Global methylmercury exposure from seafood consumption and risk of developmental neurotoxicity: a systematic review

Abstract

Objective

To examine biomarkers of methylmercury (MeHg) intake in women and infants from seafood-consuming populations globally and characterize the comparative risk of fetal developmental neurotoxicity.

Methods

A search was conducted of the published literature reporting total mercury (Hg) in hair and blood in women and infants. These biomarkers are validated proxy measures of MeHg, a neurotoxin found primarily in seafood. Average and high-end biomarkers were extracted, stratified by seafood consumption context, and pooled by category. Medians for average and high-end pooled distributions were compared with the reference level established by a joint expert committee of the Food and Agriculture Organization (FAO) and the World Health Organization (WHO).

Findings

Selection criteria were met by 164 studies of women and infants from 43 countries. Pooled average biomarkers suggest an intake of MeHg several times over the FAO/WHO reference in fish-consuming riparians living near small-scale gold mining and well over the reference in consumers of marine mammals in Arctic regions. In coastal regions of south-eastern Asia, the western Pacific and the Mediterranean, average biomarkers approach the reference. Although the two former groups have a higher risk of neurotoxicity than the latter, coastal regions are home to the largest number at risk. High-end biomarkers across all categories indicate MeHg intake is in excess of the reference value.

Conclusion

There is a need for policies to reduce Hg exposure among women and infants and for surveillance in high-risk populations, the majority of which live in low-and middle-income countries.

Introduction

The World Health Organization (WHO) considers mercury (Hg) among the top 10 chemicals of “major public health concern”.¹ Evidence of ubiquitous Hg contamination globally led to the recent Minamata Mercury Convention, a binding international treaty to control anthropogenic Hg emissions.² A principal form of Hg to which general populations are exposed is methylmercury (MeHg). Transformation of Hg emissions to organic MeHg takes place in the aquatic environment, where MeHg bioaccumulates in food webs. In human beings MeHg exposure occurs predominantly through the consumption of seafood (including freshwater and marine varieties, shellfish and marine mammals).^3–6 MeHg is a neurotoxin particularly harmful to the developing fetal brain.^3–6 A large body of research has demonstrated an association of exposure in utero with developmental neurotoxicity (e.g. deficits in fine motor skills, language and memory) among populations that consume seafood regularly.^3,7–9 Such studies have been used to develop health-based reference doses below which no appreciable risk of harm is thought to occur, including the provisional tolerable weekly intake (PTWI), established by the Joint Expert Committee on Food Additivies (JECFA) of the Food and Agriculture Organization (FAO) and WHO.^6,10 Recent research suggests harm at doses associated with relatively infrequent seafood consumption.¹¹

Seafood species vary in MeHg content depending on contamination source, trophic level and other factors.^12–14 Seafood, on the other hand, is an important source of nutrients, including neuroprotective omega-3 polyunsaturated fatty acids.¹⁵ Research on the benefits and harms of seafood highlights the importance of choosing species low in MeHg and high in these polyunsaturated fatty acids and of ensuring that consumers have sufficient information to make such choices.^15,16 Well-designed seafood advisories can be helpful to this end,^17,18 but they exist in a small number of countries, most of which are high-income.¹⁹ An estimated 400 million women of reproductive age in the world rely on seafood for at least 20% of their intake of animal protein; a large share of them live in low- and middle-income countries where access to information on MeHg content in seafood is not widely available.^20–22 Although the research conducted in the last two decades has highlighted the risk in subsistence fishing communities that practise artisanal and small-scale gold mining²³ and among Arctic peoples whose diet consists of apex marine predators such as the pilot whale,²⁴ few researchers have compared MeHg exposures globally in women who consume seafood.

Human exposure to chemical contaminants can be characterized by examining biomarkers.²⁵ Total Hg in hair (THHg) and total Hg in blood (TBHg) are both validated biomarkers of MeHg intake correlated with seafood consumption in general human populations.^4,26 Our goal was to review and synthesize the evidence from published studies reporting THHg and TBHg biomarkers to systematically compare global MeHg exposure among women and their infants from seafood-consuming populations. By identifying populations at higher risk, we aim to provide policy-makers with scientific evidence for the prioritization of risk reduction messages and targeted population surveillance.

Methods

Based on a pre-defined study protocol,²⁷ we performed a systematic electronic search of the peer-reviewed scientific literature (Box 1). Studies were selected in two stages: title and abstract screening, followed by full text review after application of exclusion criteria. We excluded studies not involving women or infants from general populations and not reporting a central THHg or TBHg biomarker estimate. When multiple articles reported on a single sample, we chose the most recent one with complete data. To ensure robust summary statistics, we excluded studies with less than 40 participants.

Box 1. Literature search strategy for global systematic review of methylmercury exposure from seafood in women and infants.

1. “fetus” OR “infant” OR “newborn” OR “maternal” OR “mother” OR “pregnant” OR “women”

2. “fish” OR “marine” OR “shellfish” OR “seafood”

3. “mercury” OR “methylmercury” OR “methyl AND mercury” OR “biomonitoring”

Combined terms: 1 AND 2 AND 3.

We extracted data on study design, population characteristics, measures of average (geometric mean or median) and high-end (90th or 95th percentile or maximum) biomarkers, exposure conditions and main covariates examined. Extracted biomarkers were organized into three subpopulation groups: non-pregnant women; pregnant women and mothers who had recently given birth; and infants (up to 12 months of age). Because biomarkers for more than one subpopulation with different levels of exposure were often reported in the same study, the subpopulation was our main level of analysis.

We stratified subpopulations into six mutually exclusive categories based on predictors of the body burden of MeHg. The most important of these predictors are seafood consumption frequency and seafood MeHg content. In most seafood species MeHg represents the largest fraction of total Hg (inorganic Hg representing a much smaller share). Thus, seafood MeHg concentration is commonly measured as total Hg in tissue.^3,4 Seafood consumption estimates were reported in some studies; data on total Hg concentrations were rarely provided. Research suggests the following general hierarchy: marine mammals, other apex marine predators and some industrially-contaminated fish [containing several parts per million (ppm)]; large marine fish [containing up to 1 or more ppm]; most commercially purchased marine and freshwater fish [often containing less than 0.5 ppm] and  most shellfish [often containing less than 0.2 ppm].^{23,24,28–31} Seafood intake is generally higher in coastal regions than inland^30,32 and seafood from globalized commercial sources predominates in many urban areas.¹⁴ We therefore generated six categories based on the following proxy predictors, reported in most studies: seafood source; seafood type; likely Hg contamination pathway; and residential context. Four categories included populations consuming seafood that was mainly self-caught and two included populations consuming seafood that was commercially purchased primarily (Table 1).

Table 1. Methylmercury exposure categories^a for women and infants from seafood-consuming populations.

