EFSA 2021 — Chronic dietary exposure to inorganic arsenic in the European population

This 50-page EFSA scientific report, approved 14 December 2020 by the CONTAM Panel at its 112th Plenary, presents the updated chronic dietary exposure assessment for inorganic arsenic (iAs) across the European population using measured iAs data from 13,608 analytical results (7,623 for drinking water; 5,985 for food), collected across 23 European countries between 2013 and 2018. This is a major methodological advance over the 2014 EFSA scientific report, which derived iAs from total arsenic (tAs) data using modelling assumptions — the primary source of uncertainty in that earlier assessment. Mean dietary exposure estimates (lower bound) were below the BMDL₀₁ range of 0.3–8 µg/kg bw per day established by the EFSA CONTAM Panel in 2009, but 95th percentile exposures for infants, toddlers, and other children fell within or at the lower end of that BMDL₀₁ range. The main contributors to iAs exposure across all age classes were rice, rice-based products, grain and grain-based products (no rice), and drinking water.

Key numbers

• Initial extraction from EFSA Data Warehouse: 67,401 analytical results on arsenic species for samples 2013–2018 (Table 1). Composition: tAs 58,013 (86.1%), iAs 6,021 (9.0%), As(V) 737 (1.1%), As(III) 730 (1.1%), MA 779 (1.2%), DMA 783 (1.2%), AB 317 (0.5%), organic As 21 (0.03%).
• After preliminary cleaning (exclusion of 1,316 duplicate/unconfirmed results from 690 samples; further 60 results from 21 samples not entering the food chain): 66,025 results retained for grouping.
• After full analytical-method, LOQ, and food-chain validation: 13,608 results used for chronic dietary exposure (7,623 drinking water; 5,985 food); 44 dietary surveys; 87,945 subjects; 23 European countries.
• Drinking-water analytical pool: of 7,623 samples, only 63 were directly analysed for iAs; the remaining 7,560 were analysed for tAs and treated as iAs per the assumption that almost all arsenic in drinking water is inorganic (US EPA 2001; FAO/WHO 2011).
• LOQ cut-off applied per EFSA 2018 guidance: 200 µg/kg for tAs food samples; 100 µg/kg for iAs food samples; 10 µg/L for drinking water (264 water samples excluded for LOQ > 10 µg/L).
• 51 of 2,621 rice/rice-based samples (<2%) exceeded applicable MLs under Commission Regulation (EU) 2015/1006 (Table 2). Distribution: rice brown n=15 (max 570 µg/kg, <4%); rice popped n=18 (<3%); rice unspecified n=9; rice long-grain n=4; rice white n=5. Mean iAs in exceeding rice popped 344 µg/kg, max 409 µg/kg.
• About 22% of “Cereal-based food for infants and young children” samples (50 of 224) had iAs > 100 µg/kg, the EU ML for rice-containing infant food.
• Mean iAs in rice by type (Table 4, µg/kg LB–UB): red 233 (n=7); brown 128–131 (n=382); parboiled 98–102 (n=131); long-grain 77–87 (n=502); rice unspecified 79–85 (n=404); white 74–79 (n=337).
• Mean iAs in rice-based products (Table 5, µg/kg LB–UB): rice cakes/wafers/crackers 146–148 (n=579; 95th 270 LB/UB); rice flakes 123–126 (n=113; 95th 267); rice flour 104–109 (n=19); breakfast rice popped (loose) 100–114 (n=62); rice bread 75–80 (n=10); noodle rice 66–76 (n=17); rice porridge 12 (n=7); rice drink 8–17 (n=43).
• Mean iAs in infant/young-children food (Table 8, µg/kg LB–UB): cereal-based with declared rice 77–79 (n=93; 95th 130); cereal-based without declared rice 31–38 (n=131; 95th 110, indicating undeclared rice presence); biscuits/rusks/cookies for children 83–94 (n=52); ready-to-eat meal meat/fish-based 8.8–39 (n=12); ready-to-eat meal cereal-based 3.3–5.8 (n=12); follow-on formula milk-based 0.08–2.1 (n=47); infant formula milk-based 0.02–4.0 (n=38).
• Mean iAs in drinking water used for exposure (Table 7, p.17 — EFSA labels the column “µg/kg” but the matrix is naturally µg/L as in Table 6 above): unspecified 2.0–2.4 LB–UB (n=1,715; 95th 8.2); tap 1.7–2.7 (n=593); well 1.5–2.7 (n=945); bottled 0.6–1.2 (n=522); still mineral 1.0–2.1 (n=1,895); carbonated mineral 0.8–1.7 (n=1,597).
• Mean exposure estimates (lower bound, LB): toddlers 0.30 µg/kg bw/day (highest among age groups); both infants and toddlers reach 0.61 µg/kg bw/day at the upper bound (UB).
• 95th percentile dietary exposure (Table 11, p.24): highest LB max in toddlers 0.58 µg/kg bw/day (infants LB max 0.52); highest UB max in infants 1.20 µg/kg bw/day (toddlers UB max 0.99). The PDF Summary frames these together as “the highest exposure estimates (LB–UB) were 0.58 and 1.20 µg/kg bw per day in toddlers and infants, respectively” — within the BMDL₀₁ range of 0.3–8 µg/kg bw/day.
• UB estimates were 2–3 times higher than LB estimates across all age groups.
• Rice-based-formula consumption scenario: mean 0.30 µg/kg bw/day; high consumers 0.39 µg/kg bw/day (at or approaching the lower end of the BMDL₀₁ range).
• Across all age classes, the main LB contributors to iAs exposure were rice, rice-based products, grain and grain-based products (no rice), and drinking water.
• Comparison to 2014 EFSA report: 2021 mean and 95th percentile estimates were 1.5–3 times lower than 2014 estimates, primarily due to use of directly measured iAs rather than modelled iAs from tAs data.
• EFSA CONTAM Panel BMDL₀₁ (2009): 0.3–8 µg/kg bw/day for 1% increased risk of cancer of lung, skin, bladder, and skin lesions.
• EU regulatory context: Commission Regulation (EU) 2015/1006 set MLs for iAs in rice effective 1 January 2016 — non-parboiled white rice 200 µg/kg, parboiled/husked rice 250 µg/kg, rice waffles/wafers/crackers/cakes 300 µg/kg, rice for production of infant/young-children food 100 µg/kg.

