Perspectives on How FDA Assesses Exposure to Food Additives When Evaluating Their Safety: Workshop Proceedings


  • Author disclosures: All authors are employed by The Pew Charitable Trusts, 901 E St., NW Washington, DC 20004, U.S.A. Funding for the workshop and preparation of this article was provided solely by The Pew Charitable Trusts.


Food additives and substances considered “generally recognized as safe” must not be allowed in food unless there is a reasonable certainty in the minds of competent scientists that the substance is not harmful under the intended conditions of use. Scientists determine safety by ensuring that the expected exposure is less than the acceptable daily intake. The U.S. Food and Drug Administration (FDA) provides guidance documents to assist safety assessors in this analysis. A November 2011 workshop sponsored by The Pew Charitable Trusts, the Institute of Food Technologists, and the journal Nature reviewed the agency's exposure assessment approaches. More than 70 experts from government (including FDA), industry, academia, and public interest organizations examined the principles underlying dietary exposure assessments for substances added to human food, and responded to questions about current methods. FDA's approach was seen as serving the agency reasonably well, but participants identified opportunities for improvement. Although reaching a consensus was not a goal, general agreements emerged that FDA should develop a science-based framework to prioritize and reassess prior safety decisions, and conduct more extensive postmarket monitoring. Participants discussed the possibility of harmonizing different approaches to assess dietary exposure. They generally agreed that collaboration, communication, and exchanging scientific information between agencies and stakeholders would help assessors use the most current information to make better decisions. Participants identified data gaps and opportunities to fill the gaps using new tools and technologies. Participants generally agreed on the need to consider all dietary sources in a cumulative dietary exposure assessment.

Executive Summary

To determine whether substances added to human food are safe, the U.S. Food and Drug Administration (FDA) asks 2 basic questions: First, how much of that substance will people consume on a daily basis over a lifetime? And 2nd, how much of that substance is considered a safe amount to eat daily based on scientific studies?

The answers to these questions are informed by the calculation of the estimated daily intake (EDI) and acceptable daily intakes (ADI), respectively. These are complex scientific principles and not easy to determine. Complicating matters, the answers may change over time as consumer behavior shifts (for example, more people may seek out specialized diets, such as meat-free or low-carb), or as business practices change (such as when a new sweetener is developed). Similarly, the answers may change as the science of toxicology advances and more information becomes available (when we learn that some substances may be more or less harmful than we once thought). Differences in diet among certain subpopulations (for example, variations in consumption by age, region, or socioeconomics) make it even more difficult to estimate dietary exposure.

In November 2011, The Pew Charitable Trusts, the Institute of Food Technologists (IFT) and the journal Nature cosponsored the 2nd in a series of 4 workshops, Perspectives on FDA's Exposure Assessment to Ensure Substances Added to Human Food Are Safe, to examine how the FDA addresses these questions and reminds current as the science advances or evolves.

The forum convened a range of stakeholders, including more than 70 experts from academia, government, industry, and public interest groups. FDA scientists provided valuable input in planning the event and were on-hand to discuss the issues and answer questions from the participants. The workshop followed an April 2011 conference—cosponsored by the same organizations and in which FDA also participated—that examined how the agency determines the levels at which consumption of certain substances may be harmful to human health.

While a consensus was not the goal, participants generally agreed that there were opportunities for FDA to improve the current system for ensuring the safety of substances in food. As a result of the discussion, the following 3 major topics emerged for improving exposure assessments:

  • Food consumption data;
  • Greater consideration of subpopulations and cumulative dietary exposure; and
  • Interagency communication, collaboration, and coordination.

Food consumption data

Today, the FDA typically relies on food consumption data collected from the National Health and Nutrition Examination Survey (NHANES). Conducted by the Centers for Disease Control and Prevention (CDC), this survey provides detailed information on what and how much, thousands of people consume by asking them to recall what they ate over a 2-d period. On a more limited basis, the FDA uses 14-d data collected by private firms. Workshop participants agreed that these longer term surveys may provide more representative estimates of consumption of food over a lifetime, and so may better estimate dietary exposure to less frequently consumed food and nutrients, enhancing the agency's ability to establish limits for substances. Participants also encouraged the FDA to work with stakeholders to identify gaps in the available data and develop priorities and options to fill those gaps, including making use of new technologies.

Participants also saw value, in some cases, in making better use of the results of blood and urine samples that NHANES collects to monitor chemicals that people are exposed to, whether in food or from nondietary sources. Where these data raise potential health questions, they pointed out that the FDA would still need to distinguish the amount of the substance that results from food consumption (versus nondietary sources, such as environmental exposure) and to develop methods to directly correlate the levels to diet.

Consideration of subpopulations and cumulative dietary exposure

Before a substance is allowed in most foods, FDA scientists typically take into account exposure for 2 groups: (1) the general population (anyone over 2 y old) who eat the food containing the additive and (2) children between the ages of 2 and 5 y old. However, other relevant subpopulations are considered, when appropriate. For example, when data indicate a substance has the potential to affect reproduction or fetal development, women of childbearing age or who are pregnant are given special consideration. Additives in infant formula are evaluated separately.

Participants noted that, while FDA currently considers different subpopulations, including those based on age, gender, region, and ethnicity, on a case-by-case basis in its dietary assessment, FDA should consider expanding its subpopulation considerations on a routine basis. As a result of the Food Quality Protection Act (FQPA), the U.S. Environmental Protection Agency (EPA) regularly assesses exposure for a wide range of subpopulations. Participants noted that this could serve as a model for FDA's exposure assessments.

It is important to consider the cumulative effects of all dietary sources as well. In estimating the amount of a substance that people consume, FDA routinely considers the amount added to food in manufacturing. When appropriate, FDA also considers other dietary sources of the substance, such as whether it is naturally occurring in food, present in tap water, consumed as part of dietary supplements, or found in food contact substances (for example, packaging). Participants generally agreed on the importance of including all dietary sources in the exposure assessment so that it accurately represents what a person may actually be exposed to.

Interagency communication, collaboration, and coordination

Today, many aspects of data gathering and exposure assessment are spread among multiple federal agencies, including the FDA, CDC, U.S. Department of Agriculture (USDA), and EPA. Participants encouraged the FDA to expand the existing communication, collaboration, and exchange of scientific information and expertise through more regular interagency meetings or sabbaticals.

Congress should provide funding to increase collaboration among agencies as well as international regulatory agencies. This collaboration will enhance transparency and FDA's accountability for its decisions. Additionally, a harmonized approach to assessing exposure would allow for easier reassessments when the use of a substance changes.

Participants also suggested the FDA consider the National Research Council's recommendations for improving risk assessments that are described in its report, “Science and Decisions: Advancing Risk Assessment.” The report's recommendations broadly apply to exposure assessments for substances that are regulated by multiple agencies.

How much of a food substance do people consume on a regular basis? And is this amount of consumption safe? Answering these 2 simple questions is a complex and critical undertaking. FDA, industry, academics, public interest organizations, and other agencies must continually work together to refine exposure assessments. Ongoing discussions will allow the FDA and other stakeholders to keep pace with the latest technological advancements in the food additive industry and ensure our food supply is safe.


Over the past 50 y, FDA has developed a complex regulatory program, based on the Food Additives Amendment of 1958 to the Federal Food, Drug and Cosmetic Act (Public Law 85–929, 72 Stat. 1784), to ensure the safety of substances that are added to or come in contact with food. See Navigating the U.S. Food Additive Regulatory Program for detailed information on this program (Neltner and others 2011). Assessors determine the safety by (1) identifying and characterizing the hazard of the substance to determine the ADI and (2) calculating dietary exposure to determine the EDI for the intended and currently authorized uses. For there to be a reasonable certainty that no harm will result from the intended use, the expected exposure (EDI) must not exceed the ADI. With more than 10000 substances allowed to be added to human food in the United States (Neltner and others 2011) and a global food supply that is nearly as diverse as those who consume it, ensuring that the dietary exposure estimates are reasonable and sufficiently protective of public health is a daunting task that needs a rigorous yet adaptable approach.

An April 2011 workshop cosponsored by The Pew Charitable Trusts, the Institute of Food Technologists, and the journal Nature brought together experts to discuss how the safety assessors determine the ADI (Maffini and others 2011).

In November 2011, the same 3 organizations hosted a workshop, Perspectives on FDA's Exposure Assessment to Ensure Substances Added to Human Food Are Safe, that included more than 70 scientists and other experts from government, industry, academia, and public interest organizations to examine the principles underlying dietary exposure assessments for substances added to human food. Organizers attempted to reach a balance of perspectives and participants affiliations in the discussions. The final breakdown of participants was: government (42%); industry (30%); public interest (15%); and academia (12%). This workshop relied on extensive small-group discussions to respond to a series of questions developed by the organizing committee with input from the participants about the current approaches used by food safety agencies to assess dietary exposure. The primary focus of the workshop was FDA's dietary exposure assessment methods and related guidance for food additives and generally recognized as safe (GRAS) substances that are used in human food. Participants also discussed approaches used by other regulatory agencies such as EPA and the European Food Safety Authority (EFSA). EPA sets tolerances for more than 500 pesticides used in the United States on raw agricultural commodities and human food (Neltner and others 2011).

Note that the terms “substances,” “chemicals,” and “additives” are used interchangeably in this article to make the document more readable. The term “substance” is generally used; however, when a speaker or an organization uses a specific term, we use that term.

The organizers invited attendees to participate in a preworkshop training webinar to familiarize them with the discussion topics and format of the workshop. This workshop relied on extensive small-group discussions to respond to a series of questions developed by the organizing committee with input from the participants about the current approaches used by food safety agencies to assess dietary exposure. The workshop (see Table 1) began with a review of FDA's approaches to assessing dietary exposure assessment methods. This was followed by presentations by representatives of 3 federal agencies, 2 public interest groups, and 2 industry stakeholders who each shared their particular perspectives. Participants then split into 6 small groups to review materials prepared by The Pew Charitable Trusts, and discussed questions on a series of topics that had been developed in advance. After the discussion, the groups reported back to a plenary session. The 1st day concluded with EFSA providing participants with an overview of how it carries out exposure estimates, and on new tools being developed to estimate exposure to substances in food.

Table 1. Workshop agenda
Day 1—November 17, 2011
 • Welcome and workshop overview
 ∘ Erik Olson, The Pew Charitable Trusts
 ∘ Bob Gravani, Institute of Food Technologists
 ∘ Dennis Keefe, FDA
 • Introduction to dietary exposure assessment for food
 ∘ Tom Neltner, The Pew Charitable Trusts
 ∘ Barbara Petersen, Exponent
 • Federal agency perspectives on exposure assessment
 ∘ Michael DiNovi, FDA
 ∘ Philip Villanueva, EPA
 ∘ Donna Rhodes, USDA
 • Stakeholder perspectives on exposure assessment
 ∘ Michael Jacobson, Center for Science in the Public Interest
 ∘ Sean Taylor, Flavor and Extract Manufacturers Association
 ∘ Gina Solomon, Natural Resources Defense Council (NRDC)
 ∘ Leon Bruner, Grocery Manufacturers Association (GMA)
 • Small-group discussions—Round 1
 • Small-group reports from Round 1
 • Global perspectives on exposure assessment
 ∘ Jordi Serratosa, EFSA
 ∘ Davide Arcella, EFSA
Day 2—November 18, 2011
 • Questions for further consideration
 ∘ Abby Dilley of RESOLVE
 • Stakeholder perspectives on questions for further consideration
 ∘ Angela Lim, Danisco/International Food Additives Council
 ∘ Michael Hansen, Consumers Union
 ∘ Nancy Rachman, GMA
 ∘ Miriam Rotkin-Ellman, NRDC
 ∘ Barbara Petersen, Exponent
 ∘ Dennis Keefe, FDA
 • Small-group discussions—Round 2
 • Small-group reports from Round 2
 • Participant perspectives
 • Review of next steps and adjourn

Day 2 began with a review of questions developed from issues raised on day 1. Representatives from FDA, industry, and public interest groups provided comments before the participants broke into 4 small groups and discussed possible solutions to the issues raised. After reports from each of the groups, participants shared their observations before the workshop was adjourned.

This article presents the proceedings of this workshop. The authors shared the draft article with all participants and incorporated their comments whenever possible. Where comments conflicted, they worked with the commenters to find acceptable language. The authors provided the journal with the responses to each comment.

This article provides an overview of the current approaches to dietary exposure assessment and includes a summary of agency and stakeholder presentations given during the workshop. The article reviews 24 questions developed prior to the workshop and discussed by participants in small-group sessions. The 24 questions fell into the broader 6 topics listed below, each of which is described in detail. The seventh topic was a catch-all category for issues not addressed elsewhere and for additional issues that arose throughout the 2 d. To provide continuity and reduce duplication, questions from the 2nd round of small-group discussions and from the individual participant perspectives have been incorporated in the appropriate topics. The topics discussed were:

  1. Consumption data sources, changes, and trends: Focused on the different sources of food consumption data commonly used for dietary exposure assessments and how the data generated can reflect changes and trends in consumption in the American diet.
  2. Consideration of all dietary sources and additional uses: Discussed the aspects of the diet that should be considered in a dietary exposure assessment, and how to evaluate substances that have been approved for use in selected foods when additional uses are proposed.
  3. Approaches to postmarket assessments: Discussed how the methods, assumptions, and data for postmarket assessments differ from premarket assessments.
  4. Food contact substance assessments: Discussed how dietary exposure assessments are performed for food contact substances (FCSs) that may migrate into food. FCSs are not added directly to food but may contact food from packaging or other sources.
  5. Substances with similar biological effects: Focused on substances that have similar chemical structures or impacts on the body, and how those similarities are considered in a dietary exposure assessment.
  6. Subpopulation considerations: Discussed how dietary exposure assessments account for subpopulations and their specific dietary habits.
  7. Additional dietary exposure assessment issues: Discussed additional issues that are important but that did not fit into the categories above.

The discussion of each of the 1st 6 topics is divided into 4 parts:

  • a background to provide necessary context for the questions assigned to the topic;
  • the answers from participants in discussion groups assigned to those topics on the 1st day, including areas of general agreement and disagreement;
  • the answers to additional questions addressed on the 2nd day from a different mix of participants; and
  • a summary of comments made by participants at the end of the workshop.

Overview of Exposure Assessment for Food Additives

How does the law define safety?

The Food and Drug Act of 1906 prohibited adulteration of food by addition of any harmful or deleterious substance. The Food Additives Amendment of 1958 provided the 1st formal definition of food additives and established a framework to ensure their use is safe before use. However, “safety” was not specifically defined. In the committee report used to support the legislation, Congress, however, described what it meant to be “safe”:

“[I]nstead of insisting on proof beyond any possible doubt that no harm will result under any conceivable circumstances from the use of a particular additive—which could, of course, occur if an individual decided to eat a pound of salt or drink 4 gallons of pure water in an hour—the test which should determine whether or not a particular additive may be used in a specific percentage of relationship to the volume of the product to which it might be added should be that of reasonable certainty in the minds of competent scientists that the additive is not harmful to man or animal … ” (U.S. Senate Committee on Labor and Public Welfare 1958)

This 54-y-old statement has become the foundation of FDA's safety assessment, and has been incorporated into the Code of Federal Regulations (CFR), as follows:

Safe or safety means that there is a reasonable certainty in the minds of competent scientists that the substance is not harmful under the intended conditions of use (21 CFR §170.3(i)).”

The regulation also contains the 3 factors, mandated by Congress, that FDA is required to consider in determining whether a use is safe:

  1. The probable consumption of the substance and of any substance formed in or on food because of its use;
  2. The cumulative effect of the substance in the diet, taking into account any chemically or pharmacologically related substance or substances in that diet; and
  3. Safety factors that, in the opinion of experts, are generally recognized as appropriate (21 CFR §170.3(i)).

How does FDA assess safety?

In a premarket context, the agency typically responds to a safety assessment performed by a manufacturer or its representative as part of a request seeking FDA review for the use of a substance in or in contact with food. Manufacturers and trade associations may alternatively conduct their own safety assessment and decide that a substance is GRAS without requesting FDA's review. In a postmarket situation, FDA conducts safety reassessments on its own initiative when warranted by new information, such as when it identifies a potential public health concern. FDA may also initiate a reassessment in response to a citizen petition or when a firm petitions or notifies FDA of new uses of an approved substance.

Since the 1958 law was enacted, FDA has developed regulations and guidance describing how premarket safety assessments for food additives, GRAS substances, and FCSs are to be performed, whether by agency or industry scientists. According to these documents, a safety assessment should include both the amount of a substance that is not expected to cause harm over a lifetime (that is, its ADI) and the predicted dietary exposure to the substance (that is, its EDI).

The ADI calculation for a substance results from a review of the substance's chemical nature, available toxicology studies, and the application of appropriate safety factors (FDA 2000). The ADI is generally expressed as milligrams per kilogram body weight per day (mg/kg bw/d). Scientists calculate the ADI after reviewing the chemical nature of the substance and from applying appropriate safety factors to the lowest relevant no-observed adverse effect level (NOAEL) in available toxicology studies. It represents an estimate of the maximum amount of a substance that consumers can eat daily over a lifetime without significant risk of harm based on the scientific data and methods available at the time.

The EDI is based on a dietary exposure assessment, which is discussed in detail below.

When the ADI is greater than the EDI—that is, when people consume less of a substance than the calculated safe limit—there is a reasonable certainty of no harm from the proposed use of the substance in food. In slightly more mathematical terms:

  • EDI < ADI = Meets safety standards
  • EDI > ADI = Does not meet safety standards

If the EDI is greater than the ADI, FDA may determine that the substance should not be allowed on the market for the intended use. The agency also may consider other options to ensure that the EDI is less than the ADI, including:

  • Making a safety decision to restrict the amount of the substance allowed in food, or to limit the types of food to which the substance can be added; or
  • Refining the dietary exposure assessment to be more representative of potential consumption by:
  • Using a more intensive analysis of the data and modifying the assumptions to predict potential consumption; and/or
  • Obtaining additional information to reduce uncertainties in and refine the available data.

How does FDA conduct dietary exposure assessments?

