Essential and Potentially Toxic Elements in Commercial Milk Formulas: Health Risk Assessment Through a Systematic Review and Meta-analysis

Overview

Largueza and colleagues at the Federal University of Rio de Janeiro State (UNIRIO) and partner Brazilian institutions conducted a PRISMA-registered systematic review and random-effects meta-analysis (REML) of essential and potentially toxic elements in commercial milk formulas (CMF) for children aged 0 to 3 years. Thirty observational studies were synthesized qualitatively and eighteen provided sufficient data for the meta-analysis, which pooled element concentrations across three Codex Alimentarius CMF categories (infant formula IF, follow-up formula for older infants FFI, and product for young children PYC) and stratified by WHO region. Non-carcinogenic risk was assessed via the EPA Target Hazard Quotient (THQ) framework using EPA RfDo values and product-specific estimated daily intake equations. The review documents Total Target Hazard Quotient (TTHQ) values above one in several countries for all three formula categories, identifies the elements driving the excess risk on a per-country basis, and concludes that continuous monitoring of CMF is necessary to safeguard children’s health. This is the first global systematic review-with-meta-analysis covering essential elements, potentially toxic elements, and the Codex follow-up and young-children categories alongside infant formula in a single THQ framework.

Key numbers

The qualitative-synthesis corpus is 30 studies (Table 1). Geographic distribution: Asia 33% (6 studies including Turkey, Iran, Saudi Arabia, Jordan, Kuwait), Americas 33% (Brazil 5, USA 6 — six U.S. studies including Frisbie 2019, Scher 2021/2024, Mitchell 2020, Da Silva 2018; one Italian-authored study), Europe 28% (Spain 2, Italy, France, Malta — 5 studies), Africa 6% (Ethiopia, Egypt; 3 studies). One study received private funding but declared no conflict of interest. Methods of determination: ICP-MS (47% of studies), AAS (20%), ICP-OES (17%); the rest used FAAS, GF-AAS, INAA, MP-AES, or PIXE/RBS per the Table 1 method column.

The meta-analysis pooled 18 studies. Nine of the 30 included studies did not report concentrations in means with standard deviation or standard error and were excluded from the meta-analysis; three more were excluded because their reported concentration values exceeded three standard deviations from the pooled distribution of study means and lacked sufficient methodological detail to confirm analytical comparability.

Table 2 — pooled weighted-mean-difference (WMD) concentrations across all included regions (µg/kg, except Fe and Zn in mg/kg as marked with *).

For infant formulas (IF, 0-6 months): aluminum (Al) 946.75 µg/kg (95% CI 169.35-1724.14; N=10 studies), inorganic arsenic (iAs) 12.2 (5.66-18.74; N=10), cadmium (Cd) 4.13 (2.86-5.41; N=9), Co 11.55 (7.26-15.84; N=19), chromium (Cr) 250.37 (148.97-351.77; N=10), Cu 1882.7 (1366.62-2398.77; N=25), Fe* 34.28 mg/kg (26.45-42.1; N=26), methylmercury (MeHg) 10.26 (0-29.52; N=3), Mn 1566.66 (774.59-2358.72; N=15), nickel (Ni) 131.14 (0-336.7; N=4), lead (Pb) 42.98 (7.6-78.36; N=8), Se 180.64 (132.21-229.06; N=19), uranium (U) 3.19 (1.61-4.77; N=16), Zn* 26.84 mg/kg (18.25-35.43; N=15). All Iˢ² values exceeded 90% indicating substantial heterogeneity (p<0.001 throughout).

For follow-up formulas (FFI, 6-12 months): Al 740.02 µg/kg (349.23-1130.82; N=7), iAs 12.59 (6.86-18.32; N=12), Cd 3.36 (1.06-5.65; N=6), Co 7.91 (4.04-11.78; N=6), Cr 261.89 (162.26-361.52; N=9), Cu 1947.63 (1166.61-2728.65; N=12), Fe* 45.07 mg/kg (27.36-62.78; N=13), MeHg 0.39 (0-1.06; N=2), Mn 1491.78 (752.54-2231.03; N=13), Ni 406.35 (0-864.59; N=4), Pb 45.5 (16.33-74.66; N=11), Se 336.88 (107.49-566.27; N=7), U 6.12 (1.72-10.51; N=6), Zn* 31.31 mg/kg (22.54-40.08; N=18).