Category/subcategory	Predominant Hg pathway to seafood	Predominant seafood type	Seafood intake range (kg per month)b	Residential context
Locally self-caught seafood is important share of diet
Arctic – Traditional diet – Mixed diet	Unique polar meteorology and Hg deposition/mobilization, Arctic food-chain (marine mammals as apex predators)	Traditional: marine fish and marine mammals Mixed: marine fish and non-seafood protein sources, few if any marine mammals	0.6–7.1	Far northern Arctic, where people rely on apex Hg-contaminated marine mammals and fish
Gold mining – Rural riverine – Urban	Artisanal and small-scale gold mining, soil lixiviation, forest fires releasing Hg emissions	Rural: high share of locally-caught freshwater fish Urban: mixed diet including non-seafood protein, low share of locally-caught freshwater fish	0.6–14.9	Rural and urban tropical areas near artisanal and small-scale gold mining, where the diet includes fish from rivers contaminated by gold mining activity
Fishing	Local and general global transport of Hg emissions	Marine and freshwater fish and shellfish	0.1–3.8	Recreational or subsistence fishing areas near rivers, reservoirs or lakes without a particular Hg contamination source
Industry	Local Hg-emitting industry (chloralkali, power generation, mining other than gold mining)	Marine and freshwater fish and shellfish	0.2–5.8	Recreational or subsistence fishing areas near water bodies with active or disused industrial facilities
Seafood consumed is mostly from commercial sources (i.e. non-self-caught)c
Coastal – Atlantic – Mediterraneand – Pacific	Local and general global transport of Hg emissions in all three regions; natural Hg emission sources in the Mediterranean	Marine and freshwater fish and shellfish	0.3–5.6	Atlantic, Mediterranean or Pacific coastal areas where seafood intake is frequent
Inland	Local and general global transport of Hg emissions	Marine and freshwater fish and shellfish	Very little–2.0	Inland areas where seafood intake is low

Hg, mercury.

^a Exposure categories based on proxy predictors reported in selected studies.

^b Estimated per capita seafood intake ranges were derived from data reported in selected studies. They were converted to kg per month for comparability.

^c Several subpopulations consume an important share of self-caught marine seafood in addition to commercially-purchased varieties.

^d Because Indian Ocean and Persian Gulf subpopulations were not numerous and reported seafood intake and total Hg biomarkers similar to those of the more numerous Mediterranean subpopulations, the former were included with the latter.

As recommended in guidelines for the systematic review of observational studies,²⁷ we evaluated study quality by examining the risk of bias in three areas: selection of participants (selection methods and reporting of exposure characteristics); exposure measurement (laboratory methods and quality control); and statistical methods and covariate analysis (evaluation of distribution shape, reporting of seafood intake and exposure to non-seafood sources of Hg).

We derived two summary distributions – central and upper bound – for each exposure category by pooling average and high-end biomarkers. For comparability, all TBHg biomarkers were converted to THHg-equivalent at a hair-to-blood ratio of 250:1.^3,5 We summarized resulting statistical distributions using medians and percentiles. To interpret results, we compared distribution medians with the THHg-equivalent value of the PTWI dose (approximately 2.2 μg/g) established by the JECFA.¹⁰ We also determined the share of subpopulations with average and high-end biomarkers over this reference. In sensitivity analysis we evaluated the impact on pooled biomarkers taking into account differences in participant selection, exposure measurement and statistical methods identified in the quality review. Given substantial heterogeneity in population exposure conditions, study designs and reporting, we did not undertake a meta-analysis. All data analysis was performed in Stata10 (StataCorp, College Station, United States of America).

Results

Selected studies

Of 3042 articles identified in the published literature, we screened 1402 non-duplicates (1379 were identified by electronic search and 23 by hand search); we excluded 1120 and we reviewed the full texts of the remaining 282, from which we excluded 118 (Fig. 1). The remaining 164 articles, which reported total Hg biomarkers for 239 distinct subpopulations, were included in this review. Selected articles report biomarker concentrations for 63 943 women and infants from 43 countries (Table 2). Most (73%) studies were cross-sectional and over half (56%) reported THHg measures; the majority (79%) were published after 2001. Studies published in 1991–2001 were conducted primarily in populations consuming self-caught seafood; since 2001, the number of studies in consumers of seafood that is predominantly commercially purchased has increased notably in both absolute and relative terms (Fig. 2). The characteristics of the selected studies are provided in Table 3 and Table 4 (both available at: http://www.who.int/bulletin/volumes/92/04/13-116152).

Fig. 1.

Selection of articles for the review of studies on methylmercury exposure in women and infants from seafood-consuming populations

Table 2. Summary of studies assessing total mercury in hair (THHg) or total mercury in blood (TBHg) among women and infants from seafood-consuming populations, by exposure category.

Study characteristics	No. of studies	Exposure categories
		Self-caught seafood					Commercially-purchased seafood
		Arctic	Gold mining	Fishing	Industrya		Coastal	Inland
Population studied
Mothers and/or infantsb	73	9	10	3	5		37	9
Women in general	91	3	19	9	15		32	13
All	164	12	29	12	20		69	22
Study design
Cross-sectional	119	10	28	9	13		44	15
Other	45	2	1	3	7		25	7
Biomarker reported
Reporting THHg biomarkersc	92	1	27	5	16		37	6
Reporting TBHg biomarkersb	72	11	2	7	4		32	16
Reporting of seafood data
Some	84	6	14	10	11		37	6
None	80	6	15	2	9		32	16
Publication date
Published in 1991–2001	34	6	10	3	4		9	2
Published in 2002–2013	130	6	19	9	16		60	20
Subpopulation studiedd
Infants	55	7	9	3	3		27	6
Pregnant women or mothers	74	10	13	2	4		35	10
Non-pregnant women	110	4	21	9	18		40	18
All	239	21	43	14	25		102	34
Study participants
Average participants per study	390	495	350	263	152		448	48
Average participants per subpopulation	268	283	236	236	121		303	316
Total no. of participants	63 943	5935	10 152	3161	3035		30 915	10 745
Countries represented	43	5	6	5	17		23	16

^a Other than gold mining.

^b Mother and infant studies include pregnant women, mothers who have recently given birth and infants (i.e. children up to 12 months of age).

^c Some studies reported both TBHg and THHg biomarkers. When both were reported, THHg biomarkers were extracted.

^d Of these studies, 48 reported on two or more distinctly-defined exposed subpopulations of more than 40 non-pregnant women, pregnant women, women who had recently given birth, or infants (i.e. children up to 12 months of age).

Fig. 2.

Number of selected studies reporting total mercury in hair (THHg) or total mercury in blood (TBHg) in women and infants from seafood-consuming populations, by predominant seafood type (local self-caught or commercially purchased) and year of publication

Table 3. Characteristics of studies assessing total mercury in hair (THHg) or total mercury in blood (TBHg) in women and infants consuming self-caught seafood, by exposure category and subcategory.

