Neurodevelopmental effects of methylmercury (MeHg): a review of epidemiological points of departure (PoDs), toxicological reference values (TRVs), and key uncertainties in human health risk assessment

Overview

Blechinger, Singh, Afghan, and Smith review how major risk-assessment organizations have derived points of departure (PoDs) and toxicological reference values (TRVs) for neurodevelopmental effects after prenatal methylmercury exposure. The paper is a Health Canada-authored peer-reviewed review in Archives of Toxicology that compares US EPA, ATSDR, Health Canada, JECFA, and EFSA approaches, then runs sensitivity analyses against published meta-analyses and Iraq-poisoning neurological-score data. It is toxicology and risk-assessment context, not seafood occurrence evidence.

Key numbers

Publication and review scope: received 2025-11-04, accepted 2026-02-12, published online 2026-03-10; Archives of Toxicology volume 100, pages 2191-2219; DOI 10.1007/s00204-026-04345-8.

Historical exposure anchors: congenital Minamata Disease is described as having up to 66 reported cases from the Japanese poisoning incidents. The Iraq poisoning incident covered 1971-1972; the 1995 US EPA RfD derivation used 81 Iraqi mother-child pairs and peak maternal-hair total Hg (MH THg) exposures ranging from 1-674 µg/g.

1990 WHO framing: the review quotes WHO’s interpretation that peak maternal hair mercury above 70 µg/g was associated with a high risk, more than 30%, of neurological disorder in offspring, while a prudent reading of the Iraq data associated 5% risk with peak maternal hair mercury of 10-20 µg/g.

Table 1 - organization-level PoDs and TRVs: the review compares these oral TRVs for prenatal MeHg neurodevelopmental effects:

Table 1 dosimetry parameters: blood-to-oral conversion used oral intake of MeHg in µg/kg-bw/day = [MB THg (µg/L)] * [(B*V)/(A*F*BW)]. Reported conversion factors were ATSDR 0.0150, Health Canada 0.0246, JECFA 0.0265, EFSA 0.0265, and US EPA 0.0186. Oral absorption was 0.94 for ATSDR and 0.95 for the other organizations. Body weights were 60 kg for ATSDR and Health Canada, 65 kg for JECFA and EFSA, and 67 kg for US EPA.

Table 2 - Faroe Islands BMD modelling values selected for US EPA comparison: Boston Naming Test cue BMDL values were 58 µg/L CB THg for the whole cohort and 71 µg/L CB THg after PCB adjustment, corresponding to oral BMDLs of 1.081 and 1.323 µg/kg/day MeHg and rounded RfDs of 0.1 µg/kg/day. Continuous Performance Test reaction-time values were 46, 49, and 28 µg/L CB THg for whole cohort, PCB-adjusted, and lowest-PCB analyses, with oral BMDLs 0.857, 0.913, and 0.522 µg/kg/day; the lowest-PCB 0.522 value was treated as a deviation corresponding to an RfD of 0.05 µg/kg/day. Finger-tapping preferred-hand values were 79, 66, and 24 µg/L CB THg for whole cohort, PCB-adjusted, and lowest-PCB analyses, with oral BMDLs 1.472, 1.230, and 0.447 µg/kg/day; the 0.447 value was also treated as a low-end deviation. CVLT delayed recall whole-cohort CB BMDL was 103 µg/L, oral BMDL 1.920 µg/kg/day, and rounded RfD 0.2 µg/kg/day, treated as a high-end deviation.

Table 3 - early cohort data coverage: New Zealand contributed 23 assessed outcome tests, 5 with reported regression coefficients, 18 without reported coefficients, 4 coefficients included in Axelrad et al. 2007, 23 tests eligible for BMD modelling with participant data, and 5 BMD estimates reported. Faroe Islands Cohort 1 contributed 17 assessed tests, 17 coefficients reported, 6 coefficients included in Axelrad et al. 2007 but collapsed to 4 entries after compositing WISC full-scale IQ, 17 tests eligible for BMD modelling, and 5 BMD estimates reported. Seychelles Main Cohort at 5.5 years contributed 6 assessed tests, 6 coefficients, no Axelrad inclusion, 6 tests eligible for BMD modelling, and 6 BMD estimates. Seychelles Main Cohort at 9 years contributed 22 assessed tests, 22 coefficients, and 5 coefficients included in Axelrad et al. 2007. Across the rows, the table totals were 68 assessed outcome tests, 50 with regression coefficients, 18 without coefficients, 62 eligible for Axelrad meta-analysis, 15 coefficients included before compositing (13 after compositing), 46 tests eligible for BMD modelling, and 16 BMD estimates reported.

Table 4 - biomarker and oral-TRV comparison: reported ranges were MH PoDs 5.6-14 µg/g THg, MH TRVs 0.9-2.2 µg/g THg, cord-blood PoDs 46-79 µg/L THg, cord-blood TRVs 4.6-7.9 µg/L THg, maternal-blood PoDs 28-79 µg/L THg, maternal-blood TRVs 4.6-9.2 µg/L THg, conversion factors 0.0150-0.0265, oral PoDs 0.41-1.48 µg/kg-bw/day MeHg, and oral TRVs 0.1-0.23 µg/kg-bw/day MeHg.

