Scientific opinion on the tolerable upper intake level for manganese

View full-text article in PMC

. 2023 Dec 8;21(12):e8413. doi: 10.2903/j.efsa.2023.8413

Author, year Country Study design Follow‐up Funding	Original cohort (N total) Exclusion criteria Study population (n)	Ascertainment of outcome	Exposure groups n/person‐years Exposure assessment method	Mean scores	Model covariates	Results
Nascimento et al. (2015, 2016) Brazil CS Funding: public	N = 90 Population sample: Children from two different areas, rural and urban recruited through school principals Exclusion criteria: NR n for NEUPSILIN‐Inf battery = 63 (rural: 43, urban: 20) n for RCPM test = 59 (rural: NR, urban: NR) Sex (% girls): 49 Age: 6–12 y	Neuropsychological functions assessed by the Brazilian Child Brief Neuropsychological Assessment Battery NEUPSILIN‐Inf, composed of Attention, Perception, Working Memory, Phonological Awareness, Written Language and Executive Function Raven's Coloured Progressive Matrices (RCPM) test (nonverbal IQ)	Mean (min‐max) water manganese (WMn) concentration (mg/L) Rural area: 0.02 (0.00003–0.28) Urban area: 0.001 (0.001–0.002) Water collected from tap in children's homes and Mn measured by ICP‐MS	Mean (SEM) scores Attention Rural: 44.7 (1.4) Urban: 49.6 (1.3) Perception Rural: 3.4 (0.2) Urban: 3.8 (0.1) Working memory Rural: 24.3 (3.9) Urban: 34.3 (3.9) Phonological awareness Rural: 5.4 (0.8) Urban: 8.4 (0.7) Language Rural: 50.9 (4.2) Urban: 59.5 (3.7) Go/no go task Rural: 54.5 (3.8) Urban: 53.7 (3.2) Non verbal IQ Rural: 39.1 (3.5) Urban: 64.2 (4.6)	IQ, age, gender and parents' education	Change in NEUPSILIN‐Inf scores by WMn increase (mg/L) Word/nonword reading: β = −0.307, p = 0.019 Word/nonword written: β = −0.253, p = 0.045 Written language: β = −0.361, n = 0.007 Language: β = −0.390, p = 0.002 Executive function: β = −0.547, p = 0.001 Visual attention, perception, working memory and phonological awareness: no association. Spearman's correlation between WMn and nonverbal IQ score: r = −0.317, p = 0.014
Bouchard et al. (2011) Canada CS Funding: public	N = 696 Population sample: Children living in the same house for > 3 months, in eight municipalities in Quebec, Canada, where presence of Mn is naturally elevated (2007–2009) Exclusion criteria: NR n = 362 (52% participation rate) Sex (% girls): 54 Age [mean (SD)]: 9.3 y (1.8)	Children's cognitive abilities assessed with WASI (Verbal IQ, Performance IQ and Full‐scale IQ)	Median/arithmetic mean/geometric mean (P5, P95) WMn (mg/L): 0.031/0.098/0.020 (0.005, 0.255) Median (P5, P95) Mn intake from water consumption (mg/kg bw per month): 0.008 (0, 0.286) Median (P5, P95) Mn intake from diet (mg/kg bw per month): 2.335 (0.840, 6.418) Water sampled directly from kitchen tap of children's homes and Mn measured with ICP‐MS Mn intake estimated from diet and water consumption with orally administered semiquantitative FFQ answered by parent and child	NR	Model 1: maternal education, maternal nonverbal intelligence, family income, home stimulation score and family structure Model 2: model 1 + sex and age of child, IQ testing session time, source of water and iron concentration in tap water	Change in IQ for a 10‐fold increase in WMn (log ₁₀ ‐transformed) β (95% CI) Full‐scale IQ Unadjusted model: −2.1 (−3.5, −0.8) Model 1: −1.9 (−3.1, −0.7) Model 2: −2.4 (−3.9, −0.9) Performance IQ Unadjusted model: −2.4 (−4.0, −0.7) Model 1: −2.3 (−3.7, −0.8) Model 2: −3.1 (−4.9, −1.3) Verbal IQ Unadjusted model: −1.4 (−2.6, −0.2)* Model 1: −1.5 (−2.6, −0.3) Model 2: −1.2 (−2.7, 0.3) p < 0.05, p < 0.01 Change in IQ per 10‐fold increase Mn intake from water consumption (log* ₁₀ ‐transformed) β (95% CI) Full‐scale IQ Unadjusted model: −1.3 (−2.5, −0.2) Model 1: −1.2 (−2.3, −0.1) Model 2: −1.2 (−2.3, −0.1) Performance IQ Unadjusted model: −1.6 (−3.0, −0.3) Model 1: −1.6 (−2.9, −0.3) Model 2: −1.9 (−3.3, −0.4) Verbal IQ Unadjusted model: −0.7 (−1.7, 0.3) Model 1: −0.6 (−1.6, 0.3) Model 2: −0.3 (−1.4, 0.7) p < 0.05, *p < 0.01 No relationship with Mn intake from diet
Oulhote et al. (2014) Canada CS Funding: public	N = 696 Population sample: Children living in the same house for > 3 months, in eight municipalities in Quebec, Canada, where presence of Mn is naturally elevated (2007–2009) Exclusion criteria: NR N = 375 Sex (% girls): 53 Age: 6–13 y	Children's cognition assessed with CVLT‐C (memory), CPT II (sustained attention), Digit Span (working memory) Children's motor function evaluated with Santa Ana Pegboard test and manual Finger tapping Children's behaviour evaluated Conners' Rating Scales completed by a teacher (CRS‐T) and a parent (CRS‐P)	Arithmetic mean/geometric mean (range) WMn (mg/L): 0.099/0.020 (0.001–2.791) Geometric mean (range) Mn intake from water consumption (mg/kg bw per month): 0.0055 (0–1.059) Mn concentration measured, by ICP‐MS, in home tap water samples. Mn intake estimated from diet and water consumption with orally administered semiquantitative FFQ answered by parents and children	NR	Child's sex, age, maternal education, maternal nonverbal intelligence, family income, maternal depressive symptoms and tap water lead concentrations	Change in outcome per 10‐fold increase in WMn (log ₁₀ ‐transformed) β (95% CI) Memory: −1.0 (−1.6, −0.4), p < 0.01 Attention: 0.5 (−0.4, 1.3), p = 0.31 Motor function: −1.2 (−2.7, 0.3), p = 0.11 Hyperactivity: −0.2 (−1.2, 0.8) p = 0.71 Change in outcome per 10‐fold increase Mn intake from water consumption (log ₁₀ ‐transformed) β (95% CI) Memory: −0.4 (−0.9, 0.1), p = 0.13 Attention: 0.1 (−0.6, 0.8), p = 0.80 Motor function: −1.13 (−2.4, −0.2), p = 0.02 Hyperactivity: 0.2 (−0.5, 0.9), p = 0.51
Dion et al. (2016) Canada CS Funding: public	N = 362 Population sample: Children living in the same house for > 3 months, in 8 municipalities in Quebec, Canada, where presence of Mn is naturally elevated (2007–2009). Among the initial sample (Oulhote et al., 2014), children were selected based on the concentration of manganese in the tap water of their house. Exclusion criteria: families who had moved or changed source of drinking wter after the initial study. n = 13 children consuming water with manganese concentration < 0.030 mg/L (‘low exposure’ (LE) group) and 10 children consuming water with manganese concentration > 0.100 mg/L (‘high exposure’ (HE) group) Sex (% girls): 60 Age [mean (SD)]: 12.5 y (1.3) and 11.9 y (1.9)	MRI measurements: a standard pallidal index and a pericranial pallidal index were calculated, based on the signal intensity ratio of the globus pallidus relative to the frontal white matter or the pericranial muscles, respectively, on T1‐weighted images	Mean (SD) WMn (mg/L) LE group: 0.0009 (0.009) HE group: 0.145 (0.054) Mean (SD) manganese intake from water consumption (mg/month/kg) LE group: 0.0006 (0.0006) HE group: 0.058 (0.087) Mn concentration measured, by ICP‐MS, in home tap water samples. Mn intake estimated from water consumption with orally administered questionnaire answered by the child with parents help		None	Mean ranks in LE group versus HE group (p‐value for group comparison by Mann–Whitney U tests) Standard pallidal index: 12.9 versus 10.8, p = 0.5 Paricranial pallidal index: 14.9 versus 8.2, p = 0.02 T1 relaxation time: 9.7 versus 15.0, p = 0.07
Dion et al. (2018) Canada PC Follow‐up time 4–5 years Funding: public	N = 380 Population sample: Children in Quebec recruited in Bouchard et al. (2011) in 2007–2009, invited for a follow‐up in 2012–2013 Exclusion criteria: NR n = 287 Sex (% girls): 53 Age [mean (SD)]: 13.7 y (1.8)	Children's cognitive abilities assessed with WASI (Verbal IQ, Performance IQ and Full‐scale IQ)	Arithmetic mean/geometric mean (range) WMn (mg/L) at follow‐up: 0.058/0.015 (0.0002–0.961) Water was sampled directly from kitchen tap of children's homes and Mn measured with ICP‐MS Parents were interviewed to collect information on water supply to the house, including changes from original study Analyses used baseline MnW and a time‐averaged MnW (TAWMn), taking into consideration changes to water supply made in‐between studies	NR	Maternal nonverbal intelligence, maternal education and family income	Change in IQ scores per 10‐fold increase in baseline WMn (log ₁₀ ‐transformed) β (95% CI) for boys/girls Full‐scale IQ Unadjusted: 2.4 (0.1, 4.8)/−2.1 (−4.0, −0.3) Adjusted: 2.4 (0.3, 4.6)/−2.3 (−4.1, −0.6) Performance IQ Unadjusted: 3.8 (1.1, 6.5)/−2.9 (−5.0, −0.8) Adjusted: 3.9 (1.4, 6.4)/−2.8 (−4.8, −0.8) Verbal IQ Unadjusted: 1.1 (−1.7, 3.9)/−1.4 (−3.6, 0.9) Adjusted: 1.0 (−1.7, 3.7)/−1.9 (−4.2, 0.3) #p < 0.1, * p < 0.05, p < 0.01 Change in IQ scores per 10‐fold increase in TAWMn (μg/L) β (95% CI) for boys/girls Full‐scale IQ Unadjusted: 0.5 (−1.9, 2.9)/−2.0 (−4.0, 0.0) Adjusted: 0.3 (−1.9, 2.5)/−2.5 (−4.5, −0.7) Performance IQ Unadjusted: 1.5 (−1.3, 4.4)/−2.7 (−4.9, −0.5) Adjusted: 1.4 (−1.2, 4.0)/−3.1 (−5.2, −1.0) Verbal IQ** Unadjusted: −0.6 (−3.4, 2.3)/−1.3 (−3.7, 1.2) Adjusted: −0.7 (−3.5, 2.0)/−2.1 (−4.4, 0.3) #p < 0.1, * p < 0.05, **p < 0.01
Bouchard et al. (2018) Canada CS Funding: unclear	N = 307 Population sample: Children from 10 schools in rural areas of New Brunswick, Canada (2010–2012) Exclusion criteria: Children whose house was connected to municipal aqueduct and children who had been living in the current house <4 months or did not drink tap water N = 259 Sex (% girls): 51 Age [mean (range)]: 9.4 y (5.9–13.7)	IQ assessed with four sub‐tests of WISC‐IV (Verbal IQ, Performance IQ and Full‐scale IQ)	Median/arithmetic mean/geometric mean (range) WMn in home water (mg/L): 0.005/0.062/0.006 (0–1.046) WMn in school's water fountains < 0.060 mg/L in all schools except one in which it was 0.532 mg/L Mn concentration measured, by ICP‐MS, in home tap water samples and school's water fountains	Mean (SD) scores Full‐Scale IQ: 101.2 (12.9) Verbal IQ: 101.4 (11.8) Performance IQ: 100.6 (13.9)	Child's age, maternal nonverbal intelligence, maternal education, family income, IQ tester	Change in IQ scores per 10‐fold increase in WMn (log ₁₀ ‐transformed) β (95% CI) Full‐Scale IQ All: −0.58 (−2.27, 1.12) Boys/girls: −6.75 (−17.22, 3.72)/−1.29 (−3.21, 0.63) Verbal IQ All: −0.99 (−2.48, 0.05) Boys/girls: −2.41 (−11.63, 6.81)/−0.88 (−2.78, 1.01) Performance IQ All: 0.12 (−1.68, 1.91) Boys/girls: 1.99 (−0.75, 4.72)/−1.13 (−3.21, 0.95)
