EFSA CONTAM 2014 — Chromium in Food and Drinking Water

The EFSA Panel on Contaminants in the Food Chain Scientific Opinion on chromium in food and drinking water, adopted 27 February 2014 (EFSA Journal 2014;12(3):3595, 261 pp.) responds to a request from the Hellenic Food Authority (EFET) and treats trivalent chromium (Cr(III)) and hexavalent chromium (Cr(VI)) on different footings: a tolerable daily intake of 300 µg Cr(III)/kg b.w. per day anchored on a NOAEL of 286 mg/kg b.w. per day from the 2-year NTP chronic oral toxicity study in rats (uncertainty factors 100 × 10 for absence of adequate reproductive/developmental data), and a margin-of-exposure approach for Cr(VI) using three reference points from the NTP 2-year sodium-dichromate-dihydrate drinking-water bioassay — a BMDL10 of 1.0 mg Cr(VI)/kg b.w. per day for combined adenomas and carcinomas of the mouse small intestine (neoplastic effects), a BMDL10 of 0.11 mg Cr(VI)/kg b.w. per day for diffuse epithelial hyperplasia of the duodenum in male mice (non-neoplastic effects), and a BMDL05 of 0.2 mg Cr(VI)/kg b.w. per day for decreased haematocrit in male rats (haematotoxic effects). Because no occurrence data on Cr(VI) in food were available and food is by-and-large a reducing medium, the Panel assumed all chromium in food to be Cr(III) and all chromium in water to be Cr(VI). Mean and 95th-percentile dietary exposure to Cr(III) across all age groups was well below the TDI; MOEs for Cr(VI) intake via drinking water indicate low concern for average consumers across all age groups (mean-exposure MOEs > 10,000 except for infants at upper-bound estimates), with a potential concern at upper-bound 95th-percentile exposure in ‘Infants’, ‘Toddlers’ and ‘Other children’.

The opinion identifies cinnamon (mean ≈ 84,250 µg/kg in four quantified Cinnamomum verum samples), cocoa powder (mean 4,345 µg/kg, n = 239), products for special nutritional use (mean 12,129 µg/kg, n = 2,131; combined vitamin/mineral supplements 23,441-23,514 µg/kg, n = 582) and sunflower oil (LB 592-UB 680 µg/kg, n = 57) as the highest-occurrence food sub-groups. Across European populations ‘Toddlers’ showed the highest mean chronic exposure to Cr(III) (2.3 LB-5.9 UB µg/kg b.w./day) and the dominant contributors to Cr(III) exposure in infants and toddlers were ‘Foods for infants and small children’ and ‘Milk and dairy products’, followed by ‘Bread and rolls’ and ‘Chocolate (cocoa) products’ in older children and adults. The Panel concluded that current dietary exposure to Cr(III) does not raise concern for public health and that current exposure to Cr(VI) via drinking water is of low concern for average consumers but warrants attention for high consumers in the youngest age classes, and recommended generation of speciated Cr(III)/Cr(VI) occurrence data for both food and drinking water and further data characterising Cr(VI) reduction in the GI tract at human-relevant doses.

Key numbers

Health-based guidance values and reference points:

Occurrence dataset:

Total-chromium occurrence (µg/kg) at FoodEx Level 1 — five highest food groups by lower-bound mean (Table 6):

Selected high-occurrence sub-groups (LB-UB unless noted):

Drinking-water occurrence (µg/L) — Table 7, with Cr in water treated as Cr(VI):

For the 88 bottled-water samples reporting Cr(VI) directly, mean Cr(VI) = 4.7 µg/L (LB = UB) across 76 quantified samples; minimum 0.1 µg/L, maximum 36.0 µg/L.

Chronic dietary exposure to Cr(III) via food across European surveys (µg/kg b.w./day), Table 8:

Median values for medians are not calculated where only two surveys were available (infants) or where dietary surveys/age classes had fewer than 60 observations.

