Learn → How Food Actually Works → Module 11

The additive playbook

Most of the 10,000+ substances allowed in U.S. food are inert. About forty matter — and a regulatory loophole means roughly nobody is checking the rest.

11 min read

The additive playbook

TL;DR. The U.S. food supply contains roughly ten thousand approved substances; another thousand sit in products without the FDA knowing they exist. About forty matter. The framework — GRAS, FDA, EFSA, the seven categories, the concern-level rubric — fits on one page. The watchlist is short: sodium nitrite, a handful of emulsifiers, most coal-tar dyes, the older synthetic preservatives, the artificial sweeteners. Most of what frightens people on a label (xanthan, lecithin, citric acid) is the wrong target. Knowing the difference is the skill.

What you'll learn

  • What "GRAS" actually means, and why ~98.7% of new additives since 2000 were never reviewed by the FDA.
  • How EFSA, JECFA, and FEMA divide the regulatory map — and why the EU bans things the U.S. allows.
  • The seven major additive categories with named examples.
  • The short list of additives GoodEnough scores at concern level 4 or 5, and the mechanisms behind them.
  • Which additives flag a NOVA-4 product but aren't themselves harmful.
  • How GoodEnough's two-axis ingredient scoring works.

The GRAS framework

The 1958 Food Additives Amendment required pre-market testing but carved out a narrow exception: substances scientists already agreed were safe (salt, vinegar) didn't need review. That exception — Generally Recognized As Safe, GRAS — was meant for a few dozen everyday ingredients.

In 1997 the FDA formalized a voluntary notification process. A company could now hire its own experts, declare a substance GRAS, add it to the food supply, and not tell the FDA. Filing was optional. Working with EDF's Tom Neltner and Maricel Maffini, Chris van Tulleken documents the consequence: 756 of 766 new food substances introduced between 2000 and 2010 — about 98.7% — were self-determined GRAS by the company selling them. EDF estimates ~1,000 substances are now in U.S. food without the FDA knowing they exist. Harvard's Food Law and Policy Clinic concluded U.S. food-additive regulation is "essentially voluntary."

Flavorings are worse. Roughly 2,600 are GRAS-certified by FEMA (the Flavor and Extract Manufacturers Association), an industry trade group whose expert panel is paid by member companies. Several substances it has cleared — including isoeugenol, which caused liver tumors in 80% of male mice in NTP testing — would not pass an independent Delaney-rule review.

The other agencies: FDA runs U.S. GRAS, color additives, and contaminants. EFSA runs the EU E-number system (E100s colors, E200s preservatives, E300s antioxidants, E400s thickeners/emulsifiers, E600s flavor enhancers), publishes OpenFoodTox, and sets an Acceptable Daily Intake per substance. JECFA (Joint FAO/WHO Expert Committee) handles international evaluations underpinning Codex.

EFSA operates per substance with a numeric ADI. The FDA operates per intended use — a substance can be safe at 0.01% of a baked good and silently appear at 0.1% of a different product without new review. Combine that with the GRAS loophole and the system produces no data on cumulative effects. Nobody is testing what happens when a person eats fourteen emulsifiers a day for forty years.

The major additive categories

Every additive does one of seven jobs.

Emulsifiers keep oil and water from separating. Lecithin is the original; industry added carboxymethylcellulose (CMC, E466), polysorbate 80 (E433), DATEM (E472e), mono- and diglycerides (E471). They let supermarket ice cream sit in a warm aisle and bread stay soft for three weeks.

Gums and thickeners mimic the texture of fat or dairy. Xanthan (bacterial exudate), guar (bean), locust bean (carob), carrageenan (red seaweed). Modified starches do similar work in dressings, fillings, frozen meals.

Preservatives stop microbes or rancidity. Sodium benzoate (E211) and parabens inhibit yeast and mold; propionates keep bread from molding. Sodium nitrite (E250) cures bacon and lunch meats. BHA, BHT, TBHQ, and propyl gallate are synthetic antioxidant preservatives slowing fat oxidation in cereals, gum, oils.

