Clean Label Project & Organic Consumers Association 2020 — Heavy metals in US collagen supplements

Summary

This is a Clean Label Project and Organic Consumers Association white paper reporting a market survey of heavy-metal contamination in 30 collagen supplement products purchased from US online and brick-and-mortar retailers. Samples were drawn from Amazon best-seller listings and the cohort tested in Clean Label Project’s earlier 2018 protein-powder investigation, then sent to two ISO-accredited analytical chemistry laboratories for total arsenic, cadmium, total mercury, and lead by ICP-MS (EPA Methods 6020A and 3015, level of quantification 8 ppb). Results were reported on a per-serving basis (µg/serving) and benchmarked against the California Office of Environmental Health Hazard Assessment Proposition 65 Safe Harbor levels rather than against concentration limits. The headline pattern is that arsenic was the most prevalent contaminant (64% of products with measurable per-serving As, range 0.09 to 4.7 µg/serving), lead was measurable in 37% of products (range 0.09 to 1.57 µg/serving, with the highest single-serving exposure ~3× the Prop 65 Pb MADL), and cadmium was less prevalent (17% measurable) but reached 9.17 µg/serving in the highest-loading product, ~2× the Prop 65 oral Cd MADL on a single-serving basis. Mercury was non-detect or below the 8 ppb LOQ in all 30 products. The report is non-peer-reviewed and structured as a consumer-advocacy document, but the underlying analytical chemistry is laboratory-grade and the per-serving exposure framing is a useful occurrence dataset for the collagen-supplement product category.

Key numbers

Source reports per-serving heavy-metal mass in µg/serving for arsenic, cadmium, and lead (Chart 1, page 10). Mercury is reported qualitatively as ND or “Trace” (Trace = detected but below the 8 ppb LOQ); no quantitative Hg values are given. Population summaries are author-reported in the Findings section (pages 3-5). Detection rate denominators are n=30 unless noted.

Worst-case per-serving findings (Chart 1):

• Total arsenic top loader: an unflavored bovine collagen-peptide product at 4.7 µg tAs/serving (20 g serving), below the 10 µg/d Prop 65 NSRL for As.
• Cadmium top loader: a chocolate collagen-protein product at 9.17 µg Cd/serving (41 g serving). The report compares this to a 4.1 µg/day Prop 65 Cd reference and notes it is “over 2 times” that benchmark.
• Lead top loader: an unflavored bovine collagen-peptide product at 1.57 µg Pb/serving (20 g serving), “over 3 times the State of California threshold” (Prop 65 Pb MADL = 0.5 µg/day).
• Lead secondary cluster: three additional chocolate flavored / keto-style collagen products each exceeded 2× the Prop 65 Pb MADL on a per-serving basis (1.10 to 1.33 µg Pb/serving).
• Mercury: the LOQ-limited “Trace” classification covered 34% of products; no individual µg/serving values are reported because all detections fell below the 8 ppb LOQ.

Context for the comparison ladders. The source white paper itself prints the numeric tAs comparison value on the face of the PDF (“None of the products tested exceeded the 10 micrograms/serving State of California threshold,” page 4), but for cadmium and lead it expresses the worst-case findings as multiples of the Prop 65 threshold (“over 2 times” Cd; “over 3 times” Pb) without naming the underlying µg/day reference values. The standard California OEHHA Proposition 65 Safe Harbor levels that those multipliers back-derive against are: tAs 10 µg/day NSRL (cancer endpoint); Cd 4.1 µg/day MADL (oral reproductive endpoint); Pb 0.5 µg/day MADL (oral reproductive endpoint). Those Cd and Pb numerics are imported here from OEHHA’s published Safe Harbor list, not from the white paper’s own text. The report does not cite a Prop 65 reference for total mercury and does not interpret the trace Hg results against any threshold.

Methods (brief)

Thirty collagen supplement products were procured from US local, national, and online retailers using Clean Label Project’s documented Consumer Chain of Custody and Sampling Methodology, intended to replicate the consumer purchasing experience. Samples were sent to two ISO-accredited analytical chemistry laboratories — HRI Labs and Ellipse Analytics — for total arsenic, cadmium, total mercury, and lead analysis by Inductively Coupled Plasma Mass Spectrometry (ICP-MS), using US EPA Method 6020A with EPA Method 3015 microwave-assisted acid digestion. The reported level of quantification was 8 parts per billion. Results were calculated on a per-serving basis (µg/serving) using each product’s label-declared serving size (2.6 g to 41 g across the cohort) and compared to California Office of Environmental Health Hazard Assessment (OEHHA) Proposition 65 Safe Harbor exposure levels. No metal speciation was performed; arsenic is reported as total As and mercury as total Hg. No replicate-sample protocol, per-lab assignment of products, or lab-blank/spike-recovery summary is disclosed in the public white paper.