FDA does not have a formal definition for “dietary exposure assessment” but prefers this term, in part, to clarify that its focus is solely on food consumption and diet, and that it does not take into account other (nondietary) potential routes of exposure. A widely recognized definition of dietary exposure assessment is from the World Health Organization (WHO) and the Food and Agriculture Organization (FAO):

“[t]he qualitative and/or quantitative evaluation of the likely intake of chemicals (including nutrients) via food, beverages, drinking water, and food supplements” (WHO 2009).

To explain its procedures and define its expectations for estimating dietary exposure, FDA has 3 guidance documents:

  • Guidance for Industry: Estimating Dietary Intake of Substances in Food (FDA 2006a)
  • Guidance for Industry: Recommendations for Submission of Chemical and Technological Data for Direct Food Additive Petitions (FDA 2006b)
  • Guidance for Industry: Preparation of Premarket Submissions for Food Contact Substances: Chemistry Recommendations (FDA 2002)

These documents provide many details that were discussed during the workshop. The guidance for direct food additives (substances added directly to food instead of indirectly through packaging or use on food handling equipment), which is also applicable to GRAS substances, identifies the basic information an assessor needs to estimate the dietary exposure to an additive, including:

  1. The specific foods in which the additive will be used;
  2. The typical and maximum levels of the additive in each food;
  3. Information on the consumption of those foods in which the additive is proposed for use;
  4. Information on subpopulations that might be particularly affected by the use of the additive (for example, infants or those on a calorie-restricted diet);
  5. Any anticipated increase in consumption due to a previously sanctioned additive's anticipated use, or the increase in the use level of a food additive in a new application or food; and
  6. Any anticipated increase in consumption from the additive's proposed use, if the food additive is also naturally occurring.

The guidance for direct food additives also explains key aspects of FDA's approach to dietary exposure assessment for substances added directly to food:

“The petitioner should provide, at minimum, a mean EDI (to represent the ‘average’ consumer) and the EDI at the 90th percentile (to represent the ‘high’ consumer), assuming a maximum use level of the food additive in all foods intended to be formulated with the additive.” Guidance for Industry: Recommendations for Submission of Chemical and Technological Data for Direct Food Additive Petitions (see Figure 1 for an explanation of mean and 90th percentile EDIs) (FDA 2006b).

Figure 1.

–Typical food consumption distribution If you graph the number of people eating a particular food as a function of the amount of that food consumed, it results in a distribution curve that typically looks like the one above. Usually, 65% to 70% of consumers eat less than the mean amount of food on a daily basis. In this situation, percentile and percent are synonyms. FDA generally bases the EDI on the 90th percentile consumer.

In other words, for premarket exposure estimates, the assessor is expected to use “conservative assumptions” that include:

  • the presence of the additive at the maximum level allowed in all types of food in which it is intended to be used;
  • saturation of the intended market for the additive; and
  • estimated intake of the additive by the 90th percentile of people eating the food.

Further, the additive is assumed to be consumed in these foods throughout the person's lifetime (chronic exposure). These assumptions are intended to provide a conservative—and thus protective of safety—estimate of dietary exposure in the premarket evaluation of the additive. Generally, FDA makes safety decisions using the 90th percentile EDI to protect “high” consumers over their lifetime.

The dietary exposure assessment for FCSs, which are not added directly to food but may come into contact with food from packaging or other sources, relies on similar assumptions. Assessors use additional factors to refine the estimate of how much of the substance might get into food from the FCSs. For more information, see Topic #4: Food Contact Substance Assessments.

How does FDA calculate an EDI for substances added directly to food?

There are 3 factors involved in estimating the EDI for an additive:

  1. Frequency: how often a person eats a particular food containing the substance on a typical day;
  2. Portion size: how much of the food a person eats; and
  3. Concentration: how much of the substance is in a given amount of the food.

The EDI is calculated by multiplying these factors as follows:

EDI = frequency × portion size × concentration

When the use of a single substance involves only a single type of food, such as fruit-flavored beverages, the analysis is relatively simple. The evaluation becomes more complex if an assessor must consider multiple variables, such as:

  • Varieties of foods in which the additive is intended to be used;
  • Different concentrations of the substance in each relevant food product;
  • Differences in food consumption among subpopulations (for example, children, pregnant women, teenage boys);
  • Cumulative contribution from other dietary sources of the substance, such as drinking water, dietary supplements, and natural occurrence in food; and
  • Interaction with substances that have similar chemical structures, pharmacological effects, or biological effects.

A systematic approach to estimating dietary exposure can help to focus resources where they are most needed. Assessors often begin by making a basic calculation to develop a “worst case,” or maximum, EDI. If this initial estimate, using the conservative assumptions mentioned above, is greater than the ADI, assessors can refine the analysis. If sufficient data are available, assessors use modeling programs to simulate potential consumption more closely. A premarket estimate tends to be more conservative (and therefore higher) because of the conservative assumptions regarding market penetration and consumption rates, than a postmarket assessment, which can be based on information about a substance's actual use, such as its market share and use level in food.

Critical variables in these estimates are how often and in what quantities people eat various foods that may contain the substance being evaluated. Assessors rely primarily on government food consumption surveys that describe which foods and beverages, and how much of each people consume.

What are the sources of food consumption data?

FDA and EPA use data from the ongoing food consumption survey conducted as part of the CDC's NHANES as the basis for dietary exposure assessments. This survey provides the most comprehensive, nationally representative food consumption data that are publicly available. NHANES relies on thousands of interviews each year with participants from across the United States who recall their food consumption over a 1 or 2 d period.

For many years, the USDA conducted a similar survey, the Continuing Survey of Food Intakes by Individuals. In 2002, USDA and CDC integrated this survey into NHANES.

Private firms also conduct food consumption surveys, with some covering consumption over 14 d. These surveys usually rely on daily diaries kept by participants. Although developed primarily for marketing research purposes, these data may also be used to support a dietary exposure estimate.

FDA also identifies 3 other data sources for estimating dietary exposure:

  1. Food/ingredient disappearance data: Developed by the USDA for 16 different types of food commodities, these data reflect the annual amount of commodities that enter the food supply. They are calculated as the amount produced plus imports minus waste and exports. These estimates are referred to as “disappearance” data because they represent the disappearance of food from producers into the marketplace. Disappearance data represent average food availability rather than actual consumption, and usually serve to confirm estimated mean consumption rates. This information is useful for identifying major trends in food consumption.
  2. Market basket studies: FDA monitors levels of substances in a wide range of foods through its ongoing market basket study (the total diet study, or TDS). These substances include pesticide residues, toxic and nutrient elements, industrial chemicals, and radionuclides. Samples of about 280 foods are collected from regions around the country and analyzed 4 times each year. Results from the TDS provide information on the actual consumption levels of these substances. The information is useful for postmarket dietary exposure assessments of substances analyzed in the study.
  3. Biomonitoring data: A routine part of the ongoing NHANES food consumption surveys includes measurements of selected biomarkers in blood and urine for contaminants, pesticides, nutrients, and other chemicals. The total amount of these substances in the body, or body burden, can be calculated from these data. While these studies are conducted on data from people chosen at random as a nationally representative population, they do not clearly reveal exposure due to food consumption. The sources of the substances—from diet, inhalation, dermal absorption, or otherwise—can be hard to determine. Therefore, biomonitoring data are difficult to use in a dietary exposure assessment. For more information, see Topic #1: Consumption Data Sources, Changes, and Trends.

For some situations, assessors have developed tailored exposure assessment methodologies. The Flavor and Extract Manufacturers Association (FEMA) Expert Panel, as part of its evaluation of flavoring ingredients, has utilized a maximized survey-derived dietary intake (MSDI) method that uses volume-based consumption information and data on the amount of flavoring substances that will be added to foods to determine whether a flavoring ingredient would be considered a GRAS substance. The Expert Panel considers that these ingredients are used in relatively small amounts, since consumers would find greater concentrations in their food unpalatable (Hall and Ford 1999; Hallagan and Hall 2009; Lambe and others 2002; Young and others 2006). The panel has made thousands of safety decisions using this method.

Similarly, FDA has developed a specific methodology to estimate the amount of a substance that may migrate into food from food packaging and FCS materials. For more information, see Topic #4: Food Contact Substance Assessments.

How does FDA use the exposure assessment when considering the data necessary to establish the ADI?

In this overview, we describe how the EDI and the ADI are related. However, the EDI may have an impact on the toxicological information required to develop the acceptable daily intake. FDA guidance and regulations allow a manufacturer to use the EDI to determine the type and number of toxicology studies required to establish safety. For direct additives at low levels of exposure FDA typically recommends in vitro toxicity tests to predict potential hazards. Negative in vitro tests indicate that no additional toxicological testing may be required at the low exposure levels, while positive results indicate that additional safety information is needed. As the exposure increases, data from in vivo animal studies are necessary to establish an ADI. The manufacturer also may use predictions of the substance's potential toxicity based on similar chemicals (NRC 1983).

For direct additives (that is, substances added directly to food) with low expected exposure levels, only in vitro gene toxicity studies may be required. If the estimates are higher, however, a set of animal studies will be needed. More information is available in FDA's Guidance for Industry and Other Stakeholders: Toxicological Principles for the Safety Assessment of Food Ingredients (more commonly known as FDA's Redbook) (FDA 2000).

For food contact substances (for example, components of materials used on food handling equipment or in packaging), no toxicology studies may be required if the exposure is estimated to be below 1.5 micrograms per person per day (μg/person/d) (FDA 2002). Because the toxicological testing requirements are keyed to the dietary intake, manufacturers generally develop a preliminary EDI as an early step in assessing the safe use of a substance in food.

How does EPA estimate dietary exposure for pesticides?

In response to the FQPA of 1996, EPA revised its dietary exposure assessment methodologies and issued new guidance for pesticide manufacturers (EPA 2000).

The main features of EPA's dietary exposure assessments are:

  • Developing estimates for short-term (acute) and long-term (chronic) exposure and comparing them to acceptable intake for each type of exposure. EPA uses the term Reference Dose (RfD) rather than ADI.
  • Using mean pesticide concentrations and mean food consumption for long-term exposure estimates.
  • Estimating short-term exposure by combining NHANES food consumption survey information with the amount of pesticides measured in foods to create a distribution of exposures. The EDI is estimated at the 99.9th percentile of the total population, not just those who eat the foods containing the pesticide residues.
  • Evaluating exposure for more than 20 selected subpopulations based on age (such as infants and children), pregnancy, race, ethnicity, and gender.

In addition, the law directs EPA to reassess its pesticide residue tolerance safety decisions every 15 y (7 U.S.C. §136a(g)(1)(A)). A tolerance is a limit on the amount of pesticide that may be in or on a specific type of food (for example, pesticide X can be used up to limit Y on strawberries). Using the agency's revised methodologies (Neltner and others 2011), EPA completed its 1st round of reassessments in 2007, reviewing 9721 tolerances for 581 pesticides.

According to the FQPA's reassessment requirement, pesticide manufacturers must submit relevant hazard and exposure data to EPA to conduct a safety assessment according to an EPA schedule. The assessment must be based on the most current scientific knowledge and exposure assessment methods. Although FDA has the authority to revise its regulations as needed, Congress has not given FDA a similar directive to conduct systematic reassessments of food additives and GRAS substances. In response to a request by President Richard Nixon, FDA undertook a systematic reassessment of many of the GRAS substances in the 1970s and early 1980s, based on scientific studies and methods available at that time. It was later expanded to include food and color additives. The agency terminated that systematic review in the early 1980s; however, as a result of this systematic review a database was established to monitor the amounts of certain additives produced. The agency continues to reassess substances added to food (food additives, GRAS substances, food contact substances) on a case-by-case basis.

Summary of Stakeholder and Agency Perspective Presentations

The workshop agenda included several panels that allowed experts to share a wide range of perspectives on the health effects of exposure to food additives, from assurances of safety to assertions of inaction. Panelists consisted of representatives from 3 U.S. federal agencies, EFSA, 3 food manufacturer trade associations, 1 consulting firm, and 3 public interest organizations.

Federal agency perspectives

Philip Villanueva, a statistician in the Office of Pesticide Programs (OPP) at EPA, noted that the agency determines the safety of pesticides in part by assessing the dietary and residential or occupational exposures and associated risks. This analysis requires that food consumption data be combined with pesticide residue concentrations to develop a distribution of dietary exposures. OPP uses a number of software models that incorporate food commodity consumption data from NHANES to perform the agency's risk assessments. Important questions that inform this modeling include: Are particular subpopulations, such as infants or children, at greater risk? How might consumption change over time? What is the impact of processing since it can remove and/or concentrate pesticides used on fruits and vegetables?

For its exposure assessment models, EPA relies on the What We Eat in America (WWEIA) survey, which is conducted as part of NHANES. The agency also uses data from its own Food Commodity Intake Database and from the USDA's Pesticide Data Program, which is a nationally representative commodity residue sampling program. These resources translate foods reported as eaten to raw agricultural commodities using a food vocabulary EPA has developed and converts more than 5000 food codes into recipes containing roughly 540 different food commodities. To produce more accurate estimates of the entire range of exposures throughout the population, the agency uses probabilistic techniques, which combine the distribution of residue levels in crops with the distribution of food consumption.

Donna Rhodes, a nutritionist with the USDA, described the WWEIA survey in more detail. The survey collects 2 d of dietary intake data for about 8000 people every 2 y. Data for the 1st day of consumption are collected by WWEIA staff in person, and data for the 2nd day are collected by telephone. Data for 2007 to 2008 are available, and data for 2009 to 2010 are on schedule for a summer 2012 release. Each survey is designed to oversample selected subgroups. In 2007 to 2008, persons 60+, African-Americans, Hispanics, and low-income persons were oversampled. The sample sizes are still small for particular subgroups, such as children.

Rhodes noted that the survey results have shortcomings in some areas. For example, it is difficult to separate homemade from more processed foods, since respondents’ definition of “homemade” or “processed” can vary. Also, more than half of the U.S. population reports eating at least one food or beverage item from a restaurant on any given day. It is difficult to know how these foods were prepared. The overall strategy for updates is to reflect marketplace changes, scientific issues, and policy needs.

Finally, Michael DiNovi, a chemist in the Division of Biotechnology and GRAS Notice Review, Office of Food Additive Safety (OFAS), FDA, pointed out that there are 3 broad categories of food additives. The 1st category includes substances with dietary exposure far below the ADI, while the 2nd includes substances with dietary exposure far above the ADI. From a regulatory standpoint, both of these cases are trivial, as it is a relatively simple matter to reach a conclusion regarding safety. The 3rd category is more complicated, and consists of substances with dietary exposure close to the ADI, where the outcome of a safety assessment is not immediately clear.

DiNovi added that regulations for such additives need to be fit for purpose. For example, food intake data are inevitably uncertain, so regulations should not be based on false precision. Therefore, it is vital to have the best data possible. While precise data are not essential, the data need to be sufficiently representative and robust to perform accurate safety assessments.

Food industry perspectives

The food industry will not be successful if its products are unsafe, said Leon Bruner, senior vice president of the Grocery Manufacturers Association (GMA), an organization that represents food manufacturers. For that reason, safety is the number one job for the industry. The safety assessment processes used today for food ingredients are effective given our current understanding of food safety science, said Bruner. He also noted that as scientific research delivers new discoveries and technological advancements, it is appropriate to update safety assessment procedures in a way that allows the use of these advancements. Also, if consumers are interested or need more information to build confidence about food safety, they should have appropriate visibility to safety assessment procedures, he added.

When compounds are shown to be unsafe for use in food products, the food industry will not use them, Bruner stated. In order to make effective assessments of ingredient safety, adequate data must be available. Programs like NHANES are critical sources of information that can be used in a safety assessment process. In addition, FDA and other agencies need adequate funding to continue producing information to protect the public and increase public confidence that foods are safe.

The FEMA Expert Panel, which is an international panel that evaluates the safety of food flavorings, uses a variety of methods to estimate exposures to flavorings added to foods, said Sean Taylor, FEMA's scientific director and a managing director at Verto Solutions. One is the MSDI method, which is an anticipated intake based on a company's premarket estimation of how much of a flavoring ingredient it expects to sell into the food supply. Another is the possible average daily intake (PADI), which relies on assumptions about how much of a given substance will be used in different categories of food. Another approach is a method similar to that used by the FAO/WHO Joint Expert Committee on Food Additives (JECFA) in its evaluation of flavoring ingredients.

Taylor noted that from an international perspective, regulatory authorities and others tasked with ingredient evaluations rely on a number of methods. These different approaches generally focus either on the use of a substance within the final food product, or the volume of a substance produced and introduced into the food supply. Taylor said that both sets of methods tend to be conservative in estimating potential exposures to flavoring substances. For example, companies routinely overestimate the amount of a newly evaluated flavoring ingredient that they will be able to sell and will be used in food, particularly when the ingredient is new to the market. Using either volume- or use-level-based approaches, said Taylor, the vast majority of flavoring ingredients on the market have a margin of safety relative to the NOAEL greater than 100, and for many substances the margin of safety is much higher. FEMA conducts both premarket evaluation and postmarket dietary exposure evaluations. In the postmarket scenario, the flavor industry conducts a volume-of-use survey for all flavoring ingredients every 5 y. The information is used to gauge the current use of the flavoring ingredient as well as how the use of flavoring ingredients has changed. As a result of these continuing surveillance methods, safety evaluations are continually revised using updated exposure information.

Public interest organization perspectives

Gina Solomon, codirector of the Occupational and Environmental Medicine Residency and Fellowship Program at the University of California, San Francisco, and a senior scientist at Natural Resources Defense Council (NRDC), emphasized the potential of new science to increase knowledge about the health effects of food ingredients. Toxicological screening methods such as the Tox21 program and biomonitoring efforts are creating much more data than have ever been available before. For example, Solomon noted that toxicological screens could challenge the assumption that all food substances are nontoxic, while biomonitoring might reveal unexpected exposures, such as the widespread exposure to bisphenol A (BPA).