Main food-group contributors to Cr(III) chronic exposure (median across surveys, % of total):

Chronic exposure to Cr(VI) via drinking water (ng/kg b.w. per day) and resulting MOEs (Table 24) — RP = BMDL10 1.0 mg Cr(VI)/kg b.w. per day for neoplastic effects, all-types drinking water:

Mean and 95th-percentile chronic exposure to Cr(VI) via all-types drinking water ranged 0.7-159.1 ng/kg b.w./day (min LB - max UB) and 2.8-320.2 ng/kg b.w./day, respectively. MOEs for non-neoplastic and haematotoxic effects, using the BMDL10 of 0.11 mg Cr(VI)/kg b.w./day and BMDL05 of 0.2 mg Cr(VI)/kg b.w./day as reference points, were 690 and 340 (mean and P95, maximum UB) for non-neoplastic effects and 1,300 and 630 for haematotoxic effects — all well above the threshold of 100 the Panel considers low concern for thresholded effects.

Existing EU legislation cited in the opinion:

The opinion notes that no maximum levels for chromium in food are laid down in EU legislation.

Methods (brief)

EFSA called for occurrence data on chromium in food and drinking water from EU Member States; by end-February 2013, 79,809 analytical results were received (27,074 food, 52,735 drinking water). Information on oxidation state was not available for food and for drinking water only the 88 bottled-water Cr(VI) results carried explicit speciation. After data cleaning and validation, applying LOQ cut-offs based on reported LOQs, 24,629 food results and 46,234 water results were used. Two European standardised methods are available for total chromium in food and four for water; two standardised methods exist for Cr(VI) in various waters based on colorimetric reactions with 1,5-diphenylcarbazide and UV-Vis detection. The Panel notes that for Cr(VI) in food no validated or standardised method is available and only limited standard or certified reference materials exist for chromium speciation. Left-censored data were treated by substitution per WHO/IPCS (2009) and EFSA (2010d) guidance — LB sets < LOQ values to zero, UB sets < LOD values to LOD and < LOQ to LOQ.

Chronic dietary exposure to Cr(III) was calculated from the EFSA Comprehensive European Food Consumption Database (26 dietary surveys in 17 European countries; only surveys with more than one reporting day per subject retained); occurrence and consumption data were linked at the lowest available FoodEx level. Food categories not represented were assigned an occurrence value from the immediate upper FoodEx level when fewer than ten samples were available. Dilution factors were applied to match occurrence in dry samples to liquid consumption (18 for coffee with separate factors for espresso = 7 and instant = 63; 100 for tea and herbal leaf; 60 for cocoa powder; 8 for follow-on and infant formulae). 95th-percentile estimates from dietary surveys/age classes with fewer than 60 observations were considered not statistically robust and were excluded. The food group ‘Products for special nutritional use’ was excluded from dietary exposure calculations because consumption information was limited and incomplete; a separate supplemental-intake scenario based on the 2010 EFSA ANS Panel chromium picolinate opinion was used for adults. Cr(VI) exposure from food was not quantified because reliable Cr(VI)-in-food data do not exist; a worst-case scenario assumed Cr(VI) in water used to prepare coffee, tea, infant dry/follow-on food, instant soup, evaporated/dried milk and dehydrated fruit juice was not reduced before ingestion, yielding up to a two-fold increase over drinking-water-only exposure.

Dose-response assessment for Cr(III) was infeasible from the most reliable animal studies because no effects were seen even at the highest doses tested; the TDI was derived from the long-term NTP rat NOAEL of 286 mg Cr(III)/kg b.w. per day with default UF 100 plus an additional factor of 10 for the absence of adequate reproductive and developmental data. Dose-response for Cr(VI) was modelled by BMD analysis (BMDS 2.4 of US-EPA and PROAST of RIVM) on the 2-year NTP carcinogenicity study of sodium dichromate dihydrate in male and female F344/N rats and B6C3F1 mice using a default BMR of 10% extra risk for neoplastic incidence and 5% for haematological continuous endpoints; combined-sex modelling was used where no statistically significant sex differences were found. PROAST Exponential and Hill models were applied to the four haematological endpoints (haematocrit, haemoglobin, MCV, MCH) on male rats at day 22 of the 2-year drinking-water exposure.