Rahman et al. (2017) MINIMat Bangladesh CS and PC Follow‐up 10 years Funding: public	N = 1607 Population sample: Mother–child cohort initially nested in a randomised food and micronutrient supplementation trial (MINIMat) conducted in pregnant women living in Matlab, in rural Bangladesh. Singleton children born between 2002 and 2003 were invited for a follow‐up examination after 10 years Exclusion criteria: NR n = 1265 Sex (% girls): 48 Age [mean (SD)]: 9.5 y (0.1)	Cognitive functions assessed with WISC‐IV (all sub‐tests) at 10 years of age Behaviour assessed using a 25‐item, parent‐reported SDQ at 10 years of age	Median (P5, P95) WMn During pregnancy (n = 1265): 0.204 (0.023, 2.494) At 5 y (n = 1162): 0.228 (0.0082, 2.605) At 10 y (n = 1530): 0.339 (0.004, 3.203) Drinking water sampled during pregnancy and after 5 and 10 years Mn measured by ICP‐MS	Mean (SD) WISC‐IV scores Full‐Scale IQ: 131 (34) SDQ scores Conduct: 2.9 (1.7) Attention: 4.5 (1.7) Relationships: 2.1 (1.3) Emotional: 1.6 (1.4) Prosocial: 6.6 (1.8)	Model 1: age and gender Model 2: mother's IQ, socioeconomic status, child age, gender, education, height for age at 10y, haemoglobin concentration, school type, Home Observation for Measurement of Environment (HOME), tester, number of siblings and urinary As concentrations at each respective timepoint (natural log–transformed)	No association between IQ and WMn exposure during pregnancy and at 5 and 10 y of age Estimates: β (95% CI) Exposure during pregnancy (n = 554) Full‐Scale IQ: 0.42 (−1.6, 2.5) Verbal: 0.070 (−0.62, 0.76) Perceptual reason: 0.16 (−0.65, 0.96) Working memory: 0.072 (−0.33, 0.47) Processing speed: 0.12 (−0.64, 0.88) In boys (n = 288) Full‐Scale IQ: −1.8 (−5.3, 1.7) Verbal: −0.62 (−1.8, 0.53) Perceptual reason: −0.22 (−1.6, 1.2) Working memory: −0.32 (−1.0, 0.39) Processing speed: −0.64 (−1.9, 0.59) In girls (n = 266/n = 27) Full‐Scale IQ: 5.2 (1.8, 8.6)/−5.4 (−13, 2.0) Verbal: 1.5 (0.31, 2.6)/−1.7 (−4.2, 0.81) Perceptual reason: 1.4 (0.03, 2.7)/1.3 (−4.2, 1.5) Working memory: 0.7 (0.08, 1.4)/−0.5 (−1.9, 0.88) Processing speed: 1.6 (0.29/3.0)/−1.9 (−4.8, 1.0) Exposure at 5y (n = 705) Full‐Scale IQ: −0.37 (−2.3, 1.5) Verbal: −0.10 (−0.73, 0.52) Perceptual reason: 0.10 (−0.64, 0.84) Working memory: −0.043 (−0.43, 0.34) Processing speed: −0.32 (−1.0, 0.40) In boys (n = 345) Full‐Scale IQ: −3.2 (−6.4, 0.065) Verbal: −0.88 (−1.9, 0.17) Perceptual reason: −1.1 (−2.3, 0.18) Working memory: −0.70 (−1.4, −0.022) Processing speed: −0.52 (−1.7, 0.67) In girls (n = 350)
						Full‐Scale IQ: 0.92 (−1.4, 3.3) Verbal: 0.26 (−0.53, 1.0) Perceptual reason: 0.65 (−0.27, 1.6) Working memory: 0.38 (−0.074, 0.84) Processing speed: −0.37 (−1.3, 0.57) Exposure at 10y (n = 801) Full‐Scale IQ: 0.18 (−1.3, 1.6) Verbal: −0.33 (−0.81, 0.15) Perceptual reason: −0.050 (−0.63, 0.53) Working memory: 0.12 (−0.17, 0.40) Processing speed: 0.44 (−0.13, 1.0) In boys (n = 406) Full‐Scale IQ: −0.56 (−3.0, 1.8) Verbal: −0.59 (−1.4, 0.17) Perceptual reason: −0.35 (−1.3, 0.60) Working memory: −0.058 (−0.53, 0.41) Processing speed: 0.44 (−0.44, 1.3) In girls (n = 395) Full‐Scale IQ: 0.70 (−1.2, 2.6) Verbal: −0.16 (−0.78, 0.46) Perceptual reason: 0.17 (−0.55, 0.90) Working memory: 0.23 (−0.13, 0.59) Processing speed: 0.45 (−0.31, 1.2) OR (95% CI) for raised conduct problems per 0.001 mg/L increase in WMn Exposure in pregnancy: 1.20 (1.04, 1.39), p = 0.013 5 y: 1.20 (1.03, 1.39), p = 0.016 10 y: 1.17 (1.04, 1.31). p = 0.007
Wasserman et al. (2006) HEALS Bangladesh CS Funding: unclear	N = 11,749 Population sample: Children who had been drinking water low in As (< 0.01 mg/L) and high in Mn. Recruited in two phases (2002 and 2004) Exclusion criteria: Children who had taken part in a previous study or had siblings (Wasserman, 2004) n = 142 Sex (% girls): 50 Age: 9.5–10.5 y	IQ assessed with WISC‐III (Verbal IQ, Performance IQ and Full‐scale IQ)	Mean (SD) WMn (mg/L): 0.795 (0.755) Mn measured in well water by ICP‐MS	Mean (SD) scores Full‐scale IQ: 71.2 (22.9) Verbal IQ: 15.9 (5.4) Performance IQ: 55.4 (18.9)	Maternal education, maternal intelligence, house type, TV access, height and head circumference	Association between IQ score and WMn (μg/L). β (unstandardised) Full‐Scale IQ Unadjusted: −5.20, p < 0.001 Adjusted: −4.35, p < 0.001 Performance IQ Unadjusted: −4.43, p < 0.001 Adjusted: −3.76, p < 0.001 Verbal sIQ Unadjusted: −0.80, p < 0.05 Adjusted: −0.63, p < 0.05
Wasserman et al. (2011) HEALS Bangladesh CS Funding: public	N ∼ 12,000 (HEALS adult cohort) Population sample: Children from adult participants of the HEALS cohort. Household were visited until ~75 children in each of four categories of WAs x WMn exposure had been enrolled (i.e. WAs < or >0.010 mg/L and WMn < or >0.500 mg/L) Exclusion criteria: twins and share the same as other participants n = 303 [Low As‐Low Mn (n = 77); Low As‐High Mn (n = 74); High As‐Low Mn (n = 73); High As‐High Mn (n = 79)] Sex (% girls): 50 Age [mean (SD)]: 9.64 (0.77)	Children's cognitive ability assessed with WISC‐IV; a Full‐Scale score (not standardised; calculated as the sum of all scores) and scores for specific cognitive domains (verbal comprehension, perceptual reasoning, working memory and processing speed) were analysed	Mean (SD) WMn (mg/L) Overall: 0.726 (0.730) Low As‐Low Mn: 0.202 (0.145) Low As‐High Mn: 1.111 (0.686) High As‐Low Mn: 0.184 (0.146) High As‐High Mn: 1.367 ± 0.692 Mn measured in drinking water wells and analysed with high resolution ICP‐MS	Mean (SD) scores Full‐Scale IQ 108.50 (31.32) Verbal Comprehension 23.80 (8.77) Perceptual Reasoning 31.24 (11.24) Working Memory 20.25 (6.06) Processing Speed 33.21 (13.48)	N/A	Mean (SD) raw scores by WAs x WMn categories Full‐Scale Low As‐Low Mn: 107.39 (30.90) Low As‐High Mn: 107.73 (27.34) High As‐Low Mn: 110.74 (32.51) High As‐High Mn: 108.22 (34.45) Verbal Comprehension Low As‐Low Mn: 25.56 (8.32) Low As‐High Mn: 22.93 (7.91) High As‐Low Mn: 23.82 (8.41) High As‐High Mn: 23.85 (10.26) Perceptual Reasoning Low As‐Low Mn: 30.21 (10.61) Low As‐High Mn: 30.69 (10.16) High As‐Low Mn: 32.86 (10.53) High As‐High Mn: 31.25 (13.30) Working Memory Low As‐Low Mn: 20.05 (5.79) Low As‐High Mn: 20.28 (6.01) High As‐Low Mn: 20.42 (6.87) High As‐High Mn: 20.24 (5.66) Processing Speed Low As‐Low Mn: 32.57 (12.12) Low As‐High Mn: 33.82 (11.35) High As‐Low Mn: 33.63 (13.46) High As‐High Mn: 32.87 (13.07)
Khan et al. (2012) Bangladesh CS Funding: public	N = 1925 Population sample: Subjects were recruited through 14 different school in Araihazar, a rural area of Bangladesh Exclusion criteria: Twins, disability orchronic disease. n = 840 [952 children from initial sample randomly selected, of which 840 children agreed to participate] Sex (% girls): 53 Age [median (range)]: 9.3 (8.0–11.0)	Children's annual scores from the academic achievement records in their school was obtained for three disciplines (Bangla, English and mathematics)	Mean (SD; range) WMn in wells (mg/L): 1.388 (0.866; 0.01–5.710) Mn measured in well water from children's homes and schools and analysed by ICP‐MS		Model 1: arsenic water concentration (Was) Model 2: log‐transformed WAs, school‐grade, maternal education, paternal education and head circumference and controlling for within‐teacher correlations in rating the children	Associations between WMn (> 0.400 vs. ≤0.400 mg/L) and academic scores β (95% CI) Bangla Language score Model 1: −0.16 (−5.45, 5.12) Model 2: −1.01 (−6.14, 4.12) English Language score Model 1: −2.33 (−6.93, 2.24) Model 2: −2.66 (−7.16, 1.83) Math score Model 1: −6.67 (−13.42, 0.25) p < 0.06 Model 2: −6.37 (−12.27, −0.46) p < 0.05 Adjusted mean (SEM) academic scores by categories of WMn Bangla Language score ≤ 0.400: 49.8 (3.1) 0.401–1.000: 48.1 (2.1) 1.001–1.440: 49 (2.1) 1.441–2.000: 48.1 (1.6) 2.001–6.000: 48.8 (2.1) English Language score ≤ 0.400: 51.3 (3.2) 0.401–1.000: 46.9 (1.9) 1.001–1.440: 49.3 (2.4) 1.441–2.000: 49.3 (1.7) 2.001–6.000: 48.8 (2.4) Math score ≤ 0.400: 58.9 (3.7) 0.401–1.000: 51.7 (1.9) 1.001–1.440: 52.8 (2.1) 1.441–2.000: 53.5 (1.7) 2.001–6.000: 53.0 (2.1)
Parvez et al. (2011) HEALS Bangladesh CS Funding: unclear	N ∼ 12,000 (HEALS adult cohort) Population sample: Children from adult participants of the HEALS cohort. Household were visited until ~ 75 children in each of four categories of WAs x WMn exposure had been enrolled (i.e. WAs < or > 0.010 mg/L and WMn < or > 0.500 mg/L) Exclusion criteria: twins and share the same as other participants. n = 303 [Low As‐Low Mn (n = 77); Low As‐High Mn (n = 74); High As‐Low Mn (n = 73); High As‐High Mn (n = 79)] Sex (% girls): 50 Age [mean (SD)]: 9.64 (0.77)	Motor function was assessed with BOT 2nd edition	Mean (SD) WMn (mg/L) Overall: 0.726 (0.730) Low As‐Low Mn: 0.202 (0.145) Low As‐High Mn: 1.111 (0.686) High As‐Low Mn: 0.184 (0.146) High As‐High Mn: 1.367 ± 0.692 Mn measured in drinking water wells and analysed with high resolution ICP‐MS	Mean (SD) Fine Manual Control 42.3 (8.6) Manual Coordination 38.9 (7.3) Body Coordination 41.3 (6.0) Strength and Agility 37.4 (3.7) Total Motor Composite 160.0 (18.5)	N/A	Mean (SD) BOT scores by WAs x WMn categories (p for ANOVA) Fine Manual Control Low As‐Low Mn: 43.7 (7.6) Low As‐High Mn: 42.8 (9.6) High As‐Low Mn: 41.7 (8.8) High As‐High Mn: 41.2 (8.2) p = 0.26 Manual Coordination Low As‐Low Mn: 39.7 (6.9) Low As‐High Mn: 39.2 (7.8) High As‐Low Mn: 38.4 (8.2) High As‐High Mn: 38.2 (6.4) p = 0.56 Body Coordination Low As‐Low Mn: 42.7 (5.9) Low As‐High Mn: 41.2 (6.0) High As‐Low Mn: 41.2 (6.5) High As‐High Mn: 40.0 (5.7) p = 0.07 Strength and Agility Low As‐Low Mn: 37.7 (3.7) Low As‐High Mn: 37.5 (3.5) High As‐Low Mn: 37.3 (3.9) High As‐High Mn: 37.2 (3.8) p = 0.81 Total Motor Composite Low As‐Low Mn: 162.5 (17.3) Low As‐High Mn: 161.2 (19.8) High As‐Low Mn: 159.2 (19.8) High As‐High Mn: 157.2 (17.2) p = 0.33
