EFSA CONTAM Panel 2024 — Risk assessment of small organoarsenic species in food

This 89-page EFSA Scientific Opinion, adopted 4 June 2024, is the first comprehensive EFSA risk assessment of small (mono- and dimethylated) organoarsenic species in food, completing the 2009 Scientific Opinion on arsenic in food where these species had been set aside for lack of data. The CONTAM Panel derived BMDL₁₀ reference points for monomethylarsonic acid MMA(V) (18.2 mg MMA(V)/kg bw per day, equivalent to 9.7 mg As/kg bw per day, from decreased body weight with diarrhoea in rats) and for dimethylarsinic acid DMA(V) (1.1 mg DMA(V)/kg bw per day, equivalent to 0.6 mg As/kg bw per day, from urinary bladder tumour incidence in male rats), applied a margin-of-exposure (MOE) approach because key toxicological data are missing, and concluded that DMA(V) raises a health concern: MOEs are below 10,000 in many dietary surveys at the 95th percentile, especially in infants, toddlers and other children whose main dietary contributors are rice and fish meat. MMA(V) does not raise a health concern across the assessed scenarios. Tri-/tetra-methylated species (TMAO, TETRA) and the thio-analogues (thio-DMA, dithio-DMA, thio-MMA) could not be assessed for lack of toxicological data.

