By Michael J. Critelli | MakeUsWell Newsletter,
For decades, U.S. food safety assessments have relied on a deceptively simple principle: if the concentration of a food additive is low enough in each product, it must be safe.
This is the logic behind the FDA’s Acceptable Daily Intakes (ADIs) and the industry’s formulation strategies. It assumes that consumers will encounter these ingredients only in modest amounts, spread thinly across a varied diet.
But real life doesn't follow regulatory assumptions. People don’t consume additives in theoretical units—they consume foods, in real households, in real patterns, driven by marketing, convenience, cost, culture, and habit.
California’s Office of Environmental Health Hazard Assessment (OEHHA) recognized this gap and did something paradigm-shifting. Instead of relying on ingredient concentrations, it relied on publicly available real-world consumption data to measure actual exposure. And the results showed that children, the most vulnerable population, were routinely consuming far more synthetic dyes than the FDA’s outdated models ever anticipated.
The lesson is clear: real-world exposure, not theoretical ingredient limits, is the only reliable measure of food safety.
The Tools California Used …and Why Industry and FDA Do Not
California did not invent new science. It used datasets and tools that are all publicly available, widely validated, and methodologically sound. Yet neither the FDA nor the industry uses these tools routinely when evaluating additive safety.
1. USDA’s What We Eat in America (WWEIA) Dietary Intake Data
Publicly available through NHANES, this dataset provides detailed, representative records of everything Americans eat in a 24-hour period, collected using USDA’s gold-standard Automated Multiple Pass Method (AMPM).
This data set tells regulators:
what foods children actually consume,
how often,
in what portions, and
with what demographic patterns?
Despite its central importance, industry does not use NHANES/WWEIA to model real-world exposure, and FDA has not systematically applied it to food dye safety in decades.
2. USDA’s Food and Nutrient Database for Dietary Studies (FNDDS)
This is a granular mapping of foods to ingredients, recipes, and nutrient profiles. California used FNDDS to classify foods by type, preparation, and composition—enabling them to match each food item to possible dye-containing categories.
Neither the FDA nor the industry typically connects ingredient safety to this level of real-world dietary detail.
3. Industry Self-Reported Dye-Use Concentration Data Submitted to FDA
Manufacturers already report the concentrations of food dyes used in different categories of foods. These data are publicly accessible and were integrated into OEHHA’s modelling.
Ironically, industry does not combine its own concentration data with real intake data, and FDA has not routinely merged these datasets to update exposure estimates, even though the tools to do so are readily available.
4. Modern Statistical Exposure Modeling (Used by OEHHA)
California combined consumption data (NHANES/WWEIA), food-code mappings (FNDDS), and dye concentrations (industry submissions) to compute:
average exposure,
90th and 95th percentile exposure, and
exposure by population subgroup.
This kind of modelling is common in environmental health, pesticide regulation, and air pollution analysis, but industry does not apply it to food additives, and the FDA uses older, simplified assumptions rather than current exposure modelling frameworks.
Why Ingredient Concentrations Alone Cannot Tell Us the Truth
Industry prefers to talk about ingredient limits, not exposure, because concentration-based safety assessments allow them to declare a product safe without asking how consumers actually behave.
But consumers, especially children, do not eat in regulatory increments.
A child may consume, in a single day:
a yogurt with dye,
a sports drink with dye,
cereal with dye,
fruit snacks with dye,
a frosted cookie with dye.
Each item, on its own, may be formulated “safely”. But cumulatively, California’s models showed exposure levels far above the FDA’s own average daily intake for multiple dyes.
This is the failure of concentration-based thinking: it treats consumers like theoretical abstractions rather than real people living in real contexts.
Why Real-World Exposure is More Accurate, More Ethical, and More Modern
1. It reflects actual behavior, not idealized assumptions.
NHANES/WWEIA shows what children truly eat, including the concentration of dye-heavy foods in low-income households.
2. It identifies the populations at greatest risk.
Exposure varies dramatically by age, income, race, geography, and food environment, realities ingredient-based models entirely miss.
3. It updates with every new dataset.
Food environments shift rapidly; ingredient concentrations alone cannot capture how consumption patterns evolve.
4. It leverages tools already available but unused.
The USDA dietary datasets, FNDDS mapping systems, and industry-submitted concentration data are sitting in plain sight. What’s been missing is the willingness or accountability to use them.
The Path Forward
In an age when AI can integrate these datasets in minutes, the fact that the FDA and industry still rely on decades-old exposure assumptions is unacceptable and unnecessary. Real-world exposure analysis, using AI, is not only more accurate; it is more aligned with modern public health, modern data science, and modern consumer behavior.
California is showing the country what responsible, evidence-based food safety looks like. The rest of the system should catch up.
Because food safety in 2025 should not be based on what industry hopes people eat but on what they actually do. Our browser-based product will be designed from day 1 to capture and integrate data sets that have not been integrated in the past because the labor required was too costly and time-consuming. With AI, that cost and time becomes manageable.