Contaminants: permitted, detected, and harmful are three different questions

In 2024, USDA's Pesticide Data Program tested 9,872 food samples and found more than 99% in compliance with EPA's legal residue limits, while 42.3% had no detectable pesticide residue of any kind — up from 38.8% in 2023 and 27.6% in 2022.^[2] Those two facts, read together, describe a food supply where legal violations are rare and instrument-detectable residue is declining as a share of samples even as the instruments themselves get more sensitive every year. They also describe a food supply that regularly makes headlines implying the opposite, because most coverage of contaminant testing collapses three separate questions into one: whether a chemical is legally permitted, whether it was actually detected, and whether the amount detected is harmful. Every other topic page on this site — romaine, beef, chicken, protein powder, chocolate protein powder — leans on this page to make that distinction once, carefully, with primary sources, rather than re-litigating it for every commodity. This page covers eight contaminant classes relevant to a US grocery basket: pesticide residues, heavy metals (lead, cadmium, arsenic), PFAS, glyphosate, microplastics, nitrates/nitrites, mycotoxins, and the packaging chemicals acrylamide, BPA, and phthalates. It treats the EWG Dirty Dozen and the IARC processed-meat classification — the two most-cited and most-misreported data points in this space — as case studies in exactly the collapse this page exists to prevent.

History

US pesticide regulation began as a labeling statute, not a safety statute. The 1947 Federal Insecticide, Fungicide, and Rodenticide Act (FIFRA) required registration with USDA and truthful labeling, but did not require proof of safety before sale.^[5][6] In 1958, Congress passed the Food Additives Amendment, giving FDA authority to require a safety assessment before a new food additive could be used — and attached the Delaney Clause, which categorically barred FDA from approving any food additive "found to induce cancer in man, or, after tests, found to induce cancer in animals," with no allowance for dose.^[7][8] Because pesticide residues in processed food were treated as food additives, Delaney's zero-risk standard collided with improving lab sensitivity: as instruments got better, more approved pesticides could be shown, at some concentration, to cause tumors in high-dose animal studies, making a 1958 statute increasingly unworkable against 1990s analytical chemistry.

Rachel Carson's Silent Spring (1962) built a public case — organized around bird die-offs and bioaccumulation up the food chain — against unrestricted DDT use. It became a bestseller, prompted Congressional hearings, and helped catalyze a 1963 President's Science Advisory Committee report affirming her central claims.^[9][10] The book is broadly credited as a catalyst, not the sole cause, of the regulatory buildout that followed: President Nixon created the EPA by executive reorganization in 1970, moving pesticide authority to it from USDA, Interior, and FDA.^[11] In 1972, Congress substantially rewrote FIFRA to require registrants prove a pesticide would not cause "unreasonable adverse effects," and EPA administrator William Ruckelshaus announced a ban on most DDT uses effective December 31, 1972, citing persistence, bioaccumulation, and wildlife harm rather than acute human toxicity.^[12]

The Delaney Clause created a two-decade absurdity: raw produce was regulated under a risk-benefit standard, while the same pesticide residue, if concentrated during processing (tomato paste made from that same tomato), was technically bound by Delaney's zero-cancer-risk rule — a pesticide could be legal on the raw crop and illegal in the processed product made from it. A 1987 National Academy of Sciences report, Regulating Pesticides in Food: The Delaney Paradox, documented the inconsistency and pushed Congress toward reform.^[13] The Food Quality Protection Act of 1996 (FQPA) resolved it by eliminating Delaney's application to pesticide residues and replacing both standards with a single risk-assessment test — while tightening protection for children by requiring EPA to consider aggregate and cumulative exposure and apply an additional tenfold children's safety factor unless data justify a smaller one.^[4][14] FQPA is the direct ancestor of the modern regime: it is why a 2020s EPA tolerance decision comes with an aggregate, cumulative, age-stratified dietary risk assessment rather than a single-chemical calculation.

Supply chain: where each contaminant class actually enters

This page's "supply chain" is really eight different entry points, since each contaminant class gets into food a different way and at a different stage.

• Pesticide residues enter at the field (foliar spray, seed treatment, soil-applied herbicide) and, for some crops, again post-harvest (fungicide dips and waxes on citrus and apples, fumigants on stored grain). Systemic pesticides move into plant tissue and cannot be washed off; contact pesticides sit on the surface and are reduced, though rarely eliminated, by washing.
• Heavy metals (lead, cadmium, arsenic) mostly enter through the soil a crop was grown in. Cadmium in cacao is the clearest example: cacao trees are unusually efficient at taking up cadmium from volcanic, tropical soils and translocating it into the bean as it grows — a geological, agronomic problem, not a manufacturing one.^[17] Lead, by contrast, is largely a post-harvest contamination problem for cacao: levels are low in freshly picked beans and rise during days-long sun-drying, when airborne lead-containing dust adheres to the wet, sticky beans.^[18] This cadmium-is-agronomic / lead-is-handling distinction is the single fact downstream pages on cacao and chocolate protein powder most need from this page. Arsenic in rice follows a related soil-uptake logic: paddy flooding mobilizes arsenic already present in many soils, including legacy arsenic from pesticides banned decades ago.
• PFAS enter primarily via biosolids (treated sewage sludge) applied to farmland as fertilizer, and via PFAS-contaminated groundwater used for irrigation or livestock drinking water; a secondary route is food-contact packaging, historically including some fast-food wrappers used as grease-proofing.
• Glyphosate enters via in-season herbicide application to glyphosate-tolerant corn, soy, and cotton, and via pre-harvest desiccation spraying on non-genetically-modified small grains — wheat, oats, barley, lentils, dry beans — to synchronize harvest.^[33]
• Microplastics enter via multiple, poorly quantified routes: breakdown of plastic packaging and food-contact surfaces, agricultural plastic mulch degradation, atmospheric deposition, and aquatic organisms filtering contaminated water.
• Nitrates/nitrites enter deliberately, as a curing agent (synthetic sodium nitrite, or nitrate-rich vegetable powders such as celery powder converted to nitrite via bacterial culture) added during meat processing — and separately, as naturally occurring nitrate taken up from soil by leafy greens and root vegetables.
• Mycotoxins enter in the field or in post-harvest storage, produced by molds (chiefly Aspergillus and Fusarium) colonizing grain and nut crops under drought stress, insect damage, or improper drying and storage moisture.
• Acrylamide is not present in raw food; it forms during high-heat cooking via the Maillard reaction between asparagine and reducing sugars — it enters at the cooking step, not the farm.
• BPA and phthalates enter via food-contact packaging: can linings (BPA, historically) and plastic tubing, gaskets, and flexible plastics (phthalates) that migrate into fatty or acidic foods during processing or storage.