Studies, by category and subcategory	Study design	Location	Seafood intakea (kg per month)	Subpopulation	n	THHg, averageb (μg/g)	THHg, High-endb (μg/g)
Gold miningc
Gold mining: rural riverine
Monrroy et al. 200833	Cross-sectional	Bolivia (Plurinational State of), Beni valley	2.2	W	163	3.9	20.0
Barbieri et al. 200934	Cross-sectional	Bolivia (Plurinational State of), Beni valley	5.1	W	77	2.5	–
Boischio et al. 199335	Cross-sectional	Brazil, upper Madeira (river)	–	W	70	10.0	125.0
Barbosa et al. 199836	Cross-sectional	Brazil, upper Madeira (river)	–	MO	98	12.8	94.7
Lebel et al. 199837	Cross-sectional	Brazil, Tapajos	6.9	W	46	11.2	26.6
Grandjean et al. 199938	Cross-sectional	Brazil, Tapajos	10.2	W	114	11.6	–
Amorim et al. 200039	Cross-sectional	Brazil, Tapajos	–	W	46	10.8	–
Boischio et al. 200040	Cross-sectional	Brazil, Madeira	–	MO	90	12.6d	28.3
Dolbec et al. 200041	Cross-sectional	Brazil, Tapajos	9.0	W	40	8.7	–
Harada et al. 200142	Cross-sectional	Brazil, Barreiras	–	W	44	16.4d	53.8
Crompton et al. 200243	Cross-sectional	Brazil, Jacareacanga	–	W	113	6.7d	–
Santos et al. 200244	Cross-sectional	Brazil, Sai Cinza	5.1	W	192	14.7	90.4
Santos et al. 200345	Cross-sectional	Brazil, Pakaanova	–	W	549	8.55	39.4
Santos et al. 200746	Cross-sectional	Brazil, Itaituba	–	IN	1510	4.2d	–
				MO	1510	2.9d	–
Passos et al. 200847	Cross-sectional	Brazil, Tapajos	–	W	121	16.3d	150.0
Grotto et al. 201048	Cross-sectional	Brazil, Tapajos	–	W	54	8.8	–
Fillion et al. 201149	Cross-sectional	Brazil, Tapajos	–	W	126	9.4	–
Dórea et al. 201250	Cross-sectional	Brazil, Bom Futuro	1.6	IN	166	1.6	–
Barcelos et al. 201351	Cross-sectional	Brazil, Tapajos	14.9	W	193	16.3	–
Marques et al. 201352	Cross-sectional	Brazil, Madeira (river)	–	IN	396	3.0	18.5
		Brazil, Madeira (river)	4.3	MO	396	12.1	130.7
		Brazil, Madeira (tin region)	–	IN	294	0.8	2.0
		Brazil, Madeira (tin region)	0.9	MO	294	4.5	11.9
		Brazil, Madeira (rural)	–	IN	67	2.0	8.8
		Brazil, Madeira (rural)	2.6	MO	67	7.8	41.1
Vieira et al. 201353	Cross-sectional	Brazil, Porto Velho (river)	4.4	MO	75	8.2	20.1
Olivero-Verbel et al. 201154	Cross-sectional	Colombia, Antioquia	–	W	757	1.4	10.0
Cordier et al. 199855	Cross-sectional	French Guiana	–	PW	109	1.6	22.0
Cordier et al. 200256	Cross-sectional	French Guiana, upper Maroni	10.2	W	90	12.7	–
		French Guiana, Camopi	–	W	63	6.7	–
		French Guiana, Awala	–	W	55	2.8	–
Fujimura et al. 201257	Cross-sectional	French Guiana, upper Maroni	8.63	W	234	9.9d	26.6
Bose-O’Reilly et al. 201058	Ecological	Indonesia, Kalimantan	–	W	64	2.5	29.6
Gold mining: urban
Hacon et al. 200059	Cross-sectional	Brazil, Alta Floresta	0.6	MO	75	1.1d	8.2
Marques et al. 200760	Cross-sectional	Brazil, Porto Velho	0.7	IN	100	0.2	–
			0.7	MO	100	0.1	–
Dorea et al. 201250	Cross-sectional	Brazil, Porto Velho	1.4	IN	82	1.8	–
Marques et al. 201352	Cross-sectional	Brazil, Madeira (urban)	–	IN	676	1.5	4.8
			1.7	MO	676	5.4	24.1
Vieira et al. 201353	Cross-sectional	Brazil, Porto Velho (urban)	0.7	MO	82	1.3	6.1
Mohan et al. 200561	Cross-sectional	Surinam, Paramaribo	–	IN	39	1.6d	19.6
			–	MO	39	0.8d	15.4
Arctice
Arctic: Traditional diet
Dewailly et al. 200162	Cross-sectional	Canada, Nunavik	–	W	284	4.2	28.0
Muckle et al. 200163	Cohort	Canada, Nunavik	–	IN	95	4.6	24.3
			–	MO	130	2.6	11.1
Lucas et al. 200464	Cross-sectional	Canada, Nunavik	4.9	IN	439	3.5	–
Butler-Walker et al. 200665	Cross-sectioal	Canada, Northwest Territories (Inuit)	–	IN	132	1.7	19.0
			3.5	MO	132	0.9	8.5
Fontaine et al. 200866	Cross- sectional	Canada, Nunavik	1.5	W	308	2.1	41.1
Grandjean et al. 199267	Cohort	Denmark, Faroe Islands	–	IN	1020	6.1	–
			2.2	MO	1020	4.5	–
Bjerregaard et al. 200068	Cross-sectional	Denmark, Greenland (Disko Bay)	–	IN	178	6.3	45.3
			7.1	MO	180	3.2	18.9
Nielsen et al. 201269	Cross-sectional	Denmark, Greenland	–	W	1040	3.7	42.5
Arctic: Mixed diet
Butler-Walker et al. 200665	Cross-sectional	Canada, Northwest Territories (Caucasian)	–	IN	124	0.3	3.2
			0.6	MO	124	0.2	1.1
Odland et al. 199970	Cross-sectional	Norway, northern (Norwegian)	–	MO	81	0.6	0.6
		Norway, northern (Russian)	–	MO	151	0.4	1.4
Hansen et al. 201171	Cross-sectional	Norway, northern	–	MO	211	0.3	0.9
Klopov et al. 199872	Cross-sectional	Russian Federation, Norilsk-Sakelhard	–	IN	42	3.1d	–
			1.5	MO	42	3.9d	–
Arnold et al. 200573	Cross-sectional	United States, Alaska	–	MO	150	0.5	6.4
			–	W	52	0.6	7.8
Industryf
Nilson et al. 200174	Cross-sectional	Brazil, Itapessuma	–	W	84	1.9d	12.5
Kuno et al. 201075	Cross-sectional	Brazil, São Paulo state	0.2	W	265	0.3	1.1
Bruhn et al. 199476	Cross-sectional	Chile, 8th district	–	PW	59	1.7	7.1
Li et al. 200677	Ecological	China, Chanchung	0.6	W	69	0.5d	10.5
Zhang et al. 200678	Cross-sectional	China, Wujiazhan	–	W	40	0.6	–
Tang et al. 200879	Cohort	China, Tongliang	–	IN	110	1.8d	9.9
Fang et al. 201280	Cross-sectional	China, Zhejiang	1.9	W	50	0.8d	3.0
Pawlas et al. 201381	Cross-sectional	China, Guiyang	–	W	49	2.2	35.0
Olivero-Verbel et al. 200882	Cross-sectional	Colombia, Cartagena (bay)	4.3	W	258	1.0	–
Madeddu et al. 200883	Case control	Italy, Sicily Augusta	–	W	100	1.2	5.0
Deroma et al. 201384	Cohort	Italy, Venice (region)	–	IN	70	0.7	–
			–	MO	79	1.2	–
Hsiao et al. 201185	Cross-sectional	Kazakhstan, Temirtau	1.1	W	174	0.4	4.6
Lim et al. 201086	Cohort	Republic of Korea, Sinha-Banud	0.4	W	852	0.7	–
Trasande et al. 201087	Cross-sectional	Mexico, Lake Chapala	–	W	91	0.5	–
Elhamri et al. 200788	Cross-sectional	Morocco, Martil	1.2	W	40	1.4	7.9
Lacayo et al. 199189	Cross-sectional	Nicaragua, Lake Xolotlan	–	W	40	3.4	–
Bravo et al. 201090	Cross-sectional	Romania, Babeni	1.5	W	38	1.0	–
Palkovicova et al. 200891	Cohort	Slovakia, eastern	–	IN	99	0.2	0.64
			–	MO	99	0.2	0.73
Pawlas et al. 201381	Cross-sectional	Slovakia, Baska Bystrica	–	W	52	0.6	3.3
Oskarsson et al. 199492	Cross-sectional	Sweden, Boliden	–	MO	124	0.3d	–
Chang et al. 200893	Cross-sectional	China, Taiwan, Tainan	5.8	W	99	3.7	–
Lincoln et al. 201194	Cross-sectional	United States, Louisiana (gulf)	1.5	W	44	0.7	3.6
Rojas et al. 200795	Case control	Venezuela (Bolivarian Republic of), Valencia	–	W	50	0.9d	4.31
Fishingg
Black et al. 201196	Cross-sectional	Botswana, Okavango delta	2.6	W	60	0.1	0.9
Girard et al. 199597	Cross-sectional	Canada, St James	–	MO	991	2.5	–
Mahaffey et al. 199898	Cross-sectional	Canada, St Lawrence	0.6	W	99	0.04	–
Belles-Isles et al. 200299	Cohort	Canada, St Lawrence	3.8	IN	40	0.5	2.8
Cole et al. 2004100	Cross-sectional	Canada, Ontario	2.2	W	38	1.5	5.4
Morrissette et al. 2004101	Cohort	Canada, St Lawrence (river)	0.6	IN	101	0.1	0.4
			0.6	MO	101	0.1	0.3
Abdelouahab et al. 2008102	Cross-sectional	Canada, St Lawrence (river)	1.2	W	87	0.4	3.9
Jenssen et al. 2012103	Cross-sectional	Norway	2.2	W	100	0.9	4.0
Johnsson et al. 2004104	Cross-sectional	Sweden, Hagfors	–	W	51	0.7	–
Stewart et al. 2000105	Cohort	United States, New York (state)	–	W	296	0.5	0.7
Knobeloch et al. 2007106	Cross-sectional	United States, Wisconsin	1.3	W	1050	0.4	5.3
Schantz et al. 2010107	Cross-sectional	United States, Wisconsin	0.1	W	79	0.4	3.3