Antioxidants specifically prevent fat oxidation. Ascorbic acid (E300) and tocopherols (E306–E309) are the benign vitamin members; BHA, BHT, TBHQ, propyl gallate are the synthetic ones. The split inside this category is sharp.

Colorants split into synthetic FD&C dyes (Red No. 3, Red No. 40, Yellow No. 5/tartrazine, Yellow No. 6, Blue No. 1) and "natural" colors (caramel, beet juice, annatto, paprika, carmine, turmeric). "Natural" does heavy lifting — Class IV caramel contains 4-methylimidazole, an animal carcinogen. FD&C dyes are coal-tar derivatives.

Flavor enhancers make food taste more savory than its protein content justifies. MSG (E621) is the headline; hidden glutamates — hydrolyzed vegetable protein, autolyzed yeast extract, soy protein isolate, "natural flavors" — appear under names that don't trigger MSG-avoidance. Disodium inosinate (E631) and disodium guanylate (E627) amplify glutamate; finding them means glutamate is somewhere else on the label.

Artificial sweeteners. Aspartame (E951), sucralose (E955), acesulfame K (E950), saccharin (E954); newer entrants stevioside (E960), neotame, advantame.

Memorize the categories and you can decode 90% of any label. The trick is knowing which entries actually matter.

The ones to pay attention to

GoodEnough scores ingredients 1–5, anchored to Winter, Lustig, van Tulleken, Moss, and primary toxicology. The watchlist — concern 4 and 5 — is short.

Trans fats / partial hydrogenation (historically 5; banned 2018). Drove cardiovascular disease for fifty years. Proof an entire additive class can be wrong. Under-0.5g/serving still rounds to "0g trans fat" on labels listing "partially hydrogenated."

Sodium nitrite (5). Cured meats — bacon, hot dogs, salami — are IARC Group 1 carcinogens, same class as tobacco. Nitrite reacts with amines under heat to form nitrosamines, which cause cancer in lab animals at minute doses. The FDA proposed a phase-out in 1980 and abandoned it for lack of a substitute. Treat cured meats as occasional.

CMC and polysorbate 80 (both 4). Chassaing and Gewirtz 2015 (Nature) showed these two emulsifiers, at concentrations comparable to common dietary exposure, induced low-grade intestinal inflammation, metabolic syndrome, and microbiome shifts in mice. A 2022 human controlled-feeding trial showed CMC altered the human microbiome after eleven days. Mechanism: emulsifiers disrupt lipid interfaces, and the gut mucus barrier is itself lipid-rich.

Most coal-tar dyes (typically 4). Red No. 3 (E127) was banned in U.S. cosmetics in 1990 for thyroid carcinogenicity in rats; it stayed in food another 35 years until the FDA finally banned it in 2025. The 2007 Southampton study tying six common dyes plus sodium benzoate to childhood hyperactivity produced an EU warning-label requirement the FDA has never matched.

BHA and BHT (both 4). Synthetic phenolic antioxidants in cereals, gum, shortenings, animal fats. NTP lists BHA as "reasonably anticipated to be a human carcinogen." Both caused rodent tumors; BHT caused brain chemistry changes in mouse pups. Banned in Japan (BHA) and the U.K. (BHT in food); GRAS in the U.S. TBHQ (4) — petroleum-derived, five grams a lethal dose. Propyl gallate (3) is the milder member.

Most artificial sweeteners (aspartame, sucralose: 4; acesulfame K: 3). Suez 2014 (Nature) showed sucralose, saccharin, and aspartame altered the gut microbiome and induced glucose intolerance in mice, replicated in a small human cohort. Lustig and Means argue sweeteners decouple sweetness from caloric content, breaking the feed-forward insulin response. NutriNet-Santé 2022 found dose-dependent cardiovascular associations with aspartame. IARC reclassified aspartame in 2023 as Group 2B. None a smoking gun; all of it points toward "the diet-soda hypothesis was wrong."