Studies & nuance

This is the heart of the page. Every contaminant discussed below gets evaluated against the same three-word framework, because collapsing it is the single most common error in food-safety journalism.

The framework: permitted, detected, harmful

An EPA tolerance under 40 CFR Part 180 is a maximum legally allowable pesticide-chemical residue on a specific commodity. EPA states that tolerances are "limits on the amount of pesticides that may remain in or on foods marketed in the USA" and that it "must make a safety finding that the pesticide can be used with 'reasonable certainty of no harm'" before establishing one.^[1] But a tolerance is set relative to good agricultural practice, not a toxicological cliff-edge: it comes from field-trial data on the highest residue that turns up when a farmer follows the label, then EPA runs a separate dietary exposure assessment to confirm that even eating the tolerance level every day for a lifetime would sit safely under the reference dose. Exceeding a tolerance makes a food legally "adulterated" and seizable — first and foremost a violation of an agricultural-practice standard, not automatically a poisoning event.^[1]

"Detected" is an empirical category, not a legal or toxicological one. PDP prepares samples the way a consumer would — washed under running water for 15–20 seconds, peeled first for produce with inedible skins — and analyzes them down to parts per billion.^[2] Because instrument sensitivity keeps improving, "percent of samples with detectable residue" trends upward over time even when farming practice does not change; labs can simply see lower concentrations than they used to. "Harmful" requires a second, separate calculation: toxicologists derive a Reference Dose from animal or human dose-response studies by finding the highest No Observed Adverse Effect Level and dividing by an uncertainty factor — typically 100 (10x animal-to-human, 10x human variability), and since FQPA in 1996, often an additional 10x for children unless data show it's unnecessary, some effective margins landing 1,000x below the No Observed Adverse Effect Level.^[3][4] EPA then multiplies a measured or tolerance-level residue by how much of that food people actually eat, by age and sex cohort, to get an estimated daily intake — the number compared to the reference dose. A detection at 1/1,000th of tolerance is a fact; whether it is a hazard depends entirely on where that exposure estimate lands relative to the reference dose, and PDP/FDA data repeatedly show that even samples at the tolerance ceiling tend to produce exposures far below it. Popular coverage skips this calculation almost every time: it reports "detected" as "harmful," and sometimes "over tolerance" as "harmful," when neither claim does that logical work alone.

Pesticide residues: what PDP and FDA actually find

The 2024 PDP Annual Summary tested 9,872 samples across 19 fresh and processed produce, nut, and fish commodities and found more than 99% in compliance with the applicable EPA tolerance; 76 samples (0.77%) exceeded a tolerance — 12 domestic, 63 imported — a pattern (imports exceeding more often than domestic) generally attributed to different approved-use lists abroad rather than more hazardous farming.^[2] 42.3% of 2024 samples had no detectable residue at all, continuing a multi-year upward trend (38.8% in 2023, 27.6% in 2022) usually read as broad compliance rather than declining contamination.^[2] An independent 30-year retrospective of PDP data found the same pattern: residues, where present, overwhelmingly well under both tolerance and reference-dose thresholds.^[15] FDA's complementary Total Diet Study measures foods as consumed — cooked and prepared, not washed-raw — valuable because it captures what home cooking does to residue levels that PDP's protocol does not.^[16]

Heavy metals: lead, cadmium, arsenic, and why cacao concentrates them

This mechanism matters well beyond this page: it is the reason a chocolate or chocolate-protein-powder product has to treat heavy metals as close to inescapable rather than a quality-control failure by one brand. A multi-disciplinary expert panel jointly convened by As You Sow (an advocacy/shareholder-engagement organization) and the National Confectioners Association (the US chocolate industry trade group) concluded in 2022 that cadmium in chocolate comes overwhelmingly from cadmium naturally present in the volcanic, tropical soils where cacao is grown — concentrated in Latin America — and is deposited in the nibs as the tree grows, not introduced during roasting, grinding, or conching.^[17] Because this is a soil-chemistry problem, it cannot be fixed by better manufacturing hygiene; the fixes under active research are agronomic (soil amendments, cadmium-excluding rootstock, blending beans from lower-cadmium regions). Lead is a different story: Consumer Reports' investigative testing, drawing on the same As You Sow-funded research program, found lead levels are low in freshly picked cacao beans and rise substantially during days-long sun-drying, when airborne lead-containing dust and soil particles adhere to the wet, sticky beans — drying practice, not the tree's biology, is the controllable variable.^[18] Cadmium is agronomic and essentially unavoidable at current soil-chemistry knowledge; lead is a handling practice and at least partly correctable. That distinction is the single most important fact to carry into any downstream page about cacao or chocolate products.

