of toxicity^5,21.

Energy. The Estimated Energy Requirement (EER) is an equation to determine energy needs based on age, sex, weight, height, and physical activity level. An active level of physical activity is recommended for long-term health and therefore we recommend estimating nutrient requirements for this level of activity. For populations and groups, reference values for weight and height are taken from child growth charts, discussed in the next section.

Macronutrients. Humans need carbohydrates, protein, and lipids in larger amounts to provide energy to the body and carry out numerous metabolic functions. There is an EAR for carbohydrates and protein that sets an absolute minimum amount required⁵. The protein EAR is calculated per kilogram of body weight and is therefore sensitive to weight in addition to age, sex, and life stage. Long-term health requires more than these minimum values. The Acceptable Macronutrient Distribution Range (AMDR) specifies a range of acceptable proportions of calories from each macronutrient. These are then converted into gram quantities based on energy needs. Carbohydrates and protein supply four calories per gram and lipids provide nine calories per gram. Therefore, the AMDR moves with the energy needs, which are sensitive to the height and weight reference parameters used and the selected physical activity level.

Reference heights and weights

Energy requirements are sensitive to the reference heights and weights, and level of physical activity selected. The 2006 DRIs reference heights and weights come from the CDC/NHCS growth references³. After the release of the WHO child growth standards, the CDC recommended the use of the WHO standards for all children 0–59 months (through 5 years) in the United States²².

The publication of the WHO child growth standards in 2006 represented an important step forward in the assessment of child growth and the definition of healthy growth trajectories. The standards were developed through a multi-center study carried out in six sites around the world, covering all continents and communities at different levels of economic development. Though it developed growth standards only for children 0–59 months, mathematical modeling permitted the extrapolation of growth references for school aged children and adolescents (ages 5–19 years) based on the child growth standards⁷. The development of the WHO child growth standards is discussed in depth in numerous other publications, but it is important for non-nutritionists to know that the rigorous multi-center global study was designed to move beyond growth references for a single population (describing how children do grow), to a set of standards for how children should grow in the condition of optimal health and growth^6,23–25. They are now considered the globally recognized source for children 0–59 months and therefore are the appropriate resource from which to draw reference heights and weights to estimate nutrient requirements^26–29.

To be consistent in the source of references for all ages and sexes, and in light of the global studies underlying the references, we recommend the use of the WHO references for all ages and sexes. The end-growth (19 year old) median heights and weights are carried forward into adulthood, which is the methodology used by the DRIs with respect to the CDC/NCHS growth charts³. The WHO growth reference study explicitly studied children age 5 years and under to determine that child growth curves are the same across cultures and populations; it did not determine the similarities or differences in observed attained adult heights across populations. We use the WHO growth standards end-growth values as the best estimate of median adult heights and weights⁶.

The appropriate reference weight for pregnant women can be estimated as the median adult weight for her age group plus a recommended amount of weight gain during pregnancy. The IOM reference heights and weights used in the DRIs incorporate CDC guidelines regarding weight gain during pregnancy. To the best of our knowledge, the WHO does not have any guidance regarding weight gain during pregnancy. To develop the reference weights for pregnant women included in these software tools, we therefore take the median adult weight as defined by the WHO growth references and add the midpoint of the CDC recommended weight gain range (13.25 kg) over the total pregnancy for a woman of healthy weight at the onset of pregnancy⁸. Researchers wishing to include this life stage in analyses may utilize the median weight gain recommended over the entire pregnancy included in these software tools; however, those particularly concerned with nutrient requirements during pregnancy are encouraged to work with a nutritionist and consider the differing needs in each trimester.

Considerations in application

We outline four likely ways in which researchers may need to modify the nutrient requirements for certain analyses.