The Panel’s uncertainty analysis (Tables 26 and 27) summarises eleven sources of uncertainty for Cr(III) in food (extrapolation from mainly one Member State, LB/UB exposure framing, dilution-factor choice, vegetarian/supplement data limitations, infant data limitations, stainless-steel food-preparation contribution, breast-milk-based infant exposure) and twelve for Cr(VI) in water (analytical measurement, mainly-one-Member-State sampling, very limited Cr(VI) sample base only on bottled water, all-Cr-in-water-is-Cr(VI) assumption, assumption that no Cr(VI) is present in food, smoking-exposure data absence, GI-tract reduction and absorption uncertainty between rodents and humans, combined adenoma + carcinoma small-intestine MOE basis). The Panel concluded the impact of uncertainties is “large” for Cr(III) in food and “very large” for Cr(VI) in drinking water.

Wiki pages updated on ingest

• chromium
• chromium-hexavalent
• water
• cocoa
• chocolate
• cinnamon
• herbs-and-spices
• sunflower-oil
• rapeseed-oil
• bread
• cereals
• milk-and-dairy
• vegetables
• tea
• coffee

Provenance notes

License public-reference-only: the EFSA Journal is published under the Creative Commons Attribution-NonCommercial-NoDerivatives licence at the EFSA Journal website; this wiki extracts numerical findings and Panel conclusions as facts and paraphrases narrative content without reproducing verbatim passages of substantial length. The opinion was published 12 June 2014 and replaces an earlier 13 March 2014 version; corrections were editorial (author name “Cubbada” corrected to “Cubadda” on pages 27 and 135; short title in headers corrected throughout to “Chromium in food and drinking water”). The Question reference is EFSA-Q-2012-00379.

The opinion uses the EFSA Comprehensive Database version 1, comprising 32 dietary surveys in 22 Member States covering more than 67,000 individuals; chronic exposure assessment retained the 26 surveys in 17 countries with more than one consumption day per subject.

Verification notes

The opinion is internally inconsistent about which sex the BMDL10 of 0.11 mg Cr(VI)/kg b.w. per day for diffuse epithelial hyperplasia of the duodenum was modelled in. The abstract (page 1), the risk-characterisation text on page 115, and the final conclusions bullet on page 126 attribute the BMDL10 to female mice; the BMD-analysis text on page 113 (“For male mice the BMD approach was only applicable to the data of diffuse epithelial hyperplasia in duodenum and resulted in a BMDL10 of 0.11”), the risk-characterisation text on page 117, the Dose-Response Conclusions bullet on page 125, and — most decisively — Table 22 (page 113) attribute it to male mice, with female mice marked “No BMDL could be determined.” The wiki uses the data-authoritative male-mice attribution from Table 22 and the supporting BMD-analysis text. Source citation: BMDL10 0.11 male mice → Table 22 page 113 and page 113 supporting text; female mice → no BMDL determinable per Table 22.

Fresh-context audit subagent (Claude Opus 4.7, 2026-06-03) returned verdict REVISE with the BMDL10-sex finding above (applied). All other checks clean: ✅ Check 1 (numerical fidelity verified across Tables 6, 7, 8, 21–25 and abstract/summary/conclusions; one minor rounding note — page 42 actual food LC is 49.1% rounded to “50%” in summary, accepted), ✅ Check 2 (all 13 ingredient slugs and both metal slugs in taxonomy snapshot), ✅ Check 3 (BMD modelling software, BMRs, endpoints accurately reported), ✅ Check 4 (Part 12 brand firewall — Latin botanical names are taxonomic identifiers, “Schmaltz” is a generic culinary term, no per-brand rankings), ✅ Check 5 (Part 2 wiki/HMTc firewall — no threshold proposals, no consumer advisories, EFSA conclusions paraphrased not synthesised).

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.