Saraiva et al. 2021 — Cr(III) and Cr(VI) speciation in bread and breakfast cereals using SS-ID HPLC-ICP-MS

This study developed and validated a method for simultaneous Cr(III) and Cr(VI) speciation in bread and breakfast cereals using species-specific isotope dilution (SS-ID) coupled to HPLC-ICP-MS, then applied it to 22 retail samples (11 breads, 11 breakfast cereals) purchased in Denmark and France. The SS-ID approach corrects for any species interconversion that occurs during sample preparation, which is the principal weakness of conventional speciation methods. Cr(VI) was below the 0.047 µg/kg quantification limit in every sample analyzed; all chromium present was Cr(III), with concentrations spanning 5.2–176 µg/kg in breads and 23.8–350 µg/kg in breakfast cereals. The Cr(III) totals agreed with independent total-Cr measurements by ICP-MS to within 87–109%. The authors also performed Cr(VI) spike-recovery experiments on bread and recovered all of the spiked Cr(VI) as Cr(III), supporting the hypothesis that the bread matrix actively reduces Cr(VI) to Cr(III) and that prior reports of Cr(VI) in such matrices are most likely analytical artefacts.

Key numbers

Cr(III) by SS-ID HPLC-ICP-MS vs total Cr by ICP-MS (µg/kg, mean ± SD, n=2 per sample; Table 7):

Range summary:

• Bread Cr(III): 5.2–176 µg/kg (spelt bread is the maximum).
• Breakfast-cereal Cr(III): 23.8–350 µg/kg (20.1% cocoa cereals is the maximum; the cocoa cereals dominate the upper tail of the dataset).
• Cr(VI): not detected in any sample (< LOQ of 0.047 µg/kg).
• Cr(III)/total-Cr ratio: 87–102% across breads with quantifiable total Cr; 90–109% across breakfast cereals — consistent with all measured Cr being Cr(III).

Cr(VI) spike-recovery experiments on French wheat bread (Fig. 1; n=4 spike levels at 5, 10, 15, 20 µg/kg):

• When the sample was spiked with Cr(III), Cr(III) was recovered quantitatively (104 ± 5%) and Cr(VI) remained undetected.
• When the same sample was spiked with Cr(VI) at the same levels, the spike was recovered entirely as Cr(III) (98 ± 7%) and Cr(VI) remained undetected.
• This is the load-bearing evidence in the paper: the bread matrix reduces added Cr(VI) to Cr(III) quantitatively, so Cr(VI) cannot be measured in bread matrices using extraction approaches that do not correct for species interconversion.

Method validation (accuracy profile, three measurement series × duplicate × 5 days over one month; Table 6):

• Trueness (bias): individual matrix/level values run from −0.31% to +0.49% (Cr(III) values 0.30, −0.02, −0.31, 0.01 across the four matrix levels; Cr(VI) values 0.03, 0.14, 0.49 across the three spike levels). The paper summarizes this as “<1% in all cases” (§3.2).
• Repeatability (CV_r): 1.3–4.4% for Cr(III); 0.6–7.9% for Cr(VI).
• Intermediate reproducibility (CV_R): 1.3–4.4% for Cr(III); 2.0–8.9% for Cr(VI).
• LOQ (IUPAC, 10× σ_0 from 21 blank replicates, 0.3 g test portion): 14 ng/kg for Cr(III); 47 ng/kg for Cr(VI). The authors note these LOQs are substantially lower than off-line speciation methods reported in the literature (Hernandez 2017/2018, Mathebula 2017).
• Cr(III) recovery factors in two bread matrices (Table 5; recovery factor = Cr(III) by SS-ID compared to total Cr by ICP-MS): 97–113% across all eight unspiked-plus-spiked cells. At the 5/10/15 µg/kg spike levels alone, recoveries are 99–113%; the 97% lower bound comes from the unspiked (0 µg/kg added) baseline cell where the genuine Cr(III) is compared against total Cr.
• NIST 8436 (Durum wheat flour) certified reference material recovered at 96–103% for the Cr(III) + Cr(VI) sum over 5 measurement days (the CRM has a certified value for total Cr only; the paper uses it to validate the sum of speciated Cr against the certified total).

Extraction-condition sensitivity (Table 3, French bread, n=2): Using the milk/meat protocol (70 °C × 25 min EDTA + 70 °C × 25 min DPC) on bread produced an apparent Cr(VI) value of 28.5 ± 2.8 µg/kg, which the optimized 90 °C × 60 min EDTA step then revealed to be an artefact: at the optimal conditions Cr(III) was 25.8 ± 0.4 µg/kg and Cr(VI) was < LOQ. The “Cr(VI)” measured under sub-optimal conditions is genuinely Cr(III) that failed to release into the EDTA step and was subsequently captured during the DPC step. This is a direct experimental demonstration of how Cr(VI) over-reporting in cereal matrices can arise from extraction-step design alone, even before considering matrix reduction.

