Updated: Feb 11
The first lesson in toxicology is that “the dose makes the poison”; meaning that all things are toxic, even water when consumed in unsafe quantities. In this article, the author will describe how calculating daily exposure to a food, using serving sizes described in 21CFR101 et seq, can result in gross exaggerations: under- or over-estimating consumption. For example, the “official” serving size for ice cream is 2/3 of a cup, although when people purchase ice cream, they tend to eat the entire pint container (2 cups) at once. Thus, estimating consumption using serving size results in egregious errors that may endanger consumers. Read on to see how such errors can be avoided.
The National Health and Nutrition Examination Survey (NHANES) provides data that is more suitable to represent the dietary intake of the U.S. population than other data sources, including but not limited to the weighed/ measured food intake, duplicated portions, and food disappearance data. For example, the 24-hour recall data collected in What We Eat in America (WWEIA), NHANES provides reliable exposure data because it includes detailed information specific to each food and beverage consumed on a recall day (e.g., what foods were eaten in combination) (CDC, 2013), and does not include food discarded as the result of spoilage (unlike food disappearance data). Burdock Group’s proprietary consumption software utilizes NHANES providing an accurate estimated daily intake (EDI) of a substance and not the inaccurate serving size data.
A serving size is the amount of food customarily consumed (i.e., typically eaten) in one sitting for that food and is determined from the Reference Amounts Customarily Consumed (RACC). Contributing to its inaccuracy is the fact that the last time the RACC was updated was 1993 (Food Insight, 2020); therefore, the RACC does not reflect shifts in eating preferences as a result of consumer behavior (e.g., people seeking out specialized diets, such as meat-free or low-carb), or business practices change (e.g., new sweetener is developed) (Alger et al., 2013). Indeed, serving size is not cited as an acceptable source of food consumption data in the literature including the FDA (2006) guidance document, “Guidance for Industry: Estimating Dietary Intake of Substances in Food”, where RACC is stated to be a measure of food portion sizes. Portion size has been shown to influence intake as much as taste; for instance, moviegoers have been shown to eat more popcorn from a large-size container than a medium-size container even when the popcorn was stale (Wansink and Kim, 2005). However, there are a multiple of factors that affect portion size (packaging, labeling, etc.) and therefore it does not sufficiently measure consumption.
Consumption Data Sources
Data collection methods used to estimate intake of substances in the diet include, but are not limited to food disappearance data, household disappearance data, dietary histories, dietary frequencies, 24-hour recalls, food records, weighed intakes, and duplicate portions (Pennington, 1991). NHANES collects dietary data from 24-hour recalls, food frequency questionnaire, and survey questionnaires (CDC, 2013). In the third National Health and Nutrition Examination Survey (NHANES III), data were collected in 81 counties from approximately 30,000 respondents across the United States (CDC, 2000). Burdock Group consumption analysis software specifically utilizes the two days of 24-hour dietary recall data collected in WWEIA, NHANES. The immediacy of the 24-recall period allows respondents to be able to recall most of their dietary intake. Respondents of the 24-hour dietary recall surveys are also more likely to be representative of the population than those who agree to do a food record over a longer period of time and therefore, the 24-hour recall method is useful across a wide range of populations (Thompson and Subar, 2017).
Aside from the duplicate portion and weighed food intake method, which is not appropriate for large-scale food consumption due to cost, food record (diary), food frequency and 24-hour dietary recall are considered the most valid methods (i.e., how closely the method measures daily intake) according to WHO (1985). The method that is least valid is food disappearance data (WHO, 1985). Food disappearance estimates can overstate actual consumption because they include waste accumulated and spoilage in the home as well as through the marketing and distribution system; in general, food disappearance estimates serve more appropriately as indicators of trends in consumption over time than as a measurements of absolute levels of food consumed (Putnam et al., 1970).
