16.1 Food, Drug & Cosmetic Regulation: FDA Framework, Delaney Clause & GRAS Evaluations

Why FDA’s statute, not a single “safety number,” sits in Domain IV

Handbook IV.12 asks you to work inside named statutes. IV.8 asks you to evaluate risk–benefit and margins of safety (MOS) for agents used to treat disease. IV.7 B asks you to evaluate extractables and leachables from packaging, food-contact materials, medical devices, and primary container-closure systems. Independent OpenExamPrep teaching in this section covers those tasks through the Federal Food, Drug, and Cosmetic Act (FFDCA) as FDA administers it: food-additive approval, the original Delaney Clause, color-additive listing, generally recognized as safe (GRAS) conclusions under the 2016 final rule and October 2016 FAQ, Redbook 2000 for food ingredients, ICH M3(R2) for drug nonclinical timing, extractables/leachables, drug MOS, and FDA’s March 2020 Safety Testing of Drug Metabolites. This chapter is independent OpenExamPrep study material. It is not an FDA, ABT, or ICH product and does not claim official approval, review, or partnership. OpenExamPrep is not FDA-approved.

Domain IV (Applied Toxicology) is 13% of the examination. Items here punish treating a food additive, a color, a GRAS use, a food-contact polymer, and an investigational drug as if they shared one safety standard.

FFDCA architecture: five engines, not one

Section 201(s) defines a food additive as a substance whose intended use may reasonably be expected to become a component of food, unless the use is GRAS, prior-sanctioned (FDA or USDA sanction before 1958), or falls into another listed exclusion (pesticide residues have their own section 408 path after FQPA). A food additive may not be used in interstate commerce until FDA issues a regulation (usually after a food additive petition) establishing safe conditions of use. The general food-additive safety standard is reasonable certainty of no harm under the intended conditions of use (21 CFR 170.3(i)).

That general standard does not finish the job when cancer data exist. The Food Additives Amendment of 1958 added section 409(c)(3)(A) — the Delaney Clause — which, as originally written and as FDA still applies it to food additives, provides that no additive shall be deemed safe if it is found to induce cancer when ingested by man or animal, or if appropriate tests find that it induces cancer in man or animal.

Prior sanction is a historical exclusion, not a modern testing waiver you invent in 2026. If the file is a new use of a new molecule in food, you are in petition, GRAS, color listing, or FCN — not “it looks like salt.”

Delaney as originally written — and where it still bites

Teach the original food-additive rule, then the color cousin, then the pesticide carve-out. Do not flatten them into one slogan.

Food additives. Delaney is a strict legal standard. FDA has stated (for example in the 9 October 2018 final rule on certain synthetic flavoring substances, 83 FR 50490) that once appropriate tests show the additive induces cancer in animals, FDA cannot find the additive safe for that food-additive use regardless of the probability of cancer at the intended dietary level and regardless of mode of action (genotoxic versus secondary). In that 2018 action FDA revoked food-additive listings for several synthetic flavorings (including benzophenone, ethyl acrylate, methyl eugenol, myrcene, pulegone, and pyridine) after animal carcinogenicity findings, even though the Agency’s own exposure analysis concluded the intended uses did not pose a public-health concern. That is the teaching point: risk assessment can say “tiny risk” and Delaney can still compel “unsafe as a matter of law.” Public Citizen v. Young (D.C. Cir. 1987) earlier rejected a de minimis escape hatch for the color-additive Delaney language.

Color additives. The Color Additive Amendments of 1960 require FDA to list a color additive for each use. There is no GRAS exemption for colors. Section 721(b)(5)(B) contains Delaney-style language. The ingested-color prong (721(b)(5)(B)(i)) deems an ingested color unsafe if it is found to induce cancer when ingested by humans or animals. FDA’s January 2025 action on FD&C Red No. 3 in food and ingested drugs is a current illustration: thyroid tumors in male rats triggered the color Delaney provision, with staged compliance dates after the final rule — not a finding that Red No. 3 failed a quantitative human risk cup.

Pesticide residues are the exception you must name. Until 1996, some concentrated pesticide residues in processed food could be treated as food additives and hit Delaney. The Food Quality Protection Act (FQPA) of 1996 established a single dietary standard for pesticide residues in food: reasonable certainty of no harm, including an extra tenfold margin for infants and children when data on them are incomplete, plus aggregate and cumulative assessment. FQPA does not repeal Delaney for intentional food additives or ingested colors. A DABT stem that mixes a pesticide tolerance with a flavoring listing is asking which statute you are in.

