Potential for Manganese-Induced Neurologic Harm to Formula-Fed Infants: A Risk Assessment of Total Oral Exposure

Abstract

Background:

High oral exposure and biological vulnerabilities may put formula-fed infants at risk for manganese-induced neurotoxicity.

Objectives:

We sought to characterize manganese concentrations in public drinking water and prepared infant formulas commonly purchased in the United States, integrate information from these sources into a health risk assessment specific to formula-fed infants, and examine whether households that receive water with elevated manganese concentrations avoid or treat the water, which has implications for formula preparation.

Methods:

Manganese was measured in 27 infant formulas and nearly all Minnesota community public water systems (CPWS). The risk assessment produced central tendency and upper-end exposure estimates that were compared to a neonatal animal-based health reference dose (RfD) and considered possible differences in bioavailability. A survey study assessed esthetic concerns, treatment, and use of water in a Twin Cities community with various levels of manganese in drinking water.

Results:

Ten percent of CPWSs were estimated to exceed the EPA health advisory level of $300\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$. Manganese concentrations in formula ranged from 69.8 to $741\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, with amino $\text{acid}>\text{soy}>\text{cow}’\mathrm{s} \text{milk}$ formula concentrations. Central tendency estimates of soy and amino acid formula reconstituted with water at the CPWS 95th percentile manganese concentration exceeded the neonatal-based RfD. Upper-end estimates of manganese intake from formula alone, independent of any water contribution, equaled or exceeded the neonatal-based RfD. In the survey study, we observed increased awareness of esthetic issues and water avoidance at higher manganese concentrations, but these concentrations were not a reliable consumption deterrent, as the majority of households with inside tap drinking water results above $300\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ reported drinking the water.

Discussion:

Excessive exposure to manganese early in life can have long-lasting neurological impacts. This assessment underscores the potential for manganese overexposure in formula-fed infants. U.S. agencies that regulate formula and drinking water must work collaboratively to assess and mitigate potential risks. https://doi.org/10.1289/EHP7901

Introduction

Manganese Toxicity and Children’s Vulnerability

Manganese is a naturally occurring metal in air, soil, water, and food worldwide. Although manganese is an essential nutrient in the diet, it is also a well-established neurotoxin at higher levels of exposure. Neurological effects were first identified in occupational workers from inhalation exposure (Blanc 2018) and later observed in humans with long-term parenteral nutrition therapy (Fitzgerald et al. 1999; Abdalian et al. 2013), impaired manganese clearance because of liver dysfunction (O’Neal and Zheng 2015), and high accidental ingestion (Holzgraefe et al. 1986; Ghosh et al. 2020).

High oral exposure and physiologic vulnerabilities make young infants a potentially vulnerable group to manganese-induced neurotoxicity. First, the formula-fed infant is the subpopulation with the highest drinking water intake per kilogram of body weight and thus has the highest risk of overexposure to manganese in water (EPA 2019; also see Figure S1). In addition, regulation of manganese intestinal absorption is not yet fully developed in young infants (Erikson et al. 2007; Ljung and Vahter 2007), and the biliary excretion system, which is the primary route of excretion, is also not fully functional (Aschner and Aschner 2005), leading to higher infant body burden. The limited available data show notably higher absorption of manganese from formula in infants than adults (Dörner et al. 1989; Davidsson et al. 1989). Animal studies support these findings in humans. Prior to weaning, mice and rats do not appear to exhibit homeostatic control of manganese levels, which is accompanied by accumulation in the brain and liver (Miller et al. 1975; Ballatori et al. 1987).