Miyake et al. (2021) KOMCHS Japan PC Follow‐up 5 years Funding: mixed	N = 1757 Population sample: Mother–child pairs who lived on eight prefectures in Japan in 2007–2008. They were recruited through hospitals Exclusion criteria: NR n = 1199 Sex (% girls): 53 Maternal age (mean): 32	Childhood behavioural problems assessed at 5 years of age using Japanese parent–report version of the Strengths and Difficulties Questionnaire (SDQ) which has five scales: emotional problems, conduct problems, hyperactivity, peer problems and prosocial scale	Median (IQR) Mn intake (mg/day): 3.6 (2.8–4.5) Quartile median Mn intake (mg/day) Q1: 2.4 Q2: 3.2 Q3: 4.0 Q4: 5.2 Dietary intake assessed with semi‐quantitative diet history questionnaire	Overall prevalence of emotional problems, conduct problems, hyperactivity problems, peer problems and low prosocial behaviour was 12.9%, 19.4%, 13.1%, 8.6% and 29.2%, respectively % cases in each quartile Emotional Problems: 13.7, 14.7, 13.3, 10.0 Conduct Problems: 19.4, 19.0, 22.0, 17.3 Hyperactivity Problems: 12.7, 16.0, 11.0, 12.7 Peer Problems: 10.4, 7.7, 9.0, 7.3 Low prosocial behaviour: 31.8, 32.0, 27.3, 25.7	Maternal age, gestation at baseline, region of residence at baseline, number of children at baseline, maternal and paternal education, household income, maternal depressive symptoms during pregnancy, maternal alcohol intake during pregnancy, maternal smoking during pregnancy, child's birth weight, child's sex, breastfeeding duration and smoking in the household during the first year of life. Model for Emotional problems is additionally adjusted for maternal intake of vitamin B2 and calcium. Model for Hyperactivity problems is additionally adjusted for maternal vitamin C, B2 and calcium. Model for Prosocial behaviour is additionally adjusted for maternal B2, vitamin V and folate.	Odds ratios (95% CI) for behavioural problems by quartile of maternal Mn intake during pregnancy. Emotional Problems Q1: ref Q2: 1.16 (0.71, 1.89) Q3: 1.03 (0.62, 1.71) Q4: 0.76 (0.42, 1.38) p per tend = 0.39 Conduct Problems Q1: ref Q2: 0.97 (0.64, 1.48) Q3: 1.28 (0.85, 1.93) Q4: 0.86 (0.56, 1.32) p per tend = 0.83 Hyperactivity Problems Q1: ref Q2: 1.45 (0.89, 2.36) Q3: 1.06 (0.62, 1.79) Q4: 1.35 (0.79, 2.32) p per tend = 0.52 Peer Problems Q1: ref Q2: 0.70 (0.39, 1.25) Q3: 0.89 (0.51, 1.55) Q4: 0.72 (0.39, 1.29) p per tend = 0.41 Low prosocial behaviour Q1: ref Q2: 1.08 (0.75, 1.56) Q3: 0.99 (0.67, 1.45) Q4: 0.98 (0.64, 1.49) p per tend = 0.83
Khan et al. (2011) HEALS Bangladesh CS Funding: public	N ∼ 12,000 (HEALS adult cohort) Population sample: Children from adult participants of the HEALS cohort; 772 children randomly selected; children who attended one of 10 elementary schools in the HEALS study area Exclusion criteria: Family had moved or wasn't home at the time of visit, shared same wells as other participants; ill or disabled children, twin siblings, irregular school attendance, family had switched wells since initial visit or refusing to participate n = 201 Sex (% girls): 50 Age [median (range)]: 9.6 y (8.0–10.9)	Children's classroom behaviour assessed with teacher reported CBLC‐TRF standardised form. Two subscales from internalising domain (anxious/depressed and withdrawn), and two from externalising domain (attention problems and aggressive behaviour) were used for the analysis	Mean/median (SD; range) WMn in wells (mg/L): 0.889/0.650 (0.784; 0.04–3.443) Mn measured in well water from children's homes and analysed by high resolution ICP‐MS	Mean (SD) score TRF total: 47.8 (18.3) TRF externalising behaviour: 31.7 (14.0) TRF internalising behaviour: 16.7 (6.1)	Model 1: water arsenic concentration (WAs) Model 2: WAs, sex, maternal education, arm circumference and log‐transformed BMI	Change in TRF scores per μg/L WMn increase β (95% CI) TRF Total scores Model 1: 2.59, (0.14, 5.02) p = 0.04 Model 2: 3.35 (0.86, 5.83), p = 0.008 TRF externalising scores Model 1: 2.04 (0.26, 3.81), p = 0.02 Model 2: 2.59 (0.81, 4.37), p = 0.004 TRF internalising scores Model 1: 0.60 (−0.11, 1.32), p = 0.10 Model 2: 0.82 (0.08, 1.56), p = 0.03
Rodrigues et al. (2016) Bangladesh PC Follow‐up max 49 months (from 1st trimester of pregnancy to 20–40‐month age) Funding: public	N = 812 Population sample: Mother–child pairs in Sirajdikhan and Pabna districts in Bangladesh recruited between 2008 and 2011; when children were 12–40 months, parents were offered to participate in this study Exclusion criteria: NR n = 524 Sex (% girls): 50 Age [median (range)]: 2.3 y (1.8–3.4)	Neurodevelopmental outcomes assessed at 20–40 months using the BSID‐III tests of cognitive function, receptive language, expressive language, fine motor and gross motor functions	Median (P25, P75) WMn (mg/L) Sirajdikhan: 1st trimester (n = 239): 0.54 (0.03, 0.96) 1 month (n = 238): 0.56 (0.039, 0.89) 12 months (n = 92): 1.15 (0.217, 2.0) 20–40 months (n = 239): 0.948 (0.164, 1.82) Pabna: 1st trimester (n = 286): 0.5 (0.27, 0.9) 1 month (n = 278): 0.545 (0.34, 1.06) 12 months (n = 99): 0.464 (0.244, 0.891) 20–40 months (n = 286): 0.515 (0.299, 0.969) Water sampled from tube wells used by the family as drinking water and Mn concentration analysed with ICP‐MS.	NR	Maternal age, maternal education, child's gender, exposure to second‐hand smoke, HOME score, maternal Raven score and child haematocrit levels. Models used age‐adjusted BSID‐III scores	Association between natural log WMn (μg/L) and cognitive scores β(SE) Sirajdikhan: 0.02 (0.02), p = 0.35 Pabna: −0.06 (0.07), p = 0.33 Association between natural log WMn and fine motor scores β(SE) Sirajdikhan: −0.04 (0.03), p = 0.21 Pabna: 0.85 (0.39), p = 0.03 No association found with children's scores on receptive and expressive language or gross motor domains (data not shown)
Schullehner et al. (2020) Denmark PC Follow‐up to 19 years Funding: public	N = 815,810 Population sample: All singletons born in Denmark between 1992 and 2007. Cohort members had to be living in Denmark on their fifth birthday and mothers had to be residents 9 months before giving birth as well as using water source with available data for > 7 months before pregnancy and > 4 years for the children Exclusion criteria: Children given a ADHD diagnoses before 5 years of age, and private well users n = 643,401 Sex (% girls): 49 Age: not reported	Clinical diagnoses of AHDH were obtained from the Danish Psychiatric Central Research Register and the Danish National Patient Register	Mn (mg/L) measured in 82,574 drinking water samples from 3509 active public waterworks. Exposure period from 1997 to 2012 Females N by exposure groups (mg/L) < 0.005: 60,501 0.005–0.019: 69,352 > 0.019–0.034: 58,403 > 0.034–0.100: 63,505 > 0.100: 61,652 Exposure groups (time‐weighted average Exposure, mg/L) n: < 0.005: 220,043 0.005–0.01: 31,832 > 0.01–0.025: 31,211 > 0.025: 30,327 Males N by exposure groups (mg/L): < 0.005: 63,144 0.005–0.019: 73,341 > 0.019–0.034: 61,454 > 0.034–0.100: 66,900 > 0.100: 65,149 Exposure groups (time‐weighted average Exposure, mg/L) n: < 0.005: 231312 0.005–0.01: 33631 > 0.01–0.025: 33112 > 0.025: 31933	Overall prevalence of ADHD diagnosis was 3.5%, or 22,730 out of 643,401 Female Exposure groups (highest exposure, mg/L) Cases/Person‐years at risk: < 0.005: 1082/535831 0.005–0.01: 1404/735445 > 0.01–0.025: 1355/695611 > 0.025: 1542/810171 Exposure groups (time‐weighted average Exposure, mg/L) n: < 0.005: 4648/2343484 0.005–0.01: 754/417242 > 0.01–0.025: 775/420688 > 0.025: 787/434323 Male Exposure groups (highest exposure, mg/L) Cases/Person‐years at risk: < 0.005: 2856/548913 0.005–0.01: 3301/765728 > 0.01–0.025: 2977/723183 > 0.025: 3168/846375 Exposure groups (time‐weighted average Exposure, mg/L) n: < 0.005: 10878/2427506 0.005–0.01: 1635/434162 > 0.01–0.025: 1654/439435 > 0.025: 1599/451072	For the main analysis – age and birth year. Fully adjusted analysis adjustment for maternal education, parental education and urbanicity	Hazard ratios (95% CI) for the association between time‐weighted average WMn during the first 5 y of life and ADHD‐Overall and ADHD subtypes. ADHD‐Overall (n = 6964) Females WMn < 0.005: ref WMn 0.005–0.01: 0.94 (0.80, 1.10) WMn > 0.01–0.025: 1.11 (0.96, 1.28) WMn > 0.025: 1.11 (0.98, 1.27) p‐trend: 0.055 Males WMn < 0.005: ref WMn 0.005–0.01: 0.96 (0.86, 1.06) WMn > 0.01–0.025: 1.07 (0.97, 1.18) WMn > 0.025 1.02 (0.93, 1.12) p‐trend: 0.33 ADHD‐Inattentive subtype Females WMn < 0.005: ref WMn 0.005–0.01: 1.16 (0.79, 1.72) WMn > 0.01–0.025: 1.53 (1.08, 2.17) WMn > 0.025: 1.31 (0.96, 1.79) p‐trend: 0.011 Males WMn < 0.005: ref WMn 0.005–0.01: 1.06 (0.78, 1.43) WMn > 0.01–0.025: 1.40 (1.06, 1.83) WMn > 0.025: 1.38 (1.08, 1.79) p‐trend: 0.0009 ADHD‐Combined subtype Females WMn < 0.005: ref WMn 0.005–0.01: 1.06 (0.87, 1.29) WMn > 0.01–0.025: 1.09 (0.90, 1.32) WMn > 0.025: 1.19 (1.01, 1.40) p‐trend: 0.038 Males WMn < 0.005: ref WMn 0.005–0.01: 0.92 (0.81, 1.04) WMn > 0.01–0.025: 1.09 (0.97, 1.22) WMn > 0.025: 0.97 (0.88, 1.08) p‐trend: 0.86

