Cr(VI) and Cr(III) in milk, dairy and cereal products and dietary exposure assessment

Hernandez F; Bemrah N; Séby F; Noël L; Guérin T

doi:10.1080/19393210.2019.1598506

Hernandez et al. 2019 — Cr(VI) and Cr(III) in milk, dairy and cereal products (France)

This A-tier peer-reviewed analytical study determines total chromium and hexavalent chromium in 68 French food samples representing milk, dairy, and cereal products — the categories that contribute 32% of total Cr dietary exposure for French adults and 42% for French children. The headline finding is that Cr(VI) was not detected in any of the 68 samples despite high method sensitivity (Cr-VI LOD 0.3-0.4 µg/kg), confirming the EFSA 2014 assessment that food matrices reduce hexavalent chromium to the trivalent form (Cr(III)). Total Cr was quantified in 58% of dairy and 76% of cereal samples; the highest mean total Cr was in breakfast cereals (269 µg/kg fresh weight) and cheese (25 µg/kg fw). The authors explicitly address older studies that reported detected Cr(VI) in food (Figueiredo et al. 2007, Ambushe et al. 2009, Soares et al. 2010, Mandiwana et al. 2011 reporting Cr(VI) at 1.3-20% of total Cr) and conclude those were “probably due to analytical artefacts or lack of specificity” given that newer LC-ICP-MS speciation finds Cr(VI) absent. Per the corrected row-fit rule (CLAUDE.md Part 6), this paper’s findings extend by chemistry mechanism to infant cereal and milk-based formula matrices even though the samples are general-population French foods.

Key Numbers

Cr(VI) — definitive finding

Cr(VI) was not detected in any of the 68 samples at the analytical LOD of 0.3 µg/kg (milk and dairy) or 0.4 µg/kg (cereal). The authors confirmed analytical sensitivity by spiked-sample recovery (80-120% across 0.05-5 µg/L) and CRM verification (ERM-CE278k with all n=5 recoveries within confidence interval).

The HMTc Cr-VI standard target across food matrices is therefore < LOD with sensitive speciation — detection alone is the threshold-violation signal, since food chemistry naturally reduces any incoming Cr(VI) to Cr(III).

Total Cr (Table 3) — by food category, UB lower-bound treatment

Food category	n	Mean total Cr (ppb)	Min (ppb)	Max (ppb)	>LOQ %
Cereal products (combined)	30	83	10	483	76%
— Bread and dried bread	8	44	20	135	50%
— Breakfast cereals	9	269 (highest)	35	483	100%
— Pasta	1	18	—	—	100%
— Rice and wheat products	5	19	10	33	60%
— Sweet and savoury biscuits and bars	7	66	20	236	71%
Milk and dairy products (combined)	38	14	1.1	101	58%
— Fluid milk	3	9.5	6.2	12	100%
— Ultra-fresh dairy	24	5.9	1.1	21	38%
— Cheese	11	25	4.8	101	36%

The highest individual-product total-Cr concentrations were in two melted cheeses (93 and 101 ppb) and a low-fat fruit yogurt (21 ppb) for dairy; in cereal products, four breakfast cereals containing chocolate (range 360-483 ppb) and a dry chocolate biscuit (mean 103 ppb). The authors note that cocoa-containing products consistently show the highest Cr concentrations, consistent with prior EFSA findings.

Cr(III) dietary exposure (Table 4) — UB scenario

Mean French adult Cr(III) exposure: 0.130 µg/kg bw/day (P95 0.282); children: 0.309 µg/kg bw/day (P95 0.712). Both well below the EFSA TDI of 300 µg/kg bw/day. Cereal products contribute 78% of adult Cr(III) exposure; milk and dairy 22%.

Cr(VI) dietary exposure (Table 4) — UB scenario

Mean French adult Cr(VI) exposure (UB scenario, treating <LOD as LOD): 0.0012 µg/kg bw/day (P95 0.0024); children: 0.0027 (P95 0.0063). Children’s exposure represents 2.7-6.3% of EFSA’s MOE for neoplastic effects (0.1 µg/kg bw/day) — so under the upper-bound assumption that all <LOD Cr(VI) actually equals LOD, exposure remains well below health-based guidance values.