Implications

• This is the Index’s per-serving occurrence dataset for the US collagen supplement market in 2019-2020, complementing the concentration-basis European marine-collagen data in cammilleri2025-marine-collagen-toxic-metals and the broader protein-powder occurrence base in bandara2020-protein-powder-heavy-metals. Together these support the supplements-protein-collagen-powders product page.
• The per-serving framing (µg/serving against Prop 65 daily reference levels) is the consumer-relevant axis for this product category, because serving size varies four-fold across products (2.6 g to 41 g) and concentration-only data understates per-use exposure for high-serving keto-style chocolate collagens.
• Lead is the most regulatorily salient finding: 37% of the surveyed market had measurable per-serving Pb, the worst single product was ~3× the Prop 65 Pb MADL on a single-serving basis, and three additional products each exceeded 2× that benchmark.
• Cadmium prevalence is lower (17%) but the high-end exposure (9.17 µg/serving) is the largest single-analyte exceedance in the cohort relative to its Prop 65 reference.
• Total mercury was not a meaningful exposure signal at the 8 ppb LOQ used; HMI ingredient/product pages should not draw a tHg signal from this source.
• Evidence fitness caveat: non-peer-reviewed industry/advocacy white paper, n=30 single-pull market sample, accredited-lab ICP-MS but no per-product replicate disclosure, no lab-blank or recovery reporting, no speciation. Treat as a B-tier per-serving occurrence anchor rather than as a definitive market-distribution estimate.

Wiki pages this source may touch

• supplements-protein-collagen-powders — first US per-serving occurrence dataset for the collagen-supplement sub-category.
• lead — measurable per-serving Pb in 37% of US collagen supplements; worst single-serving exposure ~3× Prop 65 MADL.
• cadmium — 17% measurable per-serving Cd; high-end 9.17 µg/serving (~2× Prop 65 MADL).
• arsenic — 64% measurable per-serving tAs; highest product 4.7 µg/serving (below Prop 65 NSRL).
• mercury — 34% trace-level detections; no quantitative signal above 8 ppb LOQ.

Verification notes

• All per-serving values taken from Chart 1 (page 10) of the source PDF. Population summaries (detection percentages, range statements, Prop 65 comparison statements) taken from the Findings text (pages 3-5).
• Part 12 brand firewall: the source PDF names 28 individual consumer brands by name in Chart 1 (a brand/product table) and names specific top-performing and worst-performing brands by name in the Findings narrative (“Garden of Life Unflavored Collagen Peptides … most amount of Arsenic”; “BulletProof Collagen Protein Chocolate … highest amount of Cadmium”; “Orgain Collagen Peptides, Unflavored had the most amount of Lead”; “Ancient Nutrition Keto Collagen Chocolate, Bulletproof Collagen Protein Chocolate, and Preferred Elements Keto Collagen Chocolate all had over 2 times the State of California threshold”; “top collagen products tested are: BioOptimal Collagen Powder Unflavored, Neocell Super Collagen, Puori CP1 Pure Collagen Peptides”). Per Part 12 strict reading (locked 2026-05-17), none of these brand names is reproduced in this wiki page. Worst- and best-performing products are described by product form (unflavored collagen peptides, chocolate flavored / keto-style collagen, etc.) rather than by brand name. The lab names (HRI Labs, Ellipse Analytics) are retained under the Part 12 scientific-method vendor/material exception because they identify analytical service providers, not sampled consumer products.
• Part 2 wiki/HMTc firewall: the source white paper is heavily framed around consumer advocacy (recommendations like “Vote with your dollars”, “Demand that these multi-million dollar brands do better”) and around supply-chain critique of Concentrated Animal Feeding Operations (CAFOs). Those framings are not reproduced here; only the analytical occurrence findings are recorded. No HMTc threshold proposal is drawn from this source.
• Speciation: EPA Method 6020A reports total elemental concentrations; tAs and tHg are correct labels, iAs and MeHg are not claimed. No speciation chemistry was performed.
• Units: all values are µg/serving (mass per consumer-use unit), not µg/kg (concentration). The four-fold spread in declared serving size across the cohort (2.6 g to 41 g) means per-serving rankings and per-mass concentration rankings differ; the source reports only per-serving. Any downstream conversion to ppb concentration must be done with the product’s serving-size denominator and labelled accordingly.
• Mercury LOQ: the 8 ppb level of quantification is the source’s stated LOQ; “Trace” detections in Chart 1 fall below this LOQ and are not quantitated. No µg/serving Hg values are reportable from this source.
• Evidence tier B rationale: non-peer-reviewed industry/advocacy white paper jointly authored by Clean Label Project and Organic Consumers Association; published as a downloadable PDF on cleanlabelproject.org rather than in a peer-reviewed journal; no DOI. Analytical chemistry was performed by ISO-accredited third-party labs using EPA-standard methods, which justifies B over C, but the absence of peer review, of per-product replicate disclosure, and of QA/QC reporting (blanks, spikes, recoveries) prevents A-tier classification.
• Scaffold-introducing slugs: matrices labels collagen-supplement, bovine-collagen, marine-collagen, fish-collagen, and chicken-collagen follow the convention already established by cammilleri2025-marine-collagen-toxic-metals; they are not yet in docs/gpt-collaboration/taxonomy-snapshot.md. Surfaced for the next taxonomy-snapshot regeneration.
• Audit subagent (2026-06-08) Check 1 ⚠️: flagged that Cd 4.1 µg/day and Pb 0.5 µg/day Prop 65 MADL numerics are not printed on the face of the white paper (only “over 2×” and “over 3×” multipliers are). Verified against PDF page 4 — finding is correct. Revised Key numbers context paragraph to make explicit that the 4.1 and 0.5 µg/day values are imported from OEHHA’s published Safe Harbor list rather than from this source; the 10 µg/day tAs comparison value is on the PDF face. Audit subagent Check 2 ⚠️ (matrices slugs not in snapshot) was already disclosed in this notes section before audit; no further action.

Page history

The five most recent substantive edits to this page. The full version history lives in git; when DOI minting comes online (see schema docs), each entry below will also link to a version-pinned DataCite DOI.