Dietary exposure to Cr from cereal products in DK and FR (Table 8; FAOSTAT 2018 consumption × measured Cr(III) levels; 70 kg body weight):

• Wheat: 0.017–1.168 µg/kg bw/day (DK); 0.024–1.593 µg/kg bw/day (FR).
• Maize: 0.006 µg/kg bw/day (DK); 0.014 µg/kg bw/day (FR).
• Rye: 0.044 µg/kg bw/day (DK); 0.001 µg/kg bw/day (FR).
• Oats: 0.004–0.006 µg/kg bw/day (DK); ~0.001 µg/kg bw/day (FR).
• Malt: ~0.000 µg/kg bw/day (DK); 0.001 µg/kg bw/day (FR).
• All values are several orders of magnitude below the EFSA 2014 TDI of 300 µg/kg bw/day for Cr(III) and the US IOM adequate-intake range of 20–35 µg/day for adults; cereal-derived Cr exposure is not a regulatory or toxicological concern in either country on these data.

Methods

Speciation analysis (SS-ID HPLC-ICP-MS). ~0.3 g of homogenized bread or breakfast cereal was weighed into a 50 mL polypropylene tube and spiked with isotopically enriched ⁵⁰Cr(III) (97.4% enrichment, ISC Science) and ⁵³Cr(VI) (92.4% enrichment, ISC Science) at amounts chosen to bring blend ⁵⁰Cr/⁵²Cr and ⁵³Cr/⁵²Cr isotope ratios to ~1. 16 mL of 2.4 mM EDTA (Sigma-Aldrich, 99.995%) was added and the mixture heated in a water bath at 90 °C for 60 min to extract and complex genuine Cr(III). After cooling (~15 min), 200 µL of 0.02 mM DPC (1,5-diphenylcarbazide, Sigma-Aldrich, in methanol/sulfuric acid) was added, volume brought to 20 mL with ultrapure water, and the mixture re-heated at 90 °C for 25 min to reduce any Cr(VI) to Cr(III) and complex it as Cr(III)-DPCO. Extracts were cooled to room temperature and filtered through 0.45 µm PVDF syringe filters before same-day analysis. The 90 °C × 60 min first step was experimentally established as necessary to release genuine Cr(III) from the bread matrix — the milk/meat protocol (70 °C × 25 min) does not (Table 3).

HPLC separation. Dionex IonPac AG7 anion-exchange column (2 mm × 50 mm, 10 µm) on a Dionex ICS-5000+ (at Anses, France) or ICS-6000 (at DTU Food, Denmark). Mobile phase 10 mM HNO₃ + 2.5% methanol + 0.30 mM EDTA (pH 2.1), isocratic, 0.20 mL/min, 5 µL injection, column 30 °C, injector 5 °C. The Cr(III)-EDTA and Cr(III)-DPCO complexes co-eluted as separated peaks in under 3 min.

ICP-MS detection. Thermo iCAP TQ (at DTU) for speciation; Agilent 7700 (at Anses) and iCAP TQ used for total Cr. PFA standard nebulizer, cyclonic quartz spray chamber, He collision cell (KED mode) at ~5 mL/min, 1550 W plasma power, monitoring ⁵⁰Cr, ⁵²Cr, ⁵³Cr, ⁴⁵Sc (internal standard).

Total Cr determination. Closed-vessel microwave digestion (Anton Paar Multiwave 3000): 0.3 g sample + 3 mL concentrated HNO₃ (67% m/m, Merck) + 3 mL ultrapure water in 80 mL quartz vessels; digest diluted to 50 mL with the addition of ⁴⁵Sc as internal standard. External calibration 1.0–10.0 µg/L. Performed in duplicate at Anses and DTU. Method is the accredited 31-element procedure from Chevallier et al. 2015.

Validation. Accuracy profile per AFNOR NF V03-110 / Thompson et al. 2002 IUPAC guideline: three measurement series in duplicate on 5 separate days over one month. Three bread samples spiked with Cr(VI) at 5, 10, 15 µg/kg; two bread samples spiked with Cr(III) at 5, 10, 15 µg/kg. Validity domain: LOQ to 15 µg/kg. NIST 8436 Durum wheat flour CRM analyzed alongside food samples for internal QC.

