Urinary biomarkers of exposure to toxic and essential elements: A comparison of infants fed with human milk or formula

Overview

This Environmental Epidemiology original-research paper from the New Hampshire Birth Cohort Study (NHBCS) reports associations between urinary biomarkers of exposure to seven trace elements and infant feeding mode (exclusively human milk, exclusively formula, or combination-fed) in 462 infants at approximately 6 weeks of age. The analytes span four nonessential elements (arsenic, cadmium, nickel, and uranium) and three essential elements (cobalt, molybdenum, and selenium). Urine was collected via cotton pads, and 3-day food diaries with relevant covariate data were obtained from each family. General linear models adjusted for infant age, sex, maternal smoking during pregnancy, analytical batch, and urinary specific gravity. This is the most analytically comprehensive published comparison of urinary trace-element biomarkers by feeding mode in young U.S. infants, extending and complementing earlier NHBCS arsenic-focused work (Carignan et al. 2015; Signes-Pastor et al. 2017) with a broader seven-element panel.

Study design

Prospective cohort substudy nested within the NHBCS, a rural New Hampshire birth cohort. Participants were recruited from prenatal clinics in New Hampshire beginning January 2009. Initial enrollment included the full NHBCS (n = 1,238 within the study period); the analytic sample was restricted to infants with urine data, 3-day food diary data, age at collection below 100 days, detectable element concentrations in at least one metal, complete covariate data, and specific gravity above 1.000. The final analytic sample was 462 infants (Figure 1). The mean age at sample collection was 44.6 days (range 17 to 90). At approximately 6 weeks, 339 infants (73%) were fed exclusively human milk, 80 (17%) were combination-fed, and 44 (10%) were exclusively formula-fed. NHBCS participants had higher rates of exclusive human-milk feeding than the national average. The study was approved by the Committee for Protection of Human Subjects at Dartmouth College; all participants provided written informed consent.

Analytes

Seven elements were measured in spot urine samples:

Nonessential (toxic) elements: total arsenic minus arsenobetaine (denoted As in the paper; hereafter urinary arsenic), cadmium, nickel, and uranium.
Essential elements: cobalt, molybdenum, and selenium.

Arsenic was reported as total arsenic minus arsenobetaine to remove the nontoxic dietary organic form. Mercury, manganese, lead, and other metals were not included as primary analytes; the paper notes there is limited literature on viable urinary biomarkers for manganese, iron, or zinc in infants, and urine is not generally used to assess lead exposure.

Instrument and analytical methods

Urine samples were analyzed in four batches at the Trace Elements Analysis Core at Dartmouth College (Jackson BR, principal investigator). Prior to analysis, urine was filtered through a 0.6-mL pre-washed cotton pad placed in a collection cup; 0.6 mL was transferred to a pre-weighed tube and vortexed before filtering. Specific gravity (SG) was measured in one aliquot before freezing using a handheld refractometer with automatic temperature compensation (ATAGO PAL-10S, Bellevue, WA, USA). A second aliquot was stored at minus 80 degrees Celsius until elemental analysis.

Urine samples were analyzed in four batches at the Trace Elements Analysis Core using a custom-built element ICP-MS standard (inorganic Elements Analysis Core, Gathercole et al. 2016) to create a series of standards. After analysis, those standards were used by a custom-built multi-element ICP-MS (Thermo Fisher Scientific, operated in helium and oxygen modes). Seven elements were measured in elemental mode; arsenic species were separately identified by HPLC-ICP-MS. Continuing calibration checks were run after every 10 samples; duplicates and spiked samples were analyzed. When sample volume was less than 0.6 mL, a laboratory duplicate was prepared.

The paper does not report individual instrument model numbers, specific mobile-phase gradients for arsenic speciation, or detailed digestion protocols for urine.

LOD/LOQ

The paper does not report individual detection limits by element. Values below the detection limit were assigned the detection limit divided by the square root of 2, following standard practice cited in the paper. The paper notes that all infants included in the analytic sample had detectable concentrations for at least one targeted element.