Innovations in some of these data sources have been completed to increase the accuracy and efficiency of the data collected. For instance, National Cancer Institute (NCI) has developed a new Food Propensity Questionnaire (FPQ) that has been included in the NHANES study since 2003. The FPQ, a modified version of NCI’s diet history questionnaire, is designed to enable assessment of daily lifetime (i.e., chronic) food intake by different populations by combining FFQ and 24-hour recall data (FDA, 2006). Also, the computerized method known as the USDA Automated Multiple-Pass Method (AMPM) reduces respondent burden and is used to collect interviewer-administered 24-hour dietary for WWEIA (CDC, 2013). Because NHANES uses a combination of the most valid methods that are continually improving to collect dietary data from a large number of people, it is an accurate source of food consumption data.
Using the appropriate methods to analyze the data from the consumption data source is an essential step in achieving accurate EDIs. Burdock Group’s proprietary statistical program uses WWEIA, NHANES public files which unlike the RACC, is regularly updated every two years. The data sets include two files: a total nutrient intakes file and an individual food file (dietary supplement data are available in separate files) (FDA, 2006). USDA’s Food and Nutrient Database for Dietary Studies (FNDDS) is used with the files to code foods/beverages. There are 7,000 food items (each with a unique food code) in FNDDS (USDA, 2008), and each food item belongs to one of the 150 WWEIA food categories. Like WWEIA, serving size covers about 150 food categories; however, the categories are not further divided into food items which can result in an error in daily exposure of an ingredient. For example, in the consumption scenario in Table 1, Vitamin K1 (phylloquinone) intended to be used in milk products for children will not be in all foods that fall into the milk category including “milk substitutes added to cereal” causing an underestimate of mean daily consumption when serving size data is used.
Race/ethnicity, gender and age, income, etc., can also affect consumption. For example, according to WWEIA NHANES 2015-2016 data, far fewer non-Hispanic black (34%) children drank milk than children of other race/ethnic groups (45 – 56%) (USDA, 2019); therefore, the resulting EDI of the ingredient(s) for use in milk will be lower for non-Hispanic black than other race/ethnic groups. This information is not evident with serving size data. Also, from the WWEIA NHANES 2015-2016 data, the most commonly consumed beverages among children were sweetened beverages and milk; however, milk consumption decreases with age (USDA, 2019). Also, adults were included in the serving size data which could be another reason why the mean consumption was underestimated in Table 1 for serving size.
a Estimates based on one day of dietary intake data collected in What We Eat in America (WWEIA), the dietary intake interview component of the National Health and Nutrition Examination Survey (NHANES), in 2015-2016.
b WWEIA Food Categories: https://www.ars.usda.gov/ARSUserFiles/80400530/pdf/1516/food_category_list.pdf
c Excludes milk or milk substitutes added to alcoholic beverages, coffee, tea, and/or foods such as cereal.
d (USDA, 2019)
f Typical (mean) use = One serving/day
The NHANES data files, the USDA food code list, the ingredient amount(s) corresponding to each food code (in mg/g) are entered into our proprietary statistical software. The software then runs the consumption analysis resulting in descriptive statistics of consumption of ingredient per capita (including non-eaters) and overall consumption for eaters only. The statistical analysis includes the mean, median, 90th and 95th percentile of the ingredient being consumed. The EDI of an ingredient from its proposed use in food can be calculated by multiplying the concentration of the ingredient in food (mg per g food) and the amount of food consumed (g per day). When people consume less than the calculated safe limit of a substance, there is a “reasonable certainty that no harm will result from the intended uses under the anticipated conditions of consumption” (OECD, 1993). Therefore, a substance is safe if the EDI for the intended use is less than the amount of the substance that is not expected to cause harm over a lifetime (i.e., the acceptable daily intake (ADI)). Below is a mathematical comparison of EDI to the ADI (Redbook, 2000):
EDI < ADI = Consumption is considered safe by FDA for its intended use
EDI > ADI = Consumption is not considered safe by FDA for its intended use
If the EDI exceeds the ADI, the amount of the ingredient exposure can be mitigated, either as a reduction in the amount of ingredient added to a food in a particular category or the elimination of a food category altogether (as some categories have higher consumption than others), and the consumption analysis is performed again. This process can be repeated and ingredient addition tweaked until an ingredient’s EDI falls below the ADI. Based on the categories of foods to which the ingredient will be added, the number of individuals consuming the foods and the frequency in which the foods are eaten (once per day, three timers per day, etc.), will determine the complexity of the safety assessment. This shows again why using serving size is inappropriate as an estimate of ingredient exposure.