IN, infants; MO, mothers; PW, pregnant women; W, women.

^a Seafood intake as reported in studies, converted to kg per month (assuming average meal size of 170 g if not stated) and shown for mothers if reported for both mothers and infants; not all studies reported seafood intake.

^b Biomarker concentrations shown as THHg, either as reported or as converted from TBHg using the hair-to-blood ratio of 250:1. All THHg concentrations are rounded to one decimal place. Average THHg is the geometric mean or median (unless noted with “^d”); high-end THHg is the maximum or the 95th or 90th percentile.

^c Women and infants near tropical small-scale gold mining sites who consume freshwater fish from Hg-contaminated rivers.

^d The average is the arithmetic mean and was not included in main pooling results.

^e Women and infants living in the Arctic or far-Northern regions consuming apex marine foods, including marine mammals.

^f Women and infants periodically consuming marine and freshwater fish caught locally from water bodies contaminated by mercury-emitting industry.

^g Women and infants periodically consuming marine and freshwater fish caught locally from water bodies not affected by industrial emissions.

Table 4. Characteristics of studies assessing total mercury in hair (THHg) or total mercury in blood (TBHg) in women and infants consuming seafood that is predominantly commercially purchased, by exposure category and subcategory.

Studies, by category and subcategory	Study design	Location	Seafood intakea (kg/mo)	Subpopulation	n	THHg, averageb (μg/g)	THHg, high-endb (μg/g)
Coastalc
Coastal: Atlantic
Carneiro et al. 2011108	Cross-sectional	Brazil, Porto Alegre	0.5	W	107	0.1d	–
Legrand et al. 2005109	Cross-sectional	Canada, Bay of Fundy	1.5	W	77	0.5d	0.7
Albert et al. 2010110	Risk assessment	France, north-western	–	PW	125	0.7	2.8
Drouillet-Pinard et al. 2010111	Cohort	France, Poitiers	–	IN	645	0.4	–
	Cohort		1.4	MO	645	0.5	–
Vahter et al. 2000112	Cohort	Sweden, Solna	–	IN	148	0.4	1.2
	Cohort		–	MO	148	0.2	0.7
Björnberg et al. 2003113	Cross-sectional	Sweden, Uppsala	–	IN	123	0.3	1.4
	Cross-sectional		0.8	MO	123	0.4	1.5
Rosborg et al. 2003114	Cross-sectional	Sweden (acid region)	–	W	47	0.4	3.5
	Cross-sectional	Sweden (alkaline region)	–	W	43	0.3	1.0
Brantsaeter et al. 2010115	Cohort	Norway, Baerum	1.2	MO	119	0.4	1.1
Gerhardsson et al. 2010116	Cross-sectional	Norway, Simrishamn	0.7	PW	50	0.2	–
Renzoni et al. 1998117	Cross-sectional	Portugal, Maderia	–	W	181	8.6	42.6
Ramon et al. 2011118	Cohort	Spain, Asturias	2.7	IN	340	2.7	17.3
	Cohort	Spain, Gipuzkoa	2.4	IN	529	1.9	12.5
Oskarsson et al. 199492	Cross-sectional	Sweden, Homsund	–	MO	79	0.3d	–
Björnberg et al. 2005119	Cross-sectional	Sweden	2.1	W	127	0.7	6.6
Pawlas et al. 201381	Cross-sectional	Sweden, southern	–	W	54	1.4	9.8
Bates et al. 2007120	Cross-sectional	United Kingdom	0.7	W	44	0.2	–
Dewailly et al. 2012121	Cross-sectional	United Kingdom (Bermuda)	–	MO	49	1.1	5.0
Stern et al. 2001122	Cross-sectional	United States, New Jersey	1.2	MO	143	0.3d	8.0
Ortiz-Roque et al. 2004123	Cross-sectional	United States, Puerto Rico	2.0	W	45	0.4	–
	Cross-sectional	United States, Vieques	3.6	W	41	0.3	–
Oken et al. 2005124	Cohort	United States, eastern Massachusetts	0.9	MO	135	0.1	0.6
McKelvey et al. 2007125	Cross-sectional	United States, New York City	1.5	W	1049	0.7	2.8
Karouna-Renier et al. 2008126	Cross-sectional	United States, Florida panhandle	–	PW	83	0.2	10.7
			–	W	515	0.3	22.1
Lederman et al. 2008127	Cross-sectional	United States, New York City (non-Asian)	–	IN	178	0.7	–
		United States, New York City (Chinese)	–	MO	83	1.1	–
		United States, New York City (non-Asian)	–	MO	176	0.4	–
Caldwell et al. 2009128	Cross-sectional	United States (national)	–	W	1888	0.2	1.1
Wells et al. 2011129	Cross-sectional	United States, Maryland	–	IN	300	0.3	–
King et al. 2013130	Cross-sectional	United States, Pawtucket	–	IN	538	0.1	9.8
Traynor et al. 2013131	Cross-sectional	United States, Duval County, Florida	2.1	W	698	0.3	3.0
Coastal: Mediterranean, Indian Ocean, Persian Gulf
Babi et al. 2000132	Cross-sectional	Albania, Tirana	0.3	W	47	0.6	2.0
Miklavčič et al. 2013133	Cohort	Croatia, Rijeka	0.8	IN	210	0.7	8.0
			0.8	MO	255	0.5	5.3
Gibičar et al. 2006134	Cohort	Greece, islands	1.5	PW	246	1.4	17.5
Vardavas et al. 2011135	Cohort	Greece, Heraklion Crete	–	PW	47	0.4	1.7
Miklavčič et al. 2013133	Cohort	Greece, Lesvos and Chios	1.0	MO	391	1.5	8.3
Fakour et al. 2010136	Cohort	Islamic Republic of Iran, Mahshahr	1.3	W	195	3.0d	26.5
Salehi et al. 2010137	Cross-sectional	Islamic Republic of Iran, Mahshahr	2.9	PW	149	2.0	10.0
Barghi et al. 2012138	Cross-sectional	Islamic Republic of Iran, Noushahr	3.9	PW	59	0.3	0.6
Okati et al. 2012139	Cross-sectional	Islamic Republic of Iran, Mazandaran	–	IN	93	1.9d	6.9
			1.1	MO	93	3.6d	9.0
Díez et al. 2008140	Cross-sectional	Italy, Naples	–	W	114	0.5	1.5
Maddedu et al. 200883	Case control	Italy, Sicily, Catalina	–	W	100	0.9	4.2
Miklavĉiĉ et al. 2013133	Cohort	Italy, Trieste	1.2	IN	614	1.0	8.3
	Cohort		1.2	MO	871	0.6	10.0
Bou-Olayan et al. 1994141	Cross-sectional	Kuwait	2.2	W	68	4.1d	25.0
Khassouani et al. 2001142	Cross-sectional	Morocco, Rabat	–	W	70	1.6d	–