Probability of iron deficiency in adequacy analysis. Many researchers may be interested in estimating the probability of inadequacy by comparing observed nutrient intakes to requirements. At the population level, for most nutrients, the EAR cut-point method can be used because the distribution of actual needs in the population is understood to be normally distributed. This means that any individuals consuming below the EAR are estimated to be inadequate and those above to be adequate and at the population level this yields an accurate estimate of the percentage of the population likely to be inadequate^3,30,31. However, iron requirements are not normally distributed in the population and therefore a probability estimate must be used. The probability of iron inadequacy varies by age, sex, reproductive status, and oral contraceptive use. The DRIs provide the probability of inadequacy for intakes within specified ranges by age, sex, and oral contraceptive use. For population level analyses, researchers may know that oral contraception is not available and assume an entire non-user population, or, if oral contraceptives are available and likely to be used, the researcher should estimate the probability based on a mixed-user population.

Children 6–23 months. Children who are still breastfeeding have some of their nutrient needs met from breastmilk and the WHO recommends continued breastfeeding for the first two years of life and as long thereafter as desired by mother and child³². The definitive resource on nutrient needs during this period is Dewey 2005³³ which contains comprehensive requirements for all nutrients including the quantities provided by breastmilk and needed from food. When considering this age group as a member of a family or otherwise heterogeneous groups, such as in establishing household needs³⁴, we recommend applying consistent criteria for the choice and inclusion of nutrient requirements across all groups and therefore including only those contained in the DRIs for this age range, and adjusting for the contribution of breastmilk according to Dewey 2005³³. This is the method we have adopted in these software tools. Given the complexity of estimating nutrient needs met through breastmilk vs. complementary feeding, we advise any social scientists carrying out studies of nutrient needs that are explicitly focused on children under two to collaborate with a nutritionist.

Iron and zinc bioavailability. Many populations in low- and middle-income countries consume diets high in phytates and oxalates. These compounds, present in unrefined grains and some other plant foods, bind to iron and zinc and reduce the body’s ability to absorb those nutrients. The body will only absorb some overall fraction of each nutrient, and the specific value of this fraction is dependent on the nature of the whole diet and not yet fully understood^35,36. When evaluating the adequacy of observed diets where the dietary pattern can be characterized by low bioavailability, researchers may choose to revise the iron and zinc requirements to reflect an assumption of lower bioavailability¹⁸. We stress that understanding bioavailability is an active area of nutrition research and therefore do not make an explicit recommendation regarding whether or not to adjust the iron and zinc requirements for low-bioavailability diets. If choosing to do so, we advise researchers to collaborate with a nutritionist.

Physical activity in rural agricultural populations. Researchers interested in certain populations whose daily lives are physically demanding may wish to consider energy needs at a higher level of physical activity. Where women’s daily lives include energy-demanding activities such as collecting water and firewood and pounding grain, the UN characterizes these energy needs as consistent with the “active” physical activity level in the DRIs. We therefore do not recommend any adjustments for adult women. But for men such as farmers in unmechanized settings and construction workers carrying out demanding manual tasks, the UN describes energy needs consistent with the DRIs’ “very active” physical activity level¹⁰. Therefore, in such settings, we recommend using energy requirements for the very active level of physical activity for males aged 14–59 as they more accurately reflect needs of physically demanding lifestyles^3,10.

Operation

The software tools in this article are developed to run using open source R statistical software. Code is also provided in Stata format. The extended data files can be opened with any spreadsheet software or freely available browser application. It is not required to open the extended data files separately from compiling the larger datasets with statistical software, however it is helpful to become familiar with the shape of the data to understand how they must be reshaped and transformed in order to compile a single long form dataset. The workflow in brief starts with compiling the DRIs data files, resulting in a dataset that is long form by age-sex group, nutrient, and physical activity level. Then, optionally, the energy and protein requirements can be recalculated using WHO reference heights and weights. Then, also optionally, the requirements for children 6–23 months can be reduced to only the percentage required from food. A single age-sex group at a single level of physical activity is extracted in the final step as an example for use in calculating a least-cost diet.