For products for young children (PYC, 1-3 years): Al 2052.61 µg/kg (173.67-3931.52; N=9), iAs 23.99 (10.82-37.17; N=11), Cd 6.44 (3.64-9.25; N=8), Co 0.90 (0.80-1.00; N=1), Cr 27.70 (16.39-39.01; N=1), Fe* 8.89 mg/kg (6.58-11.21; N=1), MeHg 0.14 (0-0.03; N=1), Mn 649.86 (0-1649.41; N=2), Ni 1358.2 (761.2-1955.21; N=7), Pb 16.9 (0-37.48; N=7), Se 20.00 (11.65-28.35; N=1), U 6.65 (4.19-9.11; N=6). PYC Cu and Zn were not reported in Table 2.

The text rank-order for IF was Fe>Zn>Cu>Mn>Al>Cr>Se>Ni>Pb>iAs>Co>MeHg>Cd>U. For FFI it was Fe>Zn>Cu>Mn>Al>Ni>Se>Cr>Pb>iAs>Co>U>Cd>MeHg. For PYC it was Fe>Al>Ni>Mn>Cr>iAs>Se>Pb>U>Cd>Co>MeHg.

Table 3 — pooled concentrations stratified by WHO region. Eastern Mediterranean Region (EMR) IF samples carried the highest concentrations of Al (4600 µg/kg, 95% CI 3889.87-5310.13; N=1 study), Cd (6.10), Co (55.40), Cr (588 µg/kg, CI 516.8-659.21; N=1), Ni (344 µg/kg, CI 169.54-518.46; N=1), and Zn (32.90 mg/kg). Region of the Americas (AMR) IF samples carried the highest Cu (2672.33 µg/kg, CI 1667.44-3677.22), Mn (1845.78 µg/kg, CI 832.66-2858.91), and U (8.36 µg/kg). European Region (EUR) IF samples carried the highest Pb (53.68 µg/kg, CI 0-124.26; N=4) and iAs (9.30 µg/kg, though the AMR value of 22.41 was actually higher — the text states “highest concentrations of Pb (53.68 µg kg⁻¹) and iAs (9.30 µg kg⁻¹) belonged to the European Region”; we record this as published).

For FFI, AMR samples carried the highest Al (838.25 µg/kg, CI 516.68-1159.82; N=4), iAs (21.25), Cd (4.92), Co (10), Cr (340.18), Cu (2127.23), Fe (52.94 mg/kg), Mn (1853.79 µg/kg, CI 935.79-2770.81), Se (462.5), and U (7.22). EUR FFI samples carried the highest MeHg (0.39, CI 0-1.06; N=2), Ni (406.35), and Pb (60.92 µg/kg, CI 0-135.97; N=4). African Region (AFR) FFI samples carried the highest Zn (45.43 mg/kg, CI 31.35-59.5; N=5).

For PYC, AMR samples carried the highest Al (1522.47 µg/kg) and iAs (40.03 µg/kg, CI 24.79-54.27; N=6); the EMR value of 7.4 µg/kg (CI 5.55-9.24; N=3) and the EUR value of 1.614 µg/kg (CI 0.5-2.73; N=2) were substantially lower. EUR PYC samples carried the highest Ni (550 µg/kg, CI 341.07-758.93; N=1), Cr (27.7), Se (20), Fe (8.9 mg/kg), Pb (6.96), Co (0.9), and MeHg (0.14). EMR PYC samples carried the highest Mn (1160 µg/kg) and Al (8690 µg/kg, CI 8169.47-9210.53; N=1).

Table 4 — non-carcinogenic risk (THQ) by formula type using EPA RfDo values. RfDo values used (mg/kg body weight/day): Al 1.0, iAs 0.0003, Cd 0.0001, Co 0.0003, Cr 1.5, Cu 0.04, Fe 0.7, MeHg 0.0001, Mn 0.14, Ni 0.011, Se 0.005, U 0.0002, Zn 0.3. THQ for Pb and Ti was not calculated because the EPA had not established RfDo values for these elements as of the publication. Average daily intake was assumed to be 780 mL/day for 0-3 months, 1,200 mL/day for 3-6 months, 500 mL/day for 6-12 months and 1-3 years. Reference body weights were 6.1 kg (0-6 months), 9.1 kg (6-12 months), and 12.1 kg (1-3 years). Powder per 100 kcal reference values used were IF 20.28 g/100 kcal, FFI 20.78 g/100 kcal, PYC 21.26 g/100 kcal. iAs THQ was calculated assuming total arsenic (As) represents inorganic As (a deliberately conservative worst-case interpretation, since most included studies measured total As rather than speciated As).

For IF the mean THQ values were Al 0.019, iAs 0.834, Cd 0.847, Co 0.79, Cr 0.003, Cu 0.965, Fe 1.004, MeHg 2.104, Mn 0.229, Ni 0.244, Se 0.741, U 0.327, Zn 1.835. Mean THQ values exceeding 1.0 (probable health risk): MeHg (2.104), Fe (1.004), Zn (1.835). At the worst-case (maximum value within the 95% CI): iAs (1.281), Cd (1.109), Co (1.083), Cu (1.23), Fe (1.233), MeHg (6.054), Zn (2.422) also exceeded 1.0.