CRM and recovery

Certified reference material recoveries are not reported in the paper. Quality control is described as continuing calibration checks after every 10 samples, and analysis of duplicate and spiked samples, without quantitative recovery figures.

Results

Geometric mean urinary concentrations by feeding mode

Values are from Table 2 of the paper. All units are micrograms per liter (µg/L). Values are geometric means with 95% confidence intervals; all were adjusted for specific gravity.

Element	All infants (n = 462)	Exclusively human milk (n = 338)	Combination-fed (n = 80)	Exclusively formula-fed (n = 44)
arsenic (As) (tAs minus arsenobetaine)	0.23 (0.21, 0.25)	0.15 (0.15, 0.18)	0.35 (0.35, 0.50)	0.85 (0.66, 1.09)
cadmium (Cd)	0.14 (0.13, 0.15)	0.15 (0.13, 0.16)	0.14 (0.13, 0.16)	0.12 (0.09, 0.16)
nickel (Ni)	1.82 (1.70, 1.94)	1.88 (1.74, 2.04)	1.75 (1.47, 2.08)	1.49 (1.19, 1.86)
uranium (U)	1.20 (1.12, 1.29)	1.22 (1.42, 1.86)	1.05 (0.87, 1.27)	0.74 (0.55, 0.99)
Co	0.20 (0.19, 0.21)	0.2 (0.19, 0.22)	0.2 (0.17, 0.23)	0.16 (0.13, 0.19)
Mo	2.26 (1.99, 2.57)	1.21 (1.10, 1.34)	9.07 (7.13, 11.54)	21.99 (17.44, 27.74)
Se	16.47 (15.90, 17.06)	15.19 (14.62, 15.78)	19.61 (18.19, 21.17)	22.26 (19.63, 25.24)

Note: The urinary concentration column heading in Table 2 reads “All infants (n=462).” The human-milk group n shown in Table 1 is 338 for the substudy with specific gravity above 1.000, while n = 339 is the feeding-mode count before the final exclusion.

Associations with feeding mode (adjusted models)

Values are from Table 3. The outcome is the log-transformed SG-adjusted urinary element concentration. The feeding mode effect is expressed as the back-transformed ratio (10 raised to beta) comparing formula-fed or combination-fed infants to exclusively human-milk-fed infants; the 95% confidence interval is the 99.5% CI (Bonferroni correction for seven tests). Bold entries in Table 3 indicate associations that are statistically significant after Bonferroni correction at alpha = 0.05; gray font indicates nonsignificance.

Nonessential elements:

Element	Model R-squared	Formula-fed vs human milk: ratio	95% CI	99.5% CI	Combination-fed vs human milk: ratio	95% CI	99.5% CI
arsenic (As)	0.36	5.15	(4.04, 6.58)	(3.51, 7.56)	2.51	(2.09, 3.03)	(1.92, 3.28)
cadmium (Cd)	0.08	0.81	n/a	(0.65, 1.00)	0.93	n/a	(0.79, 1.11)
nickel (Ni)	0.03	0.79	n/a	(0.61, 1.02)	0.93	n/a	(0.79, 1.10)
uranium (U)	0.08	0.59	n/a	(0.46, 0.75)	0.81	n/a	(0.67, 0.98)

Essential elements:

Element	Model R-squared	Formula-fed vs human milk: ratio	95% CI	99.5% CI	Combination-fed vs human milk: ratio	95% CI	99.5% CI
Co	0.20	0.78	(0.65, 0.95)	(0.61, 1.03)	1.00	(0.86, 1.15)	(0.81, 1.23)
Mo	0.58	19.02	(14.13, 25.59)	(12.59, 28.75)	7.59	(6.05, 9.51)	(5.46, 10.50)
Se	0.27	1.51	(1.35, 1.69)	(1.30, 1.75)	1.30	(1.20, 1.41)	(1.16, 1.46)

The 99.5% CIs are included as a conservative check for multiple comparisons. Associations that were not statistically significant at alpha = 0.05 (cadmium, nickel, uranium for formula vs human milk; cobalt for formula vs human milk after Bonferroni correction; uranium for combination vs human milk) are shown in gray font in the original Table 3.