At the same time, risk assessment is becoming more sophisticated as it focuses on vulnerable subpopulations, performs cumulative exposure assessments, and considers human variability in responses to substances. Solomon also pointed to the potential of new technologies to provide more information about exposure. For example, people can use their cell phones to take pictures of what they are about to eat, producing much more accurate consumption data than do dietary recall surveys.

The public is very concerned about what is in food, Solomon observed. Thus, scheduled reassessments of food substances are needed as science progresses. Also, national surveys of food consumption are obviously missing important information, such as foods that are uncommonly consumed or foods that are consumed in large quantities by relatively small numbers of people. EPA has done a fairly good job of considering subpopulations, said Solomon, and FDA could incorporate its expertise and experiences into exposure assessments.

Finally, Michael Jacobson, director of the Center for Science in the Public Interest, said that FDA systematically ignores or downplays evidence of risk from food additives. For example, in 1984 FDA labeled Red Number 3 dye a carcinogen, but the substance has not yet been banned. From Jacobson's perspective, other dyes containing cancer-causing contaminants also have not been banned. In addition, some types of caramel coloring contain illegal levels of cancer-causing contaminants, but FDA has not taken action. And some evidence points to chemicals in artificial sweeteners causing cancers in laboratory animals, he said.

In addition, good evidence indicates that food dyes adversely affect the behavior of children, Jacobson said. FDA held a public meeting on the issue and acknowledged that food dyes have behavioral effects, but it has not yet taken any other steps. However, these problems are minor, Jacobson said, compared with the health problems caused by salt and partially hydrogenated oils. If sodium consumption were cut in half, he said, 100000 lives a year would be saved. Yet 40 y of voluntary efforts to limit sodium consumption have been a failure, and consumption levels are higher than they were in the past.

Jacobson suggested several legislative remedies to the problems he cited. Congress should pass legislation to more tightly control GRAS substances, so that companies would be required to provide FDA with toxicological and use data for these food substances. Second, food additive legislation should specify that behavioral reactions, neurotoxicity, or severe allergic reactions should disqualify substances from being considered as GRAS substances. Finally, cyclic reviews should be conducted based on risk levels for food additives.

Global perspectives on exposure assessment

Jordi Serratosa, the European Food Safety Authority (EFSA) liaison officer to FDA, explained that risk assessment and risk management are conducted separately in Europe, unlike in the United States, where FDA performs both functions. EFSA conducts hazard identification and characterization, exposure assessment, and risk characterization, whereas the European Commission (EC) assesses policy alternatives, and selects and implements appropriate risk management options. Responsibilities for various aspects of food safety in the United States are split among FDA, USDA, CDC, and EPA, whereas all of these responsibilities are combined in the European Union (EU) under the Directorate-General for Health and Consumers (DG SANCO). The directorate makes legislative proposals to EU member states, the European Council, and the EU Parliament, conducts stakeholder meetings and impact assessments, and coordinates databases with member states.

Davide Arcella, an officer in the Risk Assessment and Scientific Assistance Directorate, Dietary and Chemical Monitoring Unit, at EFSA, presented details on how the agency conducts dietary exposure assessments. The agency is developing a Comprehensive European Food Consumption Database that contains food consumption data from 32 different dietary surveys from 22 member states for different age groups from infants to the elderly. However, individual member state surveys use different measures and methodologies, which complicates comparisons and creates information gaps. To fill these gaps, EFSA is developing a “What's on the Menu in Europe” program, which is seeking to collect food consumption data at the EU level and in each of the member states using methods that allow for comparisons of results.

EFSA has also developed a food categorization system called FoodEx to describe in detail the ingredients for about 1500 different food codes. An elaboration of the system, known as FoodEx 2, has both a core list of foods and further levels of detail. In addition, different computational models are used to combine or integrate food consumption data with residue or concentration data, depending on the purpose of the assessment and the availability of information. In the future, Arcella said, data from the comprehensive database should be used to validate and cross-check screening methods based on model diets. Also, high consumers need to be studied to calculate acute and chronic exposures, and the use of probabilistic exposure assessments needs to be expanded and tested.

Perspectives on themes emerging from 1st day discussions

The small-group discussions on the 1st day led to additional questions that were discussed on the 2nd day. Those questions were grouped into 4 themes: data needs; harmonization of methodologies; greater transparency; and need for research. FDA and the stakeholders shared the following perspectives on these themes during a plenary session.

Data needs. 

Dennis Keefe, director of the Office of Food Additive Safety, FDA, emphasized the primary importance of data. More data and better-quality data, based on the composition and consumption of food, will improve exposure assessments, he said.

Answering the seemingly simple question of how much people are exposed to a given food substance can be surprisingly complex, said Miriam Rotkin-Ellman, a scientist with the NRDC. In order to protect public health, it is important to thoroughly evaluate the exposure distribution, assess exposures for the most vulnerable, and account for cumulative exposures. Evaluating cumulative exposures must include assessment of multiple chemicals, sources, and exposure routes. Unless these complications are explored, she said, potentially important information can be missed, leaving people at risk.

Barbara Petersen, principal scientist with Exponent, Inc., observed that exposure data on particular groups, such as children, can be especially valuable. For this reason, data needs should be prioritized so that the focus is on data that will make the greatest difference. For example, testing animals at very high doses may be less relevant compared with data on long-term cumulative exposures.

Sources of data such as NHANES certainly can be improved, said Nancy Rachman, senior director of science policy for GMA, but she, too, noted that improvements have to result in better decisions, not just a better database. For example, one important question is whether regulators are being sufficiently conservative.

Food additives are very diverse and have many purposes, noted Angela Lim, senior manager of corporate regulatory affairs for Danisco. Similarly, the data gathered to ensure food safety are diverse. In some areas the data are strong, and in other areas the data have gaps. An additive like grape extract can contain many components. Having a database of the biologically active substances in such additives would be very useful, she said.

The importance of data argues for several policy changes, according to Michael Hansen of Consumers Union. Source reporting should be required so that exposures can be assessed, he said. This should cover GRAS substances as well as other food additives, even if legislative changes are necessary. Other federal agencies provide such reports, such as USDA's reports on the detailed use of pesticides on crops. As a specific example, reporting to FDA of antibiotic use is incomplete and should be expanded.

Also, exposure data should not be limited just to food, Hansen continued. Exposure can come from many different sources, such as exposure to BPA from not only food and drinking water, but consumer products and other sources. Similarly, different compounds may affect the same organs, which would require that collective exposures be assessed.

New technologies offer exciting prospects for gathering data that are not available today. For example, consumers can use smart phones to take pictures of the foods they eat, with the pictures being sent to a central location for analysis. In addition, various nongovernmental agencies have created networks that could provide resources for research and advocacy. For example, Consumers Union has a Listserv with 900000 members who can weigh in on various health questions. If used creatively, new technologies could greatly enhance public health, including the health of particular subpopulations, such as children or the high-end consumers of certain foods.

Harmonization of methodologies. 

With regard to the harmonization of exposure assessment methodologies among federal agencies, Keefe observed that the 1st question to be answered should be why the methodologies need to be harmonized. The tools, data, and goals of different agencies can differ. He asked what the alignment of methodologies is designed to achieve.

Hansen pointed out that assessments of total exposure may demand such harmonization. The scientific details would need to be worked out, but some entity in government should be looking at the larger picture. For example, some aspects of the FQPA covering pesticides could be applicable to food additives, he said. Partnerships will be needed to carry out coherent assessments, and such partnerships should include all stakeholders, not just government and industry. In particular, public interest communities should be represented, said Hansen, as should the people who are being exposed to substances.

Petersen and Rachman both observed that exposure assessments need to be fit for purpose, especially as the federal budget tightens. Thus, harmonization of methodologies needs to reflect the circumstances. For example, the use of pesticides on a crop can cause many more people to be exposed to a substance than the addition of a food additive to a narrow food category. Such differences may call for different kinds of tests and measures. In addition, assessments need to be straightforward, simple, and reliable, so that people do not try to do too much with the available data. FDA can learn from collaborations with other agencies and with international organizations, said Rachman. For example, internationally, JECFA has criteria for deciding when an additive should be reevaluated.

Greater transparency. 

Rachman also observed that the deliberate pace of decision making at FDA contributes to scientifically based judgments. However, the system could be made more responsive through improvements in transparency. FDA could provide more information about how it gathers data and makes decisions, while industry could be more transparent about how it does GRAS affirmations without compromising proprietary information.

FDA and industry deal with issues on a case-by-case basis, she continued. This requires flexibility to gather complex information and make appropriate decisions to protect public health.

Lim pointed out that demands for transparency could conflict with the need to remain competitive. For example, companies have seen competitors enter the market based on information in a GRAS notification. As a result, companies can have a disincentive to notify FDA of their use of such substances. Data exclusivity or cost sharing of data development could provide an incentive for companies to be more forthcoming with data.

Need for research. 

Finally, several speakers emphasized the great need for more research on exposure assessments and food safety in general. Within academia, food toxicology is severely underfunded, said Keefe, adding that drug safety receives much more funding than food safety, yet people eat food every day, including foods that can be harmful to health, such as sodium and trans fats. More broadly, said Rachman, research on food safety assessment is woefully underfunded. The research portfolio needs to encompass not only basic research but public health and improvements in risk assessment methods that are useful for regulators.

Topic #1: Consumption Data Sources, Changes, and Trends

The discussion focused on the different sources of dietary consumption data commonly used for dietary exposure assessments and whether they identify changes and trends in the American diet. The analysis considered the advantages and limitations of the standard methods for collecting food composition and consumption data.


A dietary exposure assessment relies on 2 types of information:

  • Food composition data: Identifies and quantifies ingredients, nutrients, contaminants, and some additives in food as served. Using a pizza as an example: How much flour is used? How much of an ingredient or contaminant is in the flour? How much calcium is in the cheese?
  • Food consumption data: Identifies and measures what types of food people eat, how much they eat, and how often they eat it. Using pizza as an example: How many slices did people eat? How large was the slice? How often was pizza consumed during the course of the survey?

Table 2 provides details on the specific sources of information in each type of data.

Table 2. –Data sources used to identify composition and consumption rates of food
Data typeData sourceBrief description
Food composition dataUSDA National Nutrient Database for Standard Reference (SR)Publicly available analytical/label ingredient data; provides the foundation for many other food composition databases
 USDA Food and Nutrient Database for Dietary Studies (FNDDS)Combines USDA's Standard Reference data with recipe data in order to provide nutrition data for dietary studies; used to process NHANES dietary intake data and estimate nutrient intake
 FDA total diet study/market basket studyAnalytical data collected by FDA on levels of specific contaminants, pesticides, and nutrients in a “market basket” of 280 prepared foods
 Food Package Label DatabasesCommercial databases providing food labels, nutrition facts, ingredient lists, and images of U.S. food products
 Food Commodity Intake Database (FCID)Developed by EPA's Office of Pesticide Programs, this database can be used to translate the foods reported in consumption surveys into recipes of food commodities
Food consumption dataCDC NHANES food consumption survey (1 to 2 d data)Publicly available consumption data source; provides dietary survey data combined with corresponding FNDDS nutrient data; updated every 2 y
 14-d consumption diariesCommercial food frequency data collected by marketing groups; usually in the format of food diaries
 CDC NHANES food frequency questionnaire (FFQ)Survey of what foods participants recall eating over the past 12 mo; developed by the NCI and used by NHANES to support recall data and estimate long-term consumption
 Market Share DataCommercial point-of-sales data from participating stores; sales information on individual products; can be combined with label information to estimate consumption of ingredients
 USDA food disappearance figuresData on the total poundage production of specific commodities that is consumed or “disappears” into the marketplace for food or other uses; annually reported to USDA/Economic Research Service (ERS)
 National Academy of Science (NAS) food ingredient disappearance dataData on the annual production (poundage) of certain food ingredients; have been collected at various times by NAS and through government contracts
 NHANES biomonitoring dataA potential exposure/monitoring data source; urine and blood sampling data collected by NHANES on a nationally representative sample of the U.S. population since the late 1990s for selected biomarkers of nutrients and chemicals

An assessor's choice of data sources is usually determined on a case-by-case basis, and dependent upon such factors as the best data that are available and whether the exposure estimate is for a premarket or postmarket assessment. As a practical matter, assessors, both within and outside of FDA, rely on food consumption survey information collected by the CDC's NHANES when estimating exposure to most additives used directly in food. They prefer NHANES because it is a large, regularly updated, publicly available database with detailed information on a statistically representative sampling of the U.S. population. EPA also uses NHANES for pesticide assessment. Other data sources are most useful for evaluating trends, identifying ingredients (for example, raw commodities) used in processed foods, and developing postmarket exposure estimates. See Table 2 for a description of each.

The NHANES food consumption survey is based on interviews with participants who recall 1 or 2 nonconsecutive days of their food consumption. Participants report to an interviewer the amount and types of food and beverages they consumed in the previous 24 h for each day of the survey. For example, a participant might tell the interviewer that he or she ate a sandwich and drank coffee for lunch. The interviewer then asks follow-up questions to gain a more accurate understanding of the serving size and the ingredients in the foods and beverages. The 2007 to 2008 survey included consumption data for more than 7800 individuals.

Assessors use the NHANES data to develop EDIs representing the mean and 90th percentile consumers who eat the food that contains or may contain the substance. Because the initial sample size is large, the assessors may also estimate consumption for subpopulations such as toddlers, pregnant women, or the elderly. When estimating exposure for subpopulations, assessors must make sure that the sample population is large enough to allow for a statistically valid estimate for consumers of the food.

The National Cancer Institute (NCI) designed a food frequency questionnaire (FFQ) that has been included in NHANES since 2003. This FFQ asks participants to recall and identify the foods they ate in the past 12 mo. Assessors can use this information to identify eating habits that may not be captured in an interview based on 1 or 2 d of consumption. However, the FFQ relies on a participant's memory over a long period of time and therefore introduces larger systematic error into the assessment. FDA prefers the NHANES 24-h recall data due to the inclusion of amounts of food consumed, the specificity of the food products, and the ability to correlate them to concentration/ingredient data.

Marketing groups such as the Market Research Corporation of America collected 14-d food frequency data from over 25000 participants in the 1980s. These studies are no longer conducted by the firm, but the archived data are still used for comparison purposes. Other firms currently collect proprietary 14-d food intake data for manufacturers and retailers to identify trends or describe the potential market for products. This costly information is not freely available to the public, so it is not known how accurate, complete, or representative it may be. FDA has recently purchased access to some of these resources to evaluate their usefulness.

The challenge of biomonitoring. 

Apart from nutrients, FDA does not typically use biomonitoring data to assess postmarket exposure to substances added to food. FDA states that “the biomarker methodology requires demonstration of a quantitative relationship between intake of a substance and the amount of the substance or a metabolite in the body tissue or fluid of humans” (FDA 2006a). This information is seldom available. Additionally, the guidance specifies that “the methodology also requires an understanding of dietary and endogenous factors that might influence the concentration (that is, uptake, clearance, excretion) of the biomarker in different body tissues.” This is particularly important when diet is not the only possible source of exposure to the substance. Assessors do not frequently use biomonitoring data in estimating dietary exposure to food ingredients because it would require a quantitative relationship between dietary exposure to the chemical and the amount of that chemical or its metabolite in the human body. Some research studies have used biomonitoring data combined with other information to back-calculate potential dietary exposures to certain chemicals (Marsee and others 2006), although these studies were not provided for discussion during the workshop. Note that CDC selects chemicals for biomonitoring for several reasons, including available scientific data on exposure, the seriousness of potential health effects associated with exposure to the chemical, the availability of test methods and participant samples, and the cost of the test (CDC 2012).

Changes and trends in dietary consumption. 

Assessors can use the dietary information described in Table 2 to monitor changes and trends in a population's dietary consumption over time. The food types and portion sizes that are available today are different from those available 50 y ago. Ongoing food consumption data will reflect the preferences and nutritional priorities of the total population and subpopulations. Using current and past consumption data, assessors can quantify differences in the American diet over time.

Substantial dietary trends and shifts can occur quickly. Consider, for example, the rapid rise in popularity of Greek yogurt. Whereas 5 y ago, Greek yogurt may have been infrequently consumed by the general population and was available only through specialty grocery stores, it now is widely consumed and is available in many traditional grocery stores. Consumption data from 5 y ago would not reflect current consumption of this specific food as distinguished from other kinds of yogurt.

Answers to questions from 1st round of small-group discussion

Question 1: Do longer term food consumption surveys (for example, 14-d surveys) provide more representative data than shorter term food consumption surveys (for example, 1- or 2-d surveys)? Both surveys have benefits and limitations, and the data are used based on their “fit-for-purpose.” The 14-d survey data are more representative for less frequently consumed foods or nutrients (for example, cranberry sauce). FDA finds that using 2-d data is acceptable to assess exposure to frequently consumed foods (for example, cereal). It also is advantageous because the raw data are publicly available, and other scientists can reproduce the analysis. Data from the 2-d surveys are used to fill data gaps in other food consumption data sources, such as serving sizes for 14-d surveys. The primary drawback of using 2-d dietary survey data to estimate long-term dietary exposure is that it tends to overestimate dietary exposure to reported foods. The analysis assumes respondents would eat the reported food every 1 or 2 d for the rest of their lives and that only these people consume the food. People's diets tend to vary more than that. Additional methods of analysis may become available in the future that allow for assessors to calculate “usual intake estimates,” which take into account covariates and assumptions that the food item is not actually being consumed every day by a participant.

By comparison, 14-d diaries often capture less frequently consumed foods and are more reflective of a diverse diet. Longer term surveys are cost prohibitive for large numbers of participants. Both short- and long-term surveys are limited by reliability of participant input (for example, people may make different dietary choices when they know that someone is monitoring their consumption). Participants in 14-d surveys, however, may also have response fatigue (noting everything consumed for 14 d straight is tedious). Lastly, these data typically contain consumption frequency but not amounts consumed (for example, a participant ate applesauce 3 times in the 14 d, but there is no information on the portion size). Exposure assessors work within the limitations of the available food consumption data because they cannot design and collect data for each individual food or additive.