Consumer Reports' December 2022 test of 28 dark chocolate bars found detectable cadmium and lead in all 28; 23 of 28 exceeded a public-health-based comparison level for at least one metal at a one-ounce daily serving, and five exceeded for both metals simultaneously.^[19] The comparison level Consumer Reports used is drawn from California's Prop 65 framework, discussed below, not directly from FDA. The National Confectioners Association, as an industry source, said lead reductions were achievable within about a year of new post-harvest handling practices, while cadmium reduction — a soil-uptake problem — would take substantially longer, consistent with the mechanism above.^[19]

Two different thresholds do two different jobs here, and neither is a bright safety line by itself. FDA's Interim Reference Levels (IRLs) for lead, last updated 2022, are 2.2 micrograms/day for children and 8.8 micrograms/day for women of childbearing age, derived by taking CDC's blood lead reference value and back-calculating, via a biokinetic model, the daily dietary intake predicted to raise blood lead by that amount, then applying roughly a 10-fold safety factor.^[20] For cadmium, FDA uses a toxicological reference value range of 0.21–0.36 micrograms per kilogram of body weight per day.^[21] FDA frames IRLs as risk-management screening tools that trigger further investigation, not legal limits — part of the "Closer to Zero" action plan FDA launched in April 2021 to sequentially lower dietary exposure to lead, arsenic, cadmium, and mercury in foods commonly eaten by babies and young children.^[22] California's Prop 65 Maximum Allowable Dose Level (MADL) is a different instrument: the daily exposure level below which a business is exempt from Prop 65's warning-label requirement, calculated with its own (typically 1,000x) margin of safety below a no-observed-effect level for reproductive/developmental toxicity — lead's MADL is 0.5 micrograms/day, cadmium's is 4.1 micrograms/day.^[23] Because the MADL is a warning-label trigger under California's own framework, and the FDA IRL is a national dietary-screening tool under a different model and endpoint, the two numbers are not interchangeable: a product can be over the Prop 65 MADL while well under FDA's IRL, or vice versa, a distinction press coverage of chocolate testing routinely blurs. Many ordinary Prop 65 warnings on grocery items reflect this low, precautionary MADL threshold rather than a level of immediate acute toxicological significance — the warning is as much a compliance artifact of California's legal regime as a health signal.

Arsenic gets its own detailed treatment on this site's carb-staples page (rice); briefly here, inorganic arsenic (the more toxic form, versus the organic arsenic species common in seafood, considered far less hazardous) accumulates in rice because paddy flooding mobilizes arsenic already present in many soils, including legacy arsenic from pesticides banned decades ago. FDA's action level for inorganic arsenic in infant rice cereal is 100 parts per billion.^[24]

PFAS: real, but concentrated in specific, traceable pathways

PFAS (per- and polyfluoroalkyl substances) are a large family of synthetic chemicals valued for grease-, water-, and stain-resistance and extreme chemical stability — the same stability that makes them accumulate in the environment and living tissue. The food-relevant pathway most tied to agriculture is biosolids: treated sewage sludge, land-applied as fertilizer, can carry PFAS from wastewater into farmland soil, taken up by crops or consumed by livestock. Maine is the only state to have systematically tested farmland for this and, after finding contamination on more than 80 farms — including one dairy operation forced to close in 2019 — banned land application of biosolids outright in 2022; Connecticut followed with its own ban in 2024, while Michigan, New York, and Wisconsin have interim limits short of a ban.^[25][26] In Texas, farmers affected by PFAS-contaminated biosolids sued EPA in 2024 over its failure to regulate PFAS in sludge; a federal court dismissed the case in 2025 and the plaintiffs have appealed.^[27] EPA has not finalized a national biosolids PFAS limit as of July 2026.^[28]

On drinking water — a related but distinct pathway from food — EPA finalized its first national PFAS rule in April 2024 (4.0 parts per trillion each for PFOA and PFOS, 10 ppt each for three other compounds), then proposed partially unwinding it in May 2026 by extending compliance deadlines and rescinding some of the individual limits — a live rollback that should be treated as unsettled.^[29][30] On direct food testing, FDA reports finding no detectable PFAS in 95% of general fresh/processed foods tested, but detected PFAS in 74% (60 of 81) of tested seafood samples in a 2022 targeted survey, leading FDA to treat seafood as a meaningfully higher-risk category than the general food supply; a first round of infant-formula PFAS testing under the "Operation Stork Speed" initiative was released in April 2026, though specific quantitative results were not located in this research.^[31][32] The net picture: PFAS in the general food supply is detected infrequently outside seafood and biosolids-affected farmland; where it is a genuine problem, it is farm-specific and traceable to a known pathway rather than diffuse.