Author, year

Country

Study design

Follow‐up

Funding

Original cohort (N total)

Exclusion criteria

Study population (n)

Ascertainment of outcome

Exposure groups

n/person‐years

Exposure assessment method

Mean scores

Model covariates

Results

Nascimento et al. (2015, 2016)

Brazil

Funding: public

N = 90

Population sample:

Children from two different areas, rural and urban recruited through school principals

Exclusion criteria:

n for NEUPSILIN‐Inf battery

= 63 (rural: 43, urban: 20)

n for RCPM test

= 59 (rural: NR, urban: NR)

Sex (% girls): 49

Age: 6–12 y

Neuropsychological functions assessed by the Brazilian Child Brief Neuropsychological Assessment Battery NEUPSILIN‐Inf, composed of Attention, Perception, Working Memory, Phonological Awareness, Written Language and Executive Function

Raven's

Coloured Progressive Matrices (RCPM) test (nonverbal IQ)

Mean (min‐max) water manganese (WMn) concentration (mg/L)

Rural area: 0.02 (0.00003–0.28)

Urban area: 0.001 (0.001–0.002)

Water collected from tap in children's homes and Mn measured by ICP‐MS

Mean (SEM) scores

Attention

Rural: 44.7 (1.4)

Urban: 49.6 (1.3)

Perception

Rural: 3.4 (0.2)

Urban: 3.8 (0.1)

Working memory

Rural: 24.3 (3.9)

Urban: 34.3 (3.9)

Phonological awareness

Rural: 5.4 (0.8)

Urban: 8.4 (0.7)

Language

Rural: 50.9 (4.2)

Urban: 59.5 (3.7)

Go/no go task

Rural: 54.5 (3.8)

Urban: 53.7 (3.2)

Non verbal IQ

Rural: 39.1 (3.5)

Urban: 64.2 (4.6)

IQ, age, gender and parents' education

Change in NEUPSILIN‐Inf scores by WMn increase (mg/L)

Word/nonword reading: β = −0.307, p = 0.019

Word/nonword written: β = −0.253, p = 0.045

Written language: β = −0.361, n = 0.007

Language: β = −0.390, p = 0.002

Executive function: β = −0.547, p = 0.001

Visual attention, perception, working memory and phonological awareness: no association.