Uncertainties affecting the estimation of food consumption
With respect to food consumption data, several uncertainties can be identified. Uncertainty is the lack of knowledge of vital parts that are needed to perform an exposure (WHO, 2008). Uncertainties include, but are not limited to sampling uncertainties, measurement uncertainties, extrapolation uncertainties, dependency and ambiguity/imprecise language (EFSA, 2006). For instance, measurement uncertainties that can affect individual food surveys based on diaries, e.g. 7 days include the following: uncertainty in recording of food types; measurement error in weights; underreporting, especially of fatty foods; and incomplete information about, e.g. packaging and processing (EFSA, 2006). The scientific staff at Burdock Group reduces uncertainty by obtaining necessary or applicable information or data (e.g., food codes, NHANES files) for the consumption analysis software – this ensures the accuracy and safety of our consumption estimates.
Through our proprietary software, the EDI of a contaminant can be determined by using the appropriate NHANES data file, the USDA food code list, and the contaminate amount corresponding to each food code (in mg/g). For example, in a hypothetical contamination of a large quantity apple juice with inorganic arsenic in an amount of 0.004 mg/g – this doesn’t seem like much, but what can the data tell us? We would then assume that all consumers would be exposed to the apple juice. Therefore, we would first select the individual food file, and enter into the software the gender (female and male) and age group (age 2 and over), along with the food code (description: apple juice, 100%; food code: 64104010) and corresponding contaminant amount (0.004 mg/g of inorganic arsenic). The consumption analysis automatically generates the mean, median, 90th and 95th percentile of the contaminate being consumed. If the amount of apple juice consumed was determined to be 252 g/day for females and males (age 2 and over) then the EDI of the contaminant in this scenario is calculated by multiplying the concentration of the contaminate in food, 0.004 mg/g, by the amount of apple juice, 100% product consumed, 252 g/day, and then dividing by the body weight (assume a body weight of 60 kg). How often the individual is exposed to the chemical during the year, can also impact EDI and should be considered when exposure does not occur daily. Under the assumption that the product will be consumed every day, 0.017 mg/kg/day is the resulting EDI. The FDA has concluded that an action level of 10 µg/kg or 0.01 mg/kg of inorganic arsenic in apple juice is adequate to protect the public health based on its risk assessment (FDA, 2013). Because the inorganic arsenic from the consumption scenario is above .01 mg/kg may result in the decision from the FDA to not allow the food to be sold. Needless to say, such a calculation could not be arrived at by using serving size.
With the use of Burdock Group’s proprietary consumption software and the expertise of our scientific staff, we are able to collect and analyze valid food consumption data resulting in EDIs that are based on reasonable consumption scenarios. Compared to other sources, the National Health and Nutrition Examination Survey (NHANES) is a valuable source of information on the distribution of usual intakes of nutrients in the U.S. diet, which is utilized in BG’s consumption analysis. BG can help companies conduct safety assessments and reassessments (e.g., when changes are made in manufacturing of the ingredient) for not only ingredients that are intentionally added to food for a technical effect (e.g., a preservative or color), but also substances that are unavoidably or unintentionally present in food as a result of the manufacturing process or from natural or environmental sources as required by FDA (FDA, 2006). With continual improvements and updates to our consumption analysis database, we are ensuring that the daily exposure estimates are reasonable and sufficiently protective of public health.
 21CFR101.12(b) Table 2
 Ramey, C. (2018). “Can’t Stop, Won’t Stop” Wall Street Journal (https://www.wsj.com/articles/cant-stop-wont-stop-eating-a-whole-pint-of-ice-cream-at-once-is-ok-now-1520611384?mod=searchresults) (site accessed 20May2020)
Alger, H.M.; Maffini, M. V.; Kulkarni, N.R.; Bongard, E.D. and Neltner, T. (2013) Perspectives on How FDA Assesses Exposure to Food Additives When Evaluating Their Safety: Workshop Proceedings. Comprehensive Reviews in Food Science and Food Safety. 12(1):90–119.