Key numbers

• Reference Points (RPs) selected by the CONTAM Panel:
  • MMA(V): BMDL₁₀ = 18.2 mg MMA(V)/kg bw per day (equivalent to 9.7 mg As/kg bw per day) — decreased body weight with diarrhoea, 2-year rat study (Arnold et al. 2003), BMR 10%.
  • DMA(V): BMDL₁₀ = 1.1 mg DMA(V)/kg bw per day (equivalent to 0.6 mg As/kg bw per day) — urinary bladder tumours in male rats, 2-year study (Wei et al. 1999), BMR 10%.
• Decision rule on MOEs: MMA(V) MOE ≥ 500 considered not of health concern (default UF 100 × additional UF 5 for database deficiencies); DMA(V) MOE ≥ 10,000 considered of low concern (genotoxic and carcinogenic substance criterion from EFSA Scientific Committee 2005).
• Occurrence data set (post-cleaning): 1,260 analytical results on DMA(V) (44% left-censored) and 988 on MMA(V) (94% left-censored), all expressed as µg As/kg whole-weight; sampling years 2012–2022; eight EU countries for DMA, four for MMA; Italy dominant (67% of DMA results, 85% of MMA results). 654 DMA and 653 MMA results lacked the µg As/kg vs µg DMA(MMA)/kg unit declaration and were assumed elemental µg As/kg per data-provider clarification.
• DMA(V) by FoodEx2 Level 1 (Table 8, µg As/kg, LB–UB weighted mean / 95th percentile): Grains and grain-based products n=533, %LC 15, mean 30.4–33.1, P95 75.7–75.7; Vegetables and vegetable products (includes seaweeds) n=174, %LC 38, mean 127.2–133.2, P95 387.3–387.3; Fish/seafood/amphibians/reptiles/invertebrates n=227, %LC 41, mean 58.4–62.9 (P95 not calculated because category contains pooled samples); Fruit and fruit products n=11, %LC 91, mean 140.0–156.5; Composite dishes n=10, %LC 50, mean 417.5–427.5; Food products for young population n=18, %LC 44, mean 14.4–19.4; Products for non-standard diets/imitates/supplements n=26, %LC 88, mean 76.0–89.8.
• DMA(V) by sub-commodity (Table 9, µg As/kg, LB–UB weighted mean): rice grain polished n=221, LB=UB=40.0; rice grain brown n=110, LB=UB=27.8; rice and similar (overall pool) n=439 LB–UB 33.5–34.1; rice flour (439 plus 2 own samples) LB–UB 39.3–40.0; rice popped LB–UB 39.2–39.9 (HRP P95 92.7); puffed/extruded rice textured bread n=31, LB–UB 32.6–43.5 (HRP 50.2); processed rice-based flakes LB–UB 39.2–39.9 (HRP 71.6); rice drink LB–UB 4.6–4.7 (HRP 10.9); rice porridge LB–UB 5.9–6.0 (HRP 13.9); noodle rice n=6, LB=UB=26.8.
• DMA(V) in fish (Table 9): marine fish n=34, LB–UB 36.5–45.0 (HRP 12.0–20.0); diadromous fish n=19, LB–UB 4.9–10.3 (all 10 Atlantic salmon samples quantified); unspecified fish (meat) n=67, LB–UB 19.9–28.3; molluscs n=102, LB–UB 63.0–66.9 (HRP 170); canned/marinated/pickled seafood n=123, LB–UB 53.0–57.1 (HRP 170); fish fingers (breaded) n=8, LB–UB 3.2–6.4.
• DMA(V) in algae and prokaryotes (Table 9): wet basis n=116, %LC 9, LB–UB 143–145, HRP 449; dried basis n=112, %LC 10, LB–UB 605–615, HRP 1,706; algae-based formulations (e.g. Spirulina, chlorella) n=19, LB–UB 104–118.
• DMA(V) in food products for young population (Table 9): biscuits/rusks/cookies for children (rice cake identified) n=7, mean LB–UB 22.1–25.2 (HRP 23.5); cereals for infants or children, reconstituted n=7, LB–UB 1.9–2.5.
• Outlier salad samples: three samples of salads containing seaweeds, LB=UB=1,356 µg As/kg — highest single category DMA value in the dataset.
• MMA(V) occurrence (Table 10): 94% of analytical results left-censored. Fish/seafood/amphibians/reptiles/invertebrates n=152, %LC 53, mean LB–UB 46.3–53.4, P95 251 (driven almost entirely by molluscs; 411 rice samples and 35 fish samples all left-censored). Grains and grain-based products n=463, 100% LC, mean LB–UB 0.0–6.0.
• MMA(V) supplemented with literature (Table 11, µg As/kg LB or LB–UB mean): unspecified fish meat (Leblanc 2006) n=415, mean 22.0; miscellaneous demersal marine fishes n=35, mean 87.7; mackerel n=20, mean 78.0; cods/hakes/haddocks n=130, mean 5.1; mussels n=48, 54.0–61.5; oysters n=27, mean 227 (all quantified); canned herring (Hackethal 2021) n=20, mean 152; smoked herring n=20, mean 397; canned mackerel n=1, 82.0; fish fingers breaded n=2, 2.7.
• Chronic dietary exposure to DMA(V) (Table 12, µg As/kg bw/day across 49 European surveys, LB–UB):
  • Infants (n=12 surveys): mean min–median–max 0.018–0.035–0.059 LB / 0.021–0.044–0.074 UB; P95 mean 0.069–0.100–0.251 LB / 0.080–0.119–0.308 UB.
  • Toddlers (n=15): mean 0.019–0.034–0.130 LB / 0.022–0.042–0.157 UB; P95 0.067–0.142–0.397 LB / 0.074–0.156–0.477 UB.
  • Other children (n=19): mean 0.012–0.026–0.084 LB / 0.016–0.032–0.098 UB; P95 0.058–0.097–0.258 LB / 0.065–0.108–0.310 UB.
  • Adults (n=22): mean 0.005–0.013–0.038 LB / 0.006–0.015–0.044 UB; P95 0.021–0.058–0.109 LB / 0.025–0.067–0.136 UB.
• Highest DMA(V) chronic dietary exposure: one Toddlers survey, LB–UB mean 0.130–0.157 µg As/kg bw/day; LB–UB P95 0.397–0.477 µg As/kg bw/day.
• Main DMA(V) food contributors at the LB (Table 13, median % of mean exposure): Fish meat 13% Infants, 29% Toddlers, 30% Other children, 31% Adolescents, 31% Adults, 44% Elderly, 47% Very elderly, 23% Pregnant women; Rice 18% Infants, 35% Toddlers, 43% Other children, 44% Adolescents, 32% Adults; Reconstituted infant/child cereals 33% Infants (max 72%); Molluscs up to median 4% Adults / max 46% Adolescents; Dried/smoked/salted fish up to median 6% Elderly / max 27%.
