EFSA CONTAM 2010 — Scientific Opinion on Lead in Food

The EFSA Panel on Contaminants in the Food Chain Scientific Opinion on Lead in Food, adopted 18 March 2010 (re-published 2013 with editorial corrections), pooled 94,126 quantified analytical results submitted by 14 Member States and Norway over 2003–2009 and concluded that the prior JECFA provisional tolerable weekly intake of 25 µg/kg body weight per week was no longer appropriate because the available human data show no threshold for the critical lead-induced effects. The Panel derived three Benchmark Dose Lower Confidence Limits from blood lead concentrations and identified developmental neurotoxicity in young children, increased systolic blood pressure in adults, and chronic kidney disease in adults as the critical endpoints. Estimated dietary exposures in European 1–7 year-old children and in some high-consumer adults sit at or above the BMDL intake values, so the possibility of effects in those subgroups cannot be excluded. This opinion is the literature backbone behind EU contaminant limits for lead and supplies the BMDL₀₁ of 12 µg/L B-Pb (≈ 0.50 µg/kg b.w. per day dietary equivalent) used as the children’s neurodevelopmental reference point throughout European regulatory practice.

Key numbers

Critical endpoints and Benchmark Dose Lower Confidence Limits (Section 8.6, Tables 32 and 34, and Section 9):

Lower-bound (LB) and upper-bound (UB) dietary lead exposures by population subgroup (Table 29, Sections 7.2–7.4):

Margins of exposure relative to the BMDLs (Table 38):

EU-pooled occurrence statistics for the largest food categories (adjusted means and 95th percentiles in mg/kg; LB and UB by GEMS/Food convention; outliers >10× next-highest excluded). Selected from Tables 5–19 and the adjusted-mean summary in Table 20:

Breast-milk and infant-formula occurrence (Section 5.7):

Toxicokinetic anchors (Section 8.1):

Non-dietary exposure summary (Table 29):

Headline regulatory conclusion (Sections 8.6.2 and 9.4): “The CONTAM Panel concluded that the present PTWI of 25 µg/kg b.w. is no longer appropriate and noted that there was no evidence for a threshold for a number of critical endpoints including developmental neurotoxicity and renal effects in adults. Therefore, a margin of exposure approach was applied to risk characterisation.” For dietary intake the conversions used were B-Pb 36 µg/L ≈ 90.0 µg/60 kg = 1.50 µg/kg b.w./day (SBP) and B-Pb 15 µg/L ≈ 37.5 µg/60 kg = 0.63 µg/kg b.w./day (kidney), per the Carlisle and Wade (1992) equation in adults, and IEUBK-derived 0.50 µg/kg b.w./day for the 12 µg/L children’s BMDL₀₁ (Table 37, p. 101).

Methods (brief)

This is a Scientific Opinion of the EFSA CONTAM Panel, not a primary occurrence study; the underlying analytical data were generated by Member-State control laboratories and submitted under the DATEX-2008-0002 call (closed July 2008). Of 139,423 submitted results, 139,113 remained after first-pass cleaning, and 94,126 were retained for the calculation of food-category lead concentrations after exclusion of pre-2003 samples and submissions with LOD > 0.1 mg/kg or LOQ > 0.3 mg/kg. Reported analytical methods were graphite-furnace atomic absorption spectrometry (GFAAS, 24.3 % of samples), ICP-MS (9.8 %), unspecified AAS, flame AAS, and ICP-AES; method was not reported for 52.5 % of samples. Median LODs varied by method (ICP-MS 0.003 mg/kg, AAS 0.006 mg/kg, GFAAS 0.010 mg/kg) and by matrix (tap and bottled water medians 0.0003–0.0007 mg/kg; solid foods 0.005–0.05 mg/kg). All results were converted to mg/kg with 1 kg ≈ 1 L for liquids. Lower- and upper-bound means were calculated per GEMS/Food convention (non-detects set to zero and to the LOD/LOQ respectively). Sampling adjustment factors (SAFs) derived from the German DONALD survey corrected the unbalanced sampling frequency to the consumption structure of the EFSA Concise European Food Consumption Database (15 broad food groups, 28 sub-groups), and adjusted means in Table 20 were combined with individual-level consumption data from 19 EU countries plus Norway to compute deterministic mean and 95th-percentile dietary exposures by country, sex, and age (probabilistic Monte Carlo simulation with the Creme system gave comparable means and was used as a cross-check). Long-term child exposure was modelled with the beta-binomial-normal (BBN) model in MCRA software (de Boer and Van der Voet, 2007) across 4–13 national surveys per age year (1–14 y, 10 countries). Dose-response modelling used the EFSA Benchmark Dose framework: for IQ, a logarithmic and a piecewise-linear (knot at 100 µg/L) regression on the Lanphear et al. (2005) individual-data pooled analysis of seven cohorts (n=1,333; Budtz-Jørgensen, 2010 commissioned by EFSA), with the piecewise-linear estimate (BMDL₀₁ = 12.0 µg/L for concurrent B-Pb) chosen because the logarithmic model has infinite slope at zero; for SBP, the arithmetic mean of BMDL₀₁ across four B-Pb studies (Vupputuri 2003; Nash 2003; Glenn 2003, 2006), giving 36 µg/L; for CKD, log-probit and multistage quantal models on NHANES (1999–2006) data (Navas-Acien et al., 2009; n=14,778), giving BMDL₁₀ = 15 µg/L. Conversion of blood-lead BMDLs to dietary equivalents used the Carlisle and Wade (1992) regression for adults (negligible air and soil/dust contribution assumed) and IEUBKwin version 1.1 for infants and children (food, soil/dust, air, drinking water). Reference materials, individual lab certifications, recovery corrections, and HORRAT compliance requirements followed Commission Regulation (EC) No 333/2007 (criteria approach).