CDC (1994) Consensus Workshop on Dietary Assessment; Nutrition Monitoring and Tracking the Year 2000 Objectives. Sponsored by: National Center for Health Statistics Division of Health Examination Statistics.
CDC (2000) Worker Health Chartbook, 2000. https://www.cdc.gov/niosh/docs/2000-127/pdfs/2000-127.pdf (site visited on May 27, 2020).
CDC (2013) NHANES Dietary Web Tutorial. https://www.cdc.gov/nchs/tutorials/dietary/SurveyOrientation/DietaryDataOverview/intro.htm (site visited on May 26, 2020).
European Food Safety Authority (EFSA) (2006) Guidance of the Scientific Committee on a request from EFSA related to uncertainties in dietary exposure assessment (January 2007). EFSA Journal. 438:1–54.
FDA (2006) Guidance for Industry: Estimating Dietary Intake of Substances in Food. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/guidance-industry-estimating-dietary-intake-substances-food (site visited on May 18, 2020).
FDA (2013) Supporting Document for Action Level for Arsenic in Apple Juice | FDA. https://www.fda.gov/food/chemical-metals-natural-toxins-pesticides-guidance-documents-regulations/supporting-document-action-level-arsenic-apple-juice#assessment_results (site visited on May 28, 2020).
Food Insight (2020) The Nutrition Facts Label: Its History, Purpose and Updates — IFIC Foundation. https://foodinsight.org/the-nutrition-facts-label-its-history-purpose-and-updates/ (site visited on May 18, 2020).
OECD (1993) Safety Evaluation of Foods Derived by Modern Biotechnology Concepts and Principles. http://www.oecd.org/science/biotrack/41036698.pdf (site visited on May 18, 2020).
Pennington, J.A.T. (1991) Methods for Obtaining Food Consumption Information. In Monitoring Dietary Intakes. Springer London, p. 3–8.
Putnam, J.J.; Allshouse, J.E. and United States. Dept. of Agriculture. Economic Research Service (1970) Food Consumption, Prices, And Expenditures, 1970-90. Washington, DC: U.S. Dept. of Agriculture, Economic Research Service.
Redbook (2000) Guidance for Industry and Other Stakeholders Toxicological Principles for the Safety Assessment of Food Ingredients. https://www.fda.gov/files/food/published/Toxicological-Principles-for-the-Safety-Assessment-of-Food-Ingredients.pdf (site visited on May 18, 2020).
Thompson, F.E. and Subar, A.F. (2017) Chapter 1 – Dietary Assessment Methodology. In Nutrition In The Prevention And Treatment Of Disease. 4th Edition. Elsevier Inc., National Cancer Institute, Bethesda, MD, United States.
USDA (2008) USDA Food and Nutrient Database for Dietary Studies, 3.0. Beltsville, MD: U.S. Department of Agriculture, Agricultural Research Service, Food Surveys Research Group. https://www.ars.usda.gov/ARSUserFiles/80400530/pdf/fndds/fndds3_doc.pdf (site visited on May 26, 2020).
USDA (2019) Beverage Choices among Children: What We Eat in America, NHANES 2015-2016. https://www.ars.usda.gov/ARSUserFiles/80400530/pdf/DBrief/22_Beverage_choices_children_1516.pdf (site visited on May 26, 2020).
Wansink, B. and Kim, J. (2005) Bad popcorn in big buckets: Portion size can influence intake as much as taste. Journal of Nutrition Education and Behavior. 37(5):242–245.
World Health Organization (WHO) (1985) Guidelines for the study of dietary intakes of chemical contaminants / text prepared under the Joint FAO/WHO Food Contamination Monitoring Programme, in collaboration with the Joint FAO/WHO Food Standards Programme. World Health Organization. https://apps.who.int/iris/bitstream/handle/10665/39255/WHO_OFFSET_87.pdf?sequence=1&isAllowed=y (site visited on May 18, 2020).
World Health Organization (WHO) (2008) PART 1: GUIDANCE DOCUMENT ON CHARACTERIZING AND COMMUNICATING UNCERTAINTY IN EXPOSURE ASSESSMENT. https://www.who.int/ipcs/methods/harmonization/areas/uncertainty .pdf (site visited on May 18, 2020).