Risk-assessment calculation. Daily exposure E_i = Σ (C_i,k × L_k) / W_i, where C is daily consumption (FAOSTAT 2018, Denmark and France), L is the measured Cr(III) level in this study, and W = 70 kg adult body weight. Compared against EFSA 2014 TDI of 300 µg/kg bw/day and US IOM 2001 adequate-intake range 20–35 µg/day.

Implications

This source is the principal A-tier methodological reference for the position that Cr(VI) is not analytically detectable in bread or breakfast cereals when measured by a method that corrects for species interconversion, and that prior literature reports of Cr(VI) presence in such matrices (Soares 2010, Mathebula 2017, both cited and discussed in §3.4) are most likely artefacts of off-line analytical procedures lacking on-line species separation and isotope-dilution correction. The matrix-reduction experimental evidence (Cr(VI) spike → recovered entirely as Cr(III), n=4) is the load-bearing finding and is what distinguishes this paper from a method-validation report.

For contributing-source synthesis on grain-based product and ingredient pages, the dataset establishes Cr(III) ranges for typical European bread and breakfast cereal samples (5–350 µg/kg, with cocoa-containing breakfast cereals and spelt bread driving the upper end) and supports the working assumption — at the synthesis level, not in this source page — that total Cr in grain matrices is essentially equivalent to Cr(III) for routing purposes. The exposure calculations in Table 8 provide a quantitative baseline against the EFSA TDI for the rare case where a synthesis page needs to bound cereal-derived Cr intake in EU adults.

Wiki pages this source may touch

• bread
• wheat
• oat
• maize
• corn-flakes
• cereals
• bread-and-baked-goods
• breakfast-cereals
• chromium
• chromium-hexavalent