These findings are of particular concern given the sensitivity of the developing brain to manganese neurotoxicity (Dorman et al. 2000; Tran et al. 2002; Kern et al. 2010; Conley et al. 2020). Oral exposure studies in animals during neonatal periods report changes in brain chemical end points as well as altered behaviors following manganese exposure, including hyperactivity (Kern et al. 2010), altered social interactions (Golub et al. 2005), transient ataxia (Tran et al. 2002), altered acoustic startle (Dorman et al. 2000), impaired learning (Kern et al. 2010), and increased stereotypic behavior (Kern et al. 2010). In general, these studies provide evidence of subtle developmental neurobehavioral effects following short-term neonatal exposure at doses $\ge 10–20\hspace{0.17em}\text{mg}/\text{kg}$ per day (WHO 2020). In epidemiologic studies, preclinical neurological effects in children following exposure to elevated levels of manganese in drinking water have also been reported, including intellectual deficits (Wasserman et al. 2006; Bouchard et al. 2011) and behavioral disinhibition (Bouchard et al. 2007; Schullehner et al. 2020). Despite individual study limitations in exposure assessment and study design, epidemiology studies collectively suggest developmental neurotoxicity as an end point of concern following exposure to excess manganese in drinking water (WHO 2020). Epidemiologic data on manganese exposure and neurodevelopmental effects were recently used in a quantitative risk assessment (Kullar et al. 2019). A benchmark concentration analysis was applied to data from two cross-sectional studies of Canadian children exposed to manganese in drinking water to assess IQ deficits. Girls were more sensitive to a drop in IQ associated with manganese in drinking water, with 1% and 5% IQ drops associated with drinking water benchmark concentrations of $78\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ and $192\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, respectively. The corresponding lower bound on the benchmark concentrations (BMCL) were $9\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ and $74\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, respectively.

Regulation of Manganese in Drinking Water

To date, the U.S. Environmental Protection Agency (EPA) has not established a level of manganese in drinking water that is safe for formula-fed infants to consume. EPA first evaluated manganese in 1998 to determine if there was a need to regulate it in public drinking water systems (PWS) under the U.S. Safe Drinking Water Act (SDWA). As part of this process, EPA developed an adult-based health reference level (HRL) of $300\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ to evaluate manganese drinking water occurrence data (EPA 2003b). This HRL was based on an adult body weight and drinking water consumption rate and the assumption that drinking water contributes 20% to total adult exposure. The EPA HRL’s reference dose (RfD) was not based on oral toxicity studies but, rather, the upper end of the estimated range of manganese intake from adult dietary studies. A modifying factor of 3 was applied to address concerns related to increased infant vulnerability and exposure and potentially higher absorption from water than diet. Despite identifying a number of community public water systems (CPWSs) with manganese levels above the HRL that serve large populations (EPA 2003b, 2004), the EPA determined that regulating manganese would not present a meaningful opportunity for health risk reduction (68 FR 428978) (EPA 2003a).

In 2004, the EPA established health advisory levels (HALs) for manganese (EPA 2004). A HAL is a nonregulatory health-based reference level of a chemical in drinking water at which there are no adverse health risks when ingested over various durations of time. EPA used the adult-based HRL value of $300\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ as the lifetime HAL. Although EPA stated that the lifetime HAL “will protect against concerns of potential neurological effects” (EPA 2004), it advised that infants should not be given drinking water above the HAL for more than 10 d. EPA did not identify a HAL that would be safe for infants to consume for greater than 10 d. In 2016, EPA decided to reevaluate the need to regulate manganese based on “new and updated health effects information and additional occurrence data” (81 FR 81099) (EPA 2016a). As such, manganese was sampled in finished water from CPWSs from 2018–2020 under the fourth Unregulated Contaminants Monitoring Rule (UCMR4), and the results will provide a basis for future regulatory decisions.

Manganese can impart a bitter metallic taste to drinking water and cause staining of laundry and fixtures, which may be noticed by water consumers. For this reason, the EPA established a nonregulatory secondary maximum contaminant level (SMCL) for manganese of $50\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ in 1979 to enhance consumer acceptance of public water (40 CFR 143.3) (EPA 1991). A common assumption is that a noticeable, unpleasant taste above the SMCL will trigger a CPWS to treat for manganese and/or will effectively dissuade people from using the water. However, there is a current lack of understanding on what level of manganese will act as a deterrent to drinking or will incentivize public water systems or residents to treat water to remove manganese.