Hidden glutamates (MSG: 4; autolyzed yeast, hydrolyzed proteins: 3–4). Whether MSG itself causes "Chinese restaurant syndrome" is contested. What isn't: free glutamate amplifies palatability and bypasses normal satiety — the umami receptor reads "protein incoming," but in a Pringle or KFC seasoning no real protein follows. Hidden glutamates matter more than MSG because labeling rules let manufacturers print "No MSG" on products full of autolyzed yeast.

High-fructose corn syrup (5). A sweetener, not an additive in the regulatory sense, but worth including. 100% of an oral fructose load is metabolized in the liver, drives de novo lipogenesis, glycates proteins 7x faster than glucose, and is biochemically unnecessary. HFCS is the dose vehicle for that pathway across the U.S. processed-food economy.

That's the watchlist — about a dozen substances and families. Avoid them and you avoid 90% of the additive-related harm in the source material.

The ones to mostly ignore

The scary-but-fine list is much longer. Xanthan gum (3) — bacterial exudate, benign at typical doses. Guar gum (3) — galactomannan from bean endosperm; mostly inert. Lecithin (3) — a phospholipid identical to egg yolk constituents; not in the polysorbate-80 class. Citric acid (2) — fermentation of Aspergillus niger; same molecule as lemons. Ascorbic acid (2) and tocopherols (2) — vitamins. Class I and II caramel color (3) — essentially toasted sugar; the ammoniated versions are the ones with 4-methylimidazole concerns.

The trap: a product with xanthan is almost certainly ultra-processed even though xanthan itself is benign. Its presence signals industrial formulation, meaning other things on the label are doing the work. A single benign additive isn't the issue; the configuration that requires fourteen of them is.

What the regulators actually do

The U.S. and EU diverge on three axes. Burden of proof: EFSA requires safety evidence; the FDA accepts self-certification. Transparency: EFSA publishes OpenFoodTox — every evaluated substance, its ADI, the underlying studies. The FDA's PAFA database is partial, dated, and missing the ~1,000 self-determined GRAS substances entirely. Speed of withdrawal: EFSA reassessed titanium dioxide in 2021 (no longer safe) and aspartame in 2023 (under review). The FDA's response to the 2007 Southampton hyperactivity findings was no change.

The Delaney Amendment — the 1958 rule that no animal carcinogen may be added to food — is why Red No. 3 was finally banned in 2025. It's also why most additives evade scrutiny: showing carcinogenicity requires a study someone has paid for, and nobody has paid for studies on the ~1,000 self-determined GRAS substances.

"Natural" colorants are a separate anomaly. The 1960 Color Additive Amendments grandfathered substances "exempt from certification" (caramel, annatto, beet juice, carmine, paprika) into a less rigorous track. Ammoniated caramel would not pass the FD&C-dye review it was exempted from. The category isn't "natural means safer." It's "we wrote the rules before we knew."

The GoodEnough scoring approach

GoodEnough scores every substance on two axes.

Concern level (1–5). A composite of toxicology, multi-jurisdiction regulatory status, and the mechanistic evidence in the anchor books. Level 1: no concern at any reasonable intake. Level 2: approved everywhere, no plausible mechanism (citric acid, ascorbic acid). Level 3: approved, mostly inert, but a marker of industrial processing or with edge-case concerns (xanthan, lecithin, caramel color). Level 4: significant evidence of harm or banned in multiple jurisdictions (BHA, BHT, polysorbate 80, CMC, MSG, aspartame, most coal-tar dyes). Level 5: demonstrated carcinogen, demonstrated metabolic harm, or banned in the U.S. (sodium nitrite, HFCS, trans fats).

NOVA-4 marker (boolean). Some additives are only used in industrial formulations. Xanthan has concern level 3 but a true NOVA-4 marker — its presence means the product was formulated, not cooked.

Concern level answers should I avoid this substance? The NOVA-4 marker answers what kind of food is this? A product can score low on concern (lecithin, xanthan, citric acid) and high on NOVA-4 marker count — the operational definition of ultra-processed. A different product can have one high-concern ingredient (nitrite in butcher's bacon) and zero NOVA-4 markers. Different decisions. Every score in the canonical database links to the underlying claim.