Question 2: Do dietary exposure assessments that rely on recall provide an adequate basis to estimate lifetime exposure? Relying on participant dietary recall can add uncertainty to the lifetime exposure estimate because participants may forget, omit, or alter their food consumption. Assessors apply conservative assumptions included in FDA's standard exposure methodology to minimize some uncertainty and reporting errors. The group determined that this question needed further discussion in later sessions of the workshop.

Question 3: What are the limits of NHANES in estimating consumption by subpopulations? NHANES is designed to be a nationally representative survey that includes more than 10000 respondents distributed among various subpopulations (for example, race, age, gender).

NHANES survey data can be analyzed by age. Dietary exposure assessment for children under age 2 is more complicated because infants may be bottle-fed or breast-fed, or eat solid foods at different stages. Food consumed by infants and young children is reported by parents or another proxy caregiver.

NHANES does not have sufficient data to address regional and seasonal differences in food consumption. The surveys are not designed to be regionally representative. NHANES does not collect narrow regional data, in part to maintain participants’ confidentiality and to ensure that their extensive health history remains private. NHANES survey data can model dietary exposure of an average American population; however, they do not have sufficient statistical power to adequately represent regional food consumption. For example, after the 2010 oil spill in the Gulf of Mexico, it would have been advantageous for assessors to be able to run an analysis of shrimp consumption for a specific location, such as Louisiana, and time of year. However, the available 2-d data cannot be narrowed down to a specific location to estimate food consumption without introducing additional assumptions into the calculation. Similarly, assessors cannot estimate differences in dietary exposure for residents of a city where a manufacturer is located, using the publicly available data, without including variables or assumptions.

Question 4: Is there a sufficient statistical power in food consumption surveys to reflect subpopulations and their eating habits (for example, children, ethnic groups, pregnant women, and the elderly)? Assessors often recognize that the data they are working with have limitations in statistical power (for example, the sample size) and representativeness (for example, how well the survey participants reflect the broader population). The smaller the sample size used to develop an estimate, the less accurate and representative the estimate becomes. An assessor may get an estimate, but the error may be too large for it to be relied upon for a safety assessment. Data from multiple NHANES survey years can be combined to yield a larger survey population. For example, combining the 2003 to 2004, 2005 to 2006 and 2007 to 2008 surveys yields a combined 2003 to 2008 dietary survey with triple the number of people who were in the individual surveys. While combining several years of data increases the sample size of a subpopulation, assessors must recognize changes in foods and eating patterns over the time frame for the combined surveys, as well as the survey methodology. Assessors do not routinely conduct dietary exposure assessments based on economic status, but they could as NHANES data captures this information.

Question 5: Are biomonitoring data available that can be appropriately linked to dietary intake? If so, should FDA be using these data in performing postmarket dietary intake estimates? Biomonitoring provides valuable information about substances that are actually present in the body and how frequently they appear in the sample population. The group reached a general consensus that there are significant biomonitoring data available, but that there are limited data currently available for direct food additives. Many of the substances measured in biomonitoring studies are environmental substances or food contact substances. With the available data and methodology, it is not always possible to distinguish if a biomarker found in urine is a result of dietary or other exposure, such as environmental factors or personal care products. The body may process the substance the same way independent of its source.

The relationship between food consumption and the biomonitoring data, however, needs to be clearly understood before the data can be directly correlated to dietary consumption and exposure. When absorption, distribution, metabolism, and excretion data from animals are available for additives, assessors could use physiologically based pharmacokinetic models to use the biomarker data to estimate exposure. As more biomarkers are developed and validated, scientific and regulatory communities will have additional valuable information for decision making about the spectrum of substances to which we are exposed.

Question 6: Under what circumstances should a population's exposure to a substance be reassessed? Should there be a reassessment when diets have noticeably changed? Should reassessments be carried out on a regular basis, and if so, how frequently? It would be impractical and resource intensive to reassess thoroughly all dietary exposures regularly (for example, with each NHANES data release). Instead, reassessment could be prioritized based on several factors, such as how close the EDI is to the ADI, whether the substance's safety limit was based on adverse effects in humans, or whether animal toxicology data indicate adverse developmental effects.

Exposure reassessments should be conducted when:

  • Diets have noticeably changed;
  • FDA receives a petition for new uses of an approved food ingredient or additive;
  • New toxicological information becomes available;
  • Increases in a substance's use are indicated by market basket studies;
  • Congress or an international regulatory agency questions the safety of an additive;
  • Changes are made in manufacturing or sourcing of the ingredient that could affect the identity; or
  • Improved measuring tools are available.

Answers to questions from the 2nd round of small-group discussion related to consumption data sources, changes, and trends

Question 7: Do dietary exposure assessments that rely on short-term recall provide an adequate basis to estimate lifetime exposure? If not, what other sources could be used to estimate lifetime exposure? FDA has recently taken steps to improve models of consumption patterns and lifetime exposure by purchasing commercial 14-d intake studies because 2-d surveys may not adequately model lifetime exposure.

The available short-term recall data could be enhanced by incorporating the biomonitoring data that are collected from NHANES studies. As discussed earlier, there needs to be a defined relationship between biomarker levels and food consumption for biomonitoring data to be useful. Additional dietary questionnaire methods, such as the NCI's FFQ 1-y surveys of consumption, could provide valuable data to supplement the short-term studies. Substances that warrant additional evaluation can be identified by comparing the results of different studies and identifying inconsistent results.

Question 8: How can new technologies be used to gather more [and better] (and more real-time) data on dietary intake, food consumption, food composition, and biomonitoring [and to what end]? Session participants amended the discussion question to include the bracketed text. NHANES surveys provide most of the detailed and nationally representative food consumption data used in dietary exposure assessments. Scientists are continually developing new technologies to supplement NHANES that potentially could be more rapid, more accurate, and less expensive, depending on the purpose. These additional exposure data sources include:

  • Research studies on a specific chemical;
  • Grocery store loyalty card databases;
  • Cards collecting information about individual habits in real time in different seasons and over a longer time; and
  • New systems of collecting information via mobile technologies such as smart phones, which can be used to collect food consumption data with links to restaurant menu data.

The primary purpose of these databases is not necessarily dietary exposure assessment, so the results would require a multistep quality control process to ensure that they provide reliable data to supplement dietary exposure assessments.

Question 9: How should FDA and stakeholders build and foster partnerships to gather data, conduct research, and assess data? NIH and EPA have models of public–private partnerships and scientific review panels that FDA and stakeholders could use as a framework to build opportunities to work together. These partnerships would not replace FDA's decision-making authority; instead, FDA and stakeholders would have access to a higher quantity and quality of data for exposure assessments.

Transparency is necessary for these partnerships to be effective. Specifically, both FDA and stakeholders need to share data. Some participants expressed concern about this level of transparency because all data and results from the partnerships with FDA might be available in the public domain, including industry trade information.

Participant perspectives raised at the end of the workshop related to data sources

At the conclusion of the workshop, participants provided the following perspectives.

  • NHANES data are the best data currently available and are the foundation of all dietary exposure assessments. These data, however, need strengthening. FDA, USDA, EPA, and CDC should strengthen the NHANES data with data collected in other federally funded health and nutrition studies (for example, the National Children's Study).
  • Assessors should supplement NHANES data with additional data sources to enhance the quality of the dietary exposure assessment by being more reflective of larger numbers of the U.S. population and subpopulations.
  • FDA, USDA, and EPA should improve interagency and stakeholder collaboration, particularly with the USDA, to capture robust data.

FDA should publish in a more easily accessible fashion exposure assessment guidance materials and models. This could include technical guidance for assessors and an accessible overview for consumers.

Topic #2: Consideration of All Dietary Sources and Additional Uses

The discussion focused on what types of exposures should be considered and how exposure assessments should be conducted when additional uses of a substance are proposed. The analysis considered whether the current approaches that are used to model cumulative dietary exposure meet the need and intent of the law.


The Food Additives Amendment of 1958 requires that FDA consider the cumulative effect of an additive in the diet over a lifetime, together with any chemically or pharmacologically related substances in such diet (21 U.S.C. §348(c)(5)). “Cumulative effects” and “cumulative exposure” are distinct, but related, concepts. “Cumulative effects” describes the combined biological effects of multiple factors, such as substances with similar biological effects, subpopulation-specific effects, and cumulative exposure. FDA uses the term “cumulative exposure” to describe the dietary intake of a substance from all sources of conventional foods that it regulates. The term is similar to but not the same as the term “aggregate exposure,” which is used by EPA (see side box). Other workshop sessions discussed subpopulations or substances with similar biological effects; this one focused on cumulative exposure.

Considering diet. 

Neither Congress nor FDA defined “diet.” Presumably, the term means all types of food and beverages, including dietary supplements. Congress defined “food” as “(1) articles used for food or drink for man or other animals, (2) chewing gum, and (3) articles used for components of any such article” (21 U.S.C. §321(f)). FDA clarified that food “includes human food, substances migrating to food from food-contact articles, pet food, and animal feed” (21 CFR §170.3(m)).

The WHO International Programme on Chemical Safety (IPCS) defines the scope of a dietary exposure assessment to include food, beverages, drinking water, and food supplements (WHO 2009). This definition of “diet” includes factors beyond FDA's regulatory authority, such as drinking water.

There are several situations where substances may be in food but are not regulated solely by FDA, such as:

  • Drinking water: EPA regulates drinking water provided from public water supplies and sets limits for about 100 chemicals (EPA 2009). FDA regulates bottled water, which may or may not be sourced from public water supplies. FDA does not typically consider exposure from drinking water used outside of the food manufacturing process to be part of the diet. FDA considers, for example, the exposure from drinking water for certain food contact substances that are used as processing aids (for example, a filter for drinking water that is used in the home).
  • Dietary supplements: FDA may consider substances used in dietary supplements in its exposure assessment, to the extent that they are relevant to the assessment. This consideration is contingent upon the agency having information on the use of the substance in dietary supplements. Information on use of dietary supplements is collected by NHANES as part of its food consumption survey.
  • Meat, poultry, and egg products: USDA is responsible for establishing the safe use of these products. FDA collaborates with USDA in evaluating the safety of substances that are added to or used in contact with these food products.
  • Pesticides (including antimicrobials): EPA regulates most pesticides in food by setting tolerances for their use or by exempting them from needing limits. FDA and USDA monitor compliance of pesticide residues in foods. FDA regulates the uses of these same substances in processed foods and in certain food contact applications. Some substances, such as chlorine dioxide or diatomaceous earth, that are regulated by FDA as food additives may also be regulated by EPA for use as pesticides. When conducting a dietary exposure assessment for a substance that is both a food additive and a pesticide, FDA does not consider the EPA-regulated pesticide use.

Side box: What is an “aggregate exposure”?

 EPA's “aggregate exposure” is analogous to, but much broader than, FDA's “cumulative exposure.” EPA defines “aggregate exposure” as the “combined exposure of an individual (or defined population) to a specific agent or stressor via relevant routes, pathways, and sources.” EPA also uses the term “cumulative risk” to describe “the combined risks from aggregate exposures to multiple agents or stressors.” (EPA 2003)

Considering additional uses. 

Consistent with the statutory requirement to consider the cumulative effect of an additive in the diet, FDA expects that assessors consider the following in a cumulative exposure assessment for (FDA 2002, 2006b):

  • Direct food additives and substances that are GRAS: any anticipated increase in consumption from the authorized use(s), if the
  • Proposal involves a new use or an increased level of use covered by an existing authorized use; and/or
  • Substance is also a naturally occurring material.
  • Food contact substances: the cumulative exposure to the substance from the proposed use(s) and already-permitted uses.

Complicating factors. 

Estimating the cumulative dietary exposure to a substance may be complicated by several factors:

  • A food manufacturer may determine that a new chemical or new use of a previously approved substance is GRAS without notifying FDA (that is, self-determined GRAS). This would result in certain self-determined GRAS substances not being declared on the food label (for example, substances used in manufacturing but removed from the final product, FCSs, some direct additives such as flavors, and chemicals that are used in food production that are not expected to be in the final product). Labeling does not provide information about the amount of a substance added to the food. Since FDA and food manufacturers rely on food labeling to identify substances directly added to food, FDA's and food manufacturers’ ability to realistically estimate the cumulative dietary exposure to the substance is limited.
  • Food contact substances, such as packaging materials and certain processing aids, do not appear on the food ingredient label. This limits FDA's and food manufacturers’ ability to calculate cumulative dietary exposure accurately.
  • A previous dietary exposure assessment may not be available, or is outdated. This situation is most likely to occur when the previous use was approved decades before.
  • Information on the actual use of the substance was not available at the time of the premarket assessment, and the new data need to be considered when it becomes available.
  • Some substances that are added to food are also natural components of food, but the information on the amounts present in food is not well known.
  • In considering the exposure to a substance from all dietary sources, there are cases where it is appropriate to assess the dietary exposure to specific components associated with a mixture of substances. FDA will also evaluate components of concern such as impurities, by-products from manufacturing, or components produced by reaction with food or by metabolism.

These limitations generated several discussion questions of whether the current approaches that are used to model cumulative dietary exposure meet the need and intent of the law.

Answers to questions from 1st round of small-group discussion

Question 1: What information should be considered in assessing the cumulative dietary exposure to a substance? How should older estimates (for example, from previously approved uses) be taken into account? Because there are a wide variety of substances added to foods, dietary exposure assessments should be adequate and appropriate to represent the intended use of the substance in the diet. This tailoring to the situation is referred to as “fit-for-purpose.” As a result, the approach used to assess dietary exposure to food ingredients may differ. However, regardless of the approach used, the cumulative dietary exposure should consider whether the substance:

  • Is present naturally in foods or in flavors;
  • Is a naturally occurring element (for example, selenium);
  • Is present in drinking water (for example, minerals and pesticide residues);
  • Is used in dietary supplements;
  • Is used in food contact applications; and
  • Has been previously approved, notified, or otherwise documented to occur in food.

Where FDA has not been notified by a food manufacturer of a GRAS self-determination, the lack of information provided to FDA makes it difficult to estimate exposure. In some instances, the food science literature may reveal information about some self-determined GRAS uses of a substance that could be used in a cumulative dietary exposure assessment.

In a premarket scenario, assessors use conservative assumptions on a case-by-case basis when estimating dietary exposure in order to provide an adequate margin of safety and consumer protection. When reviewing a previously approved substance in the postmarket scenario, assessors usually refine the dietary exposure estimate using postmarket data on the actual use of the substance instead of the premarket conservative assumptions. Information on the actual use of the substance is important to ensure that the EDI reflects realistic consumption trends.

The segment of the population that consumes a food containing the substance, as well as factors affecting the selection of the food, should also be considered. FDA typically performs dietary exposure assessments for the average consumer and the “high” (that is, 90th percentile) consumer. However, certain subpopulations (for example, based on age, gender, or ethnicity) may have different food consumption patterns, and should be taken into account, as appropriate. FDA assesses exposure for subpopulations on a case-by-case basis. Additionally, geographic, seasonal, and economic factors may affect food consumption patterns, and should be taken into account, as appropriate, if such data are available.

Question 2: How can the dietary contribution from uses that have not been reported to FDA (for example, the self-determined GRAS use of a substance) and the natural occurrence of a substance in food be considered? Ideally, a representative estimate of the cumulative dietary exposure to substances added to foods would take into consideration all dietary sources, including naturally occurring sources, GRAS uses, and any uses in food contact substances. However, as noted above, information on the natural occurrence of a substance in food or the self-determined GRAS uses of a substance may not be available, which introduces additional uncertainty into exposure assessments.

In the absence of required reporting for all uses of a substance in food (including self-determined GRAS uses and natural occurrence), other sources must be used to obtain this type of information. Additional sources of information for “nonreported” uses are:

  • Postmarket monitoring data: FDA's TDS is a postmarket tool that currently analyzes for radioactive contaminants in foods, pesticide residues, industrial chemicals, toxic elements, and nutrients. The TDS does not currently monitor substances added to food (for example, food additives and GRAS substances). The TDS could be valuable in providing a more complete picture of substances added to food. If resources were available, the TDS could be expanded to include monitoring of actual levels of substances added to food. The current analytical methods would need to be adapted and validated to measure many substances in complex food matrices. This could be very expensive, and may be impractical for monitoring all substances added to food. There are also databases that contain market information (for example, collected from supermarket scanner data) that are useful to determine if a food contains a particular substance as an ingredient. However, the substance of interest must be declared on the ingredient list of the food for these databases to be of use. Further, use level information for the substance cannot be obtained from these databases, since the use level of the substance is not reported on the ingredient list.
  • Surrogate data: It may be possible to apply the use levels of one substance to approximate the use levels of a similarly functioning substance of interest. If such surrogate information were not available in FDA's files, the relevant use information would need to be requested from industry.
  • Requesting usage information from stakeholders: Presumably, trade associations and food manufacturers maintain actual use level data for substances added to foods to maintain quality control and minimize batch-to-batch variation. FDA could partner with industry to obtain information on a substance's use that could be used in a cumulative exposure assessment. Open sharing of data between stakeholders and agencies would allow assessors to reduce uncertainty and conservative assumptions, thereby providing a more accurate cumulative dietary exposure assessment. However, stakeholders and FDA should work together to ensure that any confidential business information is handled appropriately.