• Rice-based infant formula scenario (Table 14, hydrolysed-rice powder from Meyer et al. 2018; n=7, DMA range 2–19 µg As/kg d.m., mean 7.6 µg As/kg d.m.; 1:8 reconstitution; consumption 200 mL/kg bw/day mean and 260 mL/kg bw/day high per EFSA SC 2017 defaults for infants <16 weeks): Scenario 1 (mean DMA in formulas) 0.19 mean / 0.25 high exposure µg As/kg bw/day; Scenario 2 (maximum DMA in formulas) 0.48 mean / 0.62 high exposure.
• Rice-consumer scenario (Table 15, consumers only, µg As/kg bw/day LB–UB): Other children highest mean 0.076–0.078 and highest P95 0.186–0.190; Toddlers 0.038–0.039 mean median / 0.086–0.087 P95 median.
• Fish-meat-consumer scenario (Table 16, consumers only, LB–UB): Toddlers highest mean 0.165–0.205 and highest P95 0.415–0.511 µg As/kg bw/day.
• MMA(V) exposure (Table 17, µg As/kg bw/day, consumers only): Fish meat highest infants 0.006–0.124 mean / 0.013–0.342 P95; processed/preserved fish highest in elderly 0.002–0.098 mean / 0.140–0.342 P95; molluscs adult P95 up to 0.185.
• DMA(V) MOEs (Table 18, RP 0.6 mg As/kg bw/day):
  • Mean exposure MOEs range 150,000 (lowest min LB) to 3,800 (highest max UB) across surveys; P95 exposure MOEs range 35,300 (lowest min LB) to 1,300 (highest max UB).
  • Infants maximum-exposure MOE: 10,200 LB / 8,100 UB mean; 2,400 LB / 1,900 UB P95.
  • Toddlers maximum-exposure MOE: 4,600 LB / 3,800 UB mean; 1,500 LB / 1,300 UB P95.
  • Other children maximum-exposure MOE: 7,100 LB / 6,100 UB mean; 2,300 LB / 1,900 UB P95.
  • Adolescents maximum-exposure MOE: 12,500 LB / 11,100 UB mean; 4,500 LB / 3,800 UB P95.
  • All age classes from Infants through Elderly fall below the MOE 10,000 threshold at maximum average exposure (except Adolescents, Adults, Very elderly UB averages) and at maximum P95 exposure; only Lactating women and Vegetarians P95 stay above 10,000.
• DMA(V) MOEs in the rice-based infant formula scenario: 3,200 (mean DMA + mean consumption) → 2,400 (mean DMA + high consumption) → 1,250 (max DMA + mean consumption) → 970 (max DMA + high consumption). All below 10,000.
• DMA(V) MOEs in rice-consumer scenario: 200,000 (min LB average) down to 7,700 (max UB average); 75,000 (min LB P95) down to 3,200 (max UB P95). Lower than 10,000 at maximum average exposures for Toddlers/Other children and at maximum P95 in all groups except Lactating women and Vegetarians.
• DMA(V) MOEs in fish-consumer scenario: 100,000 (min LB average) down to 2,900 (max UB average); 33,300 (min LB P95) down to 1,200 (max UB P95). Lower than 10,000 at maximum average exposures in Infants, Toddlers, Other children, Adolescents, Adults and Elderly, and at maximum P95 in all groups.
• MMA(V) MOEs (Table 19, RP 9.7 mg As/kg bw/day): fish-meat consumers 3,233,000 (min average) down to 28,000 (max P95); molluscs consumers 9,700,000 down to 52,000 (P95 max); processed/preserved fish consumers 9,700,000 down to 28,000 (max P95). All ≥ 500 → no health concern.
• Thio-DMA(V) and dithio-DMA(V) in puffed rice cakes (Colina Blanco 2024, n=80): DMA(V) 0–0.11 mean 0.013; thio-DMA(V) 0–0.26 mean 0.034; dithio-DMA(V) 0–0.34 mean 0.037 µg As/g d.m. — when rice is heated to make puffed cakes the original DMA(V) is mostly converted to dithio-DMA(V) and to a lesser extent thio-DMA(V); on 3-month storage dithio-DMA(V) re-converts to thio-DMA(V) and DMA(V). Toxicology of thio-analogues not assessed in this Opinion.
• DMA(V) in algae compositional context: brown algae carry the highest DMA(V) (Garcia-Salgado 2012 reported DMA up to 0.9 µg As/g d.m.; Taylor & Jackson 2016 reported up to 0.97 µg As/g d.m. in New England brown algae from 4–106 µg As/g d.m. tAs); green algae lowest. EFSA notes Italy provides 95% of EU-reported rice samples and Japonica/Indica split (75/25 EU production; Indica mostly imported from Myanmar/Pakistan/Thailand/Cambodia/India).
• DMA(III) is reactive, difficult to measure, and not detected in foods (Molin et al. 2015); only the pentavalent DMA(V) is stable in food matrices and is the reported quantity.
• Genotoxicity: DMA(V) shows in vivo evidence of oxidative DNA damage and DNA breaks in rodents and is a rodent carcinogen (urinary bladder); mechanisms of genotoxicity in carcinogenicity are unclear. MMA(V) is inadequately tested and there is no convincing in vivo evidence of carcinogenicity. MMA(III) and DMA(III) are clearly genotoxic in vitro but no in vivo genotoxicity studies are available.
• No data for risk assessment on TMAO, TETRA, thio-DMA(V), dithio-DMA(V), or thio-MMA(V); these species could not be included in the risk characterisation.
• Mandate context: Commission Recommendation (EU) 2015/1381 instructed Member States to monitor inorganic and total arsenic and, when possible, other relevant arsenic species in food during 2016–2018 and to provide data to EFSA by October 2018. This Opinion is one of three EFSA outputs responding to the 2024 mandate (the others: updated iAs risk assessment, EFSA Journal 2024 8488; forthcoming opinion on arsenobetaine, arsenosugars and arsenolipids).