Supplement 1: DRIs nutrient requirements⁷

The nutrients and requirements contained in the Extended data, Supplement 1¹⁹ data files follow the selection criteria denoted above and are taken directly from the nutrient requirement tables in the DRIs documents^3–5. Extended data, Supplement 1¹⁹ also includes a file with the iron probability of inadequacy by age, sex, consumption range, and oral contraceptive use³. Any further adjustments, important details for application, or data management procedures are denoted here.

Units of measurement. Each micronutrient is required in different amounts by the body and therefore the micronutrient requirements are expressed appropriately in either micrograms or milligrams. The Extended data, Supplement 1¹⁹ files make a few unit conversions to facilitate analysis. In the DRIs, the units of measurement for calcium and phosphorus differ between the EAR and the UL, so we have converted the ULs for both nutrients to milligrams to be consistent with the unit in which the EAR (and RDA) are expressed as well as the unit of measurement commonly used in food composition tables. Similarly, copper is measured in micrograms in the DRIs, but we have converted the requirement to milligrams since this is the common unit of measurement in food composition tables.

Researchers combining nutrient requirements with food composition are advised to confirm that the unit of measurement for the requirement is the same, per nutrient, between the requirement and the particular food composition table used in the analysis, as they do vary in some food composition tables. The AMDRs are expressed as a percentage of calories from each food, but the quantities are expressed in food composition tables in grams. Therefore, the percent of calories must be converted to the total grams of each nutrient, which depends on individual energy needs. Since energy needs are sensitive to all the aforementioned parameters (age, weight, height, physical activity), this conversion is done in the dataset compilation (Supplement 4³⁷). The other nutrient where units may not agree is vitamin A, which has been measured in three different units over time: retinol equivalents (RE), international units (IU), and retinol activity equivalents (RAE). The RAE is the currently accepted standard unit and is used in the USDA food composition table and the DRIs; however, other tables, especially if older, may use other units and may also separate retinol and carotenoids or provide multiple units. Conversion between units is not straightforward and non-expert researchers should not attempt to do so and could instead consider using alternate food composition data (such as USDA) where RAEs are not available.

Inclusion of the sodium CDRR. For the purposes of defining the requirements for healthy life, the sodium CDRR is functionally equivalent to the food-based ULs for other nutrients. In other words, both requirements put an upper bound on the amount of a nutrient that is considered within the healthy range for consumption. As such, for analytical convenience, the sodium CDRR is included in the same sheet as the other micronutrient ULs in Extended data, Supplement 1¹⁹. We stress that this upper bound requirement is not based on symptoms of toxicity, the criteria required to set a UL. So, while it is functionally equivalent for social science research applications, it is important for researchers to acknowledge the fundamentally different meaning of the sodium CDRR compared to other micronutrient ULs.

Supplement 2: WHO growth references, energy & protein requirements

Growth references are expressed with anthropometric measures that describe the relationships between weight, height and age and are sex specific. These measurements are explained in Table 1.

Measurements are presented in the growth charts as percentiles and z-score distributions, which are used to calculate population statistics for stunting, wasting, underweight and overweight. Only percentile charts are needed to calculate reference heights and weights by age and sex to determine nutrient requirements and only the percentiles charts are contained in the Extended data, Supplement 2¹⁹ data files. The “Notes” document describes each dataset. The three files necessary to recalculate the DRIs using the WHO growth references are are based on the extraction of the median heights and weights from the subsequent sheets. The “Weights Heights by DRI Group” file calculates the median of the 50^th (median) percentile heights and weights corresponding to the 25 age-sex groups contained in the DRIs. The “EnergyEER” file takes those calculated reference weights and uses the DRIs EER equation to calculate energy requirements for each age-sex group at all four levels of physical activity. Finally, the “ProteinEAR” file uses the DRIs requirement of protein per kilogram body weight and calculates a total protein EAR for every age-sex group using the WHO reference weights.