Spearman's correlation between WMn and nonverbal IQ score: r = −0.317, p = 0.014

Bouchard et al. (2011)

Canada

Funding: public

N = 696

Population sample:

Children living in the same house for > 3 months, in eight municipalities in Quebec, Canada, where presence of Mn is naturally elevated (2007–2009)

Exclusion criteria:

n = 362 (52% participation rate)

Sex (% girls): 54

Age [mean (SD)]: 9.3 y (1.8)

Children's cognitive abilities assessed with WASI (Verbal IQ, Performance IQ and Full‐scale IQ)

Median/arithmetic mean/geometric mean (P5, P95) WMn (mg/L):

0.031/0.098/0.020 (0.005, 0.255)

Median (P5, P95) Mn intake from water consumption (mg/kg bw per month):

0.008 (0, 0.286)

Median (P5, P95) Mn intake from diet (mg/kg bw per month): 2.335 (0.840, 6.418)

Water sampled directly from kitchen tap of children's homes and Mn measured with ICP‐MS

Mn intake estimated from diet and water consumption with orally administered semiquantitative FFQ answered by parent and child

Model 1: maternal education, maternal nonverbal intelligence, family income, home stimulation score and family structure

Model 2: model 1 + sex and age of child, IQ testing session time, source of water and iron concentration in tap water

Change in IQ for a 10‐fold increase in WMn (log ₁₀ ‐transformed)

β (95% CI)

Full‐scale IQ

Unadjusted model: −2.1 (−3.5, −0.8)

Model 1: −1.9 (−3.1, −0.7)

Model 2: −2.4 (−3.9, −0.9)

Performance IQ

Unadjusted model: −2.4 (−4.0, −0.7)

Model 1: −2.3 (−3.7, −0.8)

Model 2: −3.1 (−4.9, −1.3)

Verbal IQ

Unadjusted model: −1.4 (−2.6, −0.2)*

Model 1: −1.5 (−2.6, −0.3)

Model 2: −1.2 (−2.7, 0.3)

*p < 0.05, **p < 0.01

Change in IQ per 10‐fold increase Mn intake from water consumption (log ₁₀ ‐transformed)

β (95% CI)

Full‐scale IQ

Unadjusted model: −1.3 (−2.5, −0.2)

Model 1: −1.2 (−2.3, −0.1)

Model 2: −1.2 (−2.3, −0.1)

Performance IQ

Unadjusted model: −1.6 (−3.0, −0.3)

Model 1: −1.6 (−2.9, −0.3)

Model 2: −1.9 (−3.3, −0.4)

Verbal IQ

Unadjusted model: −0.7 (−1.7, 0.3)

Model 1: −0.6 (−1.6, 0.3)

Model 2: −0.3 (−1.4, 0.7)

*p < 0.05, **p < 0.01

No relationship with Mn intake from diet

Oulhote et al. (2014)

Canada

Funding: public

N = 696

Population sample:

Children living in the same house for > 3 months, in eight municipalities in Quebec, Canada, where presence of Mn is naturally elevated (2007–2009)

Exclusion criteria:

N = 375

Sex (% girls): 53

Age: 6–13 y

Children's cognition assessed with CVLT‐C (memory),

CPT II (sustained attention), Digit Span (working memory)

Children's motor function evaluated with Santa Ana Pegboard test and manual Finger tapping

Children's behaviour evaluated Conners' Rating Scales completed by a teacher (CRS‐T) and a parent

(CRS‐P)

Arithmetic mean/geometric mean (range) WMn (mg/L):

0.099/0.020 (0.001–2.791)

Geometric mean (range) Mn intake from water consumption (mg/kg bw per month):

0.0055 (0–1.059)

Mn concentration measured, by ICP‐MS, in home tap water samples.

Mn intake estimated from diet and water consumption with orally administered semiquantitative FFQ answered by parents and children

Child's sex, age, maternal education, maternal nonverbal intelligence, family income, maternal depressive symptoms and tap water lead concentrations

Change in outcome per 10‐fold increase in WMn (log ₁₀ ‐transformed)

β (95% CI)

Memory: −1.0 (−1.6, −0.4), p < 0.01

Attention: 0.5 (−0.4, 1.3), p = 0.31

Motor function: −1.2 (−2.7, 0.3), p = 0.11

Hyperactivity: −0.2 (−1.2, 0.8) p = 0.71

Change in outcome per 10‐fold increase Mn intake from water consumption (log ₁₀ ‐transformed)

β (95% CI)

Memory: −0.4 (−0.9, 0.1), p = 0.13

Attention: 0.1 (−0.6, 0.8), p = 0.80

Motor function: −1.13 (−2.4, −0.2), p = 0.02

Hyperactivity: 0.2 (−0.5, 0.9), p = 0.51

Dion et al. (2016)

Canada

Funding: public

N = 362

Population sample:

Children living in the same house for > 3 months, in 8 municipalities in Quebec, Canada, where presence of Mn is naturally elevated (2007–2009). Among the initial sample (Oulhote et al., 2014), children were selected based on the concentration of manganese in the tap water of their house.

Exclusion criteria: families who had moved or changed source of drinking wter after the initial study.

n = 13 children consuming water with manganese concentration < 0.030 mg/L (‘low exposure’ (LE) group) and 10 children consuming water with manganese concentration > 0.100 mg/L (‘high exposure’ (HE) group)

Sex (% girls): 60

Age [mean (SD)]: 12.5 y (1.3) and 11.9 y (1.9)

MRI measurements: a standard pallidal index and a pericranial pallidal index were calculated, based on the signal intensity ratio of the globus pallidus relative to the frontal white matter or the pericranial muscles, respectively, on T1‐weighted images

Mean (SD) WMn (mg/L)

LE group: 0.0009 (0.009)

HE group: 0.145 (0.054)

Mean (SD) manganese intake from water consumption (mg/month/kg)

LE group: 0.0006 (0.0006)

HE group: 0.058 (0.087)

Mn concentration measured, by ICP‐MS, in home tap water samples.

Mn intake estimated from water consumption with orally administered questionnaire answered by the child with parents help

None

Mean ranks in LE group versus HE group (p‐value for group comparison by Mann–Whitney U tests)

Standard pallidal index: 12.9 versus 10.8, p = 0.5

Paricranial pallidal index: 14.9 versus 8.2, p = 0.02

T1 relaxation time: 9.7 versus 15.0, p = 0.07

Dion et al. (2018)

Canada

Follow‐up time 4–5 years

Funding: public

N = 380

Population sample:

Children in Quebec recruited in Bouchard et al. (2011) in 2007–2009, invited for a follow‐up in 2012–2013

Exclusion criteria:

n = 287

Sex (% girls): 53

Age [mean (SD)]: 13.7 y (1.8)

Children's cognitive abilities assessed with WASI (Verbal IQ, Performance IQ and Full‐scale IQ)

Arithmetic mean/geometric mean (range) WMn (mg/L) at follow‐up: 0.058/0.015 (0.0002–0.961)

Water was sampled directly from kitchen tap of children's homes and Mn measured with ICP‐MS

Parents were interviewed to collect information on water supply to the house, including changes from original study

Analyses used baseline MnW and a time‐averaged MnW (TAWMn), taking into consideration changes to water supply made in‐between studies

Maternal nonverbal intelligence, maternal education and family income

Change in IQ scores per 10‐fold increase in baseline WMn (log ₁₀ ‐transformed)

β (95% CI) for boys/girls

Full‐scale IQ

Unadjusted: 2.4 (0.1, 4.8)/−2.1 (−4.0, −0.3)

Adjusted: 2.4 (0.3, 4.6)/−2.3 (−4.1, −0.6)

Performance IQ

Unadjusted: 3.8 (1.1, 6.5)/−2.9 (−5.0, −0.8)

Adjusted: 3.9 (1.4, 6.4)/−2.8 (−4.8, −0.8)

Verbal IQ

Unadjusted: 1.1 (−1.7, 3.9)/−1.4 (−3.6, 0.9)

Adjusted: 1.0 (−1.7, 3.7)/−1.9 (−4.2, 0.3)

#p < 0.1, * p < 0.05, **p < 0.01

Change in IQ scores per 10‐fold increase in TAWMn (μg/L)

β (95% CI) for boys/girls

Full‐scale IQ

Unadjusted: 0.5 (−1.9, 2.9)/−2.0 (−4.0, 0.0)