Myers et al. 1995143	Cohort	Seychelles, Mahe	–	PW	740	5.9	26.7
Channa et al. 2013144	Cross-sectional	South Africa, KwaZulu-Natal	–	IN	350	0.2	4.6
	Cross-sectional		–	MO	350	0.2	3.1
Rudge et al. 2009145	Cross-sectional	South Africa	–	IN	62	1.2	9.7
			–	MO	62	0.7	8.8
Soria et al. 1992146	Cross-sectional	Spain, Seville	–	W	50	2.9d	20.0
Ramon et al. 2011118	Cohort	Spain, Valencia	2.1	IN	554	2.4	16.5
		Spain, Sabadell	2.3	IN	460	1.6	15.0
Unuvar et al. 2007147	Cohort	Turkey, Istanbul	1.1	IN	143	0.1	–
			1.1	MO	143	0.1	–
Coastal: Pacific coast
Choy et al. 2002148	Case control	China, Hong Kong Special Administrative Region	–	W	155	1.7	–
Fok et al. 2007149	Cohort	China, Hong Kong Special Administrative Region	1.3	IN	1057	2.2	–
			1.3	MO	1057	1.2	–
Gao et al. 2007150	Cohort	China	2.9	IN	408	1.4	–
			2.9	MO	408	1.3	–
Liu et al. 2008151	Cross-sectional	China, 5 cities	2.1	W	321	0.7	8.5
Dewailly et al. 2008152	Cross-sectional	French Polynesia, Tahiti	5.6	IN	234	2.6	12.1
Nakagawa et al. 1995153	Cross-sectional	Japan, Tokyo	–	W	177	1.9	–
Iwasaki et al. 2003154	Cross-sectional	Japan, Akita	–	W	154	1.7	5.8
Yasutake et al. 2003155	Cross-sectional	Japan	–	W	1666	1.4	25.8
Arakawa et al. 2006156	Cohort	Japan, Sendai	2.6	MO	180	2.0	9.4
Ohno et al. 2007157	Cohort	Japan, Akita	–	W	59	1.5	3.6
Sakamoto et al. 2007158	Cross-sectional	Japan, 3 cities	–	IN	115	2.5	–
			–	MO	115	1.3	–
Sakamoto et al. 2008159	Biomarker valid	Japan, Fukuoka	–	IN	40	0.4	–
			–	MO	40	0.4	–
Miyake et al. 2011160	Cohort	Japan, Osaka	–	W	582	1.5	3.2
Kim et al. 2006161	Case control	Republic of Korea, Seoul	–	IN	63	1.0	5.0
			–	MO	63	0.6	7.4
Kim et al. 2008162	Cross-sectional	Republic of Korea (coastal)	4.4	W	111	0.8	–
Jo et al. 2010163	Cross-sectional	Republic of Korea, Busan	4.4	W	146	1.9	11.4
Kim et al. 2010164	Cross-sectional	Republic of Korea, 3 cities	4.4	IN	312	3.7	–
Lee et al. 2010165	Cohort	Republic of Korea, 3 cities	4.4	IN	417	1.4	6.0
			4.4	PW	417	0.8	4.6
Kim et al. 2011166	Cohort	Republic of Korea, 3 cities	–	IN	797	1.3	2.3
			–	MO	797	0.8	1.4
Kim et al. 2012167	Cross-sectional	Republic of Korea	–	W	2964	1.0	–
You et al. 2012168	Cross-sectional	Republic of Korea, Busan and Ulsan	–	W	200	4.7	–
Eom et al. 2013169	Cross-sectional	Republic of Korea (coastal)	–	W	308	1.1	–
Hong et al. 2013170	Cross-sectional	Republic of Korea, Seoul	–	W	79	1.4d	–
Kim et al. 2013171	Cross-sectional	Republic of Korea (urban)	1.5	W	117	0.9	–
		Republic of Korea (coastal)	1.5	W	114	0.9	–
		Republic of Korea (rural)	1.5	W	105	0.7	–
Marsh et al. 1995172	Cohort	Peru, Mancora	–	MO	131	7.1	28.5
Hsu et al. 2007173	Cross-sectional	China, Taiwan, Taipei	–	IN	65	2.3	7.0
			1.9	MO	65	2.2	5.3
Chien et al. 2010174	Risk assessment	China, Taiwan (northern)	1.5	W	263	1.7	16.3
Sato et al. 2006175	Cross-sectional	United States, Honolulu, Hawaii	0.6	IN	188	0.7d	5.0
Tsuchiya et al. 2009176	Cohort	United States, Washington state (Koreans)	1.8	W	108	0.6	–
		United States, Washington state (Japanese)	1.8	W	106	1.2	–
Inlande
Gundacker et al. 2006177	Cross-sectional	Austria, Vienna	–	W	78	0.6d	–
Rudge et al. 2011178	Cross-sectional	Brazil, São Paulo state	–	MO	155	0.2	1.1
Rhainds et al. 1999179	Cross-sectional	Canada, southern Quebec	–	IN	109	0.2	3.4
Pawlas et al. 201381	Cross-sectional	Croatia, Koprivnica	–	W	60	0.4	7.6
Puklová et al. 2010180	Cross-sectional	Czech Republic	0.5	W	163	0.2	2.3
Cerna et al. 2012181	Cross-sectional	Czech Republic	–	W	494	0.2	0.7
Pawlas et al. 201381	Cross-sectional	Czech Republic	–	W	51	0.9	8.0
Khassouani et al. 2001142	Cross-sectional	France, Angers	–	W	62	0.9	–
Huel et al. 2008182	Cohort	France, Paris	–	MO	81	1.2	2.9
Deroma et al. 201384	Cohort	Italy, northern	–	IN	58	0.9	–
			–	MO	72	0.9	–
Eom et al. 2013169	Cross-sectional	Republic of Korea (inland)	–	W	886	0.8	–
Pawlas et al. 201381	Cross-sectional	Morocco, Fez	–	W	50	1.0	9.1
Anwar et al. 2007183	Cross-sectional	Pakistan, Lahore	0.7	W	75	0.2	2.5
Jędrychowski et al. 2007184	Cross-sectional	Poland, Krakov	–	IN	313	0.1	–
		Poland	0.7	MO	313	0.2	–
Pawlas et al. 201381	Cross-sectional	Poland, Wroclaw	–	W	51	0.7	2.9
Al-Saleh et al. 2006185	Case control	Saudi Arabia	–	W	185	0.9d	5.4
Al-Saleh et al. 2008186	Case control	Saudi Arabia, Riyadh	–	W	434	0.9d	7.6
Al-Saleh et al. 2011187	Cross-sectional	Saudi Arabia, Riyadh	–	IN	1561	0.6	1.9
		Saudi Arabia, Riyahd	–	MO	1574	0.5	2.2
Al-Saleh et al. 2013188	Cross-sectional	Saudi Arabia	–	MO	150	0.3	–
Miklavčič et al. 2011189	Cohort	Slovenia, Ljubljana	–	IN	446	0.4	–
	Cohort	Slovenia, Ljubljana	0.8	MO	574	0.3	–
Miklavčič et al. 2013133	Cohort	Slovenia, Ljubljana	1.3	MO	446	0.4	3.5
Pawlas et al. 201381	Cross-sectional	Slovenia, Ljubljana	–	W	50	0.7	13.0
Díez et al. 2009190	Cohort	Spain, Madrid	1.4	IN	57	1.5	5.1
Díez et al. 2011191	Case control	Spain, Toledo	2.0	W	64	2.5	–
Bjermo et al. 2013192	Cross-sectional	Sweden	–	W	145	0.2	0.7
Gerhardsson et al. 2010116	Cross-sectional	Sweden, Hasselholm	0.4	PW	50	0.2	–
Knobeloch et al. 2005193	Cross-sectional	United States, 12 states	0.7	W	414	0.3	1.6
Xue et al. 2007194	Cohort	United States, Michigan	0.6	MO	1024	0.1	–
Pollack et al. 2011195	Cross-sectional	United States, western New York state	–	W	252	0.3	–
Pollack et al. 2012196	Cross-sectional	United States, Buffalo	–	W	248	0.4	–