Frequently asked questions

Is aspartame OK?

Probably not. IARC reclassified it as Group 2B (possibly carcinogenic) in 2023; the 2022 NutriNet-Santé cohort found dose-dependent cardiovascular associations; Suez 2014 showed microbiome disruption. None is a smoking gun; the pattern points toward harm. Concern level 4. Drink water.

Is xanthan gum bad?

Mostly no. Concern level 3. Its presence is a NOVA-4 marker — the product is industrial — but the gum itself, at typical doses, has no consistent evidence of harm. Avoid the product because of what xanthan signals, not the xanthan.

Is "no artificial colors" a meaningful claim?

Sometimes. It means no FD&C dyes. It does not exclude caramel color (4-methylimidazole), annatto, carmine, or titanium dioxide (banned in EU food, allowed in U.S.). More meaningful than "natural"; less than "no added color of any kind."

What's the deal with E-numbers?

European regulatory identifiers. E300 is ascorbic acid, E621 is MSG, E471 is mono- and diglycerides. The system makes EU labels parsable because every substance ties to a public database. U.S. labels use full names, which scan worse. An E-number isn't a danger signal — vitamin C has one.

Why does the U.S. allow what the EU bans?

Burden of proof (EFSA requires safety evidence; the FDA accepts self-certification), speed of response (EFSA is less captured by industry funding), and political will (EU member states label and restrict). The FDA's food side has been chronically under-resourced relative to its drug side since the 1980s.

Are FDA approvals trustworthy?

Drug approvals, mostly. Food approvals, no — most since 2000 are self-determined GRAS declarations the FDA never reviewed. The agency does meaningful work on color additives, food contact, and contaminants. It does little on the bulk of new additives because the law no longer requires it to.

Should I avoid all emulsifiers?

No. Lecithin in moderation is fine; mono- and diglycerides at typical doses are likely fine; egg yolks and milk are emulsifiers. The two the evidence calls out are polysorbate 80 and CMC. The useful heuristic isn't "avoid emulsifiers" — it's "avoid products that require emulsifiers to exist."

Sources

  1. Winter, R. A Consumer's Dictionary of Food Additives, 7th ed. Three Rivers Press, 2009. Entry-by-entry toxicology and regulatory status for ~12,000 substances.
  2. van Tulleken, C. Ultra-Processed People. W. W. Norton, 2023. The GRAS loophole, the FEMA arrangement, the Maffini-Neltner evidence, the emulsifier-microbiome literature.
  3. Lustig, R. H. Metabolical. Harper Wave, 2021. The eight subcellular pathologies; the case for removing fructose from GRAS.
  4. Moss, M. Salt Sugar Fat: How the Food Giants Hooked Us. Random House, 2013. How industry deploys these substances at the formulation, marketing, and policy levels.
  5. Chassaing, B., et al. (2015). "Dietary emulsifiers impact the mouse gut microbiota promoting colitis and metabolic syndrome." Nature, 519, 92–96.
  6. Chassaing, B., et al. (2022). "Randomized controlled-feeding study of dietary emulsifier carboxymethylcellulose reveals detrimental impacts on the gut microbiota and metabolome." Gastroenterology, 162(3), 743–756.
  7. Suez, J., et al. (2014). "Artificial sweeteners induce glucose intolerance by altering the gut microbiota." Nature, 514, 181–186.
  8. IARC Monograph 114 (2015). Red and Processed Meat — Group 1 carcinogen classification.
  9. Debras, C., et al. (2022). "Artificial sweeteners and cancer risk: NutriNet-Santé cohort." PLOS Medicine, 19(3): e1003950.
  10. EFSA. OpenFoodTox database.
  11. Neltner, T. & Maffini, M. (2014). "Generally Recognized as Secret: Chemicals Added to Food in the United States." Environmental Defense Fund.

Related modules

  • ← C8: The ultra-processed food problem
  • C12: Reading labels (the working method)
  • C13: How food gets engineered →

Related glossary terms

Related ingredients