GRAS: scientific procedures versus history of use (2016 rule and FAQ)

GRAS is a use, under intended conditions, generally recognized as safe among experts qualified by scientific training and experience. FDA’s 17 August 2016 final rule (81 FR 54960; effective 17 October 2016) and the October 2016 Frequently Asked Questions about GRAS (the handbook study-list document) are the current teaching texts.

Two statutory routes remain:

1. Scientific procedures (21 CFR 170.30(b)). General recognition requires the same quantity and quality of scientific evidence as would be required to obtain approval of a food additive. The evidence must be generally available and generally accepted; it is ordinarily published. Unpublished reports may corroborate but cannot, by themselves, create “general recognition.” Experts must actually recognize the safety of this use, not merely agree that some study exists in a drawer.
2. History of common use in food (21 CFR 170.30(c)). This route is available for experience based on common use in food prior to 1 January 1958, ordinarily from generally available information, and without having to assemble a full FAP-grade toxicology package. An ingredient not in common food use before that date can become GRAS only through scientific procedures. Common use primarily outside the United States can support the history route only if the information establishes safety under the intended U.S. conditions of use.

The 2016 rule replaced the old voluntary GRAS affirmation petition with a voluntary GRAS notice. FDA may respond that it has no questions, that the notice provides an insufficient basis, or that it has ceased to evaluate. Silence from FDA is not a listing. A manufacturer may still market a GRAS use without notifying — the statute does not require a notice — but the safety conclusion remains the firm’s, and FDA can still take enforcement action if the use is not GRAS. Food-contact substances are not run through this voluntary GRAS-notice program; FCN is the usual mandatory path after FDAMA.

Redbook 2000: food-ingredient testing, not an IND

Toxicological Principles for the Safety Assessment of Food Ingredients (Redbook 2000), with the July 2007 updates cited in the ABT study list, is guidance. It tells petitioners how FDA generally evaluates direct food additives, color additives used in food, and, with modifications, food-contact substances. It is not a regulation and not the playbook for an investigational drug.

FDA first places the additive in structure category A, B, or C (low, intermediate, or high toxicologic potential from structure). Estimated cumulative dietary exposure (often as ppb in the total diet) then assigns Concern Level (CL) I, II, or III, which maps to a minimum recommended test battery:

CL I typically recommends genetic toxicity tests and short-term rodent studies. CL II adds subchronic studies in rodent and non-rodent, reproduction/developmental studies, and metabolism/pharmacokinetic data. CL III adds chronic toxicity/carcinogenicity. FDA recommends a genetic toxicity battery for all three concern levels for direct food additives and food-use colors. Novel additives may be asked for a CL III-like package regardless of a low structure score. Metabolites, degradants, and impurities can drive extra work. Petitioners may justify reduced testing; the burden is theirs.

Drug nonclinical timing: ICH M3(R2), not Redbook

ICH M3(R2) (Nonclinical Safety Studies for the Conduct of Human Clinical Trials and Marketing Authorization for Pharmaceuticals) times the package to clinical duration in two species (one non-rodent). Chapter 1 already used M3(R2) to choose an IND playbook; here you need the duration tables as regulatory instruments.

To support clinical trials (M3(R2) Table 1): up to 2 weeks of dosing in humans is generally supported by 2-week rodent and non-rodent studies (the United States allows extended single-dose toxicity as an alternative for single-dose human trials). From 2 weeks to 6 months, animal duration matches the clinical trial. Longer than 6 months of clinical dosing is generally supported by 6-month rodent and 9-month non-rodent studies.

To support marketing (M3(R2) Table 2): indicated human use up to 2 weeks → 1-month studies in both species; >2 weeks to 1 month → 3-month studies; >1 month to 3 months → 6-month studies; >3 months → 6-month rodent and 9-month non-rodent.

Other M3(R2) timing that DABT items mix with food logic: a gene-mutation assay (Ames) generally supports single-dose FIH; the complete genotoxicity battery is expected before Phase II. Safety pharmacology (ICH S7A/S7B) is needed before FIH for small molecules. Embryo–fetal development timing differs by region for women of childbearing potential; do not quote Redbook CL II reproduction studies as the IND gate. Carcinogenicity is a chronic-use marketing question (ICH S1), not a 14-day FIH question.

Extractables and leachables (IV.7 B)

Extractables are organic or inorganic species that can be pulled from a material under exaggerated laboratory conditions (solvents, time, temperature). Leachables are species that actually migrate into the drug product, food, or device-contact matrix under real storage and use. The toxicologist qualifies patient or consumer exposure to leachables, not the longest laundry-list of extractables.