Routing to HMTc subcategories

Subcategory	Route	n_a_tier impact for Cr-VI
baby-cereals-dry-rice-based	Indirect via matrix-chemistry: 9 breakfast cereals including puffed rice (Cr(VI) <LOD, total Cr in puffed-rice subset specifically is in the 35-483 ppb range across the breakfast cereal category); food-matrix Cr(VI) reduction mechanism applies to infant rice cereal.	Closes Cr-VI cell from data gap to Path A — Cr-VI <LOD; food matrix reduces Cr(VI) to Cr(III) (Hernandez 2019 + EFSA 2014a chemistry); n_a_tier=1 with high-quality method sensitivity. Soares 2000 detected-Cr(VI) values in milk-formula are reinterpreted as likely method artefacts per Hernandez authors’ direct discussion.
baby-cereals-dry-non-rice	Indirect via matrix-chemistry: 9 breakfast cereals + 8 bread + 1 pasta + 5 rice/wheat + 7 biscuits cover non-rice cereal matrix.	Same: Cr-VI <LOD; n_a_tier=1.
infant-formula-powder-non-soy	Indirect via matrix-chemistry: 3 fluid milk samples cover the dairy matrix; chemistry mechanism extends to milk-based powdered formula.	Closes Cr-VI cell: Hernandez 2019 finds Cr-VI <LOD in fluid milk; supersedes Soares 2000’s detected Cr(VI) in milk-formula (the older paper used ETAAS after ion exchange, a method Hernandez authors indicate is prone to artefacts). n_a_tier=1 (Hernandez 2019 primary). Soares 2000 retained as historical context only.
infant-formula-rtf-liquid-non-soy	Indirect via matrix-chemistry.	Cr-VI <LOD; n_a_tier=1.
mixed-meals-rice-containing	Indirect via matrix-chemistry.	Cr-VI <LOD; n_a_tier=1.
mixed-meals-non-rice	Indirect via matrix-chemistry.	Cr-VI <LOD; n_a_tier=1.
teething-and-snacks-rice-based	Indirect via matrix-chemistry: biscuits and rice/wheat products cover the matrix.	Cr-VI <LOD; n_a_tier=1.
teething-and-snacks-non-rice	Indirect via matrix-chemistry.	Cr-VI <LOD; n_a_tier=1.

Resolution of the Soares 2000 Cr(VI)-detected discrepancy

Soares et al. 2000 reported detected Cr(VI) in 20 powdered milk infant formulas at means of 24 ppb (infant n=7), 12 ppb (follow-up n=5), and 33 ppb (dietetic n=8) using electrothermal atomization atomic absorption spectrometry (ETAAS) after ion exchange on an NH2 column. Hernandez et al. 2019 directly addresses this disagreement (p. 5):

“Our results confirm the assumptions previously made, indicating that the detected Cr(VI) in previous studies was probably due to analytical artefacts or lack of specificity (Novotnik et al. 2013, 2015; Vacchina et al. 2015). The absence of Cr(VI) in the investigated samples may also be due to its instability in food, which is by and large a reducing medium (EFSA 2014a; Novotnik et al. 2015; Vacchina et al. 2015).”

This is a methodologically-grounded resolution per CLAUDE.md Part 6 / Part 14: the modern LC-ICP-MS speciation with anion+cation exchange separation supersedes older ETAAS-based methods because LC-ICP-MS is matrix-validated for food samples and avoids the on-column reduction artefacts that ETAAS+ion exchange is prone to. Per Part 14, Cr-VI must not be treated as a separate species unless the source speciates hexavalent chromium with a method validated for the food matrix; Hernandez 2019 meets this bar, Soares 2000’s older ETAAS approach does not.

The Soares 2000 source page is retained as historical context. Its Cr(VI) values are documented as superseded by Hernandez 2019’s finding plus the EFSA 2014a + Novotnik 2013/2015 + Vacchina 2015 confirmation that food-matrix Cr(VI) reduces to Cr(III) and earlier-detected-Cr(VI) was method artefact.