Methods (brief)

Occurrence data: EFSA Data Warehouse extraction of 2,766 analytical results on DMA (n=1,524) and MMA (n=1,242) on food for sampling years 2012–2022 (Annex E). Data cleaning removed 248 duplicated TDS results (124 DMA + 124 MMA), 248 prepared-TDS results, 11 non-confirmed high values (5 MMA + 6 DMA), 2 reported-twice results, 9 DMA suspect-sampling samples; further 264 DMA + 254 MMA samples excluded during cleaning. Final set: 1,260 DMA and 988 MMA analytical results, all expressed as µg As/kg whole-weight. Recovery information missing or correction-status unspecified for 36% of DMA and 22% of MMA results. Of the final set, 654 DMA and 653 MMA results lacked the µg As/kg vs µg DMA(MMA)/kg unit declaration and were assumed elemental µg As/kg per data-provider feedback. No LOQ cut-off applied because for many food categories 100% of MMA data were left-censored and applying a cut-off would have eliminated the category.

Analytical methods reported (final set): DMA detection by mass spectrometry in 87% of cases, predominantly ICP-MS coupled to liquid chromatography (72%) or ICP-MS alone, lowest LOQ 2 µg As/kg in a fish sample; MMA detection by mass spectrometry in 96% of cases, LC-ICP-MS in 90%, lowest LOQ 1.6 µg As/kg in a processed/preserved fish sample. Sampling strategy: 55–62% Selective sampling, 12% Objective sampling.

Occurrence values for processed commodities not present in the DMA submission were derived from raw-primary-commodity data via the EFSA Raw Primary Commodity model (EFSA 2019), assuming no DMA losses during processing — applied to rice-and-meat meal, rice-and-vegetables meal, rice pudding, rice-meat-vegetables meal with seaweed, rice flakes, rice flour, rice porridge, rice popped, rice drink, fish soup with rice, fish balls, fish gratin, fish pâté, fish soup, prepared fish salad, and breaded fish fingers (derived from cods/hakes/haddocks). MMA occurrence values for fish meat and processed/preserved fish were derived from external literature (CALIPSO Leblanc 2006; Hackethal 2021; Zmozinski 2015) because EFSA-submitted MMA samples on rice and fish were essentially all left-censored.

Consumption data: EFSA Comprehensive European Food Consumption Database — 49 dietary surveys, 22 European countries, including dedicated population groups Infants, Toddlers, Other children, Adolescents, Adults, Elderly, Very elderly, Pregnant women, Lactating women, Vegetarians. Chronic dietary exposure calculated LB–UB per standard EFSA methodology; survey/age-class P95 estimates not reported when fewer than 59 observations available. Specific exposure scenarios for rice consumers and fish-meat consumers (“consumers only”) followed the EFSA approach to high-frequency commodity-specific intake. Rice-based infant formula scenario: hydrolysed-rice powder DMA data from Meyer et al. (2018), n=7, reconstituted 1:8, consumed at EFSA Scientific Committee 2017 default volumes for infants <16 weeks (200 mL/kg bw/day mean, 260 mL/kg bw/day high).