For FFI the mean THQ values were Al 0.011, iAs 0.632, Cd 0.506, Co 0.397, Cr 0.003, Cu 0.733, Fe 0.969, MeHg 0.059, Mn 0.16, Ni 0.556, Se 1.014, U 0.461, Zn 1.571. Mean THQ values exceeding 1.0: Se (1.014), Zn (1.571). Maximum-value THQ exceeding 1.0: iAs (0.919 — just under), Cd (0.851), Cu (1.027), Fe (1.35), Ni (1.183), Se (1.705), Zn (2.011).

For PYC the mean THQ values were Al 0.013, iAs 0.496, Cd 0.399, Co 0.019, Cu — (nd), Fe 0.079, MeHg 0.009, Mn 0.029, Ni 0.765, Se 0.025, U 0.206. Maximum-value THQ exceeding 1.0 was Ni (1.102) alone.

Table 5 — Total Target Hazard Quotient (TTHQ) and per-country average THQ. TTHQ was calculated as the sum of individual THQs for the elements reported per country. For IF the country ranking by TTHQ was Jordan (12.839) > Brazil (10.423) > Turkey (8.02) > Spain (5.311) > Malta (4.423) > Italy (3.008) > Iran (1.019) > Kuwait (0.63) > France (0.305) > United States (0.025). For FFI: Brazil (8.834) > Malta (4.675) > Turkey (4.222) > Ethiopia (2.393) > Spain (1.513) > Saudi Arabia (0.422) > France (0.305) > Iran (0.107); Jordan, Italy, Kuwait, and the United States do not have FFI rows reported in Table 5. For PYC: Brazil (2.28) was the only country with TTHQ > 1; France (0.19), Italy (0.36), Saudi Arabia (0.049), and Spain (0.284) were below 1.

Per-country THQ > 1 for individual elements (IF): iAs in Turkey (2.938) and Brazil (1.532); Cd in Jordan (1.641) and Iran (1.019); Co in Jordan (3.787); Cu in Jordan (1.241), Brazil (1.37), and Malta (1.707); Fe in Jordan (1.91), Turkey (1.797), and Brazil (1.258); MeHg in Spain (3.15); Ni in Spain (1.171); Se in Brazil (1.413); Zn in Jordan (2.249), Brazil (2.073), Turkey (2.032), and Malta (1.862). For FFI: iAs in Brazil (1.066), Cu in Malta (1.268), Fe in Brazil (1.139), Ni in Malta (1.122), Se in Brazil (1.393), and Zn in Ethiopia (2.28), Turkey (1.738), Malta (1.656), and Brazil (1.435). For PYC: only Brazil’s TTHQ of 2.28 was driven by Ni (0.84), Cd (1.019), and iAs (0.827).

The risk-assessment narrative emphasizes that the iAs THQ values use the worst-case 100% inorganic-As assumption (most included studies measured total As without speciation), and that the actual inorganic-As fraction in dairy-derived foods is plausibly lower (perhaps 50%, though the precise fraction varies with raw material and reconstitution water). Average dietary inorganic-As fraction in the meta-analysis: IF=12.2 µg/kg and FFI=12.59 µg/kg were below the EU Commission Regulation cited maximum (20 µg/kg), while PYC=24 µg/kg was slightly above the cited 20 µg/kg limit. Average dietary Pb in IF (42.98 µg/kg) and FFI (45.5 µg/kg) substantially exceeded the EU 2015/1005 maximum level for infant and follow-on formula (20 µg/kg).

Risk of bias. 80% (24/30) of included studies were classified as high methodological quality on the JBI Critical Appraisal Checklist, 16.7% (5/30) as moderate, and 3.3% (1/30) as low. Confounding identification was rated “Yes” in only 30% of studies; confounding strategies “Yes” in 63.3%; objective measurement criteria “Yes” in 80%; statistical analysis appropriateness “Yes” in 76.7%.

Methods (brief)

Systematic review-with-meta-analysis following PRISMA 2020, protocol registered on Open Science Framework on 2022-06-14 (https://doi.org/10.17605/OSF.IO/2YNKB). Population-Exposure-Comparison-Outcome (PECO) question: “Do CMFs contain levels of potentially toxic elements that could harm children’s health?” Literature search across Embase, Scopus, PubMed, Web of Science, Livivo, and Lilacs conducted 2024-02-20 by two researchers; restricted to the seven years preceding the search (2017-2024); no language, publication-status, or country restriction. Reference lists of included studies were searched for additional eligible studies. Screening via Rayyan QCRI; two independent researchers screened titles/abstracts and then full texts.