Glyphosate: a real disagreement, stated plainly

Glyphosate is the active ingredient in Roundup and the most heavily used herbicide in US agriculture, applied both in-season to glyphosate-tolerant corn, soy, and cotton, and sprayed pre-harvest on wheat, oats, barley, lentils, and dry beans as a desiccant to synchronize harvest — specifically why glyphosate shows up more consistently in oat products than in sweet corn: oats are frequently desiccated shortly before harvest, applying the herbicide directly onto the grain that will be milled.^[33]

In March 2015, IARC classified glyphosate as Group 2A, "probably carcinogenic to humans," based on limited human evidence of non-Hodgkin lymphoma, sufficient evidence in animals, and strong mechanistic evidence of genotoxicity.^[34] IARC's system is hazard-based: it asks whether an agent is capable of causing cancer under some set of conditions, independent of real-world exposure — it groups glyphosate in the same hazard category as working a night shift or eating red meat, which surprises readers who assume the categories rank magnitude of danger. They do not. In February 2020, EPA's interim registration review concluded glyphosate is "not likely to be carcinogenic to humans" at real-world dietary and occupational exposures.^[35] EPA's framework is risk-based: it combines the hazard question with actual exposure data. The Ninth Circuit subsequently found EPA's 2020 finding procedurally and substantively flawed and ordered a revisit.^[36] The EU, after its own multi-year review, renewed glyphosate's approval for ten years through December 15, 2033.^[37] This is a genuine, still-live disagreement between IARC's hazard classification and EPA/EFSA's risk conclusions, and both are real regulatory/scientific positions rather than one being "the science" and the other "corporate capture" — the disagreement traces to different methodological questions (can it cause cancer under any condition, versus does realistic exposure produce meaningful risk), not to different data.

EPA's chronic Reference Dose for glyphosate is 1.75 mg/kg body weight/day, and the EPA tolerance for glyphosate residue on oats is 30 parts per million.^[38] Advocacy-group testing by EWG has repeatedly found glyphosate in oat-based cereal at concentrations in the hundreds to low thousands of parts per billion — up to roughly 2,837 ppb (2.8 ppm) in one 2018 sample — well under the 30 ppm tolerance, though EWG's own health-based benchmark for children (160 ppb) is calculated independently of, and far more conservatively than, EPA's.^[38] Run the exposure math and even 2.8 ppm in modest daily portions is a small fraction of the 1.75 mg/kg/day reference dose for anyone above infant body weight — the detected levels do not approach EPA's reference dose even though they drove substantial news coverage. This is not a dismissal of EWG's underlying data, which appears real; the disagreement is about which reference point — EPA's, or EWG's independently set, more conservative one — is the right one to judge a detection against, a policy/values question this page cannot resolve.

Litigation is a separate track: Bayer, which acquired Monsanto in 2018, has faced more than 60,000 Roundup lawsuits alleging failure to warn of a non-Hodgkin lymphoma risk, with juries returning large verdicts.^[39] In June 2026, the US Supreme Court ruled 7–2 that federal pesticide law preempts state-law failure-to-warn claims against Bayer, expected to curtail future verdicts though not overturning finalized judgments.^[40] Litigation outcomes are a legal and jury-persuasion question, not evidence about the underlying toxicology.

Microplastics: real, detected, and genuinely unresolved

This is the least mature contaminant on this page: the literature is young and largely descriptive rather than causal. What is reasonably well established: microplastics (typically under 5mm) have been detected in a very wide range of foods and in human tissue samples including blood, stool, urine, placenta, and liver, and dietary ingestion appears to be the dominant human exposure route, ahead of inhalation.^[41] What is not established: a validated dose-response relationship, standardized measurement methods that let different studies' numbers be meaningfully compared, or direct causal evidence in humans linking dietary exposure to any specific disease outcome. In vitro and animal studies have associated exposure with oxidative stress and inflammation at the cellular level, but translating that into a human health-risk estimate is not currently possible with available data. FDA's own position is that current evidence does not demonstrate that the levels detected in foods pose a risk to human health — a statement about the current state of evidence, not a permanent finding.^[42] The honest summary: this is real, it is detected, and science does not yet know what, if anything, it means for health at current dietary exposure levels.

Nitrates/nitrites, "uncured" bacon, and the IARC Group 1 processed-meat classification

Sodium nitrite is FDA/USDA-approved as a curing agent primarily because it inhibits Clostridium botulinum spore germination — a food-safety intervention, not a coloring gimmick, though it also produces the characteristic pink color and tangy flavor of cured products. USDA regulations recognize only synthetic sodium nitrite and potassium nitrite as "curing" agents in the formal sense; when a manufacturer instead uses celery powder, fermented by a bacterial culture that converts its naturally high nitrate content into nitrite, the product delivers essentially the same curing chemistry but must be labeled "Uncured" and "No Nitrates or Nitrites Added," because celery powder is not on the approved synthetic-curing list.^[43] In other words: "uncured" bacon is, functionally, cured — with a vegetable-derived nitrite source instead of a synthetic one — and the "no nitrates or nitrites added" claim is a regulatory technicality (no synthetic nitrite was added) rather than a chemical reality, since nitrite is present, sometimes at levels comparable to or exceeding conventionally cured products because celery-powder curing is harder to dose precisely. A 2019 petition from the Center for Science in the Public Interest pushed FSIS to require more accurate labeling; as of mid-2026 a final rule had not been confirmed as issued, and the "uncured" terminology remains in practice.^[44]

In October 2015, IARC classified processed meat — meat transformed through salting, curing, fermentation, smoking, or similar processes: bacon, hot dogs, ham, sausages, deli meat — as Group 1, "carcinogenic to humans," based on what its working group judged to be sufficient evidence in humans specifically for colorectal cancer. Red meat (unprocessed beef, pork, lamb, goat) was separately classified Group 2A, "probably carcinogenic," based on limited human evidence plus strong mechanistic evidence.^[45] This is the single most misreported fact in food journalism, and it is worth stating the correction as plainly as IARC itself has: the Group classification describes the strength and consistency of the evidence that an association exists — IARC's confidence that the exposure can cause cancer at all — not the size of the risk, and not a ranking against other Group 1 agents. Processed meat, tobacco smoking, and asbestos are all Group 1 because the evidence linking each to cancer meets IARC's bar for "sufficient in humans," not because they carry comparable risk; a pack-a-day smoking habit and a daily bacon habit are not remotely comparable in absolute cancer risk, even though both sit in Group 1.^[45]

That absolute increase is for someone eating an extra 50 grams of processed meat every single day, for life, relative to someone who eats none. It is a real, non-zero, biologically plausible effect — nitrite curing does produce N-nitroso compounds implicated in colorectal carcinogenesis, and heme iron in red/processed meat is separately implicated in oxidative damage to the colon lining — but it is a small absolute effect building on a modest baseline risk, not remotely comparable in scale to the absolute risks associated with heavy tobacco use. Reporting "processed meat is a Group 1 carcinogen, same category as cigarettes" without this absolute-risk context is technically true and functionally misleading.