Risk characterisation: BMD modelling using model averaging at BMR 10% for dose–response selection. MOE benchmark of 500 for MMA(V) (default 100 inter/intra-species × additional 5 for database deficiencies per WHO 1994/1999 and EFSA SC 2012 guidance); MOE benchmark of 10,000 for DMA(V) per EFSA SC 2005 guidance for genotoxic carcinogens. Uncertainty analysis quantified probability that MOE exceeds 10,000 at mean and P95 exposures for DMA(V) (Section 3.9.2). Combined exposure across small organoarsenic species not assessed because grouping is not supported by available data.

Limitations

Toxicological database is incomplete for both compounds: reproductive toxicity and neurotoxicity data are missing, and the additional UF of 5 for MMA(V) reflects this. Mechanisms of DMA(V) genotoxicity in carcinogenicity (DNA-reactive vs threshold) are unclear — the Panel notes the MOE 10,000 benchmark may be too conservative if genotoxicity is not the driving mode of action, but in absence of mode-of-action evidence the conservative approach is retained. No in vivo genotoxicity data for MMA(III), DMA(III), or any thio-analogue. No data allow estimation of absorption rates for small organoarsenic species other than MMA(V) and DMA(V). No biomarker for human exposure to small organoarsenic species exists.

Occurrence dataset is heavily Italy-weighted (67% DMA, 85% MMA), constraining EU-representativeness; sampling strategy was predominantly Selective (55–62%); for many food categories MMA data are 100% left-censored, so LB–UB ranges differ markedly. Thio-DMA(V) is being underestimated in foods owing to conversion to DMA(V) during chemical analysis (Dai et al. 2022), meaning reported DMA(V) is likely an overestimate. Conversely, when rice is processed into puffed cakes the DMA(V) reported by the laboratory likely under-represents the actual thiolated species present in the consumed product (Colina Blanco 2024). Processing factors for derived occurrence values assume no DMA losses, which is unverified.

Dietary surveys for Infants and Toddlers are limited in number and sample size, leaving the highest-exposure subpopulations the most uncertain. No EU monitoring data on small organoarsenic species in breast milk, despite the documented placental transfer of DMA(V) and detection in breast milk in animal/human studies. Combined exposure to multiple small organoarsenic species was not assessed because grouping criteria are not satisfied by available toxicological evidence.

Implications

• Certification: This is the first EU regulatory-grade risk characterisation for organic arsenic species relevant to HMT&C scope. The Panel concludes that DMA(V) exposure raises a health concern at 95th-percentile exposures across Infants, Toddlers, Other children, Adolescents, Adults and Elderly via rice and fish, and at all exposure levels in the rice-based infant formula scenario (MOEs 970–3,200, all below 10,000). MMA(V) does not raise a concern. The rice-based infant formula scenario MOE of 970 (max DMA + high consumption) is directly relevant to any HMT&C work on rice-protein infant formulas. Occurrence values for rice grain polished (LB=UB 40.0 µg As/kg) and brown rice (LB=UB 27.8 µg As/kg) are direct evidence for rice-based product categories. Algae values up to 605–615 µg As/kg LB–UB (dried) feed seaweed/seaweed-kelp-foods occurrence baselines. The Opinion supplies the first EU regulatory-grade occurrence and toxicology dataset on DMA(V) in rice-based infant formula and seaweed.
• Courses: Foundational 2024 document for teaching organoarsenic toxicology and the methylation pathway. The two reference points (MMA(V) 9.7 mg As/kg bw/day and DMA(V) 0.6 mg As/kg bw/day) and their decision-rule thresholds (MOE 500 vs 10,000) demonstrate how database-deficiency uncertainty factors interact with the genotoxic-carcinogen criterion. The thio-analogue analytical issue (Colina Blanco 2024: thio-DMA forming in heated/stored puffed rice cakes; Dai et al. 2022: thio-DMA underestimated by standard methods) is a strong teaching case on speciation-method limitations. The pairing of this 2024 Opinion with the 2024 iAs update and the forthcoming arsenobetaine/arsenosugar/arsenolipid Opinion is the current EFSA-arsenic curriculum spine.
• App: For ingredient-list scoring of rice-containing products, this Opinion supplies organoarsenic occurrence weighting in addition to the inorganic-arsenic signal already in place. Rice grain polished (40 µg As/kg DMA) and brown rice (27.8 µg As/kg) are the relevant point estimates. Rice-based infant formula carries a documented MOE below 1,000 in the high-DMA-high-consumption scenario, which is a high-priority signal for infant-product scoring. For fish-containing baby foods and other fish/seafood products, marine fish DMA(V) means of 36.5–45.0 µg As/kg and mollusc means of 63.0–66.9 µg As/kg are the relevant baselines. For seaweed-containing products, dried algae DMA(V) up to 605–615 µg As/kg LB–UB drives a strong organoarsenic exposure signal.
• Microbiome: Not primary topic. The Opinion notes presystemic thiolation and reduction of small organoarsenic species and references in vitro evidence (without identifying gut microbial pathways); relevant cross-link is the broader gut-arsenic methylation literature already on HMI.