Implications

Certification: This opinion is the European literature anchor for B-Pb-based BMDLs and supplies the children’s 12 µg/L B-Pb / 0.50 µg/kg b.w./day dietary-equivalent reference point used implicitly in subsequent EU maximum-level setting for lead. It contributes occurrence data for cereals, vegetables (leafy and root in particular), potatoes, fruits, meat (especially game and offal), bivalve molluscs, infant formula, milk and dairy, tea, cocoa, herbs, and bottled and tap water across an aggregated EU-wide sampling frame. Margins of exposure below 1 for children 1–7 years and high-consumer adults relative to the kidney BMDL₁₀ are direct inputs to category-level threshold work for infant and follow-on food, ready-to-consume infant formula, leafy and root vegetables, cereal grains, and game-derived foods. Use of the IEUBK model and the BMDL₀₁ of 12 µg/L B-Pb in this opinion is consistent with the framework EFSA carried into the 2013 update opinion and into the 2023 EU maximum-level revision for lead in infant and young child foods.

Courses: Useful for the EU regulatory-history modules and for the dose-response section of any neurodevelopmental-lead unit. The opinion explains why a threshold-based PTWI was abandoned and how the margin-of-exposure framework is applied to a non-threshold genotoxic-carcinogen-adjacent contaminant; the piecewise-linear vs logarithmic IQ-model choice is a textbook example for benchmark-dose pedagogy.

App: Provides population-anchored mean and high-consumer dietary lead exposure estimates by age band that the app can use as European baselines for adult, child, and infant scenarios. Per-category mean and P95 occurrence values supply default contamination values where no narrower source has populated the ingredient page. Breast-milk Pb of 0.0016 mg/L and ready-to-consume infant formula 0.0020–0.0047 mg/kg are usable defaults for infant-exposure modelling when no brand- or product-specific value is available.

Microbiome: Not directly addressed by this opinion.