Key adjusted-model findings summarized from the abstract and Table 3:

Urinary arsenic was 5.15 times higher (95% CI: 4.04, 6.58) in exclusively formula-fed compared with exclusively human-milk-fed infants.
Urinary molybdenum was 19.02 times higher (95% CI: 14.13, 25.59) in exclusively formula-fed compared with exclusively human-milk-fed infants.
Urinary selenium was 1.51 times higher (95% CI: 1.35, 1.69) in exclusively formula-fed compared with exclusively human-milk-fed infants.
Urinary uranium was lower in formula-fed infants (ratio 0.59; 95% CI: 0.46, 0.75) compared with human-milk-fed infants; this was statistically significant after Bonferroni correction.
Urinary cobalt was lower in formula-fed infants (ratio 0.78; 95% CI: 0.65, 0.95) compared with human-milk-fed infants; this did not meet the Bonferroni-corrected threshold (99.5% CI includes 1.0).
Urinary cadmium and nickel did not differ significantly by feeding mode.

Arsenic and household tap water

Household tap water arsenic (SG-adjusted log concentration) was positively associated with infant urinary arsenic concentration across all three feeding modes, with a steeper slope for formula-fed infants than for human-milk-fed infants (Figure 3). Among infants in households with known water sources, 87% of the formula-fed and combination-fed infants used tap water during the 3-day study period. Among combination-fed infants, urinary arsenic decreased with the number of human milk feedings (eAppendix 2: eFigures 2 and 3). The paper explicitly implicates the water used to reconstitute formula as a source of arsenic for formula-fed infants.

Associations in combination-fed infants

Among the 42 combination-fed infants with formula ounces data available (Figure 4), increasing ounces of formula consumed was positively associated with urinary arsenic (significant at the 5% level, solid regression line), urinary molybdenum (significant), and urinary selenium (significant), but not with urinary cadmium, nickel, uranium, or cobalt (nonsignificant, dashed lines).

Evidence fitness

This is an A-tier prospective cohort study reporting directly measured urinary biomarkers across seven elements in 462 infants with characterized feeding mode and household water data. The biomarker approach provides exposure evidence at the infant body-burden level and is not subject to the input-distribution assumptions required by dietary intake models. The study is the largest published U.S. cohort study of multi-element urinary biomarkers stratified by infant feeding mode. Limitations acknowledged by the authors include: (1) the spot urine design rather than 24-hour urine collection, with the note that selenium biomarker reliability is established for 24-hour samples rather than spot samples; (2) the NHBCS population overrepresents exclusive human-milk feeding (73% at 6 weeks) relative to U.S. national rates (44 to 47% at 3 months), limiting generalizability and the statistical power for formula-fed subgroups; (3) the formula-fed subgroup was relatively small (n = 44), reducing precision for formula-fed estimates; (4) feeding mode was recorded from 3-day food diaries and classified at a single time point; (5) variability in formula composition by brand, type (soy vs non-soy), and preparation was not characterized, so the formula-fed estimates represent average exposure across brands present in the NHBCS market; (6) molybdenum feeding mode effects explained only 4% of the variance in urinary molybdenum concentrations (Table 3A), suggesting substantial between-infant variability.

Limitations

The paper explicitly acknowledges the following limitations: (1) spot urine samples are subject to within-day variability; selenium biomarker reliability is established for 24-hour integrated samples, not spot samples, so the selenium finding should be interpreted with caution; (2) NHBCS demographics (older, college-educated, predominantly exclusive human-milk feeding) differ from U.S. national demographics, particularly in the first 3 months of infancy; (3) the study population is from rural New Hampshire, with private well water, so tap water arsenic patterns may differ from nationally representative populations; (4) filter use was not assessed prospectively — participants reported filter use with their household tap water results after the study period — and 87% of formula- and combination-fed infants were fed formula reconstituted with tap water during the 3-day study period; (5) the formula-only subgroup (n = 44) is small relative to the human-milk group (n = 338); (6) feeding mode explains only a fraction of the variance in urinary concentrations for most elements, indicating other exposure sources contribute substantially; (7) the study cannot isolate formula-powder arsenic contribution from reconstitution-water arsenic contribution.