IN, infants; MO, mothers; PW, pregnant women; W, women.

^a Seafood intake as reported in studies, converted to kg per month (assuming average meal size of 170 g if not stated) and shown for mothers if reported for both mothers and infants; not all studies reported seafood intake.

^b Biomarker concentrations shown as THHg, either as reported or as converted from TBHg using the hair-to-blood ratio of 250:1. All THHg concentrations are rounded to one decimal place. Average THHg is the geometric mean or median (unless noted with “^d”); high-end THHg is the maximum or the 95th or 90th percentile.

^c Women and infants living in coastal regions and consuming marine and freshwater seafood mainly purchased from local and global markets.

^d The average is the arithmetic mean and was not included in the main pooled results.

^e Women and infants living inland and consuming marine and freshwater seafood mainly purchased from local and global markets.

Pooled biomarker concentrations

For 43 subpopulations of women and infants living near small-scale gold mining sites in Bolivia (Plurinational State of),^33,34 Brazil,^{35–53,59,60} Colombia,⁵⁴ French Guiana,^55–57 Indonesia⁵⁸ and Surinam⁶¹ the pooled central distribution median THHg biomarker concentration was 5.4 µg/g (upper bound median: 23.1) (Table 5). Values were higher (8.2 µg/g; upper bound: 27.5) in the subgroup of rural riverine dwellers reliant on local freshwater fish and lower (1.4 µg/g; upper bound: 11.8) among urban dwellers consuming less fish. For 21 subpopulations from Arctic regions, including in Canada,^62–66 Denmark (Greenland and the Faroe Islands),^67–69 Norway,^70,71 the Russian Federation⁷² and the United States (state of Alaska),⁷³ the pooled central distribution median result was 2.1 µg/g (upper bound: 9.8); values were higher (3.6 µg/g; upper bound: 24.3) for marine mammal and other self-caught seafood consumers and lower (0.4 µg/g; upper bound: 1.4) among those with a diet including less seafood and less reliant on these traditional foods.

Table 5. Pooled total THHg biomarker distributions in women and infants from seafood-consuming populations, by exposure category and subcategory.

Category and subcategory	No. of subpopulations	No. of participants	Central distributiona			Upper bound distributiona
			THHg (μg/g)b 25th, 50th 75th, 95th percentile	Percentage > PTWIc		THHg (μg/g)b 25th, 50th, 75th, 95th percentile	Percentage > PTWIc
Gold mining	43	10 152	1.80, 5.36, 10.00, 14.70	77		11.94, 23.07, 39.40, 125.00	98
Rural	34	8 283	2.50, 8.24, 11.20, 14.70	85		18.53, 27.45, 53.80, 130.70	97
Urban	9	1 869	0.19, 1.41, 1.80, 5.36	44		6.09, 11.80, 19.60, 24.14	100
Arctic	21	5 935	0.47, 2.09, 4.18, 6.33	52		2.30, 9.76, 26.13, 45.25	81
Traditional	12	4 958	2.34, 3.61, 4.56, 6.33	75		18.90, 24.25, 41.08, 45.25	100
Mixed diet	9	977	0.31, 0.40, 0.55, 0.64	11		0.93, 1.38, 6.35, 7.82	56
Industry	25	3 035	0.25, 0.75, 1.27, 3.54	32		3.04, 4.62, 9.93, 35.00	89
Fishing	14	3 161	0.13, 0.38, 0.70, 2.50	6		0.70, 2.75, 4.00, 5.38	71
Coastal	102	30 915	0.36, 0.82, 1.51, 3.70	23		2.83, 6.76, 10.65, 26.46	86
Atlantic	35	9 675	0.27, 0.35, 0.69, 2.70	16		1.16, 2.93, 9.75, 22.14	76
Mediterranean	27	6 536	0.29, 0.65, 1.45, 5.90	32		4.18, 8.53, 16.50, 26.46	96
Pacific	40	14 704	0.85, 1.34, 1.94, 4.66	23		2.83, 6.03, 10.65, 28.50	98
Inland	34	10 745	0.31, 0.38, 0.77, 1.47	18		1.93, 2.90, 7.59, 13.00	79
Total	239	63 943	–	34		–	86

PTWI, provisional tolerable weekly intake; THHG, total mercury in hair.

^a Central distribution reflects pooling of geometric mean and median biomarkers from reported studies; upper bound distribution reflects pooling of 90th, 95th percentiles and maximums from reported studies.

^b Biomarkers measuring total mercury in blood converted to THHg equivalent at a hair-to-blood ratio of 250:1.

^c Share of total subpopulations with a reported average or high-end biomarker greater than the PTWI equivalent of 2.2 μg/g of THHg.