Naturally occurring nitrate is a separate, mostly reassuring story: leafy greens (spinach, lettuce, celery, beets) take up nitrate from soil and fertilizer and can contain far more nitrate, gram for gram, than a cured-meat serving delivers as nitrite — yet vegetable nitrate is not associated with the same cancer signal, and some research suggests dietary nitrate from vegetables may even be cardioprotective via nitric-oxide-mediated blood vessel effects. The leading explanatory hypothesis in the nutrition-science literature centers on the co-occurrence, in cured meat, of nitrite with amines and heme iron under conditions (high heat, gastric acidity) that favor N-nitroso compound formation — a chemical environment vegetables don't replicate; this mechanistic detail is commonly stated in secondary nutrition-science sources but was not independently verified against a specific peer-reviewed study in this research.

Mycotoxins: potent, but caught upstream by testing

Mycotoxins are toxic secondary metabolites produced by molds that colonize crops in the field or in storage; the family most relevant to the US supply includes aflatoxins (chiefly on corn, peanuts, and tree nuts, especially under drought stress), deoxynivalenol and other Fusarium toxins (chiefly on wheat and small grains in wet years), and ochratoxin A. FDA treats aflatoxin as a "poisonous or deleterious substance" under its action-level authority rather than a pesticide-style tolerance: total aflatoxins above 20 parts per billion in peanuts and peanut products render the food adulterated, with separate, higher feed action levels — up to 300 ppb for finishing beef cattle, reflecting a judgment that mature cattle destined for slaughter can tolerate higher levels than dairy cows or young animals.^[48] Aflatoxin is a Group 1 IARC carcinogen (liver cancer) with a well-established, dose-dependent basis — a useful contrast against the processed-meat Group 1 discussion below, since both share the label but rest on very different strength and character of evidence. Grain and nut buyers in the US test incoming lots and reject or divert aflatoxin-exceeding lots before they reach retail, which is the primary reason acute aflatoxin poisoning is rare in the US despite the toxin's potency.

Acrylamide, BPA, and phthalates: packaging and processing, not farming

Acrylamide does not occur in raw food; it forms during high-heat cooking (frying, baking, roasting above roughly 120°C/248°F) via the Maillard reaction between asparagine and reducing sugars, most associated with fried/baked potato products, roasted coffee, and some baked goods and cereals. FDA issued final, non-binding industry guidance in 2016 on reducing acrylamide formation.^[49] Prop 65 lists acrylamide as a carcinogen, which produced years of litigation over whether coffee required a cancer warning; that question was substantially resolved when IARC downgraded coffee from Group 2B ("possibly carcinogenic") to Group 3 ("not classifiable") in 2016, and California exempted acrylamide from ordinary coffee roasting from Prop 65 warnings — an unusually clean example of a scare that the evidence, on reassessment, did not support.^[50]

BPA and phthalates are food-contact-material issues rather than agricultural-input ones. FDA's BPA posture has moved narrowly: it granted industry petitions to remove BPA-based materials from its approved list for baby bottles and infant formula packaging specifically because manufacturers had already voluntarily abandoned those uses — "the market moved first, the rule caught up" — while its ongoing safety review continues to state available evidence supports BPA's safety in remaining approved uses like can linings.^[51] On phthalates, FDA finalized a 2022 rule revoking authorization for 23 abandoned phthalates, narrowing the approved list to nine; it reaffirmed this in October 2024, denied a consumer-advocacy petition to restrict the remaining phthalates further, and in May 2026 proposed a future cumulative risk assessment for four of them rather than acting individually.^[52] Advocacy pressure continues — Senators Markey and Booker urged FDA in 2024 to ban remaining food-contact phthalates outright — but FDA's position as of mid-2026 is to keep evaluating rather than move to a full ban.^[53]

The EWG Dirty Dozen, treated fairly

EWG's "Shopper's Guide to Pesticides in Produce," known by its top-12 "Dirty Dozen" and bottom-15 "Clean Fifteen" sub-lists, is built directly from the same USDA PDP data discussed throughout this page — a real, citable, derivative artifact of a legitimate government dataset, not an independent testing program and not a fabrication.^[54] The methodological critique, most rigorously stated in a peer-reviewed 2011 paper by Carl Winter and Josh Katz of UC Davis, is that EWG's ranking weights indicators like the percentage of samples with any detectable residue and the average number of pesticides found, but historically only one of six indicators incorporated residue concentration at all, and none related the resulting exposure to an actual toxicological reference point. Winter and Katz ran consumer exposure estimates for the ten most frequently detected pesticides on each Dirty Dozen commodity using the same PDP data and concluded that consumer exposures "are at negligible levels" and that EWG's methodology "is insufficient to allow any meaningful rankings among commodities" on health-risk grounds.^[55] In plain terms: a commodity can rank high on the Dirty Dozen because it frequently shows some detectable pesticide, even at trace levels a small fraction of tolerance, without that reflecting relative health risk at all — dose is exactly the variable the ranking leaves out, precisely the collapse this page's framework exists to prevent.