Adjusted: 0.3 (−1.9, 2.5)/−2.5 (−4.5, −0.7)

Performance IQ

Unadjusted: 1.5 (−1.3, 4.4)/−2.7 (−4.9, −0.5)

Adjusted: 1.4 (−1.2, 4.0)/−3.1 (−5.2, −1.0)

Verbal IQ

Unadjusted: −0.6 (−3.4, 2.3)/−1.3 (−3.7, 1.2)

Adjusted: −0.7 (−3.5, 2.0)/−2.1 (−4.4, 0.3)

#p < 0.1, * p < 0.05, **p < 0.01

Bouchard et al. (2018)

Canada

Funding: unclear

N = 307

Population sample:

Children from 10 schools in rural areas of New

Brunswick, Canada (2010–2012)

Exclusion criteria:

Children whose house was connected to municipal aqueduct and children who had been living in the current house <4 months or did not drink tap water

N = 259

Sex (% girls): 51

Age [mean (range)]: 9.4 y (5.9–13.7)

IQ assessed with four sub‐tests of WISC‐IV (Verbal IQ, Performance IQ and Full‐scale IQ)

Median/arithmetic mean/geometric mean (range) WMn in home water (mg/L): 0.005/0.062/0.006 (0–1.046)

WMn in school's water fountains < 0.060 mg/L in all schools except one in which it was 0.532 mg/L

Mn concentration measured, by ICP‐MS, in home tap water samples and school's water fountains

Mean (SD) scores

Full‐Scale IQ: 101.2 (12.9)

Verbal IQ: 101.4 (11.8)

Performance IQ: 100.6 (13.9)

Child's age, maternal nonverbal intelligence, maternal education, family income, IQ tester

Change in IQ scores per 10‐fold increase in WMn (log ₁₀ ‐transformed)

β (95% CI)

Full‐Scale IQ

All: −0.58 (−2.27, 1.12)

Boys/girls: −6.75 (−17.22, 3.72)/−1.29 (−3.21, 0.63)

Verbal IQ

All: −0.99 (−2.48, 0.05)

Boys/girls: −2.41 (−11.63, 6.81)/−0.88 (−2.78, 1.01)

Performance IQ

All: 0.12 (−1.68, 1.91)

Boys/girls: 1.99 (−0.75, 4.72)/−1.13 (−3.21, 0.95)

Rahman et al. (2017)

MINIMat

Bangladesh

CS and PC

Follow‐up 10 years

Funding: public

N = 1607

Population sample:

Mother–child cohort initially nested in a randomised food and micronutrient supplementation trial (MINIMat) conducted in pregnant women living in Matlab, in rural Bangladesh.

Singleton children born between 2002 and 2003 were invited for a follow‐up examination after 10 years

Exclusion criteria:

n = 1265

Sex (% girls): 48

Age [mean (SD)]: 9.5 y (0.1)

Cognitive functions assessed with WISC‐IV (all sub‐tests) at 10 years of age

Behaviour assessed using a 25‐item, parent‐reported SDQ at 10 years of age

Median (P5, P95) WMn

During pregnancy (n = 1265): 0.204 (0.023, 2.494)

At 5 y (n = 1162): 0.228 (0.0082, 2.605)

At 10 y (n = 1530): 0.339 (0.004, 3.203)

Drinking water sampled during pregnancy and after 5 and 10 years

Mn measured by ICP‐MS

Mean (SD)

WISC‐IV scores

Full‐Scale IQ: 131 (34)

SDQ scores

Conduct: 2.9 (1.7)

Attention: 4.5 (1.7)

Relationships: 2.1 (1.3)

Emotional: 1.6 (1.4)

Prosocial: 6.6 (1.8)

Model 1: age and gender

Model 2: mother's IQ, socioeconomic status, child age, gender, education, height for age at 10y, haemoglobin concentration, school type, Home Observation for Measurement of Environment (HOME), tester, number of siblings and urinary As concentrations at each respective timepoint (natural log–transformed)

No association between IQ and WMn exposure during pregnancy and at 5 and 10 y of age

Estimates: β (95% CI)

Exposure during pregnancy (n = 554)

Full‐Scale IQ: 0.42 (−1.6, 2.5)

Verbal: 0.070 (−0.62, 0.76)

Perceptual reason: 0.16 (−0.65, 0.96)

Working memory: 0.072 (−0.33, 0.47)

Processing speed: 0.12 (−0.64, 0.88)

In boys (n = 288)

Full‐Scale IQ: −1.8 (−5.3, 1.7)

Verbal: −0.62 (−1.8, 0.53)

Perceptual reason: −0.22 (−1.6, 1.2)

Working memory: −0.32 (−1.0, 0.39)

Processing speed: −0.64 (−1.9, 0.59)

In girls (n = 266/n = 27)

Full‐Scale IQ: 5.2 (1.8, 8.6)/−5.4 (−13, 2.0)

Verbal: 1.5 (0.31, 2.6)/−1.7 (−4.2, 0.81)

Perceptual reason: 1.4 (0.03, 2.7)/1.3 (−4.2, 1.5)

Working memory: 0.7 (0.08, 1.4)/−0.5 (−1.9, 0.88)

Processing speed: 1.6 (0.29/3.0)/−1.9 (−4.8, 1.0)

Exposure at 5y (n = 705)

Full‐Scale IQ: −0.37 (−2.3, 1.5)

Verbal: −0.10 (−0.73, 0.52)

Perceptual reason: 0.10 (−0.64, 0.84)

Working memory: −0.043 (−0.43, 0.34)

Processing speed: −0.32 (−1.0, 0.40)

In boys (n = 345)

Full‐Scale IQ: −3.2 (−6.4, 0.065)

Verbal: −0.88 (−1.9, 0.17)

Perceptual reason: −1.1 (−2.3, 0.18)

Working memory: −0.70 (−1.4, −0.022)

Processing speed: −0.52 (−1.7, 0.67)

In girls (n = 350)

Full‐Scale IQ: 0.92 (−1.4, 3.3)

Verbal: 0.26 (−0.53, 1.0)

Perceptual reason: 0.65 (−0.27, 1.6)

Working memory: 0.38 (−0.074, 0.84)

Processing speed: −0.37 (−1.3, 0.57)

Exposure at 10y (n = 801)

Full‐Scale IQ: 0.18 (−1.3, 1.6)

Verbal: −0.33 (−0.81, 0.15)

Perceptual reason: −0.050 (−0.63, 0.53)

Working memory: 0.12 (−0.17, 0.40)

Processing speed: 0.44 (−0.13, 1.0)

In boys (n = 406)

Full‐Scale IQ: −0.56 (−3.0, 1.8)

Verbal: −0.59 (−1.4, 0.17)

Perceptual reason: −0.35 (−1.3, 0.60)

Working memory: −0.058 (−0.53, 0.41)

Processing speed: 0.44 (−0.44, 1.3)

In girls (n = 395)

Full‐Scale IQ: 0.70 (−1.2, 2.6)

Verbal: −0.16 (−0.78, 0.46)

Perceptual reason: 0.17 (−0.55, 0.90)

Working memory: 0.23 (−0.13, 0.59)

Processing speed: 0.45 (−0.31, 1.2)

OR (95% CI) for raised conduct problems per 0.001 mg/L increase in WMn

Exposure in pregnancy: 1.20 (1.04, 1.39), p = 0.013

5 y: 1.20 (1.03, 1.39), p = 0.016

10 y: 1.17 (1.04, 1.31). p = 0.007

Wasserman et al. (2006)

HEALS

Bangladesh

Funding: unclear

N = 11,749

Population sample:

Children who had been drinking water low in As (< 0.01 mg/L) and high in Mn. Recruited in two phases (2002 and 2004)

Exclusion criteria:

Children who had taken part in a previous study or had siblings (Wasserman, 2004)

n = 142

Sex (% girls): 50

Age: 9.5–10.5 y

IQ assessed with WISC‐III (Verbal IQ, Performance IQ and Full‐scale IQ)

Mean (SD) WMn (mg/L): 0.795 (0.755)

Mn measured in well water by ICP‐MS

Mean (SD) scores

Full‐scale IQ: 71.2 (22.9)

Verbal IQ: 15.9 (5.4)

Performance IQ: 55.4 (18.9)

Maternal education, maternal intelligence, house type, TV access, height and head circumference

Association between IQ score and WMn (μg/L).