The Minnesota Groundwater Protection Act authorizes the Minnesota Department of Health (MDH) to derive health-based guidance values (HBGVs) to assist risk managers in identifying water sources with contaminants at levels of potential health concern (Minnesota 103H.201) (Minnesota Statute 2020). A health-based value represents a concentration of a chemical in drinking water that is likely to pose little or no health risk to humans (MDH 2008), including vulnerable or highly exposed subpopulations. In 2012, the MDH issued a short-term HBGV of $100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ for manganese based on risks to formula-fed infants (reissued in 2018; MDH 2018). The MDH’s 2012 assessment concluded that a lower concentration was needed for formula-fed infants because of lower dietary needs (IOM 2001) and greater sensitivity and exposure. The MDH’s point of departure of $25\hspace{0.17em}\text{mg}/\text{kg}/\mathrm{d}$ manganese (MDH 2018), based on cognitive and behavioral effects and altered dopamine receptor levels in neonatal rats (Kern et al. 2010), was adjusted by a total uncertainty factor of 300, resulting in an RfD of $83\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$ manganese (MDH 2018). The total uncertainty factor of 300 consisted of factors of 10 for potential interspecies differences, 10 for intraspecies variability, and 3 for extrapolation from a lowest observable adverse effect level (LOAEL).

Derivation of the guidance value used a 95th percentile water intake rate for young infants and a relative source contribution (RSC) factor of 50% to acknowledge that powdered formula is an independent source of manganese (MDH 2008, 2018).

Recently, Canada implemented a drinking water standard for manganese in public drinking water of $120\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, based on the protection of infants (Health Canada 2019). The World Health Organization (WHO) has also recently proposed lowering their health-based guideline value for manganese from 400 to $80\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, based on protecting formula-fed infants, the subpopulation most susceptible to manganese exposure (WHO 2020). Both the Canadian and WHO values relied upon the same end point and key study (Kern et al. 2010) that was used by MDH to derive its HBGV.

Because manganese is an unregulated contaminant in the United States, comprehensive data on manganese levels in U.S. drinking water are lacking. More than one-third of the U.S. population relies on groundwater for drinking water (https://www.usgs.gov/news/quality-nation%E2%80%99s-groundwater), which is more likely to have elevated manganese concentrations than surface water (Kohl and Medlar 2006). The U.S. Geological Survey has characterized manganese groundwater concentrations based on data from 43,334 U.S. wells (McMahon et al. 2019). As shown in Figure 1, occurrence of groundwater manganese concentrations above the MDH HBGV of $100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, as well as above the EPA HAL of $300\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, is common and widespread. A total of 24% and 13% of U.S. groundwater samples exceeded the MDH’s HBGV and the EPA’s HAL, respectively.

Figure 1.

Manganese concentrations in U.S. groundwater. Data are from McMahon et al. 2019 supporting information.