EWG's own methodology notes are explicit that the Dirty Dozen does not perform a full quantitative risk assessment, does not weight pesticides by toxicity, and does not factor in EPA legal limits by design; its stated position is that the list measures pesticide presence and frequency, not risk, and exists to help shoppers who want to minimize exposure do so, not to identify unsafe foods. EWG spokespeople have also argued "legal does not necessarily mean safe," since EPA tolerances can be outdated relative to newer toxicology — a genuinely contestable but not unreasonable regulatory-critique position, separate from the narrower Winter/Katz critique.^[56] EWG revised its methodology in 2025 to add a toxicity-weighting component, a direct response to years of that critique, though whether it fully resolves the dose-blindness problem was not independently assessed here. EWG also states, in every year's Shopper's Guide, that the health benefits of eating fruits and vegetables — organic or conventional — outweigh pesticide-exposure concerns, a caveat that is easy to lose in headline coverage.^[56] The Alliance for Food and Farming, an industry-funded produce-marketing group, has separately criticized the Dirty Dozen for potentially discouraging produce consumption among price-sensitive shoppers who might otherwise skip the pricier organic alternative — a distinct critique that should itself be read with awareness that the Alliance's funding comes substantially from the conventional produce industry.^[57]

The fair summary: the Dirty Dozen is real, is derived from real government monitoring data, and is a legitimate tool for a shopper who has already decided, for whatever reason, that they want to minimize pesticide exposure regardless of whether it is toxicologically significant. It is not a food-safety ranking, has been criticized in the peer-reviewed literature for exactly the dose-blindness this page's framework is built to flag, and EWG itself does not claim it is a risk assessment. Both things are true at once.

Preparation, safety, and storage

Washing reduces but does not eliminate residues, and effectiveness varies enormously by pesticide chemistry and produce surface. PDP's own sample-prep protocol — rinsing under running water for 15–20 seconds, or peeling for produce with inedible skins — is itself evidence that "detected" already means "after a home-style wash," since that is literally the prep step before PDP's lab analysis.^[2] Beyond that baseline wash, controlled studies of additional washing methods show a wide range of effectiveness depending on the specific pesticide's chemistry: contact/surface pesticides wash off more readily than systemic pesticides the plant has absorbed into its tissue, which washing cannot remove at all. One study using a 2% cornstarch soak followed by a 5% baking-soda soak removed roughly 94% (without a surfactant) to 92% (with one) of the specific pesticide thiabendazole from tested fruit — a strong result, but for one specific chemical under lab conditions, not a general claim about all pesticides.^[58] Separately, more skeptical recent research, widely reported in 2024, concluded plain washing "cannot wholly remove pesticides," especially systemic ones.^[59] Peeling is the more consistently effective intervention for produce with an edible-but-removable skin, such as apples: pesticide residue is concentrated in the outer peel and immediately underlying pulp layer, and removing that layer eliminates nearly all residue for many tested pesticides, at the cost of the fiber and micronutrients concentrated in the peel itself.^[60] The net practical guidance the literature supports: washing under running water measurably helps, is worth doing, and is what PDP's baseline data already reflects; it is not a guarantee of zero residue, especially for systemic pesticides; peeling, where the produce and the recipe tolerate it, does more, at the cost of nutrients and yield concentrated in the skin.

Cooking and curing affect nitrite/nitrate and acrylamide chemistry directly. High-heat cooking (grilling, frying, roasting) of cured, nitrite-containing meat increases formation of N-nitroso compounds relative to lower-heat cooking, part of the mechanistic story behind the IARC processed-meat classification discussed above — though there is no clean primary-sourced number here for how much cooking method changes the risk estimate, and that is flagged as a gap. Acrylamide is a cooking-created contaminant entirely: raw potatoes and raw coffee beans do not contain it; frying and roasting create it, and FDA's industry guidance on reducing it (lower fry temperatures, potato varieties with lower reducing-sugar content, shorter roast times) is a processing-level intervention, though a home cook frying potatoes to a lighter golden color rather than deep brown similarly reduces acrylamide formation, following the same Maillard-reaction logic.

Storage matters most for mycotoxins. Because Aspergillus and Fusarium mold growth, and the resulting mycotoxin production, is strongly favored by moisture and warm storage temperatures, proper drying before storage and dry, cool storage conditions are the primary controllable variable for mycotoxin risk in home-stored grains, flour, and nuts — a food-safety point distinct from, and arguably more actionable for, a home cook than most of the other contaminants on this page, most of which are fixed by the time the product reaches the shelf.

Cost

• USDA's PDP program itself is publicly funded through USDA AMS's budget, and its data is free to the public via ams.usda.gov — there is no direct cost to a consumer for the underlying monitoring this whole page relies on.
• Organic certification, often marketed as a lower-pesticide-residue choice (a claim PDP data broadly supports for many though not all pesticide classes — organic produce still shows some synthetic-pesticide detections via drift and legacy soil contamination, and organic-approved pesticides are not tested by PDP with the same rigor as conventional-only compounds), commands a well-documented price premium at retail; this project's cost-per-gram-of-protein and other topic pages carry store-specific organic-vs-conventional pricing, and this page defers to those for current numbers rather than duplicating them.
• Third-party heavy-metals testing (Consumer Reports' 28-bar chocolate test, for instance) is a cost borne by the testing organization, not the consumer, and its results are published free; a consumer who wants to independently verify a specific product's heavy-metal content would need to pay for private lab testing, which commercial labs typically price in the $50–150-per-sample range for a basic heavy-metals panel — a figure drawn from general knowledge of commercial food-testing lab pricing, not confirmed against a specific lab's current rate card in this research, and flagged as such.
• Prop 65 compliance (reformulation, warning labels, or litigation defense) is a real cost borne by food manufacturers and, ultimately, passed through to shelf price to some degree, though this research did not locate a specific, sourceable estimate of that pass-through cost and does not want to guess at a number.