β (unstandardised)

Full‐Scale IQ

Unadjusted: −5.20, p < 0.001

Adjusted: −4.35, p < 0.001

Performance IQ

Unadjusted: −4.43, p < 0.001

Adjusted: −3.76, p < 0.001

Verbal sIQ

Unadjusted: −0.80, p < 0.05

Adjusted: −0.63, p < 0.05

Wasserman et al. (2011)

HEALS

Bangladesh

Funding: public

N ∼ 12,000 (HEALS adult cohort)

Population sample:

Children from adult participants of the HEALS cohort. Household were visited until ~75 children in each of four categories of WAs x WMn exposure had been enrolled (i.e. WAs < or >0.010 mg/L and WMn < or >0.500 mg/L)

Exclusion criteria: twins and share the same as other participants

n = 303 [Low As‐Low Mn (n = 77); Low As‐High Mn (n = 74); High As‐Low Mn (n = 73); High As‐High Mn (n = 79)]

Sex (% girls): 50

Age [mean (SD)]: 9.64 (0.77)

Children's cognitive ability assessed with WISC‐IV; a Full‐Scale score (not standardised; calculated as the sum of all scores) and scores for specific cognitive domains (verbal comprehension, perceptual reasoning, working memory and processing speed) were analysed

Mean (SD) WMn (mg/L)

Overall: 0.726 (0.730)

Low As‐Low Mn: 0.202 (0.145)

Low As‐High Mn: 1.111 (0.686)

High As‐Low Mn: 0.184 (0.146)

High As‐High Mn: 1.367 ± 0.692

Mn measured in drinking water wells and analysed with high resolution ICP‐MS

Mean (SD) scores

Full‐Scale IQ

108.50 (31.32)

Verbal Comprehension

23.80 (8.77)

Perceptual Reasoning

31.24 (11.24)

Working Memory

20.25 (6.06)

Processing Speed

33.21 (13.48)

N/A

Mean (SD) raw scores by WAs x WMn categories

Full‐Scale

Low As‐Low Mn: 107.39 (30.90)

Low As‐High Mn: 107.73 (27.34)

High As‐Low Mn: 110.74 (32.51)

High As‐High Mn: 108.22 (34.45)

Verbal Comprehension

Low As‐Low Mn: 25.56 (8.32)

Low As‐High Mn: 22.93 (7.91)

High As‐Low Mn: 23.82 (8.41)

High As‐High Mn: 23.85 (10.26)

Perceptual Reasoning

Low As‐Low Mn: 30.21 (10.61)

Low As‐High Mn: 30.69 (10.16)

High As‐Low Mn: 32.86 (10.53)

High As‐High Mn: 31.25 (13.30)

Working Memory

Low As‐Low Mn: 20.05 (5.79)

Low As‐High Mn: 20.28 (6.01)

High As‐Low Mn: 20.42 (6.87)

High As‐High Mn: 20.24 (5.66)

Processing Speed

Low As‐Low Mn: 32.57 (12.12)

Low As‐High Mn: 33.82 (11.35)

High As‐Low Mn: 33.63 (13.46)

High As‐High Mn: 32.87 (13.07)

Khan et al. (2012)

Bangladesh

Funding: public

N = 1925

Population sample:

Subjects were recruited through 14 different school in Araihazar, a rural area of Bangladesh

Exclusion criteria:

Twins, disability orchronic disease.

n = 840 [952 children from initial sample randomly selected, of which 840 children agreed to participate]

Sex (% girls): 53

Age [median (range)]: 9.3 (8.0–11.0)

Children's annual scores from the academic achievement records in their school was obtained for three disciplines (Bangla, English and mathematics)

Mean (SD; range) WMn in wells (mg/L): 1.388 (0.866; 0.01–5.710)

Mn measured in well water from children's homes and schools and analysed by ICP‐MS

Model 1: arsenic water concentration (Was)

Model 2: log‐transformed WAs, school‐grade, maternal education, paternal education and head circumference and controlling for

within‐teacher correlations in rating the children

Associations between WMn (> 0.400 vs. ≤0.400 mg/L) and academic scores

β (95% CI)

Bangla Language score

Model 1: −0.16 (−5.45, 5.12)

Model 2: −1.01 (−6.14, 4.12)

English Language score

Model 1: −2.33 (−6.93, 2.24)

Model 2: −2.66 (−7.16, 1.83)

Math score

Model 1: −6.67 (−13.42, 0.25) p < 0.06

Model 2: −6.37 (−12.27, −0.46) p < 0.05

Adjusted mean (SEM) academic scores by categories of WMn

Bangla Language score

≤ 0.400: 49.8 (3.1)

0.401–1.000: 48.1 (2.1)

1.001–1.440: 49 (2.1)

1.441–2.000: 48.1 (1.6)

2.001–6.000: 48.8 (2.1)

English Language score

≤ 0.400: 51.3 (3.2)

0.401–1.000: 46.9 (1.9)

1.001–1.440: 49.3 (2.4)

1.441–2.000: 49.3 (1.7)

2.001–6.000: 48.8 (2.4)

Math score

≤ 0.400: 58.9 (3.7)

0.401–1.000: 51.7 (1.9)

1.001–1.440: 52.8 (2.1)

1.441–2.000: 53.5 (1.7)

2.001–6.000: 53.0 (2.1)

Parvez et al. (2011)

HEALS

Bangladesh

Funding: unclear

N ∼ 12,000 (HEALS adult cohort)

Population sample:

Children from adult participants of the HEALS cohort. Household were visited until ~ 75 children in each of four categories of WAs x WMn exposure had been enrolled (i.e. WAs < or > 0.010 mg/L and WMn < or > 0.500 mg/L)

Exclusion criteria: twins and share the same as other participants.

n = 303 [Low As‐Low Mn (n = 77); Low As‐High Mn (n = 74); High As‐Low Mn (n = 73); High As‐High Mn (n = 79)]

Sex (% girls): 50

Age [mean (SD)]: 9.64 (0.77)

Motor function was assessed with BOT 2nd edition

Mean (SD) WMn (mg/L)

Overall: 0.726 (0.730)

Low As‐Low Mn: 0.202 (0.145)

Low As‐High Mn: 1.111 (0.686)

High As‐Low Mn: 0.184 (0.146)

High As‐High Mn: 1.367 ± 0.692

Mn measured in drinking water wells and analysed with high resolution ICP‐MS

Mean (SD)

Fine Manual Control

42.3 (8.6)

Manual Coordination

38.9 (7.3)

Body Coordination

41.3 (6.0)

Strength and Agility

37.4 (3.7)

Total Motor Composite

160.0 (18.5)

N/A

Mean (SD) BOT scores by WAs x WMn categories (p for ANOVA)

Fine Manual Control

Low As‐Low Mn: 43.7 (7.6)

Low As‐High Mn: 42.8 (9.6)

High As‐Low Mn: 41.7 (8.8)

High As‐High Mn: 41.2 (8.2)

p = 0.26

Manual Coordination

Low As‐Low Mn: 39.7 (6.9)

Low As‐High Mn: 39.2 (7.8)

High As‐Low Mn: 38.4 (8.2)

High As‐High Mn: 38.2 (6.4)

p = 0.56

Body Coordination

Low As‐Low Mn: 42.7 (5.9)

Low As‐High Mn: 41.2 (6.0)

High As‐Low Mn: 41.2 (6.5)

High As‐High Mn: 40.0 (5.7)

p = 0.07

Strength and Agility

Low As‐Low Mn: 37.7 (3.7)

Low As‐High Mn: 37.5 (3.5)

High As‐Low Mn: 37.3 (3.9)

High As‐High Mn: 37.2 (3.8)

p = 0.81

Total Motor Composite

Low As‐Low Mn: 162.5 (17.3)

Low As‐High Mn: 161.2 (19.8)

High As‐Low Mn: 159.2 (19.8)

High As‐High Mn: 157.2 (17.2)

p = 0.33

Miyake et al. (2021)

KOMCHS

Japan

Follow‐up 5 years

Funding: mixed

N = 1757

Population sample:

Mother–child pairs who lived on eight prefectures in Japan in 2007–2008. They were recruited through hospitals

Exclusion criteria: NR

n = 1199

Sex (% girls): 53

Maternal age (mean): 32

Childhood behavioural problems assessed at 5 years of age using Japanese parent–report version of the Strengths and Difficulties Questionnaire (SDQ) which has five scales: emotional problems, conduct problems, hyperactivity, peer problems and prosocial scale

Median (IQR) Mn intake (mg/day): 3.6 (2.8–4.5)

Quartile median Mn intake (mg/day)

Q1: 2.4

Q2: 3.2

Q3: 4.0

Q4: 5.2

Dietary intake assessed with semi‐quantitative diet history questionnaire

Overall prevalence of emotional problems, conduct problems,

hyperactivity problems, peer problems and low

prosocial behaviour was 12.9%, 19.4%, 13.1%, 8.6% and 29.2%, respectively

% cases in each quartile

Emotional Problems: 13.7, 14.7, 13.3, 10.0

Conduct Problems: 19.4, 19.0, 22.0, 17.3

Hyperactivity Problems: 12.7, 16.0, 11.0, 12.7

Peer Problems: 10.4, 7.7, 9.0, 7.3

Low prosocial behaviour: 31.8, 32.0, 27.3, 25.7

Maternal age, gestation at baseline, region of residence at baseline, number of children at baseline, maternal and paternal education, household income, maternal depressive symptoms during pregnancy,

maternal alcohol intake during pregnancy, maternal smoking during pregnancy, child's birth weight, child's sex, breastfeeding duration and smoking in the household during the first year of life. Model for Emotional problems is additionally adjusted for maternal intake of vitamin B2 and calcium. Model for Hyperactivity problems is additionally adjusted for maternal vitamin C, B2 and calcium.

Model for Prosocial behaviour is additionally adjusted for maternal B2, vitamin V and folate.

Odds ratios (95% CI) for behavioural problems by quartile of maternal Mn intake during pregnancy.