Verification notes

• Merge-enhanced 2026-06-03 from the prior 2026-04-25 page by reading the full PDF (151 pp.) end-to-end. The prior page had a thin Key numbers table covering only the three BMDLs and the country-range exposures; this revision adds the EU-pooled occurrence statistics by food category, the margin-of-exposure table, the toxicokinetic anchors, the non-dietary pathway estimates, and the breast-milk / infant-formula concentration data, all keyed to the source tables and page numbers.
• The pre-2026-05-14 page used legacy ## Summary and ## Wiki pages updated on ingest headings; replaced with the current opening-prose convention and removed the trailing wikilink stub (routing is handled by the audit layer per Part 5b).
• The Kimi PDF copy (raw/Manual Fetch Kimi /June 3 Folder/EFSA Journal - 2010 - - Scientific Opinion on Lead in Food (1).pdf, sha 61e2e32d…) is recorded as a near-duplicate of the existing curated raw file at raw/reports/scientific-opinion-on-lead-in-food-2010.pdf (sha 897e8bc3…). Both are the same EFSA opinion, DOI 10.2903/j.efsa.2010.1570, EFSA Journal 2010;8(4):1570; the second download came from the Wiley Online Library mirror and the bit-level mismatch is the Wiley re-publication wrapper, not a content difference.
• Prior products: [] was a Phase 1 defect under CLAUDE.md Part 5; populated with the product-category slugs the opinion supplies measurement data for (infant formula powder, infant and child foods master, bottled and mineral water, breakfast cereals, fresh and canned fish, processed meats, game meats). No new product slugs were proposed.
• Frontmatter scope deliberately kept broad: infant-formula-powder rather than soy/non-soy splits because the EFSA category “infant and follow-on formulae” does not separate them, and wild-mushrooms rather than narrower fungi sub-classes because Table 8 reports the broader category. Routing audit fans broad scopes to sibling pages.
• Game-meat numbers retain the source’s own attribution as “high consumer” exposures rather than category means — the 867 mg/kg maximum is for wild pig muscle near a shot wound (sample retained by EFSA after removal of a 3,090 mg/kg outlier). This is the cleanest defensible reading of Table 14 and the surrounding paragraph.
• No brand names appear in this source (it is a pooled regulatory occurrence assessment), so the Part 12 firewall is not engaged.
• No HMTc threshold proposals or consumer-translation language added; opinion’s published conclusions on PTWI inappropriateness are quoted but not extended.

2026-06-03 audit-subagent corrections applied

Fresh-context general-purpose audit subagent (verdict REVISE) flagged six numerical / attribution defects; all six were independently re-verified against the source PDF and corrected:

• BMDL₀₁ = 8.1 µg/g tibia bone moved from the developmental-neurotoxicity row to the SBP row. Verified against PDF p. 99 (“Effects on SBP in adults: BMDL₀₁ = 8.1 µg/g (TB-Pb)”) and Table 34 p. 97 (TB-Pb BMDL₀₁ derived from Cheng 2001 = 6.5 and Glenn 2003 = 9.7, arithmetic mean 8.1, both for SBP not IQ). The 12 µg/L B-Pb is the only BMDL for developmental neurotoxicity.
• Dry herbs mean corrected from 0.5500/0.5500 (LB/UB) to 0.8809/0.8896. The 0.5500 figures are the medians per Table 18 p. 44; the means are 0.8809 LB and 0.8896 UB.
• Spices mean corrected from 0.1380/0.2000 to 0.3313/0.3644. The 0.1380/0.2000 figures are the medians per Table 18; the means are 0.3313 LB and 0.3644 UB.
• Spices P95 corrected from 0.6036/0.9592 to 0.9592/0.9592. The 0.6036 value is the P95 LB for the Salt row in Table 18, not Spices.
• Adult pooled exposure sample size corrected from “n = 41,628” to “n = 37,599”. The latter is the actual sum of country N values in Table 22 (p. 51); 41,628 was not supported by the source.
• Diet/water-Pb→B-Pb conversion row reworded from “Soil→B-Pb conversion (child) +1.6 µg/L per 1 µg/day from soil” to “Diet- or water-Pb→B-Pb conversion: +1.6 µg/L per 1 µg/day in children; +0.4 µg/L per 1 µg/day in adults (Carlisle and Wade, 1992, cited p. 68).” The original wording misattributed the pathway as soil; the source p. 68 specifies diet or water for the 1.6 / 0.4 µg/L coefficients.
• Bone-burden row attribution reworded to cite Section 8.1.2 p. 66 (where the 70 % figure appears with no inline citation) and to clarify that Aufderheide and Wittmers (1992) is cited there only for the trabecular-vs-cortical distribution.

Subagent’s minor-stylistic concerns about children’s survey-count attributions (“9 surveys” for 1–3 y, “9–13 surveys” for 4–7 y) and vegetarian “5 countries” were retained as-is after independent verification: Table 24 actually shows surveys per age 1=6, 2=8, 3=9 and 4=13, 5=13, 6=13, 7=9, so the wiki ranges reflect endpoint counts, which is conventional rather than incorrect; and Section 6.3 p. 48 confirms 65 lacto-ovo-vegetarians from 5 countries (DK was the sixth language-screened country and contributed zero). Subagent’s audit did not flag any wiki/HMTc, brand-firewall, or speciation issues; verdict was REVISE only on numerical fidelity.

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.