Question 3: What are the limits to the data currently available for estimating dietary exposure, and what additional data from all sources would be useful to improve the estimates? The data that are currently available have the following limitations:

  • Serving sizes: The reference amounts customarily consumed for foods, or serving sizes, are reference values only. An individual's actual intake is captured only through dietary recall surveys, and his or her actual serving size may differ significantly from the labeled serving size. Further, different subpopulations (for example, based on age, gender) may consume different amounts of a particular food, which are not reflected in the serving size. Furthermore, what is considered as a “serving” has changed over time. For example, the size of a cookie may have increased to actually be 2 servings, according to the reference amount customarily consumed; however, a consumer may view one cookie as one “serving,” regardless of the size of the cookie.
  • Standardized recipes: A food reported to NHANES is deconstructed into its individual ingredients using standardized recipes and USDA Standard Reference nutrient data. For example, a “slice of pizza” reported to NHANES is translated to a standardized recipe of flour, oil, water, tomato, and so on, using the WWEIA Food Commodity Intake Database. The associated nutrient data for each ingredient are taken from the USDA Standard Reference. In reality, these standardized recipes may not fully represent the range of recipes used to manufacture a food product. For example, a “slice of pizza” may vary among manufacturers, which may use more or less of a particular ingredient, or a different ingredient, than specified in the standardized recipe. Therefore, standardized recipes are only useful as a representation of the “typical” composition of a particular food.
  • Defining population dietary patterns for the basis of exposure assessment: Even the most comprehensive databases of food consumption patterns for the U.S. population must rely on study designs that restrict observation periods to short periods of time (for example, 2 to 14 d of recall or recording) and practical limitations on participant numbers. This short-term information forms the basis for estimating lifetime consumption patterns (chronic exposure estimation), food choices (menu variety and frequency), intraindividual variation throughout life stages (for example, dietary profile changes for an individual growing from infancy to elderly), and other subpopulation dietary profiles. It is difficult to extrapolate from short-term observations within the scope of the survey designs to these broad exposure-related elements. Thus, consumption may be overestimated and may miss important trends. Keeping these databases current presents yet another challenge.
  • Per capita exposure: In the absence of consumption data and use information for a substance in food, assessors rely on per capita estimates. These estimates, which are based on poundage or “disappearance” data, reflect the average per-person consumption across the entire population. Per capita estimates do not distinguish between individuals who do or do not consume the food, and do not readily account for individuals who may be “high” consumers of a food.
  • Food contact substances: Exposure estimates for a food contact use of a substance rely on different assumptions than exposure estimates for the substance added directly to food. The assumptions used for a FCS are discussed in Topic #4: Food Contact Substance Assessments.

The current food consumption data could be improved by:

  • Incorporating new methods and unconventional data sources to complement and enhance existing food consumption surveys. These methods and sources could include validated biomonitoring data, marketing information, mobile technologies that capture real-time food consumption and portion size, and proprietary information such as customer loyalty cards.
  • Including information that accurately reflects regional food access (for example, rural, maritime, urban areas) and the socioeconomic effects of food access, which could be used as confirmation of food consumption patterns reported in existing databases.
  • Requesting usage data for substances added to food from industry along the supply chain, similar to EPA's Toxic Substances Control Act Chemical Data Reporting Rule (EPA 2011).
  • Using novel approaches (for example, modeling) to make better use of the available data.
  • Accessing global data (for example, food consumption databases, usage of substances) to complement domestic data.
  • Supplementing the existing NHANES food consumption data with additional information sources to provide a more robust model of food consumption by the U.S. population in general, and subpopulations in particular.

Question 4: Under what circumstances is it appropriate to reassess the estimated dietary exposure to a substance? FDA's current procedure for reassessing the dietary exposure to substances added to food can be described as an ad hoc process in response to identified health concerns or the availability of new information regarding the safety or use of a substance. In contrast, EPA is mandated to conduct systematic reviews of all pesticides. Cyclic reviews are likely to be resource intensive, and it may not be practical to automatically mandate periodic reassessments for all food additives. Workshop participants discussed the value of reassessments and identified several scenarios when the dietary exposure of a substance added to food should be reassessed. These include when:

  • The EDI approaches the ADI;
  • The substance is added to sole-source foods, such as infant formula;
  • There is a significant change in food consumption trends (for example, a change in soft drink consumption);
  • A change is made in the method of manufacturing the substance that could introduce new components into the diet (for example, contaminants, manufacturing by-products);
  • New toxicity data or analytical data become available for a previously approved substance (for example, when JECFA identifies a substance as carcinogenic, or improved analytical methods allow for detection of a substance at a lower level);
  • A public health concern (for example, sodium and trans fats) is identified;
  • There is an increase in the public awareness or concern about a substance;
  • A subpopulation is identified as being of concern based on a unique response or high exposure to a substance; or
  • A new use of the substance or an expansion of a previously approved use (for example, increase in the use level) is proposed.

FDA responds to safety issues that arise in the situations above, reassesses the safety data, and takes an appropriate action if dietary exposure should be reassessed. However, if manufacturers voluntarily discontinue the use of a substance, which may occur as a result of innovations in food technology or in the manufacture of FCSs, they are not required to alert the agency. In this case, the estimated exposure to a substance may overestimate the actual exposure from the current use of the substance.

Answers to questions from 2nd round of small-group discussion related to all dietary sources and additional uses

Question 5: How do you define cumulative effects and what is a trigger for cumulative exposure assessment? Participants identified 2 factors that confound the discussion around cumulative dietary exposure assessments:

  • Clarity.: Several concepts in dietary exposure assessment have different definitions depending on the agencies or assessors (for example, “cumulative” has different meanings at FDA, EPA, and FAO/WHO).
  • Implications: There are consequences and implications of the different technical calculations used to estimate exposure (for example, FDA and EPA scientists will derive different dietary exposure values using their respective data and calculations). This is due, in part, to the fact that different exposure assessments may be needed for different purposes in different agencies. Nevertheless, any exposure assessment should be “fit for purpose.”

FAO/WHO distinguishes cumulative and aggregate effects. Cumulative refers to the effects of multiple chemicals, whereas aggregate refers to the effects of multiple exposure sources of the same chemical (WHO 2009). For instance, when toxicology data demonstrate that 2 or more distinct food additives produce a common adverse effect, these substances would require a cumulative exposure assessment. A complete safety assessment would include data demonstrating these cumulative effects, as well as aggregate exposure for substances with multiple modes of exposure (for example, dietary and environmental). The National Research Council (NRC) report “Phthalates and Cumulative Risk Assessment” identified 3 factors to consider in cumulative risk assessments: (1) chemicals that are structurally similar; (2) chemicals that have common mechanisms of action; and (3) chemicals that cause common adverse outcomes (NRC 2008). These 3 factors can contribute to cumulative and aggregate effects of substances.

The group suggested that any of these 3 characteristics are sufficient to trigger cumulative exposure assessments for any substance. For instance, a cumulative exposure assessment would consider 2 or more distinct additives when toxicology data demonstrate that they produce a common adverse effect. Cumulative effects do not draw a distinction between hazard and exposure in safety decisions.

Question 6: How should exposure methods and data sources be better aligned between FDA and EPA? Should FDA follow the Food Quality Protection Act? By 1st answering the question of why methods for estimating exposure and data sources should be better aligned between FDA and EPA, a clearer answer of how these methods can be aligned becomes apparent. Fundamentally, the public expects that exposure assessments include all sources of exposure. Consumers do not assume that exposure to minerals in food is calculated independently of exposure to the same minerals in drinking water or dietary supplements. The number of substances with both dietary (FDA-regulated) and nondietary (regulated by EPA or Consumer Product Safety Commission [CPSC]) exposures is unknown.

Aligning multiple methods for estimating exposure would be resource-intensive. The standard methods for estimating exposure used by the different agencies are considered “fit-for-purpose” for each agency and include the agency's specific standard assumptions (for example, eaters-only or per capita; 90th, 95th, or 99.9th percentile; probabilistic or deterministic methods). Interagency collaboration would require assessors to develop a prearranged and well-defined process to assess exposure from dietary and nondietary sources, including a series of standard assumptions.

As discussed in the overview, the FQPA requires EPA to include additional safety considerations infants and children, such as an extra tenfold safety margin that is generally to be used unless data show it is not necessary. It also mandates a complete review of existing pesticide tolerances in an exposure assessment. The discussion participants did not address whether FDA should follow the FQPA requirements.

Question 7: How important is it that FDA extends exposure assessments beyond dietary exposure? How would FDA accomplish this? Participants generally agreed that it is important for FDA to consider nondietary exposure in an assessment. Further, they suggested that it is probably more important to consider nondietary exposure for FCSs. EPA has set a regulatory precedent for relative source contribution, which FDA could follow, that would include nondietary sources of exposure. In this model, EPA allots specific portions of the total allowable intake of a contaminant to certain uses (for example, drinking water). FDA could use this approach to identify substances for a more thorough assessment; however, the criteria for using such an assessment would need to be developed.

Many substances have food and nonfood uses, which means that there are multiple agencies regulating the uses. The group raised the question of which agency or organization would be responsible to assemble and maintain the use data to permit a cumulative exposure assessment and suggested that further discussion is necessary.

Participant perspectives raised at end of the workshop related to diets

There needs to be a standardized definition of “cumulative” exposure because the public expects that assessors consider exposure from all sources, including dietary supplements and drinking water.

Topic #3: Approaches to Postmarket Assessments

This discussion focused on how postmarket dietary exposure assessments are conducted. The analysis considered when and how dietary exposure should be reassessed, and the criteria for postmarket monitoring.


The methods, assumptions, and data for postmarket assessments can differ from premarket assessments. Premarket dietary exposure assessments rely on conservative assumptions, including that the substance will saturate the market and will be used at the maximum use levels for all types of foods in which it is intended to be used. These assumptions can result in an overestimate of consumer exposure. The postmarket exposure to the substance may be different than the premarket assessment due to several factors. These can include the actual use level of the substance being lower than the estimated maximum concentration or the consumption patterns of the foods containing the substance varying from the initial assumptions. In certain cases, postmarket monitoring may be used to reassess the intake of a substance.

After a substance has entered the marketplace and is consumed by the public, manufacturers must ensure the continued safety of the substance for its intended use. For substances that are the subject of GRAS notifications to FDA, FDA reiterates this responsibility in its correspondence with manufacturers.

Although there is no regulatory requirement that FDA or manufacturers reassess the safety of an approved substance's use after it is introduced in the market, such reassessments may occur. The following are examples of situations that may prompt a postmarket assessment:

  • A manufacturer seeks to expand uses in new types of foods or packaging, or to change current limits on existing uses;
  • There is a change in food consumption trends, with more of the food containing the substance being consumed than was originally considered in the premarket assessment;
  • Through its periodic volume of use surveys, FEMA collects data that indicate whether the use of a flavor ingredient has significantly increased since a previous survey, which can trigger review by the industry-sponsored, independent FEMA Expert Panel; or
  • Toxicological data identify a concern with the substance or with the cumulative effect of substances with similar biological impacts.

FDA conducts postmarket assessments on a case-by-case basis and can use these assessments to obtain a more accurate estimate of the dietary exposure for a substance than is possible in a premarket exposure estimate. Assessors can use various data sources (see Topic #1: Consumption Data Sources, Changes, and Trends) to determine how common the substance is in the marketplace. Assessors can combine this information, when available, with actual use levels to calculate a more realistic exposure estimate.

The primary difference between a premarket and a postmarket exposure assessment is that assessors can obtain a more accurate estimate of actual exposure once a substance is used in products in the market. In 2011, FDA expanded its existing efforts to monitor changes in food consumption through contracts with market research firms (that is, marketing and label databases) with expertise in the food industry. FDA also monitors levels of a wide range of substances (for example, pesticide residues, toxic elements) in foods through its TDS. This study is a survey of about 280 nationally representative foods that consumers might find in their own market baskets, which are collected and analyzed 4 times each year. The TDS measures the actual levels of selected substances in foods as consumed, and thus can generate more refined and precise estimates of dietary exposure to the substance than those based on presumed concentrations or use levels. Although the TDS does not currently monitor levels of food additives and GRAS substances, it might provide a vehicle for postmarket monitoring of food additives and other new or naturally occurring food ingredients.

Routine reassessments occur for some substances’ uses. For example, the FQPA of 1996 requires that EPA reassess its safety decisions every 15 y for all registered pesticides. The reassessment is based on the most current scientific knowledge and data, including exposure information, available at the time of the assessment. The FEMA Expert Panel also regularly reassesses its safety decisions if new toxicology data are available, if a manufacturer seeks to expand uses, or if the estimated consumption of an ingredient substantially increases. On the other hand, FDA does not have the regulatory mandate or the resources that would be required to systematically review dietary exposure assessments for all regulated substances.

Answers to questions from 1st round of small-group discussion

Question 1: Are there cases when dietary exposure should be reassessed? There was a general agreement that dietary exposure of a substance should be reassessed when the data that were the basis for the original exposure estimate do not reflect all current uses and dietary sources of exposure, or when there is a public health concern related to that substance. Similarly, dietary exposure estimates should be updated when the use of a substance is expanded, such as when a substance is initially approved to be used at 100 ppm and the substance is then proposed to be used at 500 ppm. This does not necessarily mean that a new safety evaluation must also be undertaken. New applications and new use levels will not always significantly increase the EDI or result in an EDI that is close to the ADI.

Dietary exposure could be reassessed when:

  • significant new toxicological data suggest a potential health concern;
  • enhanced validated analytical methods (for example, improved technologies to measure substance concentrations in foods) make possible a more refined assessment of estimated daily intake relative to an acceptable level of intake; or
  • health monitoring techniques and data (for example, biomonitoring) are developed that provide further information on potential risks for a substance.

CDC has selected chemicals for biomonitoring in NHANES for a variety of reasons, including adequacy of laboratory tests to measure the chemical or a level of concern from detecting a chemical in a significant portion of the population (CDC 2012). Dietary exposure could also be reassessed when a trend in intake impacts a federal initiative on public health (for example, sodium). An updated exposure assessment would be essential in developing regulatory options to address the public health concern.

Changes in the specifications or the composition of a substance (for example, as a result of a change in the manufacturing process) could also provide a basis for reassessing dietary exposure to ensure that such a change does not cause a health concern. Dietary exposure could be reassessed for substances that are used in foods for which there has been a change in consumption trends over a period of time (that is, the food is consumed in greater quantities than originally presumed). For example, a manufacturer could introduce into the marketplace a product that contains a substance whose safety decision was based on the food consumption data available at the time of the assessment. If the food product exceeds market expectations, and consumption significantly increases, then the original assessment may underestimate dietary exposure for the substance due to increased consumption of the food. Increases in consumption may or may not warrant an updated assessment if the EDI remains below the ADI.

Question 2: If there was a regulatory requirement that FDA conduct postmarket monitoring of food additives, what approaches should FDA use to prioritize its monitoring? The session participants agreed that FDA should prioritize the substances in which a postmarket dietary exposure estimate would be conducted based on various situations that could affect public health. Substances would have a high priority for reassessment when

  • The EDI is close to the ADI;
  • The substance has shown previously unrecognized toxicological effects or effects on specific subpopulations that should be further evaluated;
  • There is a known increase in use that could lead to a public health concern due to increased exposure; and
  • For a given substance, there are other chemicals with similar modes of action, or a health concern about biological effects from the cumulative effect of the substance.

High-throughput screening tests, if properly validated, may provide useful information for identifying potential risks, which could help assessors in the prioritization process. A possible tool for prioritizing substances for reassessment is the application of structure-activity relationships. This process could identify structurally similar substances that are likely to exhibit comparable physical and chemical properties. However, consideration also needs to be given to relevant biological information, including absorption, distribution, metabolism, and excretion data. If exposure estimates from other credible sources (for example, EPA, EFSA, and JECFA) indicate a significantly greater exposure than was originally estimated by FDA, that substance should be given priority for reassessment of dietary exposure. A substance could also receive higher priority for reassessment if a new exposure pathway (for example, inhalation) is determined. Environmental exposure, for instance, is not typically considered in a dietary exposure assessment. However, environmental exposure could increase the magnitude of exposure for some substances to a level that could cause concern.

Question 3: In what situations might food additives be monitored through the TDS? FDA's TDS collects and analyzes a wide range of foods that represent the average American diet; the same 280 foods are collected 4 times a year. The TDS is designed primarily for monitoring levels of substances that are widely distributed among many foods (for example, pesticide residues, nutrient elements). It is less effective for monitoring levels of food additives, which are generally approved for use in specific products; thus, monitoring of food additive levels would require more targeted sampling and more robust sample sizes. While the TDS would not be the best avenue for the postmarket monitoring of food additives, it has the advantage of being an ongoing study with an established infrastructure. Therefore, it could be used as a means for collecting additional targeted samples for food additive monitoring.

An industry survey of food additive usage would probably be more useful and cost-effective than collecting and analyzing samples for food additive levels via the TDS. These data could be gathered through trade associations. Such food additive usage data from industry could be used for estimating postmarket dietary exposure for a substance.

Question 4: When and how should subpopulations (for example, children, elderly, ethnic groups, vegetarians) be considered in a postmarket assessment? Typically, FDA's premarket dietary exposure assessments include estimated dietary exposure for the all-ages (aged 2+ y) group and for children aged 2 to 5 y. Additional subpopulations (for example, teenage boys, elderly) are assessed on a case-by-case-basis. When animal data indicate that a substance has potential developmental or reproductive toxicity concerns, women of childbearing age and pregnant women are a key subpopulation to consider in dietary exposure assessments because their food consumption can directly affect fetal development.

In the postmarket exposure assessment, actual use data for a specific food or substance may be available, as well as new information regarding the susceptibility of specific subpopulations (for example, children, pregnant women, elderly). These specific subpopulations may also be included in a postmarket dietary exposure estimate, as appropriate.

It may be important to consider regional and ethnic subpopulations when conducting an exposure assessment if there is reason to suspect that their consumption patterns would be substantially different from the population as a whole, leading them to be more susceptible to any potential effects from the consumption of a given substance. However, it may not always be possible to obtain consumption information on these subpopulations. NHANES includes participants throughout the United States, but the regional source of the data is not published to maintain the privacy of participants. The ethnic groups reported by NHANES may not be sufficiently specific. In addition, if marketing of products containing a substance is targeted to specific subpopulations (for example, young children, teenagers), the actual exposure to the substance for those subpopulations may be greater than the original exposure estimate. If this increase in exposure would raise a health concern, specific subpopulations could be addressed in a postmarket dietary exposure assessment.

Answers to questions from 2nd round of small-group discussion related to postmarket assessments

Question 5: What steps can be taken to help FDA obtain label, ingredient, and actual food additive (including direct and indirect and GRAS) use and use level data from industry? Manufacturers are not required to report or continually update FDA on which and how much of approved substances are actually used in foods. Additionally, the agency does not have the resources or technical capabilities to monitor all foods for all approved and unapproved substances added to foods. In the absence of actual use data, FDA scientists must rely on standard assumptions to estimate dietary exposure.