For 25 subpopulations whose self-caught fish from local waterways is affected by Hg-emitting industries in Brazil,^74,75 Chile,⁷⁶ China,^77–81 Colombia,⁸² Italy,^83,84 Kazakhstan,⁸⁵ Mexico,⁸⁷ Morocco,⁸⁸ Nicaragua,⁸⁹ Norway,¹¹⁵ the Republic of Korea,⁸⁶ Romania,⁹⁰ Slovakia,^81,91 Sweden,⁹² Taiwan, China,⁹³ the United States⁹⁴ and Venezuela (Bolivarian Republic of),⁹⁵ the pooled central THHg median biomarker was 0.8 µg/g (upper bound: 4.6). In 14 subpopulations consuming fish periodically from non-industry-contaminated waters in Botswana,⁹⁶ Canada,^97–102 Norway,¹⁰³ Sweden¹⁰⁴ and the United States,^105–107 the value was 0.4 µg/g (upper bound: 2.8).

For 102 coastal or island-dwelling subpopulations consuming seafood that is predominantly commercially purchased, the combined central median THHg concentration was 0.8 µg/g (upper bound: 6.8). On the Atlantic coast, the pooled result for 35 subpopulations in Brazil,¹⁰⁸ Canada,^99,109 France,^110,111 Norway,¹¹⁵ Portugal,¹¹⁷ Spain,¹¹⁸ Sweden,^{81,92,112–114,119} the United Kingdom of Great Britain and Northern Ireland^120,121 and the United States^122–131 was 0.4 µg/g (upper bound: 2.9). For 27 subpopulations from the Mediterranean, Persian Gulf and Indian Ocean (combined because of similar THHg ranges and referred to as “Mediterranean”) in Albania,¹³² Croatia,¹³³ Greece,^133,135 the Islamic Republic of Iran,^136–139 Italy,^83,133,140 Kuwait,¹⁴¹ Morocco,¹⁴² Seychelles,¹⁴³ South Africa,^144,145 Spain¹⁴⁶ and Turkey,¹⁴⁷ the pooled central THHg concentration was 0.7 µg/g (upper bound: 8.5). For 40 Pacific coast subpopulations in China,^148–151 Japan,^153–160 Peru,¹⁷² the Republic of Korea,^161–171 Taiwan, China¹⁷⁴ and the United States,^175,176 the pooled result was 1.3 µg/g (upper bound: 6.0).

For 34 subpopulations living in inland regions of Austria,¹⁷⁷ Brazil,¹⁷⁸ Canada,¹⁷⁹ Croatia,⁸¹ the Czech Republic,^81,180,181 France,^142,182 Italy,⁸⁴ Morocco,⁸¹ Pakistan,¹⁸³ Poland,¹⁸⁴ the Republic of Korea,¹⁶⁹ Saudi Arabia,^186–188 Slovenia,^81,189 Spain,^190,191 Sweden¹⁹² and the United States,^193–196 the pooled central TTHg median was 0.4 µg/g (upper bound: 2.9).

Comparison with provisional tolerable weekly intake

The median of the pooled central THHg biomarker distribution for women and infants in rural riverine communities near tropical gold mining sites reached nearly four times the FAO/WHO PTWI reference level of 2.2 ug/g (Fig. 3), while the upper-bound median reached more than 10 times this reference. Some individual high-end biomarkers exceeded 50 µg/g, the lower end of the range found in the neurological syndrome known as Minamata disease,⁴ associated with accidental industrial Hg poisoning in Japan in the 1950s and 1960s (Fig. 4). The median of the central THHg biomarker distribution in Arctic traditional food consumers exceeded the reference by 63%, while the upper bound median was over 10 times the value. For women and infants in the industry and fishing categories, central estimate medians were below the international reference, although the industry central median was twice that of the fishing category; most high-end biomarkers were above it. For those in the Pacific coastal subcategory, the 75th percentile approached the reference value; the upper bound median was nearly three times this value and nearly all high-end biomarkers exceeded it. Central biomarkers were below the PTWI in the Atlantic. However in many subpopulations in the Mediterranean they exceeded this reference, while the upper bound median was nearly four times the reference and most high-end biomarkers exceeded it. For the inland category, the central estimate median was well below the reference, but nearly 80% of the high-end biomarkers exceeded it.

Fig. 3.

Distributions of central estimate for total mercury in hair (THHg) reported in selected studies of women and infants from seafood-consuming populations, by exposure category

PTWI, provisional tolerable weekly intake.

Fig. 4.

Distributions of upper-bound total mercury in hair (THHg) reported in selected studies of women and infants from seafood-consuming populations, by exposure category

PTWI, provisional tolerable weekly intake.

Note: High-end biomarkers in the gold mining, Arctic and coastal categories reach into the range associated with observable neurological damage.

Study quality

A majority (78%) of selected studies were based on convenience samples taken from seafood-consuming populations. Some details of the seafood context were provided in most (71%) studies, but in the others this information was sparse. Laboratory protocols for THHg and TBHg detection were nearly universally reported (91%). Most (82%) protocols were based on cold vapour atomic absorption spectrometry (CV-AAS) or inductively-coupled plasma mass spectrometry (ICP-MS) and a majority (74%) reported laboratory quality control procedures. In 86% of studies, distributions were transformed to lognormal scale and summarized using geometric means or medians. More than half (55%) of the studies reported maximums as high-end estimates, while the remainder reported 90th or 95th percentiles. Only 51% of studies reported some seafood intake data and 25% evaluated non-seafood sources of Hg.

Discussion

We found that biomarkers of MeHg intake were of greatest health concern among three categories of seafood-consuming women and their infants: (i) rural riverside dwellers living near tropical small-scale gold mining with diets dependent on locally-caught freshwater fish; (ii) those in Arctic regions for whom apex food-chain marine mammals are a dietary staple; and (iii) coastal inhabitants, particularly in the Pacific and Mediterranean, who probably consume seafood that is primarily commercially sourced. In the first group, average Hg biomarkers suggest MeHg intake exceeds by several fold the level considered by WHO and FAO to pose no substantial risk of developmental neurotoxicity. In the second group, average biomarkers suggest MeHg intake well over the reference value. In the third group, biomarkers suggest an important share of the population approach or exceed the reference level. High-end biomarkers in all three groups indicate body burdens of MeHg in the range associated in epidemiological studies with observable neurological damage. While average biomarkers in other groups suggest that MeHg intake is below the recommended level, most upper bound biomarkers in these categories exceed the reference, which shows that even in groups with lower average exposure certain populations are at risk.

Before this study, few researchers had systematically compared the global exposures and risks linked to MeHg intake from seafood. Brune et al. reviewed Hg biomarker studies – published from 1976 to 1990 – of general populations exposed through various sources and found the highest values among seafood consumers in Greenland and Japan.¹⁹⁷ Sioen et al. estimated contaminant and nutrient intake in general populations based on global seafood availability data and found the estimated MeHg intake to be highest in Japan and the Pacific islands, followed by the Nordic and Mediterranean regions.¹⁹⁸ A recent European regional study examining biomarkers showed the highest MeHg exposure to be in Mediterranean countries.¹⁹⁹ Our findings are broadly consistent with these studies and with the literature describing MeHg exposure and risk in specific subsistence fishing communities. This review adds to the evidence by synthesizing the findings from the two most recent decades of published international Hg biomarker data specifically for women and infants and by examining, in a single study, MeHg exposure in populations consuming self-caught and commercially purchased seafood.