Methods (brief)

Occurrence data: EFSA Data Warehouse (DWH), extracted 30 April 2020, covering annual national monitoring submissions 2013–2018 per Commission Recommendation 2015/1381/EU plus seaweed-specific monitoring under Recommendation 2018/464. Analytical methods reported for iAs in food (n=5,985): ICP-MS dominant (n=3,521; ~59%); AAS n=489; AES n=294; AFS n=55; potentiometry n=13; unspecified spectroscopy n=8; ~25% of samples reported no analytical method. Separation technique: HPLC reported for ~40% of food samples. Lowest food LOQ 0.002 µg/kg (HPLC-ICP-MS for water molluscs); rice LOQs ranged 0.008 µg/kg (HG-AAS) to 4 µg/kg (HPLC-HGAFS). Drinking-water analyses (n=7,623) used MS, AAS, or AES; lowest LOQ 0.0002 µg/L (ICP-MS). Consumption data: EFSA Comprehensive European Food Consumption Database (2020 update), 44 dietary surveys, 87,945 subjects, 23 EU countries; only surveys with more than one reporting day per subject were retained for chronic exposure. Food classification: FoodEx (with FoodEx2 facets where available) enabling separation of e.g. puffed rice cereal vs rice cakes/waffles. Exposure calculation: chronic dietary exposure with lower bound (LB, below-LOD replaced by 0) and upper bound (UB, below-LOD replaced by LOD/LOQ) per standard EFSA methodology; 95th percentile estimates suppressed for surveys/age classes with fewer than 60 observations. All analyses run in SAS Enterprise Guide 7.15.

Limitations

Large LB–UB gap (factor of 2–3) primarily driven by the high proportion of left-censored data (non-detects), especially in food categories with low iAs content. Consumption and occurrence data missing for some iAs-containing ingredient-specific food groups. Food preparation effects on iAs levels (cooking rice changes iAs content) inadequately captured. No specific data on iAs in gluten-free products or food for coeliac patients, who likely consume more rice. UK included as EU member during data collection period. Drinking-water iAs assumption (tAs treated as iAs for 7,560 of 7,623 samples) introduces no resolution between trivalent and pentavalent species. Approximately 25% of food samples lacked any analytical method detail in the submission, limiting independent quality screening.

Implications

• Certification: This is the current European exposure baseline for iAs in rice and rice-based foods. The infant 95th-percentile dietary exposure sits within the BMDL₀₁ range of 0.3–8 µg/kg bw/day (0.52 LB max, 1.20 UB max per Table 11), and the rice-based-formula scenario reaches 0.30 µg/kg bw/day at the mean and 0.39 µg/kg bw/day for high consumers. The age-class exposure pattern (toddlers highest LB mean; infants and toddlers tied at 0.61 µg/kg bw/day UB max) is the relevant input for any HMT&C threshold work that segments by age. Table 5’s mean iAs of 146–148 µg/kg (95th 270) for rice cakes/wafers/crackers and Table 8’s 83–94 µg/kg for children’s biscuits/rusks/cookies are direct occurrence evidence for category-specific work.
• Courses: Foundational document for teaching European regulatory context for arsenic in rice. The 2021 vs. 2014 comparison (1.5–3× lower estimates once measured iAs replaces modelled-from-tAs) is a strong teaching point on why speciation methodology matters. The Table 1 species composition (only 9% of submissions are direct iAs; 86% are tAs) illustrates the size of the speciation gap in European monitoring infrastructure.
• App: Main dietary iAs contributors for European population: rice > rice-based products > grain-based products (no rice) > drinking water. Relevant weighting for ingredient-list scoring of rice-containing products. Drinking-water mean of ~2 µg/L tap water (LB) is a useful baseline. The 22%-of-infant-cereal-above-ML signal is high-value as an occurrence weighting for the infant-cereal category.
• Microbiome: Not primary topic.