For orally inhaled and nasal drug products (OINDP), the Product Quality Research Institute (PQRI) recommended a safety concern threshold (SCT) of 0.15 µg/day (below which identification of a leachable is generally not required from a safety standpoint) and a qualification threshold (QT) of 5 µg/day for non-mutagenic leachables. USP <1663> and <1664> describe extractables and leachables practices. Parenteral programs often discuss a different SCT discussion (PQRI PDP work has used 1.5 µg/day in some recommendations). ISO 10993-18 (chemical characterization) and 10993-17 (toxicological risk assessment) frame medical devices. Food-contact work uses food simulants, dietary concentration, and the FCN / Threshold of Regulation machinery — not a copied OINDP SCT as if it were 21 CFR. Analytical evaluation thresholds convert an SCT into a concentration the chemist must identify. Cohort-of-concern structures (aflatoxin-like, N-nitroso, alkyl-azoxy) do not get a free pass because they sit under an SCT.

Drug risk–benefit, MOS (IV.8), and metabolites (FDA 2020)

Food law asks whether the use is safe. Drug law asks whether benefits outweigh risks for a labeled indication in a defined population. IV.8 B names drug risk–benefit analysis and MOS calculations. As Chapter 14 already distinguished, pharmaceutical MOS is typically animal exposure at the NOAEL (or HNSTD) divided by clinical exposure, often as AUC; it is not an EFSA genotoxic-carcinogen MOE of 10,000 and not an EPA HQ. A cytotoxic oncology drug can proceed with an MOS near 1 if the clinical benefit is life-extending and the nonclinical findings are monitorable. A chronic non-oncology drug with a 20-fold MOS on a monitorable liver enzyme may still fail if the indication is modest. IV.8 A (side effects and mitigation) is the clinical half of the same idea: boxed warnings, REMS, dose caps, therapeutic-drug monitoring, and antidotes (N-acetylcysteine for acetaminophen, naloxone for opioids) mitigate labeled risk; they do not convert a food additive into a drug.

FDA Safety Testing of Drug Metabolites, Revision 2 (March 2020) tells sponsors when a human metabolite needs its own nonclinical characterization. Disproportionate drug metabolites are those identified only in humans or present at higher plasma levels in humans than in any animal species used in standard toxicology. Human metabolites that can raise a safety concern are those formed at greater than 10 percent of total drug-related exposure at steady state (generally compared as AUC; Cmax may be more appropriate in some cases, as ICH M3(R2) also notes). That 10% of total drug-related material trigger matches ICH M3(R2); it is not the older “10% of parent” shortcut. Phase II conjugates are generally lower concern. If the clinical dose is <10 mg/day, ICH M3 questions-and-answers note that a larger fraction than 10% may be a more appropriate trigger. Unique or reactive metabolites can still matter case by case below 10%. Late discovery delays Phase 3: either synthesize and dose the metabolite in animals or identify a species that forms it at adequate exposure. Characterization is expected to support large-scale (generally Phase 3) trials, not to be invented the week before an NDA.

Scenario

A company wants a new antioxidant in a beverage, a matching ink color, a PET bottle, and a chewable pediatric drug that uses the same antioxidant as an excipient. The antioxidant has a positive rodent bioassay. The color is a new synthetic dye. The bottle leachate includes an unidentified oligomer at 2 µg/day. The drug’s major human circulating species is a metabolite at 18% of total drug-related AUC that the rat does not make.

Four different engines apply. The beverage use is a food additive (or a scientific-procedures GRAS use with FAP-grade, generally available evidence). Delaney can block the food-additive listing despite a tiny calculated risk. The color needs 721 listing, not a GRAS memo. The bottle is food-contact (FCN/migration), not OINDP SCT theater unless you are qualifying an inhalation product. The drug excipient is evaluated in the pharmaceutical risk–benefit/MOS frame (FDA 2005 excipient guidance sits on the ABT study list), and the 18% unique human metabolite is a 2020 MIST / M3(R2) problem for the active, not a Redbook CL II checkbox.

Traps

• Reciting the ABT website sample stem as if it were the only Delaney fact; teach the original 1958 food-additive prohibition, the 1960 color clause, FQPA’s pesticide carve-out, and FDA’s no-discretion-on-risk application.
• Treating GRAS as a substance property or as a lower evidence bar than a food-additive petition for scientific-procedures conclusions.
• Putting colors on a GRAS notice.
• Using Redbook to time an IND, or ICH M3(R2) to size a flavoring petition.
• Qualifying extractables instead of leachables, or pasting 0.15 µg/day onto a food-contact polymer as if it were 21 CFR.
• Using 10% of parent after the 2020 metabolite guidance, or applying Delaney to a drug MOS discussion.
• Implying this OpenExamPrep chapter is FDA-approved.