Emotional Problems

Q1: ref

Q2: 1.16 (0.71, 1.89)

Q3: 1.03 (0.62, 1.71)

Q4: 0.76 (0.42, 1.38)

p per tend = 0.39

Conduct Problems

Q1: ref

Q2: 0.97 (0.64, 1.48)

Q3: 1.28 (0.85, 1.93)

Q4: 0.86 (0.56, 1.32)

p per tend = 0.83

Hyperactivity Problems

Q1: ref

Q2: 1.45 (0.89, 2.36)

Q3: 1.06 (0.62, 1.79)

Q4: 1.35 (0.79, 2.32)

p per tend = 0.52

Peer Problems

Q1: ref

Q2: 0.70 (0.39, 1.25)

Q3: 0.89 (0.51, 1.55)

Q4: 0.72 (0.39, 1.29)

p per tend = 0.41

Low prosocial behaviour

Q1: ref

Q2: 1.08 (0.75, 1.56)

Q3: 0.99 (0.67, 1.45)

Q4: 0.98 (0.64, 1.49)

p per tend = 0.83

Khan et al. (2011)

HEALS

Bangladesh

Funding: public

N ∼ 12,000 (HEALS adult cohort)

Population sample:

Children from adult participants of the HEALS cohort; 772 children randomly selected; children who attended one of 10 elementary schools in the HEALS study area

Exclusion criteria:

Family had moved or wasn't home at the time of visit, shared same wells as other participants; ill or disabled children, twin siblings, irregular school attendance, family had switched wells since initial visit or refusing to participate

n = 201

Sex (% girls): 50

Age [median (range)]: 9.6 y (8.0–10.9)

Children's classroom behaviour assessed with teacher reported CBLC‐TRF standardised form.

Two subscales from internalising domain (anxious/depressed and withdrawn), and two from externalising domain (attention problems and aggressive behaviour) were used for the analysis

Mean/median (SD; range) WMn in wells (mg/L): 0.889/0.650 (0.784;

0.04–3.443)

Mn measured in well water from children's homes and analysed by high resolution ICP‐MS

Mean (SD) score

TRF total: 47.8 (18.3)

TRF externalising behaviour: 31.7 (14.0)

TRF internalising behaviour: 16.7 (6.1)

Model 1: water arsenic concentration (WAs)

Model 2: WAs, sex, maternal education, arm circumference and log‐transformed BMI

Change in TRF scores per μg/L WMn increase

β (95% CI)

TRF Total scores

Model 1: 2.59, (0.14, 5.02) p = 0.04

Model 2: 3.35 (0.86, 5.83), p = 0.008

TRF externalising scores

Model 1: 2.04 (0.26, 3.81), p = 0.02

Model 2: 2.59 (0.81, 4.37), p = 0.004

TRF internalising scores

Model 1: 0.60 (−0.11, 1.32), p = 0.10

Model 2: 0.82 (0.08, 1.56), p = 0.03

Rodrigues et al. (2016)

Bangladesh

Follow‐up max 49 months (from 1st trimester of pregnancy to 20–40‐month age)

Funding: public

N = 812

Population sample:

Mother–child pairs in Sirajdikhan and Pabna districts in Bangladesh recruited between 2008 and 2011; when children were 12–40 months, parents were offered to participate in this study

Exclusion criteria: NR

n = 524

Sex (% girls): 50

Age [median (range)]: 2.3 y (1.8–3.4)

Neurodevelopmental outcomes assessed at 20–40 months using the BSID‐III tests of cognitive function, receptive language,

expressive language, fine motor and gross motor functions

Median (P25, P75) WMn (mg/L)

Sirajdikhan:

1st trimester (n = 239): 0.54 (0.03, 0.96)

1 month (n = 238): 0.56 (0.039, 0.89)

12 months (n = 92): 1.15 (0.217, 2.0)

20–40 months (n = 239): 0.948 (0.164, 1.82)

Pabna:

1st trimester (n = 286): 0.5 (0.27, 0.9)

1 month (n = 278): 0.545 (0.34, 1.06)

12 months (n = 99): 0.464 (0.244, 0.891)

20–40 months (n = 286): 0.515 (0.299, 0.969)

Water sampled from tube wells used by the family as drinking water and Mn concentration analysed with ICP‐MS.

Maternal age, maternal education, child's gender, exposure to second‐hand smoke, HOME score, maternal Raven score and child

haematocrit levels. Models used age‐adjusted BSID‐III scores

Association between natural log WMn (μg/L) and cognitive scores

β(SE)

Sirajdikhan: 0.02 (0.02), p = 0.35

Pabna: −0.06 (0.07), p = 0.33

Association between natural log WMn and fine motor scores

β(SE)

Sirajdikhan: −0.04 (0.03), p = 0.21

Pabna: 0.85 (0.39), p = 0.03

No association found with children's

scores on receptive and expressive language or gross motor domains (data not shown)

Schullehner et al. (2020)

Denmark

Follow‐up to 19 years

Funding: public

N = 815,810

Population sample:

All singletons born in Denmark between 1992 and 2007. Cohort members had to be living in Denmark on their fifth birthday and mothers had to be residents 9 months before giving birth as well as using water source with available data for > 7 months before pregnancy and > 4 years for the children

Exclusion criteria:

Children given a ADHD diagnoses before 5 years of age, and private well users

n = 643,401

Sex (% girls): 49

Age: not reported

Clinical diagnoses of AHDH were obtained from the Danish Psychiatric Central Research Register and the Danish National Patient Register

Mn (mg/L) measured in 82,574 drinking water samples from 3509 active public waterworks.

Exposure period from 1997 to 2012

Females

N by exposure groups (mg/L) < 0.005: 60,501

0.005–0.019: 69,352

> 0.019–0.034: 58,403

> 0.034–0.100: 63,505

> 0.100: 61,652

Exposure groups (time‐weighted average

Exposure, mg/L) n:

< 0.005: 220,043

0.005–0.01: 31,832

> 0.01–0.025: 31,211

> 0.025: 30,327

Males

N by exposure groups (mg/L):

< 0.005: 63,144

0.005–0.019: 73,341

> 0.019–0.034: 61,454

> 0.034–0.100: 66,900

> 0.100: 65,149

Exposure groups (time‐weighted average

Exposure, mg/L) n:

< 0.005: 231312

0.005–0.01: 33631

> 0.01–0.025: 33112

> 0.025: 31933

Overall prevalence of ADHD diagnosis was 3.5%, or 22,730 out of 643,401

Female

Exposure groups (highest exposure, mg/L)

Cases/Person‐years at risk:

< 0.005: 1082/535831

0.005–0.01: 1404/735445

> 0.01–0.025: 1355/695611

> 0.025: 1542/810171

Exposure groups (time‐weighted average

Exposure, mg/L) n: < 0.005: 4648/2343484

0.005–0.01: 754/417242

> 0.01–0.025: 775/420688

> 0.025: 787/434323

Male

Exposure groups (highest exposure, mg/L)

Cases/Person‐years at risk:

< 0.005: 2856/548913

0.005–0.01: 3301/765728

> 0.01–0.025: 2977/723183

> 0.025: 3168/846375

Exposure groups (time‐weighted average

Exposure, mg/L) n: < 0.005: 10878/2427506

0.005–0.01: 1635/434162

> 0.01–0.025: 1654/439435

> 0.025: 1599/451072

For the main analysis – age and birth year.

Fully adjusted analysis adjustment for maternal education, parental education and urbanicity

Hazard ratios (95% CI) for the association between time‐weighted average WMn during the first 5 y of life and ADHD‐Overall and ADHD subtypes.

ADHD‐Overall (n = 6964)

Females

WMn < 0.005: ref

WMn 0.005–0.01: 0.94 (0.80, 1.10)

WMn > 0.01–0.025: 1.11 (0.96, 1.28)

WMn > 0.025: 1.11 (0.98, 1.27)

p‐trend: 0.055

Males

WMn < 0.005: ref

WMn 0.005–0.01: 0.96 (0.86, 1.06)

WMn > 0.01–0.025: 1.07 (0.97, 1.18)

WMn > 0.025 1.02 (0.93, 1.12)

p‐trend: 0.33

ADHD‐Inattentive subtype

Females

WMn < 0.005: ref

WMn 0.005–0.01: 1.16 (0.79, 1.72)

WMn > 0.01–0.025: 1.53 (1.08, 2.17)

WMn > 0.025: 1.31 (0.96, 1.79)

p‐trend: 0.011

Males

WMn < 0.005: ref

WMn 0.005–0.01: 1.06 (0.78, 1.43)

WMn > 0.01–0.025: 1.40 (1.06, 1.83)

WMn > 0.025: 1.38 (1.08, 1.79)

p‐trend: 0.0009

ADHD‐Combined subtype

Females

WMn < 0.005: ref

WMn 0.005–0.01: 1.06 (0.87, 1.29)

WMn > 0.01–0.025: 1.09 (0.90, 1.32)

WMn > 0.025: 1.19 (1.01, 1.40)

p‐trend: 0.038

Males

WMn < 0.005: ref

WMn 0.005–0.01: 0.92 (0.81, 1.04)

WMn > 0.01–0.025: 1.09 (0.97, 1.22)

WMn > 0.025: 0.97 (0.88, 1.08)

p‐trend: 0.86

Abbreviations: ADHD, Attention‐deficit hyperactivity disorder; BSID‐III, Bayley Scales of Infant and Toddler Development, 3rd edition; BOT 2nd, Bruininks‐Oseretsky test, 2nd edition; CBLC‐TRF, Child Behaviour Checklist‐Teacher's Report Form; CI, Confidence Interval; CPT, Conners' Continuous Performance Test II; CS, Cross‐sectional; CRS, Conners' Rating Scale; CVLT‐C, California Verbal Learning Test–Children's Version; FFQ, food frequency questionnaire; HE, high exposure; ICP‐MS, Inductively coupled plasma mass spectrometry; IQ, Intelligence quotient; IQR, Interquartile range; LE, low exposure; Mn, manganese; NR, not reported; PC, Prospective cohort; RCPM, Raven's Coloured Progressive Matrices; SD, Standard deviation; SDQ, Strengths and Difficulties Questionnaire; SEM, Standard Error of the Mean; WAs, water arsenic concentration; WMn, water manganese concentration; WASI, Wechsler Abbreviated Scale of Intelligence; WISC, Wechsler Intelligence Scale for Children y, year.