Industry typically maintains production data and use levels for substances in finished foods and could provide these data directly to FDA—while potentially protecting confidential business information through the agency's existing authorities or through a 3rd-party aggregator. This process could be complex. There are at least 2 incentives for industry to provide these data to the agency:

  • Prioritization: FDA scientists could generate a prioritization list for substances to be reassessed based on actual use provided by industry. If the EDI is less than the ADI using precise data, then the substance may be a lower priority for reassessment than a substance where the EDI approaches the ADI; and
  • Conservative assumptions: Premarket exposure assessments typically assume that the substance is used in all foods in a category at the maximum allowable levels (for example, all baked goods). When exposure is reassessed postmarket, the exposure calculation is often based on actual use levels in foods, not the conservative maximum-use levels. Finished food manufacturers do not commonly provide actual use levels. FDA can collect information from labels and product testing to include in a revised postmarket assessment. The revised exposure calculation is less conservative when based on actual levels and all uses, but potentially more accurate.

Question 6: What criteria/framework should be used to identify and prioritize substances for reassessment? Methods to reassess exposure have been established (for example, JECFA, FEMA, and EPA) that provide a foundation for designing a new framework for postmarket assessments. The group discussed an approach to design a transparent, 2-tiered system for reassessments. Under this system, reassessments would be prioritized based on risk.

Compounds would be slated for the more rigorous 2nd tier of reassessment under several circumstances:

  • There is a marked increase in the use of the substance;
  • There are significant manufacturing changes that alter the substance's specifications or purity (for example, different contaminants as a result of scaling up); or
  • There are significant new hazard data on the substance (for example, improved quality or longer duration studies, a new public health concern, or identification of a susceptible subpopulation or life stage).

This risk-based prioritization for toxicity reassessments would allow for more efficient postmarket analysis.

Participant perspectives raised at end of the workshop related to postmarket assessments

Throughout the workshop, there was discussion about the triggers and criteria for postmarketing monitoring. At the conclusion of the workshop, participants provided the following perspectives.

  • Compliance assurance is an important factor in postmarket safety.
  • FDA has an incomplete picture of everything added to foods, which makes it difficult to monitor foods and identify potential problems for investigation and postmarket monitoring.
  • Imported foods and a global food supply add to this challenge.
  • There do not seem to be sufficient means to ensure compliance or to trigger postmarket monitoring for substances added to foods.
  • The agency could minimize uncertainty by implementing a standard, priority-based program for postmarket monitoring and exposure reassessments.

Topic #4: Food Contact Substance Assessments

The discussion focused on dietary exposure assessment for food contact substances, and considered whether the current approaches that are used to model dietary exposure from food contact substances meet the need and intent of the law.


A food contact substance (FCS) is any substance that is intended to be a component of materials used in the manufacturing, packing, packaging, transporting, holding, preparing, or consuming of food if the use is not intended to have any technical effect (for example, flavor, stability, consistency, or color) on the food itself. A food contact substance is formerly known as an indirect food additive or, in some cases, a secondary direct additive. Dietary exposure to an FCS occurs because the substance or an impurity migrates into the food. In some uses, antimicrobials are FCSs. FDA and EPA have a complex division of jurisdiction over antimicrobials, including those used on food contact articles. FDA has developed a decision tree that outlines this division of authority. Further collaboration between agencies could clarify exposure assessment scenarios for food contact substances (FDA 2011).

FCSs are a significant part of the workload at FDA's Office of Food Additive Safety, especially in the past decade. More than one-third of the more than 10000 substances currently allowed to be used in or on human food are FCSs. From 2001 to 2010, almost two-thirds of the agency's food additive safety decisions dealt with FCSs (Neltner and others 2011).

A dietary exposure assessment for an FCS generally involves 4 factors:

  1. Type of food contact articles (for example, packaging or equipment) containing the FCS;
  2. Type of food (for example, fatty, aqueous, acidic, or alcoholic) that contacts the articles;
  3. The levels of FCS that migrate from the article to the food; and
  4. The amount of food consumed that contacts the articles.

FDA has developed assessment methods tailored to types of dietary exposure presented by articles such as packaging and equipment that are expected to be used once or many times. The agency classifies single-use food packaging materials into various types (for example, glass, paper, metal cans, and different types of plastic). It has also developed packaging factors that assessors use, in combination with estimated migration rates for an FCS, to calculate an EDI consistent with the proposed use. The agency evaluates articles, such as conveyor belts that are intended for repeated use, and substances that contact these items, like lubricating oils and cleaning chemicals, on a case-by-case basis.

FDA provided its recommendations for the exposure-related information that should be submitted in a food contact notification, a food additive petition, or a Threshold of Regulation exemption request for an FCS in the agency's 2002 Guidance for Industry: Preparation of Premarket Submissions for Food Contact Substances: Chemistry Recommendations and the updated guidance in 2007 (FDA 2002). The guidance describes testing for migration into food, and explains how to calculate the EDI based on the migration studies.

Testing for migration into food. 

FDA expects assessors to provide the agency with enough information to allow its scientists to estimate the dietary concentration of the FCS in foods. FDA guidance allows assessors to:

  • Assume that 100% of the FCS migrates into food;
  • Model the migration using recognized scientific principles of diffusion in polymers; or
  • Conduct migration tests.

FDA has published migration testing guidelines, which include recommendations to simulate the migration of FCSs into different types of food. FDA expects assessors to explain and justify variations from the guidance.

The guidelines describe testing procedures and conditions for migration studies, including duration and temperature that simulate the conditions under which food will be processed and stored in contact with the FCS. These studies estimate the potential migration to each food type, in parts per million (ppm) or milligrams of FCS per kilogram of food (mg/kg), using different food simulants. The studies may be conducted in the actual food that will be used in lieu of using food simulants. These complex mixtures, however, lead to problems in the analysis, and simulants are most often used for ease of analysis. The recommended simulants include:

  • 10% ethanol to represent aqueous, acidic, and low-alcohol (up to 15% alcohol) foods;
  • to 4% acetic acid for substances that may migrate at higher levels into an acidic food in comparison to 10% ethanol;
  • 50% ethanol to represent high-alcohol food with above 15% alcohol (beer and wine generally are covered by 10% ethanol); and
  • A food oil such as corn oil, or a synthetic fat, to represent fatty food. FDA's guidance lists other simulants that may be used as well.

Calculating the EDI based on migration studies and information on uses of the article. 

FDA provides assessors with tools to estimate consumer exposure for a wide variety of foods that FCSs may contact. For common types of single-use packaging, the guidance provides the following factors:

  • Consumption factors describing the fraction of the total daily diet the agency expects to contact specific types of packaging materials (listed in appendix IV, table I of FDA's guidance). For example, the consumption factor is 0.13 for high-density polyethylene (HDPE) polyolefin plastic containers. FDA based this consumption factor on its estimate that 13% of food that an average consumer eats is packaged in HDPE.
  • Food-type distribution factors (fT) reflecting the fraction of the 4 food types (aqueous, acidic, alcoholic, and fatty) expected to contact each packaging material (listed in appendix IV, table II of FDA's guidance (FDA 2002)). For aqueous food (for example, milk) in HDPE, the factor is 0.67 based on FDA's estimate that 67% of HDPE packaging is used to package milk and other aqueous foods.

FDA provides the following equations for assessors to use when estimating an EDI for the FCS:

Equation 1:

Weighted concentration <M> of FCS in the foods it contacts (mg/kg) = sum of (food-type distribution factor × migration level) for each type of food

Equation 2:

EDI = 3 kg food and beverage/person/d × <M> (mg/kg) × consumption factor for the packaging material type (for example, 0.13 for HDPE)

FDA expects that assessors also consider all intended or permitted uses of the FCS to develop a cumulative exposure estimate, or CEDI.

A concrete example based on a substance added to HDPE to make a milk container more flexible helps illustrate how this method works in practice:

  • Assume migration concentration <M> is 0.1 mg/kg based on laboratory migration tests of the FCS in the HDPE plastic using 10% ethanol to simulate an aqueous food.
  • Determine the food-type distribution factor for polyolefins in aqueous food from a table provided by FDA. The factor is 0.67 (that is, 67% of HDPE packaging is used to package aqueous food, which includes milk).
  • Determine the consumption factor for HDPE, 0.13, based on a table provided in FDA's documents (that is, 13% of food an average consumer eats is packaged in HDPE).
  • Calculate the weighted-average concentration of the FCS in the foods it contacts using Eq. 1: <M> = 0.1 mg/kg x 0.67 = 0.067 mg/kg.
  • Calculate the corresponding concentration in the diet (DC) using Eq. 2 and the consumption factor for HDPE: DC = 0.067 mg/kg x 0.13 = 0.0087 mg/kg food.
  • Calculate the EDI using Eq. 3: EDI = 3 kg/person/d x 0.0087 mg/kg = 0.026 mg/person/d. This is equivalent to 26 μg/person/d.

The resulting EDI calculation includes the assumption that the substance will be used in all HDPE containers holding aqueous foods, not just milk. FDA's guidance provides some examples of ways in which assessors may refine the exposure estimates when calculating the EDI:

  • Setting limits on the amount of material manufactured for food contact use, container size distribution, or ratios of weight of food packaged to weight of the packaging.
  • Characterizing the foods that might contact the container or equipment, along with supporting documentation, and the likely fT food-type distribution factor values.
  • Providing information that would demonstrate that only a fraction of a packaging or resin category would be affected by the coverage sought.
  • Identifying technological limitations that could affect the type of food contacted or the fraction of the diet that might be contacted.

Exposure and toxicological studies. 

As noted in the Overview, the EDI has a direct impact on the required toxicological information to support safety of the proposed use. FDA identifies the studies it recommends for FCSs using criteria based on the EDI (FDA 1999) as illustrated in Table 3 below. Generally, the assessor bases the exposure estimate on all known uses of the FCS. However, when the estimate for a new use is below 1.5 μg/person/d, FDA allows the EDI to be based on only the proposed additional use, excluding existing uses.

Table 3. –Recommended studies for food contact substances
ExposureRecommended studies
Less than 1.5 μg/person/dNo safety studies recommended, but available information on carcinogenicity needs to be provided
Between 1.5 and 150 μg/person/dGenetic toxicity tests
Between 150 and 3000 μg/person/dGenetic toxicity tests; subchronic toxicity studies with rodents; and subchronic toxicity studies with nonrodents
Greater than 3000 μg/person/dHandle by a food additive petition

In our example of an FCS in HDPE used for milk with an EDI of 26 μg/person/d, FDA expects the manufacturer to conduct genetic toxicity tests (see Table 3). Additional tests may be required based on the results of those studies. While the guidelines define the minimum amount of toxicology testing recommended for the various levels of exposure, all available toxicology data relevant to the potential dietary exposure of an FCS must be provided to the agency to support premarket clearance (for example, a literature review).

Potential refinements to FDA's guidance for FCSs. 

In October 2011, Mike Adams of FDA's Office of Food Additive Safety spoke at a food packaging conference (see Food Chemical News, October 7, 2011: “FDA chemist says agency's food contact advice is ‘showing its age’” (Clapp 2011)). He identified the following areas where FDA anticipates refining its guidance in the coming years, based on new analytical techniques, new products, and new markets:

  • Enhancing migration modeling by expanding the information on experimentally determined migration rates of new food contact polymers from packaging into foods at various temperatures for a range of molecular weights for the polymer.
  • Determining and updating consumption factors for packaging based on information from several independent market surveys.
  • Including migration tests for packaging adhesives, especially hot melts and rubber adhesives, moving away from the current default assumption of 50 parts per billion migration to food.
  • Assessing the effects of high-pressure processing on migration for pasteurizing foods in flexible pouch packages.
  • Including migration protocols for evaluating migration of polymers from packaging into dry food.

Answers to questions from 1st round of small-group discussion

Question 1: What improvements would you recommend in FDA's approach to estimating cumulative dietary exposure for FCSs? Do the assumptions and calculations recommended in FDA's guidance for FCSs result in a realistically conservative dietary exposure assessment? FDA does not have the unlimited resources that would be necessary to conduct an in-depth analysis for all FCS assessments, whether it is a premarket or postmarket assessment. FDA and stakeholders could work together to develop a comprehensive catalog of the substances, their uses in multiple food contact articles, and sources of nondietary exposure. This work would optimize the use of available resources and provide accurate cumulative exposure estimates. Currently, assessors calculate exposure using conservative assumptions to give a premarket estimate (for example, they tend to overestimate exposure by adding all uses at the highest levels).

Assessors can refine the exposure estimates for an FCS by using probabilistic statistics calculations or by excluding the conservative assumptions when concerns arise during the initial exposure calculation. An example is when the EDI approaches the ADI using the conservative assumptions. FDA considers that there is no reason to refine the estimate if dietary exposure for an FCS is consistent with the criteria in Table 3.

Many FCSs also have nonfood uses. In this situation, the current approach to estimating cumulative dietary exposure would not include exposure from the environment, consumer products, drinking water, and medical devices. While the statute does not require FDA to consider these exposures, the assessment would be more comprehensive if it did.

Postmarket safety assessments are not routinely carried out for FCSs or for substances added directly to foods. Allocating resources to track and monitor postmarket uses would improve their exposure assessments.

Question 2: What are the options to refine dietary exposure assessments for FCSs associated with repeated use equipment, such as conveyor belts? The current predictive migration models are designed to be “worst-case scenarios.” FDA assumes that the migration of the FCS from the equipment's material (for example, a gasket or polymeric tube) is constant throughout the lifetime use of the equipment. FDA is seeking more and better tools to model repeat use articles.

A refined scenario that models actual migration levels of the FCS into foods could be more reflective of actual dietary exposure (for example, the one hundredth cookie onto a conveyor belt might have much lower migration of the FCS than the 1st cookie onto the conveyor belt). Using inert but measurable tracers in food handling equipment, and postmarket measurements of actual levels found in products, would provide opportunities to refine dietary exposure assessments.

Question 3: What options does FDA have when considering a new FCS that has a similar biological effect or presents similar hazards to one already in use? Currently, the burden is on the assessor to confirm that his or her substance has similar biological effects to other substances in food and determine that a cumulative exposure assessment must be done. One limitation to doing so is the lack of toxicological information for many FCSs (for example, when there are no data to group substances in an assessment or to link toxicity between 2 substances). The agency asks assessors to identify cumulative effects, but agency scientists rely on their own information and judgment in the evaluation.

FDA could enhance this process by providing guidance to identify categories of common biological effects that should be considered in exposure assessments, such as decreased body weight or reproductive effects. The agency could prepare a broad list of common biological effects that are considered relevant for grouping substances (such as biomarkers) and for assessing safety, similar to the Redbook list of toxicological effects. This list of effects could be publicly vetted, to encourage transparency and include stakeholder expertise. The groups would need to be specific (for example, more specific than “phenol group”) and demonstrate a link with a specific toxicity outcome.

Assessors would use this standard list when submitting future FCS notifications to identify which categories of biological effects their substance is likely to affect. Providing these data to the agency would allow for a quantitative link to toxicity because cumulative exposure assessments would be based on biological effect.

Question 4: When FDA revises FCS dietary exposure assessment guidelines, what is the most appropriate approach to ensuring that previously authorized FCSs using older methods remain safe? Several potential refinements to FDA's guidance to industry for estimating dietary exposure to FCSs have been suggested (Clapp 2011). Refinements would include improved methods and criteria for assessing the safety of all future FCSs, such as assessing migration into dry food. Based on the group's discussion, previously approved substances could be reassessed under the revised guidelines to ensure their safety. Of course, it may be impractical to reassess all previous FCS decisions using revised guidelines, but all previously approved FCSs must remain safe for use in human foods. There are substances that may need to be reassessed under revised safety guidance.

To reduce the burden on agency and industry scientists, stakeholders could develop a prioritized list of substances to reassess based on factors such as increased exposure based on market surveillance studies or biomonitoring. The list of prioritized substances, which would be publicly available and transparent, could rank previously approved substances based on concerns for subpopulations and highly used FCSs. Additionally, substances could be prioritized for reassessment under the revised guidelines based on data developed using new technologies (for example, Tox21, quantitative structure activity relationship [QSAR] analyses) that are currently being tested and validated. For example, previously approved substances that are identified in Tox21 as structurally/functionally similar to new or high-priority FCSs could be reassessed under the revised guidelines sooner than a very infrequently used FCS.

Answers to questions from 2nd round of small-group discussion related to food contact substance assessments

Question 5: What types of data are needed to improve the migration models used to estimate exposure to food contact substances? Agency scientists have identified fundamental limitations in the current methods for estimating exposure to some FCSs (for example, modeling FCS migration into dry foods, improving the food simulants and temperature ranges to better reflect actual food uses, and developing analytical tools to measure migration from paper).

The exposure models assume that an FCS is used under a defined set of conditions. The models do not account for actual consumer use or misuse (for example, from microwaving food in a nonmicrowave-safe container). Likewise, there are limitations in the models and data for estimating dietary exposure to repeated-use items where manufacturers have less control over how the consumer uses the material in the home. Until there are improved analytical methods, manufacturers should include assumptions to overestimate migration of the FCS into the food.

Data are needed to improve polymer migration models. The standard methods for estimating exposure to FCSs do not account for the chemical differences in larger polymers. These larger polymers are often used at high temperatures and may migrate into foods differently than other smaller FCSs. Data are also needed for modeling of repeated-use articles and paper/paperboard products. The current models tend to overestimate migration of paper/paperboard FCSs at the intended use level because the commonly used assumption is that 100% of the material migrates to the food. The available tools are limited in their ability to get more accurate estimates.

Participant perspectives raised at the end of the workshop related to food contact substances

At the conclusion of the workshop, participants highlighted challenges associated with assessing exposure to FCSs. Many of the exposure assessment methods rely on a series of assumptions, and some participants suggested that assessors need more accurate data for exposure assessments for FCSs, particularly for different kinds of food applications.

Topic #5: Substances with Similar Biological Effects

The discussion focused on substances with similar biological effects and their consideration in dietary exposure assessments. The analysis considered the tools available to identify these substances, and the circumstances under which biological effects can be added together.