Regulatory context

The U.S. EPA maximum contaminant level for arsenic in public drinking water is 10 micrograms per liter; private well water is unregulated. All NHBCS participants used private groundwater wells. The FDA action level for inorganic arsenic in infant rice cereal is 100 parts per billion (dry weight; FDA 2020). No U.S. regulatory maximum exists for arsenic specifically in infant formula powder or reconstitution water specifically intended for infant formula. For molybdenum, there are no established regulatory limits in infant formula or water applicable to this study population. For selenium, the Institute of Medicine adequate intake for infants 0 to 6 months is 15 micrograms per day, consistent with the observed urinary selenium gradient. For uranium, U.S. EPA drinking-water maximum contaminant level is 30 micrograms per liter; no infant-specific limit exists.

Key conclusions

In this prospective cohort of 462 rural New Hampshire infants at approximately 6 weeks of age, infant feeding mode was associated with urinary biomarkers for arsenic, molybdenum, selenium, uranium, and cobalt in adjusted models. Urinary arsenic was 5.15 times higher (95% CI: 4.04, 6.58) and urinary molybdenum was 19.02 times higher (95% CI: 14.13, 25.59) in exclusively formula-fed compared with exclusively human-milk-fed infants. Urinary selenium was 1.51 times higher (95% CI: 1.35, 1.69) in formula-fed infants. By contrast, urinary uranium was significantly lower in formula-fed infants (ratio 0.59; 95% CI: 0.46, 0.75), implying human milk as the principal uranium source, not formula. Urinary cobalt was lower in formula-fed infants (ratio 0.78; 95% CI: 0.65, 0.95) though not significant after Bonferroni correction. Urinary cadmium and nickel did not differ significantly by feeding mode. Arsenic and molybdenum exposure increased dose-dependently with formula consumption among combination-fed infants. The tap water arsenic association with infant urinary arsenic was present in all feeding modes and was steepest in formula-fed infants, implicating reconstitution water as a key arsenic source. These findings establish that the direction and magnitude of feeding-mode effects on infant trace-element exposure are element-specific and cannot be assumed to favor formula or human milk uniformly.

Implications

Certification: Primary urinary biomarker evidence for feeding-mode-stratified infant arsenic and molybdenum exposure in a U.S. prospective cohort. The arsenic finding (5.15-fold higher in formula-fed vs human-milk-fed) is directly relevant to the infant formula arsenic exposure case that underpins HMTc infant-formula category scoping. The molybdenum finding is a secondary concern given molybdenum’s current non-toxic status at dietary levels in infants. The uranium finding (lower in formula-fed infants) complicates any blanket claim that formula is a higher-exposure pathway for all nonessential elements.
Courses: This paper is the best available multi-element urinary biomarker reference for feeding-mode comparison at 6 weeks. The element-specific directionality (arsenic and molybdenum higher in formula-fed; uranium lower) is a teachable illustration of how formula composition and reconstitution water interact differently for different elements.
App: Supports the formula vs human milk per-element exposure comparison. Arsenic and molybdenum are the most defensible comparisons for the app given the effect sizes and statistical significance. Cadmium and nickel should not be presented as feeding-mode-driven elements based on this study.
Microbiome: The paper does not address metal-microbiome interactions directly. The feeding-mode comparison is primarily relevant to heavy-metals-and-diet exposure context rather than microbiome.

Provenance notes

The PDF is the open-access author’s copy distributed under CC BY-NC-ND 4.0 via PMC (PMC10852378; PMID 38343736). SHA-256 of the deposited PDF: d9e043c2dd8632da82d1800371a5ce0447e7acf1645bc91efee0e340d425d297. The paper was part of the NHBCS infant environmental health program (NIEHS grants R21 ES018675, R01 ES034958, P42 ES007373; USEPA grants P01 ES022832, P42 ES007373; NIH grants from ORIP).

Update history

No substantive edit history is available in this build. The full commit record is available in git.