Further reading

• USDA AMS, Pesticide Data Program datasets and annual summaries (ams.usda.gov/datasets/pdp) — the primary source underlying essentially every "detected" claim on this page and on the produce-specific pages that follow it.
• FDA, Total Diet Study program pages (fda.gov/food/total-diet-study) — the complementary "as consumed" monitoring program, useful for readers who want cooked/prepared-food exposure data rather than PDP's washed-raw protocol.
• EPA, "About Pesticide Tolerances" and "Setting Tolerances for Pesticide Residues in Foods" (epa.gov/pesticide-tolerances) — the primary explanation of what a tolerance is and how it's derived, worth reading directly rather than through secondary paraphrase.
• Winter, C.K. & Katz, J.M. (2011), "Dietary Exposure to Pesticide Residues from Commodities Alleged to Contain the Highest Contamination Levels," Journal of Toxicology — the peer-reviewed rebuttal of Dirty Dozen methodology; short, readable, and a good model of how to run an actual exposure calculation from PDP data.
• As You Sow / National Confectioners Association joint expert panel report on lead and cadmium in chocolate (2022) — the primary technical source for the cacao heavy-metals mechanism section above; useful because it was jointly commissioned by an advocacy organization and the industry it was investigating, an unusually collaborative model for this kind of research.
• FDA, "Closer to Zero: Reducing Childhood Exposure to Contaminants from Foods" (fda.gov) — FDA's own framing of its heavy-metals-in-children's-food regulatory program, useful for tracking a slow-moving regulatory process as it develops past this page's July 2026 snapshot.
• IARC/WHO, "Q&A on the carcinogenicity of the consumption of red meat and processed meat" (iarc.who.int) — IARC's own plain-language explanation of the Group classification system, the best primary-source corrective to the "same category as cigarettes" misreading.