Several limitations affect the interpretation of our results. Our goal was to compare MeHg exposure across various international groups of women and infants from seafood-consuming populations. However, incomplete reporting prevented us from evaluating the share of non-consumers of seafood in each study. Furthermore, most studies used convenience samples that may not have been representative of the populations from which they were taken. In sensitivity analysis we pooled biomarkers excluding the several large representative population surveys (which have a higher share of non-consumers of seafood than other studies). However, this did not alter our findings. Physiological differences in MeHg metabolism and elimination by life stage are well known²⁰⁰ and the FAO/WHO reference dose was established based on maternal biomarkers. Thus, in sensitivity analysis we also combined biomarkers excluding infants. This resulted in slightly lower medians for the Arctic and gold mining categories and higher ones for the coastal and inland categories.

TBHg is a better indicator of recent MeHg exposure than THHg, which is a better measure of longer-term MeHg exposure.^3,4,6 Although this difference may be important among sporadic seafood consumers, the majority of our subpopulations were regular seafood consumers. Conversion of TBHg biomarkers to THHg equivalents is likely to have resulted in some measurement error. However, the range of hair-to-blood ratios reported in our studies was similar to the range on which the standard conversion ratio is based, which minimizes this bias.⁵ When we pooled only THHg biomarkers, medians were slightly higher across most categories (although some categories had few observations). Despite the use of laboratory methods that relied on commonly employed protocols, detection techniques are subject to variation^3,11 and quality control practices were not uniformly reported. Sensitivity analysis examining only studies using CV-AAS or similar procedures resulted in slightly higher biomarkers for the Arctic category.

Population Hg biomarker distributions are often skewed to the right, so that central tendency is best captured by geometric means or medians.³ Thus, in reporting our main results we chose to exclude the small number of studies reporting only arithmetic means. Including arithmetic means yielded higher results for the inland category. To give greater weight to estimates from larger samples, we pooled biomarkers using sample-size weighting. Doing so yielded higher summary biomarkers in the Arctic and coastal categories. Variations in the share of MeHg in total Hg have been reported, both among frequent and infrequent seafood consumers,^23,201 depending in part on exposure to Hg sources other than seafood (such as elemental Hg in dental amalgams or inorganic Hg compounds in skin-lightening creams).^3,29 Most of the one quarter of selected studies examining non-seafood sources of Hg assessed the presence of dental amalgams, mainly in infrequent consumers of seafood; while this inorganic Hg source is best measured with urinary biomarkers, in cases where this exposure is important TBHg biomarkers may overestimate MeHg.²⁶ We eliminated high outlier biomarkers due to suspected non-seafood sources wherever these were noted by authors (most were in subpopulations where skin-lightening creams were used). Nevertheless, other sources of Hg exposure influencing high-end measures cannot be excluded. These limitations in the underlying data suggest that our findings should be interpreted cautiously. However, most sensitivity analyses resulted in higher biomarker summary statistics than the main findings we report; we chose conservative assumptions for our main results.

Estimated IQ losses in infants born to seafood consuming mothers serve as an alternative means of characterizing the public health impact of MeHg exposure. As an illustration, we applied a dose–response relationship (0.18 infant IQ point lost for every ppm increase in maternal THHg)²⁰² that has been used to estimate the economic costs associated with Hg contamination^203,204 to our pooled upper bound biomarkers. The resulting interquartile range of estimated IQ loss spanned from 1 to 13 points for the gold mining, Arctic and coastal subpopulation categories. IQ losses at the higher end of this range may be sufficient to contribute to mild mental retardation, defined as an IQ between 50 and 69 points. Among subsistence fishing populations in the Amazon, an assessment of global burden of disease showed an incidence of mild mental retardation of up to 17.4 cases per 1000 infants²⁰⁵ and separate research identified MeHg-associated deficits in memory and learning in adults.²⁰⁶ IQ losses in the lower end of the range may contribute to borderline intellectual functioning, characterized by memory and executive function deficits.²⁰⁷ Although such minor losses in IQ may go unnoticed in an individual, they can cause an important shift in intellectual capacity at the population level, as documented in the case of lead.²⁰⁸ IQ loss represents only one facet of the neurological harm resulting from MeHg; our analysis did not include recent research suggesting neurological effects at lower dose¹¹ or other documented effects, such as adverse cardiovascular outcomes.²⁰⁹

Systematic reviews provide an opportunity to identify gaps in a body of research. Small-scale gold mining is practiced in 70 countries,²¹⁰ but we found Hg biomarker studies meeting our criteria in only six. We identified studies in 23 coastal countries, although per capita seafood consumption data suggest that many other such countries warrant study.²⁰ Although reviews of subsistence fishing populations in the Amazon and Arctic are available, few have been conducted for coast-dwelling frequent seafood consumers (e.g. in south-eastern Asia or the Mediterranean) or for fishing populations near abandoned chloralkali plants and other aquatic sources of Hg contamination. We found population-based Hg biomonitoring surveys in only a handful of countries; most are high-income and have relatively low per capita seafood consumption.

It was beyond the scope of this review to assess time trends in Hg biomarkers. Without major policy changes, projections indicate that global anthropogenic Hg emissions are likely to increase.²¹¹ Moreover, modelling suggests that any reduction in Hg emissions is likely to take time to translate into reduced MeHg in seafood.²¹² Declines in Hg biomarkers in humans have been observed in association with changes in seafood consumption habits in various populations. This finding reinforces the importance of carefully designed public health messages intended to reduce MeHg exposure.^199,212 In subsistence fishing populations, the cultural importance of seafood harvesting and the scarcity of alternative animal protein sources suggest the existence of complex tradeoffs in guiding seafood consumption and the need for well-targeted messages. In predominantly urban seafood-consuming coastal populations, commercial seafood advisories may be an appropriate choice for reaching at-risk populations.¹⁹ Because of seafood’s important nutritional benefits, all such messages should aim to encourage a shift away from large apex predator species and towards those with lower MeHg and higher polyunsaturated fatty acid content, rather than to reduce seafood intake.

Conclusion

In this review of biomarkers of MeHg intake in women and infants from 164 studies across 43 countries, we found a very high risk in tropical riverine populations near gold mining sites and in traditional Arctic populations. In both groups, biomarkers suggest average MeHg intake exceeds the FAO/WHO recommendation, although their share of the global total of seafood-consuming women and infants is likely to be fairly small. We also found an elevated risk among seafood consumers in the coastal regions of south-eastern Asia, the western Pacific and the Mediterranean; a large share of the world’s seafood-consuming women and their infants is likely to be found in this group because of its large population. In other populations for whom data were available, average indicators of risk were lower and generally within international intake recommendations. However, women and infants with high exposure to MeHg were evident across all exposure categories. Although sources of bias were present, these results should help to set broad priorities for preventive policy and research.

The findings of this review underscore the importance of WHO’s call for enhanced population monitoring and risk communication to women of reproductive age regarding healthful seafood choices.¹ One of the provisions of the Minamata Convention aims to protect vulnerable populations from Hg exposure through public education and other measures.²¹³ The Convention is a potentially important strategic tool to reach the populations at highest risk through development of seafood advisory risk messages for commercial seafood consumers, targeted community-based interventions for subsistence fishing groups and regular population surveillance.