Verification notes

• 2026-05-25 merge-enhance (Claude Code, autonomous ingest skill v2.0). The pre-existing page (updated 2026-05-14) had multiple schema-era and content defects, all corrected in this pass:
  • raw_handle: manual-fetch-kimi (generic parent-folder placeholder) → MFK_p0320-speciation-analysis-of-criii-and-crvi-in-bre per the v2 manual-fetch handle convention.
  • raw_path was a non-existent stub P0320.pdf → corrected to the full filename present in the source folder (P0320_Speciation analysis of Cr(III) and Cr(VI) in bread and breakfast cereals.pdf).
  • Added no_doi_assigned: false, access_url, and raw_sha256 fields (absent in prior version).
  • ingredients previously included [[ingredients/breakfast-cereals]], which is not a slug in the current taxonomy snapshot (2026-05-18). Replaced with the canonical grain-ingredient slugs that route this study to the right ingredient pages: bread, wheat, oat, maize, corn-flakes, cereals (umbrella).
  • products had only [[products/breakfast-cereals]]; added [[products/bread-and-baked-goods]] because the 11 bread samples are direct evidence on that product category.
  • sample_n: null → 22 (11 breads + 11 breakfast cereals), with sample_population rewritten to enumerate the samples by flour type / composition and to note the NIST 8436 Durum CRM used as internal QC.
• Legacy heading removed. Prior section ## Wiki pages updated on ingest (Part 5b system-rule violation — the model does not maintain routing by hand) replaced with the current ## Wiki pages this source may touch form.
• Wiki/HMTc firewall (Part 2). Prior ## Implications section contained multiple certification-facing synthesis claims that crossed the wiki/HMTc firewall:
  • “This finding … has direct implications for how HMT&C should interpret total-Cr findings in grain products: total Cr in grain matrices is essentially all Cr(III), not Cr(VI).”
  • “For the HMT&C analyte vocabulary (which lists Cr-VI separately), this means the relevant concern in bread/cereal matrices is whether any Cr-VI is present at all — and the best current evidence says it is not…”
  • “Testing claims for Cr-VI in these matrices should therefore be interpreted with awareness of method artefact risk.”
  • Consumer-app advisory (“For grain-based ingredients, the app can treat total Cr as Cr(III) when source data does not specify speciation, consistent with this evidence. Cr-VI risk in grain products should default to ‘not detected’ …”). All deleted in this pass. The rewritten Implications section reports what the data are useful for (synthesis input, methodological reference for prior-literature artefact hypothesis) and the load-bearing matrix-reduction finding, without proposing HMTc thresholds, consumer translations, or course-facing language. The synthesis-level inference about “total Cr in grain matrices ≈ Cr(III) for routing purposes” is flagged as belonging at the synthesis level, not in this source page.
• Missing data added. The prior page summarized the Cr(III) and Cr(VI) ranges but omitted (a) the per-sample contamination table (Table 7), (b) the Cr(VI) spike-recovery stability experiment that is the load-bearing finding of §3.4 (Fig. 1), (c) the extraction-condition sensitivity experiment (Table 3) demonstrating that the milk/meat protocol gives a false-positive Cr(VI) signal of 28.5 µg/kg on bread that disappears under the optimized 90 °C × 60 min EDTA step, (d) the NIST 8436 Durum wheat-flour CRM internal-QC result (96–103% recovery over 5 days), and (e) the FAOSTAT-based dietary exposure assessment (Table 8). All added.
• Brand firewall (Part 12, strict reading locked 2026-05-17). The paper’s sample list (Table 2) uses generic product descriptors throughout; brand identifiers are anonymized by the authors themselves for the two cocoa-cereal samples (“brand I”, “brand II”). The only sample name that maps to a single commercial product is “All bran” (the Kellogg’s trademark; the paper does not disambiguate). Rephrased to “high-bran flake cereal” in the sample table to comply with the strict-reading brand-firewall convention. All other Table 2 entries are generic product-form descriptors (Danish rugbrød, malt flour bread, spelt bread, baguette, granola, muesli, etc.) and are retained verbatim. Method-section vendor names (Anton Paar Multiwave 3000, Thermo iCAP TQ, Agilent 7700, Dionex IonPac AG7, Sigma-Aldrich EDTA/DPC, ISC Science isotope-enriched spikes, NIST 8436 CRM, ED JULABO water bath, Merck HNO₃) retained per the 2026-05-17 method-vendor exception (scientific reproducibility).
• Speciation convention. The paper measures both species of Cr separately by SS-ID and reports them as Cr(III) and Cr(VI). The wiki frontmatter metals field uses the HMT&C vocabulary symbols Cr (for the total / Cr(III) measurement, which is the dominant species and is equivalent to total Cr in this matrix per the paper’s own finding) and Cr-VI (for the dedicated Cr(VI) speciation measurement, even though all values were below LOQ). Cr-III is not a standalone HMT&C analyte; it is captured under Cr.
• Jurisdictions. Samples purchased in Birkerød (DK) and Maisons-Alfort (FR); exposure assessment uses Denmark and France FAOSTAT consumption data; comparison anchored to EFSA 2014 TDI and IOM 2001 DRI. jurisdictions: [FR, DK, EU] reflects all three layers.
• DOI and license preserved from the prior page (10.1016/j.jfca.2021.103991; CC BY 4.0, open-access Elsevier publication confirmed against the PDF’s title page).
• 2026-05-25 audit application (fresh-context subagent audit; verdict REVISE). Two ❌ findings against Check 1 (numerical fidelity) verified against PDF Table 5 and Table 6 and applied; one ⚠️ left as-is.
  • Audit ❌ on trueness range: prior text “0.31% to 0.49% across all tested levels” misrepresented Table 6. PDF Table 6 reports individual bias values of 0.30, −0.02, −0.31, 0.01 (Cr(III) across four matrix levels) and 0.03, 0.14, 0.49 (Cr(VI) across three spike levels); the range is −0.31% to +0.49% with the paper’s own summary “<1% in all cases” (§3.2). Verified against PDF; corrected as flagged.
  • Audit ❌ on Cr(III) spike-recovery range: prior text “Cr(III) spike-recovery in two bread matrices at 5, 10, 15 µg/kg: 97–113%” mis-attributed the 97% lower bound. PDF Table 5 shows recoveries of 99/106/105 (Bread 1) and 101/113/107 (Bread 2) at the 5/10/15 µg/kg spike levels (range 99–113%); the 97% lower bound comes from the unspiked (0 µg/kg added) baseline cell where the genuine Cr(III) is compared against total Cr — these baseline values are 97 ± 1 (Bread 1) and 101 ± 2 (Bread 2). The PDF prose “97–113%” in §3.1 includes all eight cells (unspiked + spiked × 2 matrices); the wiki claim restricting the range to “5, 10, 15 µg/kg” was incorrect. Verified against PDF; corrected to report the full range with the spike-level-only sub-range broken out separately.
  • Audit ⚠️ on matrices: [bread, breakfast-cereal, grain-product]: the subagent noted that grain-product may not be in the controlled matrices vocabulary file. Reviewed: matrices is a free-form descriptor field across the source-page corpus (e.g., cereal-grain, whole-food, grain-product all appear across recent A-tier source pages), not a closed taxonomy like ingredients/products/metals. Left as-is.

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.