Sources

1. EPA. About Pesticide Tolerances and Setting Tolerances for Pesticide Residues in Foods. epa.gov/pesticide-tolerances. ↩
2. USDA Agricultural Marketing Service (2026). USDA Publishes 2024 Pesticide Data Program Annual Summary, press release, ams.usda.gov, Jan. 2026. ↩
3. EPA. Reference Dose (RfD): Description and Use in Health Risk Assessments. IRIS. ↩
4. PMC7011289. Application of the Food Quality Protection Act children's health safety factor to the dietary risk assessment of pesticides. ↩
5. EPA. Summary of the Federal Insecticide, Fungicide, and Rodenticide Act. ↩
6. Congressional Research Service. Pesticide Law: A Summary of the Statutes, RL31921. ↩
7. Institute for Integrative Toxicology, Michigan State University. Food Additives: Delaney Clause. ↩
8. NCBI Bookshelf. Legislative History of the Pesticide Residues Amendment of 1954 and the Delaney Clause of the Food Additives Amendment of 1958. ↩
9. Environment & Society Portal. The US Federal Government Responds. ↩
10. Silent Spring Institute. Pesticides then and now. ↩
11. Center for Food Safety. History of Pesticide Regulation. ↩
12. EPA Administrator William Ruckelshaus's 1972 DDT cancellation order, cited via the Environment & Society Portal and multiple encyclopedic summaries. Commonly stated; the order's exact Federal Register text was not independently re-verified in the underlying research. ↩
13. National Academy of Sciences (1987). Regulating Pesticides in Food: The Delaney Paradox. Cited via secondary summaries. ↩
14. Encyclopedia.com. Food Quality Protection Act of 1996. ↩
15. PMC10541321. Pesticide data program: 30 years of food residue data and trends. ↩
16. FDA. FDA Total Diet Study (TDS): Design and Implementation; Taylor & Francis chapter summary. ↩
17. As You Sow and National Confectioners Association joint expert panel report on lead and cadmium in cacao, Aug. 2022, cited via natlawreview.com summary. (As You Sow is an advocacy/shareholder-engagement organization; the National Confectioners Association is the US chocolate industry trade group — this was a jointly commissioned panel.) ↩
18. (third-party testing — Consumer Reports) How Lead and Cadmium Get Into Dark Chocolate. consumerreports.org. ↩
19. (third-party testing — Consumer Reports) Lead and Cadmium in Dark Chocolate, Dec. 2022. consumerreports.org. Includes a cited response from (industry source — National Confectioners Association). ↩
20. FDA. Lead in Food and Foodwares; ScienceDirect, Updated interim reference levels for dietary lead to support FDA's Closer to Zero action plan, 2022. ↩
21. FDA. Cadmium in Food and Foodwares. ↩
22. FDA. Closer to Zero: Reducing Childhood Exposure to Contaminants from Foods. fda.gov. ↩
23. OEHHA. Proposition 65 Maximum Allowable Dose Level (MADL) fact sheets and tables — lead and cadmium. ↩
24. Multiple secondary sources citing FDA's infant rice cereal inorganic-arsenic action level (100 ppb). Commonly stated; the primary FDA guidance document title was not independently re-verified in the underlying research. ↩
25. Inside Climate News. Maine Was First To Ban Spreading PFAS-Contaminated Sludge on Farmland; The New Republic, One State's War on Forever Chemicals in Milk, 2025/2026 reporting. ↩
26. The New Lede and trade-press summaries of state biosolids policy (Connecticut, Michigan, New York, Wisconsin), 2024–2025. ↩
27. PEER v. EPA litigation over PFAS in biosolids, reported via The New Lede and Waste Dive, 2024–2025. ↩
28. C&EN/ACS (2026). EPA moves to cut risks from PFAS in fertilizer, pans Biden-era analysis. ↩
29. EPA. PFAS National Primary Drinking Water Regulation, Federal Register, April 2024. ↩
30. EPA press materials and Federal Register notices, May 2026; White & Case LLP, EPA partially rolls back PFAS drinking water rule. ↩
31. FDA. Testing Food for PFAS and Assessing Dietary Exposure and Analytical Results of Testing Food for PFAS from Environmental Contamination. ↩
32. FDA constituent update on PFAS testing of infant formula under "Operation Stork Speed," April 2026. Specific quantitative results from this testing round were not located in the underlying research; flagged as a gap. ↩
33. Ensia. Why farmers are using glyphosate to kill their crops; Soil Association, What is pre-harvest desiccant use of glyphosate? ↩
34. IARC Monograph Volume 112 (2015), cited via multiple secondary summaries. ↩
35. EPA (2020). EPA Issues Interim Registration Review Decision for Glyphosate, cited via Bergeson & Campbell summary. ↩
36. National Agricultural Law Center. Ninth Circuit Orders EPA to Revisit Conclusion That Glyphosate is "Not Likely" to Cause Cancer. ↩
37. European Commission (2023). Renewal of the approval of glyphosate: Questions and Answers; EFSA glyphosate topic page. ↩
38. (advocacy source — Environmental Working Group). Roundup for Breakfast, Part 2 (2018) and Going, Going, Gone: EWG Finds Glyphosate Levels Drop in Oat Products (2023). ewg.org. The 1.75 mg/kg/day EPA reference dose and 30 ppm oat tolerance figures are drawn from multiple secondary/technical summaries; the original EPA RED and IRIS documents establishing these figures were not directly re-verified in the underlying research. ↩
39. Multiple legal/news trackers, e.g. Sokolove Law and TorHoerman Law "Roundup Lawsuit" trackers, July 2026 updates; Missouri Independent, May 2025. Specific dollar figures per outlet were not cross-verified against original court judgments in the underlying research; treat as approximate. ↩
40. CNBC (2026). Supreme Court limits Roundup cancer suits against Bayer's Monsanto, June 25, 2026. ↩
41. Multiple 2024–2026 reviews: MDPI, Microplastics and Human Health; Annual Reviews, Microplastics in the Food System: Should We Worry?; PMC11697325, rapid systematic review. ↩
42. FDA. Microplastics and Nanoplastics in Foods. ↩
43. FSIS regulatory summary via National Agricultural Law Center, To cure or not to cure: Groups petition USDA labeling requirements; Niche Meat Processor Assistance Network, Uncured Bacon. ↩
44. (advocacy source — Center for Science in the Public Interest) CSPI petition to FSIS, Aug. 2019; SupplySide Food & Beverage Journal, USDA plans to change "no nitrate or nitrite added" regulations. ↩
45. WHO/IARC (2015). IARC Monographs evaluate consumption of red meat and processed meat, press release, Oct. 2015; WHO, Q&A on the carcinogenicity of the consumption of red meat and processed meat. iarc.who.int. ↩
46. Harvard T.H. Chan School of Public Health (2015). WHO report says eating processed meat is carcinogenic: Understanding the findings, Nov. 2015. ↩
47. American Cancer Society. Colorectal Cancer Statistics; SEER Cancer Stat Facts: Colorectal Cancer. ↩
48. FDA. Sec. 683.100 Action Levels for Aflatoxins in Animal Food; FDA, CPG Sec 570.375 Aflatoxins in Peanuts and Peanut Products. ↩
49. FDA (2016). FDA Issues Final Guidance for Industry on How to Reduce Acrylamide in Certain Foods. ↩
50. Real Estate, Land Use & Environmental Law Blog. Prop 65 Warnings and Acrylamide in Food; OEHHA, Acrylamide. ↩
51. FDA. Bisphenol A (BPA): Use in Food Contact Application. ↩
52. FDA. FDA Update on Phthalates in Food Packaging and Food Contact Applications, Oct. 2024, and subsequent 2026 update. ↩
53. Sen. Ed Markey press release. Senators Markey and Booker Urge FDA to Ban Toxic Phthalate Chemicals. ↩
54. (advocacy source — Environmental Working Group). EWG's Shopper's Guide to Pesticides in Produce. ewg.org/foodnews. ↩
55. Winter, C.K. & Katz, J.M. (2011). Dietary Exposure to Pesticide Residues from Commodities Alleged to Contain the Highest Contamination Levels. Journal of Toxicology, 2011, Article ID 589674. PMC3135239. ↩
56. (advocacy source — Environmental Working Group) statements collected in secondary reporting, e.g. food-safety.com, EWG Publishes 2026 "Dirty Dozen" List...; Genetic Literacy Project summary. ↩
57. (industry-funded source — Alliance for Food and Farming) criticism of the Dirty Dozen, collected via multiple secondary sources. The Alliance's industry funding structure is widely reported, but a specific funding-breakdown primary source was not independently verified in the underlying research. ↩
58. PMC11764615. Efficacy of Household and Commercial Washing Agents in Removing the Pesticide Thiabendazole Residues from Fruits. ↩
59. The New Lede (2024). Peel those apples: New study confirms washing doesn't remove pesticide residues, Aug. 2024. ↩
60. Multiple studies synthesized via secondary reporting on apple peeling and pesticide residue. A specific percentage removed by peeling for apples was not independently pinned to one primary study in the underlying research. ↩