The Food Additives Amendment of 1958 states that safety decisions must consider “the cumulative effects of such additive in the diet of man or animals taking into account any chemically or pharmacologically related substances in such diet” (21 U.S.C. §348(c)(5)(B)). FDA incorporated this requirement into its definition of safety (21 CFR §170.3(i)(2)). As a result, manufacturers must consider these effects in their petitions and notifications to the agency.

For food additives, the agency adopted a rule explaining how it would set maximum allowable levels (referred to as “tolerances” in 21 CFR §170.18) for substances that cause similar or related pharmacological effects. The agency stated that it would:

  • Treat related substances as a class.
  • In the absence of evidence to the contrary, consider the substances as having additive toxic effects.
  • Consider setting maximum allowed levels to limit the amount of biological activity (such as cholinesterase inhibition) that may be present, or the total amount of related food additives that may be present in the diet.
  • If 2 or more substances in the same class are present in or on a food, use the lowest numerical allowed level for the substances in the class unless:
  • There are available methods that permit quantitative determination of the amount of each substance present; or
  • A higher use level is reasonably required for the combined substances to accomplish the physical or technical effect for which such combined substances are intended, and the higher use level is safe.

FDA specifically addressed the issue in its proposed rule for notification of substances as GRAS. The agency intended for notifiers to “establish safety, including a consideration of the probable consumption of the substance and the probable consumption of any substance formed in or on food because of its use and the cumulative effect of the substance in the diet, taking into account any chemically or pharmacologically related substances in such diet” (62 Fed Reg 18937, April 17, 1997, Proposed 21 CFR §170.36(c)(4)(i)(A)). While the agency has not finalized the rule, it has used the proposal as the basis of its evaluations since 1997. In the preamble to the GRAS proposed rule, FDA stated:

[L]ikewise, in the case of a chemically synthesized substance that is structurally identical to a naturally occurring substance in commonly consumed food, technical evidence of safety would include consideration of whether the cumulative exposure to both the synthetic and the natural substance exceeds the exposure to the natural substance and whether the combined exposure presents a basis for a safety concern.

The term “similar biological effects” for this workshop session referred to substances added to food that may have related impacts on the human body. Assessors need to consider these impacts in a safety assessment of that substance(s).

FDA has set regulatory limits for related substances in some cases. For example, the agency established maximum use levels for vitamins D2 and D3 based on the toxicity of their common metabolite (21 CFR §184.1950, 172.379, and 172.380). Similarly, the agency evaluated the safety of extracts of the leaves of the stevia plant, for use as a high-intensity sweetener. Stevia plant extracts contain several steviol glycosides that are structurally related, but are chemically distinct. Different combinations of the individual steviol glycosides may be added to food. An exposure assessment of the extracts would be based on that of their common metabolite, steviol, which is present in the extract, as well as formed in the human intestine. The above examples represent substances with additive effects. In addition, some substances may have antagonistic effects in a cumulative exposure assessment. For example, FDA required that food formulated with a fat substitute that limited the body's ability to absorb fat also include fat-soluble vitamins to offset expected reductions in the absorption of these vitamins in the human intestine.

Recognizing situations where substances have similar modes of actions is challenging. The assessor must have a thorough understanding of how the body absorbs and metabolizes a substance, and how other substances in the food may interfere with the metabolism or have an additive or synergistic effect. To assist in this analysis, FDA uses computational tools to identify similar chemical structures and predict potential health effects. This predictive, computational approach is most useful where scientists have a good understanding of how the substance acts in the body based on biochemistry and physiology. However, the computational tools may be limited due to incomplete knowledge, especially regarding biological responses to substances that may impact reproduction, development, behavior, and endocrine systems.

When dealing with substances with similar modes of action, the assessor needs to understand the interactions between related chemicals in food and the amounts that are being consumed in the diet. Note that the National Research Council recently addressed the issue of substances with similar modes of action in cumulative assessments in “Phthalates and Cumulative Risk Assessment. The Task Ahead” (NRC 2008). The committee noted that there is a distinction between a mechanism of action, which it noted “is used to describe the pathway at the molecular level,” and a mode of action, which is “used to describe the key events along the pathway.” The committee found that:

[T]he focus in cumulative risk assessment should be on the health outcomes and not on the pathways that lead to them, whether defined as mechanisms of action or modes of action. Multiple pathways can lead to a common outcome, and a focus on only a specific pathway can lead to too narrow an approach in conducting a cumulative risk assessment.

Answers to questions from 1st round of small-group discussion


The group identified several important themes that were considered during the small-group discussion. Primarily, common definitions and terminology are needed for several factors considered in safety assessments, such as adverse outcomes, biological effects, cumulative effects, and substitutional use. The group also noted that assessors include structurally different chemicals in a cumulative assessment when they are considering a similar mode of action or substances with similar biological effects.

EPA and FDA have different approaches for estimating exposure to pesticides and food additives, respectively. When estimating dietary exposure, FDA does not have a statutory mandate to consider nondietary sources. The group emphasized important roles for hazard characterization and identification, and dietary exposure in risk assessment and risk management. Additionally, there are substances added to food without consumers’ or the agency's knowledge, and dietary exposure assessments are only as robust as the data available to the scientists.

Question 1: What tools are available for an assessor to identify substances with similar biological effects or that can lead to common health outcomes, especially when there are few similarities in the chemical structures? A variety of tools are available for assessors to identify substances with similar biological effects; however, they all have different limitations. Therefore, dietary exposure from substances added to food must often be considered on a case-by-case basis. Assessors may have in-depth knowledge of a substance and its biological effects, and can perform a primary literature search to identify additional chemicals to consider in dietary exposure assessments.

Assessors routinely estimate exposure for each new food additive and its impurities. If FDA reviewed the food additive in the past, then agency scientists also consider previous dietary exposure estimates in the dietary exposure calculation. This method provides an estimate of the dietary exposure that takes into account all uses, previous and new, of the food additive under review. Impurities in a food additive are treated in a similar manner. FDA calls this a cumulative exposure estimate because it accounts for all known uses of the food additive under review.

Assessors can use several tools to identify substances that may have similar toxicological profiles by combining and comparing data collected from:

  • Structural alerts of a substance's potential for an adverse effect (for example, carcinogenicity);
  • Animal studies, when available;
  • Expert opinions; and
  • Published information.

Some of these tools are currently available. Scientists are developing additional tools that will be available for use in future studies. High-throughput screening methods will allow assessors to ask better questions, but should not be the only source of data for a safety assessment. The following computational modeling programs can predict substances that may have similar chemical structure or similar biological effects (for example, carcinogenicity or mutagenicity):

  • EPA's ToxCast program;
  • Predictive models such as QSAR; and
  • High-throughput screening assays, such as Tox21.

Question 2: How should different substances with similar toxicological endpoints be considered in a cumulative dietary exposure assessment? How does an exposure assessment take into account structurally related substances or substances with similar biological effects that have different bioavailability? The group did not directly discuss the 1st question. EPA provides several documents for considering similar toxicological endpoints in a cumulative dietary exposure assessment (EPA 2002a, 2002b, 2012a, 2012b). There are at least 2 unique situations to consider when assessing chemicals with similar biological effects or similar toxicological endpoints: chemicals with the same target use and chemicals with different uses. In both cases, the chemicals could have similar or very different chemical structures and produce similar biological effects.

When multiple companies submit notifications for the same substance and identical uses, then the use is considered substitutional because there is no net increase in estimated dietary exposure. In this case, the submissions are assumed to share the market for the food additive.

When a food additive is regulated or authorized for specific uses and a company intends to expand its uses, the new uses are “added” to the previous uses to obtain a cumulative exposure from all known uses. In a case where different additives are used for the same purpose, such as antioxidants, the exposure estimates are initially calculated individually for each additive. The cumulative exposure estimate is then prepared for the substances if there are known similar biological effects and endpoints. This is done on a case-by-case basis only after review of the toxicity data for each of the additives.

FDA's guidance documents for industry are available on its Web site with a detailed description of how dietary exposure estimates can be developed for both direct and indirect food additives (for example, food contact substances). FDA scientists are available to assist stakeholders in developing exposure estimates.

Bioavailability is an important consideration in estimating dietary exposures. FDA routinely assumes that 100% of the substance is absorbed. Relative potency factors, however, can be developed to account for different absorptions (for example, for multiple substances that are absorbed differently) and bioavailabilities (for example, for multiple substances that are available and active in the body differently). Standard exposure methods and those designed on a case-by-case basis could be made clear and open to the public to increase transparency in decision making.

Question 3: Under what circumstances is it reasonable to assume that biological effects can be added together? Is synergy between substances common enough to be handled by more than a case-by-case basis? Generally, assuming an additive approach to dietary exposure (that is, 1 + 1 = 2), such as when the actions of 2 substances can be summed, is reasonable in most cases where multiple substances have similar modes of action. Synergy, such as when the actions of 2 substances are greater than the sum of the 2 substances alone (that is, 1 + 1 > 2), may not be very common for substances added to foods. It may represent an issue for hazard characterization and identification. As science advances, standard methods for cumulative exposure assessments will have to be enhanced or developed.

Answers to questions from the 2nd round of small-group discussion related to substances with similar biological effects

Question 4: How can exposure assessments prepare for and keep up with the Tox21's implications, analyses, and opportunities? Tox21 is a large-scale collaboration between FDA, EPA, and many branches of NIH designed to characterize common toxicity pathways and identify classes of chemicals that affect these pathways. This collaboration will conduct high-throughput testing on more than 10000 substances—including those added to foods—and generate a publicly available database of toxicity data for these chemicals. Tox21 data will provide opportunities to:

  • Identify substances or groups of chemicals with similar biological effects to be considered in cumulative exposure assessments and safety decisions;
  • Include extensive predictive toxicity and safety data for informed decision making;
  • Identify and prioritize chemicals for additional toxicological testing or postmarket reassessment; and
  • Limit the use of research animals in safety decision making because certain biological effects may be inferred based on known effects of several similar chemicals.

Tox21 will undoubtedly have data gaps, including how to use the data to conduct hazard assessments and to inform exposure assessments. As the program evolves, the database will include information from thousands of known or related chemicals. The toxicity profiles may be useful in future exposure and risk assessments. Tox21 modeling may designate substances as a high priority for reassessment based on structural similarity to a known chemical of concern, safety data gaps, or increased production or use in foods.

Topic #6: Subpopulation Considerations

The discussion focused on how dietary exposure assessments account for subpopulations and their specific dietary habits. The analysis considered options that could enhance FDA's approach to subpopulations, and the limits of NHANES data to estimate consumption by subpopulations.


Eating habits are as diverse as the population and are influenced by many factors, including age, ethnicity, the desire to lose weight, the need to alleviate a medical condition, or simply a decision to stay in line with popular culture. These habits may make people increase their exposure to some additives in food. In addition, some subpopulations such as children or the elderly may be unusually vulnerable to potential effects from the consumption of a substance. According to the 1993 NRC report “Pesticides in the Diets of Infants and Children,” the most widely recognized subpopulations are children and infants because they have a different diet than adults, and their rapidly changing bodies may make them vulnerable to some substances (NRC 1993).

In its 2006 Guidance of the Scientific Committee on a Request from EFSA Related to Uncertainties in Dietary Exposure Assessment, EFSA succinctly identified the challenges of accounting for subpopulations in its analysis:

[T]he European society is increasingly complex from cultural, ethnic and religious standpoints. It is not just a matter of different amounts consumed, but also of the distribution of consumption of particular types of foods. Consequently there may be higher uncertainty in dietary exposure assessments for particular population groups, than for the so-called “general” population (EFSA 2006).

Based on FDA's Guidance for Industry: Estimating Dietary Intake of Substances in Foods and the 1997 Presidential Executive Order #13045 on protection of children from environmental health risks and safety risks, the agency expects assessors to evaluate exposure to the general population of people 2 y of age and older and to children aged 2 to 5 y (Clinton 1997; FDA 2006a). If the substance may be used in infant formula or baby food, the analysis must include infants and children younger than 2 y of age.

If research (for example, animal toxicology tests or human studies) suggests that certain subpopulations, such as women of childbearing age, might be unusually susceptible to exposure to a substance, FDA recommends that assessors develop an EDI for those subpopulations. Assessors may also identify additional subpopulations that might be particularly affected by a substance being evaluated by consulting FDA's Guidance for Industry: Recommendations for Submission of Chemical and Technological Data for Direct Food Additive Petitions (FDA 2006b).

In its dietary exposure assessments, FDA considers food consumption data for the portion of the population or subpopulation that eats the food containing the substance in question (often referred to as “eaters-only”). It relies on the NHANES food consumption surveys, which are conducted on an ongoing basis with the results released in 2-y data sets. Where a subpopulation is small, an assessor can combine the data from multiple survey periods to ensure a representative sample.

Generally, FDA uses a 90th percentile estimate of dietary exposure to represent the high-level consumers in the population who consume foods that may contain the additive (see Figure 1). If a subpopulation is identified as being unusually susceptible to exposure or consists of children 2 to 5 y old, the agency would ensure that the EDI is lower than the ADI for that group, as well as for the general population.

Where a subpopulation consumes unusual amounts of a particular food, as is common for teenage boys, for example, FDA generally does not give them special consideration because they would be represented by the high-level consumers within the general population.

Except for infants and children younger than 5 y, FDA does not consider targeted marketing of foods in developing its exposure estimates because the agency makes assumptions in a premarket exposure assessment. For example, the agency may allow health claims for soluble dietary fiber and reduced risk of coronary heart disease (21 CFR §101.81). People who want to reduce their risk for this disease might actively seek out foods they had not consumed before or consume higher-than-anticipated levels of soluble fiber in response to this health claim.

There are unusual situations where a subpopulation is small but can be seriously impacted by the exposure to food or its ingredients. Some examples include allergies to specific substances, such as gluten, nuts, or casein. Additionally, people with genetic disorders such as phenylketonuria are an example of a subpopulation that could be uniquely sensitive to exposure to the artificial sweetener aspartame, because they are unable to metabolize the substance. In regulating the overall safe consumption of food containing these ingredients, FDA requires information statements to be placed on the labels of food products by the ingredient list to alert the consumers, such as “Phenylketonurics: Contains Phenylalanine.”

  • While both EPA and FDA use NHANES, EPA's approach to pesticide dietary exposure assessment for subpopulations differs from FDA in several aspects. For example, EPA:
  • Requires assessments for all pesticides that will be used on food and routinely assesses exposure for more than 20 subpopulations, including those defined by age (such as infants and children), pregnancy, race, ethnicity, and gender. FDA evaluates fewer groups.
  • Bases its exposure estimate on all members of the subpopulation, including eaters and noneaters of the food. FDA considers only those who eat the food or “eaters-only.”

EPA's approach is based on mandates in the FQPA of 1996 and recommendations in the 1993 National Research Council of the National Academies report on children's exposure to pesticides mentioned earlier. In this report, the NRC committee “considered the development of children from the beginning of the last trimester of pregnancy (26 wk) through 18 y of age, the point when all biological systems have essentially matured.” It concluded that “infants and young children may consume much more of certain foods, especially processed foods, than do adults. And water consumption, both as drinking water and as a food component, is very different between children and adults.” It found that “age-related differences in exposure patterns for 1- to 5-y-old children were most accurately illuminated by using 1-y age groupings of data on children's food consumption.” It recommended that “[b]ecause the composition of a child's diet changes dramatically from birth through childhood and adolescence to maturity, ‘market basket’ food consumption surveys should include adequate samples of food consumption by children at 1-y intervals up to age 5, by children between the ages of 5 and 10 y, and by children between 11 and 18 y. Food consumption surveys should be conducted periodically to ascertain changes in consumption patterns over time.” (NRC 1993)

Answers to questions from 1st round of small-group discussion

Question 1: Are options available to enhance FDA's approach to subpopulations, including either those with unique susceptibility or those who consume large amounts of food, or both? Exposure assessments for subpopulations are done on a case-by-case basis, particularly where the ADI is based on a sensitive group such as children or pregnant women. The ADI is an estimate of “safety,” based upon data on relevant toxicological endpoints, where there is a reasonable certainty that no harm will result from the substance's use. Assessors calculate the ADI based on the results of experiments (for example, laboratory experiments in animals) and safety factors to account for variations between lab animals and humans, and between humans. EPA must apply an additional safety factor when calculating the ADI for pesticides to account for children's special sensitivities, unless there are empirical data to indicate that this additional factor is not necessary (EPA 2002c).

The discussion group considered 5 subpopulations:

  • Children represent the most easily identifiable subpopulation because their eating habits are highly variable and their rapidly changing physiology is unique compared with adults. There is a large gap in publicly available consumption data for children up to 2 y old because their dietary exposure is not routinely assessed. Diet is widely variable for toddlers in this age group because some children are still nursing, bottle-fed, eating solid foods, or a combination of the three. Exposure and safety assessments for children 2 to 5 y old could be conducted in 1 y increments to account for the differences in the physiology and eating habits of toddlers and 5-y-old children. FDA could take an active role in designing dietary exposure studies for children in the National Children's Study to supplement the available NHANES data.
  • Adolescents often have very different food consumption patterns than the general U.S. population, and exposure could be assessed in 3 to 5 y increments for substances in foods that these populations are likely to consume regularly. This would account for puberty, appetites, and periods of rapid growth (for example, teenage boys are often between the 90th and 95th percentiles for carbonated beverage consumption).
  • Pregnant women are uniquely vulnerable because of the developing fetus. In addition, they often have different eating habits because of advice from health care providers and federal agencies, particularly concerning alcohol and fish, or because of cravings or intolerances for certain foods. Previous NHANES surveys oversampled pregnant women so the consumption surveys were more representative; the most recent survey did not oversample pregnant women.
  • People living with chronic diseases (for example, cancer, heart disease, diabetes, and immune diseases) or conditions that predispose a person to disease (such as obesity and poverty) may be uniquely susceptible to the effects of some substances. They may have notably different eating habits, compared with the general U.S. population. These subpopulations are not homogenous, and the representative NHANES data may be limited. Those data could be enhanced by incorporating habits and practices data (for example, eating scenarios, food frequency questionnaires, and portion size information). Similarly, including dietary supplement intake data in dietary exposure assessments would more accurately identify subpopulations, such as people living with chronic diseases, who may regularly ingest very high levels of certain dietary ingredients.
  • Socioeconomic status was considered as a potential subpopulation. Poverty, low-educational status, and aggressive marketing can limit knowledge of or access to healthier foods, which can compound the challenge in assessing dietary exposure for these subpopulations. There was significant disagreement about whether this was a legitimate consideration because it may be stigmatizing.