Methods (brief)

Total Cr: 0.2-0.6 g sample digested in 6 mL HNO₃ (67-69% m/m) in closed microwave at 80 bar (Multiwave 3000 Anton-Paar), then ICP-MS (Agilent 7700x with collision cell, He) at standard plasma conditions. Total Cr LOD 5 µg/kg, LOQ 10 µg/kg.

Cr(VI): 0.3 g sample extracted with 6 mL 32.5 mmol/L NH4OH in ultrasonic bath at 100% for 30 min; centrifuged at 10,000 rpm 15 min; filtered through 0.45 + 0.22 µm PVDF; 100 µL injected onto LC-ICP-MS with CS5A anion+cation exchange column (250 × 4 mm, 9 µm) coupled to ICP-MS iCAP Q in KED mode (He collision). Quantification by external calibration. LOD/LOQ for Cr(VI): 0.3/0.6 µg/kg in milk and dairy, 0.4/0.8 in cereals.

Method validation: AFNOR NF V03-110 accuracy profile procedure; total Cr method is the only one currently accredited. CRMs: ERM-CE278k (LGC Standards) for total Cr (n=5 recoveries within confidence interval of 0.73 ± 0.22 mg/kg). Cr(VI) accuracy verified by spiked-sample recovery 80-120% at 0.05-5 µg/L.

Evidence Fitness

EF-2 reconstructable A-tier evidence at the food-category × matrix level. Per-sample Cr-VI values are uniformly <LOD; per-sample total Cr values are summarized by category mean/min/max but per-sample tabulation is not in the published Tables 3-4 (only category-level statistics). Sample-level extraction would require contacting the authors. The Cr(VI) finding is unambiguous at the n=68 level: detection itself is the failure mode.

Limitations

General-population French samples, not infant-specific. The Cr-VI <LOD finding extends to infant matrices by chemistry mechanism (food-matrix reduction is a general property of food chemistry per EFSA 2014a) but is not direct infant-formula or infant-cereal evidence.
Total Cr in pasta is n=1, insufficient for distribution.
Method accreditation only covers total Cr currently; Cr(VI) speciation is in-house validated but not third-party accredited (paper notes this explicitly).
Sample collection 2015 — older than ideal for current French market but the underlying chemistry is matrix-property invariant.
The paper’s UB scenario (treat <LOD as LOD) provides the maximum-possible Cr(VI) exposure estimate; under LB scenario (<LOD = 0), Cr(VI) exposure is zero.

Implications

Certification: This paper resolves the long-standing Cr-VI question for HMTc Category 1 food matrices. The HMTc Cr-VI threshold target across food matrices is “below LOD with sensitive speciation” — detection of Cr(VI) at modern method LOD (~0.3-0.4 µg/kg) signals matrix-anomaly or sample-prep contamination, not actual Cr(VI) presence. For all 14 IandC subcategory pages with Cr-VI cells, the cell can move from data gap to Path A primary occurrence with Hernandez 2019 as the n_a_tier=1 source plus EFSA 2014a chemistry mechanism as supporting evidence. Soares 2000’s detected Cr(VI) in milk-formula is reinterpreted as method artefact and retained as historical-context only.

Courses: Strong teaching example for the difference between speciation and total measurement, and for how analytical method generation affects historical interpretation. The Hernandez “previous studies’ Cr(VI) was probably analytical artefact” framing is the canonical resolution.

App: Cr-VI risk in infant foods is essentially zero at modern detection limits; total Cr (effectively all Cr(III)) provides the relevant risk signal at much lower toxicity.

Microbiome: No direct microbiome endpoint, though the food-matrix Cr(VI) reduction chemistry connects to organic-matter chemistry that gut microbiome dynamics also operate within.

Provenance Notes

Karen externally fetched this paper on 2026-05-09 and dropped it at raw/external-fetch/hernandez2019.pdf. Originally identified on the data-gap wishlist as the highest-leverage Cr-VI gap-filler under the misattributed name “Mathebula 2019” (the wishlist’s PMID 30931809 was correct; the first-author label was wrong). Actual first author is Fanny Hernandez (Université Paris-Est, ANSES). Published in Food Additives & Contaminants Part B: Surveillance under standard Taylor & Francis paywall; author-shared copy used internally only. Wiki cites the article record (DOI 10.1080/19393210.2019.1598506).

Heavy Metal Index

Index