Verification notes

• Merge-enhanced 2026-06-03 against the manual-fetch PDF at raw/Manual Fetch Kimi /June 3 Folder/EFSA Journal - 2021 - - Chronic dietary exposure to inorganic arsenic.pdf (5,796,506 B), which is byte-near-identical to the previously-recorded raw/reports/efsa2021-chronic-exposure-iAs.pdf (5,796,501 B; same DOI, same content, 5 B PDF-metadata delta). Recorded in near_duplicates; existing sha256 for the raw/reports file preserved.
• Verified every prior Key number against the PDF: 13,608 results / 7,623 water / 5,985 food (Summary p.3); 67,401 starting set with 9.0% iAs / 86.1% tAs (Table 1 p.8); 66,025 post-cleaning (p.8); 51 of 2,621 rice samples > ML, rice brown n=15 max 570 µg/kg, rice popped n=18 max 409 µg/kg (Table 2 p.9 and surrounding paragraph); ~22% of cereal-based infant food > 100 µg/kg (p.10, 50 of 224 samples); LOQ cut-offs 200 µg/kg tAs / 100 µg/kg iAs food / 10 µg/L water (p.9); LB/UB toddler/infant means and 95th percentiles (Summary p.3); rice-based-formula scenario 0.30 mean / 0.39 high consumer (p.4); 1.5–3× lower than 2014 (p.3). All verified.
• Expanded Key numbers with Table 4 (iAs by rice type), Table 5 (iAs in rice-based products), Table 7 (drinking water by type), and Table 8 (infant/young-children food) statistics, none of which were on the prior revision; this strengthens the page as direct evidence for downstream rice and rice-based-product synthesis.
• Fixed broken regulation wikilink: [[regulations/efsa-contam-bmdl01-arsenic-2009]] (no such page) → [[regulations/efsa-arsenic-contam-2009]] (canonical slug, verified in wiki/regulations/).
• Replaced legacy body wikilink [[products/infant-rice-cereal]] with the frontmatter-consistent [[products/baby-cereals-dry-rice-based]] for routing alignment; both product pages exist, but the frontmatter taxonomy uses baby-cereals-dry-rice-based and the body should match what the routing audit fans out from frontmatter.
• Migrated legacy section heading ## Wiki pages updated on ingest → ## Wiki pages this source may touch per the merge-enhance defect list in docs/gpt-collaboration/verification-checklist.md.
• Added sample_population and near_duplicates fields to frontmatter; both were absent on the prior revision.
• No brand names appear in this source (EFSA aggregate exposure assessment, not a brand-level survey); Part 12 brand firewall not triggered.
• Implications section reports the page’s contribution to threshold work without proposing HMTc threshold values; Part 2 wiki/HMTc firewall observed.
• Audit subagent (2026-06-03, general-purpose fresh-context) flagged ❌ that the prior wording “95th percentile LB exposure: toddlers 0.58 µg/kg bw/day; infants 1.20 µg/kg bw/day” mis-attributed the infant figure to LB; verified against Table 11 p.24 — confirmed: the 1.20 µg/kg bw/day is the infant 95th UB max, while infant 95th LB max is 0.52. Page 25 disambiguates the Summary’s ambiguous “(LB–UB) … in toddlers and infants, respectively” wording: “highest 95th percentile exposure at the LB was estimated in toddlers (0.58 µg/kg bw per day) and at the UB in infants (1.20 µg/kg bw per day).” Corrected the Key numbers line to track Table 11 explicitly and revised the dependent Implications sentence (which had read “fourfold above the BMDL₀₁ lower bound”) to PDF-aligned framing (“within the BMDL₀₁ range”).
• Audit subagent flagged ⚠️ that Table 7 unit label “µg/kg” is an EFSA labeling slip for drinking water (Table 6 in the same PDF uses µg/L); verified by direct PDF read — confirmed. Added a parenthetical noting the unit context in the Key numbers Table 7 line.
• Audit subagent flagged ⚠️ “priority population / less urgent” framing in Implications as leaning toward HMTc policy language; verified against the source — agreed, the source itself describes the age-class pattern but does not prioritize categories. Replaced with neutral age-class-pattern description.
• Audit subagent flagged ⚠️ “consumer-facing risk lens” phrasing in the App bullet; replaced with “occurrence weighting” to track app-input semantics without consumer-advisory framing.
• Audit subagent flagged ⚠️ matrices: dietary-intake as more analysis-type than sample-matrix; left as-is — false-positive against this repo’s matrices vocabulary, which already admits dietary-intake across the EFSA exposure-assessment corpus and the value was on the page pre-merge.
• Audit subagent verdict REVISE; one ❌ applied with PDF citation, four ⚠️ adjudicated (three applied as text revisions, one left as-is with rationale); zero QUARANTINE triggers.

Wiki pages this source may touch

• arsenic
• rice
• baby-cereals-dry-rice-based
• rice-bulk-grain
• other-grain-products
• eu-2015-1006-iAs-rice
• efsa-arsenic-contam-2009

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.