Question 2: Should FDA take into account the potential marketing of a substance for its health or functional claims in its exposure assessment? If the EDI is based on a broad enough category of food (for example, all baked goods and not just cookies), then the marketing of a substance should not be a problem. However, if the EDI is based on a narrow product category, such as energy drinks, the EDI would be underestimated if the product becomes popular as a result of a marketing campaign. For example, people may drink energy drinks instead of carbonated beverages, and assessors can use the food consumption data for carbonated beverages as a surrogate to estimate dietary exposure from energy drinks. Therefore, it is important for the assessor to define the product category broadly enough to accommodate a growing market share. If narrowly drawn, the exposure will need to be reassessed periodically, which can be difficult if FDA does not notice these marketplace changes or have full information on actual use of these substances.

FDA does not routinely consider food advertising in a premarket assessment, unless the notifier provides sales or marketing goals. The agency may underestimate potential exposure when manufacturers market a substance based on health or function claims—for example, “low fat diets rich in fiber-containing grain products, fruits, and vegetables may reduce the risk of some types of cancer, a disease associated with many factors.” While the popularity of many novel food ingredients may wax and wane, similar to a seasonal or short-term dietary exposure, consumption of ingredients with approved health or functional claims is unlikely to fade. Dietary exposure will then be chronic for these commonly consumed ingredients (for example, calcium, vitamin D, and dietary fiber).

Question 3: What are the limits of using NHANES data to estimate consumption by subpopulations? Are other databases available for subpopulations, such as the National Children's Study, that may be used to more accurately estimate consumption patterns and trends? NHANES collection data are designed to be a robust, nationally representative snapshot of the general U.S. population. However, the food consumption data are limited by memory and bias from the participants or parents who respond on behalf of children. For instance, people may not describe all of the foods and how much they actually ate.

The data are also limited, as identified above, for estimating exposure to the widely varying diets of infants and children up to age 18. Specifically, data are collected for a small sample of children and their daily diet assumed to represent all U.S. children in that age group. For example, if there are 950 children aged 0 to 5 in a survey, then it is assumed that all U.S. children eat a similar diet. The sample sizes would be even smaller if food consumption for children up to 5 y old is analyzed incrementally by year. Using the example above, there may be only 250 4-y-old children in a study; of these 250 children, there may be only 50 or fewer children who ate a specific food. In these studies, the statistical power is decreased as more precise questions are asked. In 1998, NHANES conducted a supplemental survey to oversample and provide more robust data for children; however, there has not been a similar survey since. Additionally, pregnant women are no longer oversampled in NHANES studies, reducing the degree of certainty for exposure to developing fetuses.

There are additional sources of dietary consumption data that are not currently used, but could be combined with NHANES data to give a broader view of dietary exposure assessment, especially for children's food consumption. The following are 2 prime examples:

  • The National Children's Study is a large-scale study that is designed to monitor dietary intake, among many other factors, during childhood. Data collected in this study could supplement the NHANES food consumption data, especially for newborns to 5-y-olds.
  • Data from the Women, Infants, and Children (WIC) Program are included in EPA's dietary exposure assessment of pesticides. Data from USDA's school lunch programs also provide valuable food consumption data that can overcome some of the biases mentioned earlier.

Answers to questions from 2nd round of small-group discussion related to subpopulations

Question 4: What populations should be oversampled by NHANES? What steps can be taken to get better data in NHANES (for example, socioeconomic, ethnicity, regional)? What should be done now when a subpopulation is not represented or is underrepresented? Subpopulations may consume a diet dramatically different from that of the general American population. NHANES is designed to collect data to represent the health and nutrition of the general American population, and the data are used to monitor subpopulations as well. The surveys collect data for infants and young children; however, these data are frequently incomplete because the report may not include all foods consumed (for example, the reporting proxy may not know all foods that were consumed at day care). NHANES methods could be modified to oversample highly variable and highly vulnerable subpopulations and get an adequately representative sample of the food consumption for the subpopulation.

Assessors could consider or design additional surveys to ask questions and collect data specific to a subpopulation's needs. Other data sources should also be included to enhance NHANES data. There are challenges in adapting or designing questionnaires that are tailored to specific subpopulations, such as these:

  • What data should be gathered?
  • Which subpopulations should be asked the questions?
  • Would the data be valuable to collect from the general population as well?

Oversampling populations in food consumption surveys provides a better understanding of the different exposures and potential risks. Typically, each NHANES study oversamples a specific subpopulation in each survey to be more representative of important groups. For example, in the past, it has focused on adolescents, pregnant women, Spanish-speaking people, or low-income Americans.

Participant perspectives raised at the end of the workshop related to subpopulations

At the conclusion of the workshop, participants highlighted some challenges in estimating dietary exposure for subpopulations using the standard dietary exposure models and methods. These include:

  • Difficulty in expressing uncertainty for consumers in the tails of the consumption curve (for example, populations that consume the upper 10th percentile of a common food ingredient).
  • Niche products are an additional source of uncertainty in exposure calculations, because consumption of their ingredients might also be above normal or extreme.
  • The field of exposure science could be improved by improving models and methods to extrapolate in vitro doses and human body burden. Assessors can use these postmarket, biomonitoring data to determine if certain subpopulations are susceptible to overexposure for certain dietary ingredients.

Topic #7: Additional Dietary Exposure Assessment Issues

The section summarizes the questions, issues, and comments not directly related to the 6 topics discussed above.

Answers to questions from 2nd round of small-group discussion related to additional dietary exposure assessment issues

Question 1: How do you harmonize exposure assessment methodologies for compounds that are regulated by both EPA and FDA? or CPSC? All chemicals regulated by CPSC are also under the jurisdiction of EPA. The 3 agencies differ in their regulatory authorities, exposure assessment methods, and levels of caution. There are, however, many substances with uses that already fall under the jurisdiction of multiple government bodies that are used both as pesticides (EPA) and food additives (FDA), or materials that are used in children's toys (CPSC and EPA) and food containers (FDA).

Regular interagency meetings or sabbaticals would increase interagency communication, collaboration, and exchange of scientific information. This opportunity may result in assessors using the most up-to-date methods. For example, for substances that are regulated by both FDA and EPA, agency staff could work together to coordinate, or harmonize, their exposure methods to ensure that public health is adequately protected. These staff meetings could be used to:

  • Forecast overlapping substances that are already evaluated by multiple agencies and do side-by-side comparisons of these substances to determine the feasibility of broadly coordinating or aggregating exposure estimates between agencies;
  • Develop a strategic plan for harmonizing methods to assess exposure from multiple sources; and
  • Identify substances where the exposure from multiple sources might become a health concern.

Participants maintained that it would not be necessary to use the harmonized exposure approach for substances that are exclusively under the jurisdiction of one agency, such as a substance that is exclusively used in food. However, assessors ideally would use a single exposure method for all substances they consider. Additionally, a unified approach could allow for easier reassessments when a substance's use changes.

These interagency meetings would also allow for further discussion between agency scientists to discuss how to implement the recommendations made by the National Research Council (NRC) in its “Science and Decisions: Advancing Risk Assessment” (the “Silver Book”) report (NRC 2009). The NRC directed its recommendations for improving risk assessments to EPA. The report includes recommendations such as developing a unified approach to safety assessments, and using available biomonitoring and surveillance data to address public health and environmental health questions. The report could broadly apply to exposure assessments for substances that multiple agencies regulate. Funding and public support are also key factors to develop a common data platform for cumulative exposure evaluation of substances that are regulated by multiple agencies.

Question 2: Is it necessary that FDA be notified for all self-determined GRAS substances? How can you do a cumulative exposure assessment when there may be self-determined GRAS substances? GRAS self-determination complicates cumulative exposure assessment, as no one assessor may have access to information on all uses of a substance in food. Not all food components are required to be listed on a label, so ingredient label review does not necessarily provide an alternative means of gathering this information. In the absence of labeling, the agency would have to do postmarket surveillance to detect these ingredients in the food supply. The group did not reach a consensus on whether it should be necessary to notify FDA of all GRAS decisions, but they did note that changes to make the GRAS notification system mandatory would require legislative action. Some participants suggested that notification may not be necessary in all cases.

Notifying the agency of all GRAS substances would give assessors greater knowledge of what is in the food supply. This level of transparency would help to ensure accurate and comprehensive exposure assessment, but stakeholders and the agency need to discuss the best method to achieve it. Participants discussed the idea of developing categories for GRAS decisions that may or may not require FDA notification. They did not conclude how the agency could ensure compliance or consider the substance's use in a cumulative exposure assessment when the agency is not aware of all uses of the substance. In the absence of complete use information, the agency cannot conduct cumulative exposure assessments.

Question 3: How do we inspire research in food additives exposure, including direct, indirect, and GRAS (for example, MRM [multiresidue method], monitoring, groundwater, drinking water)? There are 2 predominant challenges to stimulating research in food additives and food safety: funding and awareness. Stakeholders are the primary drivers for inspiring research and for motivating scientists to research food additives by:

  • Funding information and education initiatives to increase awareness of the importance of food additives and food safety research;
  • Encouraging food additives research at national meetings, in academia, and by philanthropic groups, think tanks, and universities; and
  • Developing public–private partnerships with funding agencies to increase research grants and fellowships available to scientists interested in food additive research.

One of the central challenges to inspiring and funding food additives research is the need to strike a balance between ownership, exclusivity, and transparency. Some participants felt that the current notification system could limit progress in food additives research because there is a lack of ownership and exclusivity, which can reduce a company's return on investment. For example, food manufacturers are less likely to invest in innovation and development of new food additives if their data will be available to, and can be relied on by, their competitors.

Topics that workshop participants identified which warrant further discussion

Further discussion topic 1: Defining cumulative exposure. 

Currently, dietary exposure assessors are not required to consider environmental exposure in their calculations, even though many substances can have food and nonfood uses (for example, food containers can be made of the same materials as medical devices or children's toys). Participants did not reach a consensus on how necessary it would be for scientists to routinely consider nondietary sources for substances (for example, inhalation, drinking water, pesticides, and medicine) when assessing dietary exposure. The participants also did not agree on which models and methods would be best to assess cumulative exposure from all dietary and nondietary sources.

Similarly, participants questioned whether lifetime exposure is a relevant concept today, considering that the food market regularly changes and product cycles may be much shorter than the average person's lifetime. The topic was discussed in detail, but many participants felt that lifetime exposure warranted further exploration.

Further discussion topic 2: Increasing collaboration. 

Collaboration was another topic that weaved throughout the workshop. Many organizations and agencies conduct dietary exposure assessments, often relying on the same basic data set for their results. While these groups are often asking different scientific questions, merging their data into one exposure assessment could provide a more realistic reflection of cumulative exposure. At the conclusion of the small-group sessions, some participants discussed whether or not there is a way or an opportunity to build a single, cross-agency platform for overseeing, managing, and linking exposure data.

Participant perspectives raised at the end of the workshop related to additional dietary exposure assessment issues

At the conclusion of the workshop, participants highlighted some additional challenges in estimating dietary exposure using the standard dietary exposure models and methods not addressed in prior sections. These include the following:

  • NHANES food consumption surveys need continued federal support because they are the foundation of dietary exposure assessments in the United States and provide the most current and robust data for food additive safety determinations.
  • FDA has limited resources and authority to implement all of the workshop participants’ suggestions, such as regular reassessment or compliance monitoring. FDA could increase agency resources by collecting user fees similar to EPA's fees to maintain registration of a pesticide and its formulated products. However, this would require legislative changes. Moreover, FDA may need to revise the GRAS notification program to limit its use to the notifier.
  • It is not practical to reassess all substances and uses immediately. FDA may not be able to identify or nominate substances for reassessment, but the agency and stakeholders can work together to develop a prioritized list to rank substances for reassessment based on factors such as proximity of the EDI to the ADI and the availability of toxicological data.
  • Stakeholders can work together to enhance the standard methods used to assess dietary exposure and to incorporate additional methods, such as validating biomonitoring data or developing tools for broader postmarket monitoring. These revised methods would be the foundation to reassess substances and their uses.
  • There are gaps in the available exposure and toxicology data. Stakeholders can identify and prioritize experiments to fill those gaps. Scientists can design experiments so that the doses tested in laboratory experiments match the human body burden, so the toxicology tests reflect actual exposure. Toxicologists can expand testing to include other scientific areas, such as looking at the effects of food ingredients on epigenetics (for example, soy methyl donors) and multigenerational effects.
  • Increasing transparency and stakeholder involvement may enhance dietary exposure assessments. There should be a public forum for stakeholders to review and comment on any changes or decisions allowing new uses or substances added to food or restricting existing uses. Some workshop participants said that the practice of industry GRAS self-determinations without notifying FDA should stop. Industry participants would be more willing to provide more data (for example, actual use level and food ingredients) if data could kept confidential to keep a level playing field.


In November 2011, more than 70 scientists from industry, academia, government, and public interest organizations met for 2 d in Washington, D.C., to discuss FDA's current approach to dietary exposure assessment. While the goal was to examine the principles underlying dietary exposure assessment and identify potential enhancements, organizers were not focused on reaching a consensus. Workshop participants did, however, appear to be in general agreement that FDA needs to develop a science-based framework to prioritize and reassess prior safety decisions.

Small-group discussions reached a general agreement (without objection in plenary sessions) to recommend that the agency take the following actions:

  • Continue to collect longer term (14-d) food consumption data to complement the standard 2-d NHANES food consumption surveys. These longer term surveys may provide more representative estimates of less frequently consumed foods or nutrients.
  • Use biomonitoring results from NHANES to monitor when substances in food are found in participants’ samples. Currently these data indicate that a chemical is present in the body, but they do not distinguish whether the chemical is from food or from nondietary sources such as the environment. The relationship between biomonitoring data and food consumption needs to be clearly understood before such data can be directly correlated to diet.
  • Perform a cumulative dietary exposure assessment, as defined by the FAO/WHO, that considers exposure to substances occurring naturally in food, present in drinking water, consumed as part of dietary supplements, and also found in FCSs or other consumer products.
  • Take into account subpopulations (for example, based on age, gender, or ethnicity) that may have different food consumption patterns, and consider, when available, geographic and seasonal factors that may affect food consumption patterns.
  • Refine its guidance to industry for estimating dietary exposure to FCSs by updating consumption factors and migration models.
  • Consider the 3 factors identified by the NRC report “Phthalates and Cumulative Risk Assessment” as a trigger for a cumulative exposure assessment for substances that: are structurally similar, have a common mode of action, and have common adverse outcomes.
  • Increase communication, collaboration, and exchange of scientific information through regular interagency meetings or sabbaticals with EPA and other agencies. Dietary exposure assessments for food additives rely on continued research to provide the most current information, and this opportunity may result in assessors using the most up-to-date exposure methods.
  • Review recommendations for improving risk assessments made by the NRC in its report “Science and Decisions: Advancing Risk Assessment” (the “Silver Book”). These recommendations were directed toward EPA; however, the report can broadly apply to exposure assessments for substances that are regulated by multiple agencies.
  • Pursue continued funding to increase collaboration, transparency, and accountability between agencies (for example, FDA, CDC, USDA, and EPA).
  • Support continued federal funding for NHANES, which provide robust food consumption data for food additive safety determinations. The data can illustrate changes in the population's food consumption.
  • Work with stakeholders to develop methods to identify gaps in the available exposure and toxicology data. Together, these scientists can prioritize where data needs are most pressing and conduct the experiments to fill those gaps.
  • Develop a science-based framework to prioritize and reassess prior safety decisions.

Participants also generally supported the following observations:

  • NHANES does not have sufficient data to address regional and seasonal differences in food consumption. The surveys are not designed to be representative of states or communities.
  • FDA should consider new technologies developed by scientists to supplement NHANES. These methods could provide more rapid, more accurate, and less expensive exposure information, depending on the purpose.
  • FDA and stakeholders could develop a comprehensive catalog of all substances added to food, their uses in multiple food contact articles, and sources of nondietary exposure. This database would optimize the use of available resources and provide more accurate cumulative exposure estimates. This could include building a single, cross-agency platform for overseeing, managing, and linking exposure data.
  • Allocating resources to track and monitor postmarket uses would improve postmarket safety assessments because assessors could rely on actual use information.
  • High-throughput screening and computational modeling programs are methods that could allow assessors to ask better questions, but should not be the only source of data for a safety assessment.
  • FDA could prepare a publicly vetted list of common biological effects that are considered relevant for grouping substances and assessing safety.
  • FDA does not routinely consider the impact of advertising on food consumption unless the notifier provides sales or marketing goals for the exposure assessment. However, these types of data are especially important for additives that are associated with health or function claims, where consumers may view advertisement as an incentive to consume products containing such additives.
  • FDA could work with the National Children's Study; the Women, Infants, and Children's program; and the USDA's school lunch program to supplement the NHANES food consumption data to give a broader view of dietary exposure assessment for children.
  • Ideally, assessors would use a harmonized approach to assessing dietary exposure for all substances whether they are considered by EPA or FDA.
  • The agency relies on stakeholders to maintain compliance standards because it does not have the resources or authority to monitor adherence to all previous decisions.


The authors thank the more than 70 workshop participants, sponsors, small-group discussion moderators, facilitators, and speakers for their participation. These participants were given the opportunity to review and comment on the workshop proceedings. We especially wish to thank the Office of Food Additive Safety at FDA for its valuable contribution to workshop planning and review of workshop materials. The thoughtful and thorough review of these experts was valuable to producing the most accurate account of the workshop. We also thank Steve Olson, Lani Sinclair, and Katherine Portnoy for editorial assistance in preparing this workshop summary. The authors are grateful to Erik Olson and Shelley Hearne for their support and advice. This work was solely supported by The Pew Charitable Trusts.