While only the “Weights Heights by DRI Group”, “EnergyEER”, and “ProteinEAR” files are required to compile the nutrient needs dataset using the software code in Supplements 4 and 5³⁷, the additional files provide the calculations and full data. For children 0–59 months, only the weight-for-age (WFA) and the length/height-for-age (LHFA) are needed and the 50^th (median) percentiles are extracted from the full tables and provided for each month of age, by sex, in the “Median Weight Height Month 0–19” file. For children 5–19 years, the growth reference is expressed in terms of body mass index (BMI) and height-for-age, so an extra arithmetic step is required to determine the weights. This calculation is provided in the “Median Weight Height Month 0–19” file for all children over age 5 since the child and adolescent percentiles are expressed per month of age as well. The “Median Weight Height Year 5–19” file provides the resulting calculation of the median of the 50^th (median) percentile weights and heights per year of age for all children 5–19 to provide a single value per year.

Supplement 3: Nutrient requirements during continued breastfeeding

From 6–23 months, during continued breastfeeding, the infant’s nutrient requirements change dramatically as does the relative proportion of breastmilk to food in the total diet. The DRIs divide infants in this period into only two groups: 6–11 months and 12–23 months. However, the proportion of nutrients needed from foods are much lower for 6–8 months than for 9–11 months of life. The variable bf_match_id matches the DRIs age and sex groups in the “Reference Values” file to the nutrient needs required from foods in the “6–23mo_FoodNeeds” file, effectively dividing each of the DRIs age-sex groups aged 6–11 months (age_sex_grp=2) and 1–2 years (age_sex_grp=3) into two groups. The bf_match_id variable makes possible the adjustment carried out in the data compilation below, provided with the important cautions noted above regarding nutritional vulnerability and nutrient needs during this life stage.

Supplements 4 and 5: Dataset compilation in Stata or R

We demonstrate the compilation of requirements and data management to shape the data into the format required for a least-cost diet application. The final data are in long form by age-sex group, nutrient, and physical activity level and wide form by nutrient requirement. The process is divided into five parts as follows:

Supplement 6: Harmonized requirements data files

The proposition of harmonized nutrient requirements across scientific bodies marks an important milestone for global applications of human nutrient requirements. Our use case (the primary motivation for this software tools paper) uses only the IOM DRIs values, but other related work uses the harmonized values for global analysis (e.g. Herforth et al.³⁸). Given that the harmonized values may be of interest for researchers to apply, we also provide a data files with the recently published harmonized requirements by Allen et al. (2020)¹⁸. Extended data, Supplement 6¹⁹ provides the harmonized average requirements (H-ARs) and harmonized upper levels (H-ULs) contained in Tables 2 and 6 of the aforementioned paper. We include all the nutrients covered in that paper that are commonly found in food composition tables, excluding vitamin D, iodine, pantothenic acid, biotin, and choline. The harmonized requirements do not cover energy or macronutrients beyond an average requirement for protein, and therefore researchers wishing to apply the harmonized requirement will need to choose how to define the energy and macronutrient requirements. We do not provide specific guidance and suggest working with nutritionists to make the appropriate decisions relevant to a particular study. We would, however, like to note that the DRIs age-sex disaggregation for which the EER formulas are defined do not perfectly align with the harmonized groups. The EFSA average requirements for energy would be more straightforward to use. After selecting an energy requirement, the AMDRs do not present a conflict between the requirements and the harmonized age-sex groups (with the exception of lipids for 18-year-olds) and can be applied if desired based on percentage of calories from each macronutrient.

Underlying data

All data underlying the results are available as part of the article and no additional source data are required.

Extended data

Harvard Dataverse: Data files for "Software tools for practical application of human nutrient requirements in food-based social science research". https://doi.org/10.7910/DVN/5GIMLE¹⁹.

This project contains the following extended data:

Extended data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).

Software availability

Source code in Stata and R files available from: https://github.com/KateSchneider-FoodPol/Nutrient-Requirements-Software-Tools.

Archived source code at time of publication: https://doi.org/10.7910/DVN/FBNIKO³⁷.

License: Apache License, Version 2.0.