Sensitivity analyses: the review recalculated supplementary PoDs from published meta-analyses and Iraq data. Axelrad et al. 2007 yielded BMDL05 values 16.5 µg/g MH THg using the central pooled coefficient -0.18 IQ points per 1 µg/g MH THg and a z-distribution 90% LCL of -0.331, or 15.8 µg/g with a t-distribution 90% LCL of -0.344. Frequentist validation using {metafor} gave 90% LCLs -0.502 and -0.481, corresponding to BMDL05 values 10.9 and 11.3 µg/g MH THg. Kopylev and Segal 2025 Boston Naming Test pooled coefficients corresponded to BMDL05 values 6.8-9.2 µg/g MH THg, while their language/verbal subdomain coefficient produced a much higher BMDL05 of 88.6 µg/g MH THg. Coull et al. 2003 reported pooled BMD 21.0 µg/g and BMDL 7.7 µg/g MH THg. Updated Bayesian model-average BMD modelling of abnormal child neurological scores in the Iraq poisoning data yielded BMDL05 values 12.9 and 17.5 µg/g MH THg depending on whether abnormal neurological score was defined as >=4 or >=5.

Uncertainty discussion: the review reports that the 16 BMD estimates in NAS 2000 represented 34.8% of the 46 total eligible outcome tests in the three early cohorts, and that Axelrad et al. 2007 included 15 of 62 eligible coefficients (24.2%) before compositing. The authors identify remaining uncertainties around reliance on older cohorts, selected subsets of test scores, limited access to individual participant data, benchmark-response selection, point-estimate biomarker conversion, and animal-to-human extrapolation. Candidate animal-data TRV ranges discussed were 0.01-0.05 µg/kg-bw/day from non-human primate data in US EPA 2001 and 0.009 and 0.04 µg/kg-bw/day from mouse and rat studies in ATSDR 2024, with the caveat that default uncertainty factors of 100-1000 drove those lower values.

Methods (brief)

This is a peer-reviewed review article plus sensitivity-analysis paper. The authors summarize MeHg neurodevelopmental PoDs and TRVs used by US EPA, ATSDR, Health Canada, JECFA, and EFSA; compare PoDs across maternal hair, cord blood, maternal blood, and oral-intake equivalents; and conduct supplementary PoD calculations from published meta-analyses and BMD modelling. The paper does not collect new food samples, measure mercury in fish, or generate a new HMTc product benchmark pool.

Speciation discipline: the review is about methylmercury toxicological reference values, but most human epidemiology exposure biomarkers were total mercury in maternal hair, maternal blood, or cord blood. The authors explicitly state that high percentages of total Hg in these biomarkers are MeHg (80-90%) and strongly correlated with MeHg, but the biomarker measurements remain total Hg. Food occurrence values are not reported.

Implications

Certification: This page documents literature-side toxicology and risk-assessment context for methylmercury, including how existing public-health TRVs were derived. It does not propose or justify any HMTc threshold and should not be used as a product-row percentile input.

Courses: Useful for teaching the distinction between measured exposure biomarkers, toxicological PoDs, uncertainty factors, and TRVs. The paper is also a compact map of why the Faroe Islands, Seychelles, New Zealand, and Iraq datasets recur in MeHg risk assessment.

App: Supports explanatory context for MeHg as a neurodevelopmental hazard, especially for prenatal exposure, but supplies no food-category concentration data and no consumer-product ranking.

Microbiome: No microbiome findings.

Wiki pages updated on ingest

• Methylmercury
• Mercury, Total

Verification notes

Identity checks. On 2026-07-27, DOI grep for 10.1007/s00204-026-04345-8, DOI-slug grep for s00204-026-04345-8, raw-handle grep for RPHG_10-1007-s00204-026-04345-8, checksum grep for 925d67f9713b31fb292477aec4532e79d16561473eab8eac65ddc1326b60ac0d, author grep for Blechinger, and cite-key collision checks found no existing wiki/sources/ page. This is a fresh source page.

Raw artifact. The PDF has 29 pages, file size 2066516 bytes, and SHA-256 925d67f9713b31fb292477aec4532e79d16561473eab8eac65ddc1326b60ac0d. pdfinfo reports Adobe InDesign/PDF Library metadata and no encryption. pdfimages -list found no embedded image objects; the tables are text/vector layout. Rendered checks of the Table 1, Table 2, Table 3, and Table 4 pages matched the extracted text used above.

Evidence tier. A. Although this is a review, it is a peer-reviewed risk-assessment review from Health Canada authors, compares authoritative agency/public-health derivations, and includes explicit sensitivity analyses. It is not primary food-occurrence evidence.

Firewall check. The page reports TRVs and PoDs because the article reports them. It does not translate those values into HMTc thresholds, certification pass/fail limits, clean/dirty product rows, or product-category percentiles. Products and ingredients are intentionally empty because the source does not measure seafood or any consumer product.

Number verification. Numbers in the Key numbers section were checked once against the extracted full-PDF text and a second time against the rendered table pages or the surrounding article text before commit. Table 1 values were checked against rendered PDF pages 5-6; Table 2 against rendered page 9; Table 3 against rendered page 10; Table 4 against rendered page 17; the sensitivity-analysis values were checked against the Discussion and sensitivity-analysis text.

Update history

The five most recent substantive edits to this page, classified major (evidence or structure moved), correction (a published value or statement was wrong and has been fixed), or minor (narrative rewritten without changing the underlying evidence). Each description is derived from what the edit did to this page; the linked commit is the authoritative record, routine regeneration passes are excluded, and 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.