Inclusion criteria: observational studies that investigated the concentration of essential and PTE in CMF, expressed as mass per mass (mg/kg or µg/kg), in IF, FFI, or PYC. Exclusion criteria: studies of non-fortified animal milk (cow, goat, buffalo, other), solid or fermented milk products (yogurt, cheese), studies focusing on PTE measurement methodology (validation, optimization, development), studies of dairy products obtained under abnormal conditions (e.g., environmental disasters), review studies, conference abstracts, personal opinions, books, case reports, research protocols, clinical trials, qualitative studies, in vivo (animal), and in vitro studies. Risk of bias by JBI Critical Appraisal Checklist for Analytical Cross-Sectional Studies.

Meta-analysis: random-effects model using Restricted Maximum Likelihood (REML) in STATA v16.0; pooled effect size = weighted mean difference (WMD) representing the level of elements weighted by sample size per study. Chi² test (p<0.1) for heterogeneity supplemented by I-square test; I² > 50% with visual forest-plot confirmation = substantial heterogeneity. Studies were stratified by CMF type (IF/FFI/PYC) and WHO region (AFR African, AMR Americas, EMR Eastern Mediterranean, EUR European, WPR Western Pacific). Concentrations expressed in mg/kg were converted to mg/100 kcal using powder-per-100-kcal reference values for comparison against international compositional standards.

Estimated daily intake (EDI) was calculated as EDI = (Cm × Cd) / bw, where Cm is the meta-analysis pooled concentration in mg/g powder, Cd is the recommended daily mass of powder to prepare one daily consumption following label recommendations (125.1 g/day at 0-6 months, 137 g/day at 0-12 months, 75 g/day at 1-3 years), and bw is body weight (6.1 kg at 0-6 months, 9.1 kg at 6-12 months, 12.1 kg at 1-3 years; WHO Growth Standards reference). EDI was compared against Institute of Medicine Dietary Reference Intakes for Cr, Cu, Fe, Mn, Se, and Zn (adequacy assessment) and against Tolerable Upper Intake Levels for Fe, Zn, and Se where UL had been established. Average formula intake was assumed at 780 mL/day for 0-3 months, 1,200 mL/day for 3-6 months, and 500 mL/day for FFI (6-12 months) and PYC (1-3 years).

Target Hazard Quotient (THQ) was calculated per EPA Risk Assessment Guidance for Superfund as THQ = EDI / RfDo, with RfDo values (mg/kg body weight/day): Al 1.0, iAs 0.0003, Cd 0.0001, Co 0.0003, Cr 1.5, Cu 0.04, Fe 0.7, MeHg 0.0001, Mn 0.14, Ni 0.011, Se 0.005, U 0.0002, Zn 0.3. THQ for Pb and Ti was not calculated because the EPA had not established RfDo values for these elements as of the publication; the authors note this gap explicitly. iAs THQ was calculated treating total As as if it were 100% inorganic As (deliberately conservative; the authors note that for dairy and infant-formula matrices the iAs fraction could be substantially lower). THQ > 1 indicates potential non-carcinogenic health risk. TTHQ (Total Target Hazard Quotient) was the sum of all individual element THQs per country; the same THQ < 1 / >1 interpretation applies. THQ minimum and maximum were also calculated from the 95% CI of the meta-analysis to capture heterogeneity-driven uncertainty.

Reported limitations. The TTHQ approach assumes additivity of element-specific hazards and may overestimate risk where antagonistic interactions attenuate combined effects; conversely, limited exposure data and incomplete reporting may also lead to underestimation. Aggregated country-level data do not capture intra-country regional variability. Average consumption rate assumptions were used rather than per-study measured intakes. The use of total As as a proxy for inorganic As is conservative. Reference-dose values for Pb and Ti are not yet established by competent regulatory bodies, so these elements were not included in THQ calculations.

Related evidence

• Infant Formula Powder
• Infant Formula, Powder (Non-Soy)
• Infant Formula, Powder (Soy-Based)
• Non-soy infant formula
• Soy-based infant formula
• Aluminum
• Arsenic, Inorganic
• Arsenic, Total
• Cadmium
• Cobalt
• Chromium
• Copper
• Iron
• Lead
• Manganese
• Methylmercury
• Nickel
• Uranium
• Zinc
• EU Regulation 2023/915 — Lead Maximum Levels for Infant and Young-Child Foods
• EU Regulation 2023/915 maximum levels for contaminants in food
• EPA IRIS — Cadmium Oral Reference Doses
• EPA IRIS — Mercuric Chloride Oral Reference Dose
• EPA IRIS — Methylmercury Oral Reference Dose
• EFSA — Nickel Tolerable Daily Intake (13 µg/kg b.w./day)

Update history

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