Verification notes

• Wrote 2026-06-03 from a fresh read of the EFSA 2024 Scientific Opinion PDF at raw/Manual Fetch Kimi /June 3 Folder/EFSA Journal - 2024 - - Risk assessment of small organoarsenic species in food.pdf (SHA256 25e6a3c…3e3a). No prior wiki source page existed for DOI 10.2903/j.efsa.2024.8844 (confirmed by grep over wiki/); the sibling efsa2024-risk-assessment-iAs-update.md is a different 2024 EFSA Opinion (DOI 10.2903/j.efsa.2024.8488, inorganic arsenic).
• Verified all reference points, MOEs, and occurrence values against PDF pages 5–7 (Summary), p.49 (RP derivation), p.52–57 (Tables 8–11), pp.59–64 (Tables 12–17), pp.65–68 (Tables 18–19), and p.62 Table 14 (rice-based infant formula scenario). Per-table footnotes preserved (HRP = highest reliable percentile, Table 9; 95th percentile reported only when n ≥ 59, Tables 8/10/12).
• metals: [tAs] selected because the paper is about organoarsenic forms (MMA, DMA) of total arsenic; the wiki abbreviation vocabulary does not have separate iAs/oAs slugs, and iAs would be incorrect because this Opinion explicitly excludes inorganic arsenic from its scope (which is covered by the companion 2024 EFSA Opinion).
• products: includes the rice-based infant formula scenario (umbrella infant-formula-powder) because the Opinion derives a dedicated scenario with documented MOEs for it. The narrower scope infant-formula-powder-non-soy was avoided per the verification checklist’s “scope is broad, not narrow” rule, since hydrolysed-rice formulas are neither dairy nor soy-based and the umbrella covers all powder forms.
• No brand names appear in the body or Key numbers (EFSA aggregate Opinion; brand names not used in EFSA outputs). The single literature reference to a commercial seaweed dietary supplement category (Spirulina, chlorella) names ingredient categories, not brands.
• Implications report contribution to HMT&C scope and app weighting without proposing certification thresholds; Part 2 wiki/HMT&C firewall observed.
• Audit subagent (2026-06-03, general-purpose fresh-context) verdict PROMOTE with one ⚠️ Check 5 finding on the Implications Certification bullet (“strengthens the case for monitoring DMA(V) separately…” reading as a directive to HMT&C on analyte scope). Verified against Part 2 rules — agreed, this is exactly the directional language Part 2 prohibits. APPLIED: reworded to neutral literature-contribution framing (“The Opinion supplies the first EU regulatory-grade occurrence and toxicology dataset on DMA(V) in rice-based infant formula and seaweed”). All Check 1 numerical fidelity, Check 2 slugs, Check 3 speciation/methods, Check 4 brand firewall ✅ clean.

Wiki pages this source may touch

• arsenic
• arsenic-total
• rice
• fish
• molluscs
• seaweed
• baby-cereals-dry-rice-based
• rice-bulk-grain
• other-grain-products
• rice-snacks-crackers
• rice-beverages-rice-milk
• infant-formula-powder
• fresh-fish
• canned-fish
• shellfish
• seafood
• seaweed-kelp-foods
• 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.