Regulation of Manganese in Infant Formula

Formula feeding is commonplace in the United States, with less than 50% of infants exclusively breastfed through the first 3 months of life and about 25% exclusively breastfed through the first 6 months of life (CDC 2020). By U.S. law, infant formula must contain a minimum of $5\hspace{0.17em}\mu \mathrm{g}/100\hspace{0.17em}\text{kcal}$ of manganese to meet nutritional needs (21 CFR 107.100) (FDA 2020a). This concentration is equivalent to $34\hspace{0.17em}\mu \mathrm{g}$ manganese per liter of water assuming $20\hspace{0.17em}\text{kcal}$ per ounce of formula, which is standard for most formulas (O’Connor 2009). The U.S. Food and Drug Administration (FDA), however, has not established a legal maximum manganese concentration in formula despite the recommendations of science advisory groups. The Life Sciences Research Office (LSRO) Expert Panel first recommended a maximum level for infant formula of $100\hspace{0.17em}\mu \mathrm{g}/100\hspace{0.17em}\text{kcal}$ ($676\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$) because of “considerable evidence that exposure to high intakes of manganese can be toxic” (Raiten et al. 1998). This value was not based on quantitative risk assessment but, rather, that it is “significantly below the LOAEL in adults for manganese in water and is far beyond the range likely to be encountered in milk-based formulas.” According to Ljung and Vahter (2007), the recommended value of $100\hspace{0.17em}\mu \mathrm{g}/100\hspace{0.17em}\text{kcal}$ was incorrect because of a calculation error and should have been $50\hspace{0.17em}\mu \mathrm{g}/100\hspace{0.17em}\text{kcal}$. In contrast to the LSRO Expert Panel, the Institute of Medicine (IOM) chose not to publish a manganese-tolerable upper intake level for infants because of lack of sufficient data in this age group (IOM 2001). However, they advised that “the only source of intake for infants should be from food or formula” alone because of “concern about the infant’s ability to handle excess amounts” of manganese. In 2004, an international expert group (IEG) was formed to advise the Codex Committee on Nutrition and Foods for Special Dietary Uses on nutrient levels in formula. The IEG recommended a maximum manganese content of $50\hspace{0.17em}\mu \mathrm{g}/100\hspace{0.17em}\text{kcal}$ ($338\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$), “which is equivalent to that of unsupplemented soy formula, and about 60 times higher than breast milk levels” (Koletzko et al. 2005). The IEG cautioned that higher manganese concentrations should be avoided because of immature manganese excretion in infants, which “may cause accumulation in tissues including brain and might induce potential adverse effects, such as neurodevelopmental abnormalities observed in newborn animals” (Koletzko et al. 2005).

The Need for Multisource Risk Assessment of Infant Exposure to Manganese

It is critical to assess the health risks of manganese exposure to formula-fed infants, a highly exposed and vulnerable population, as described above. Previous studies focusing on manganese in infant formula have not considered all sources of exposure nor conducted a formal risk assessment based on total oral exposure. One exception is a risk assessment by Brown and Foos (2009). The authors noted unacceptable risks to formula-fed infants across several central and upper-end exposure scenarios. However, this assessment did not use health reference values derived from neonatal toxicity studies but, rather, an adult-based RfD and the IOM’s adequate intake (AI), which may under- and overestimate risk, respectively. Other limitations included lack of risk assessment at upper percentile formula concentrations and reliance on labeled, rather than measured, concentrations of manganese in formula. A recent study by Mitchell et al. (2020) also included a risk assessment based on manganese concentrations measured in a variety of infant formulas purchased in the United States and France. However, this assessment only considered an assumed water concentration of $250\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ and focused on the extremes of the measured ranges of formulas.

The aim of this report is to a) characterize manganese concentrations in a robust data set of public drinking water; b) measure manganese concentrations in prepared infant formulas commonly purchased in the United States; c) integrate information from these sources into a quantitative health risk assessment specific to formula-fed infants; and d) examine whether households that receive water with elevated manganese concentrations have esthetic concerns about the water and use, avoid, or treat the water.

Methods

In order to fully characterize manganese exposure and risk, we integrated information from three recent MDH studies as our main data sources: a) the General Water Chemistry Project, an unpublished study which measured manganese concentrations in nearly all Minnesota CPWSs (described here); b) the Wells and Increased Infant Sensitivity and Exposure Study, a study conducted by the MDH and Dakota County, Minnesota, which included a household survey and measurement of several contaminants in private well water (Scher and Demuth 2017); and c) an assessment of manganese concentrations in prepared infant formulas (described here).

Data Collection and Analytical Methods

Public drinking water.

The majority of manganese samples used in this analysis were collected as part of the MDH’s General Water Chemistry Project, a study conducted from 2010–2014 to establish baseline chemistry data for CPWS sources statewide. In addition to manganese, samples were analyzed for several other contaminants that present health risks (e.g., nitrate and arsenic), lead to esthetic issues (e.g., iron), or affect water quality (e.g., ammonia and total organic carbon). For nearly all CPWSs, MDH engineers sampled up to three CPWS wells or surface water intakes (raw water) and/or three entry points to the distribution system (finished water) considered representative of the water system. The MDH Public Health Laboratory (PHL) used EPA Method 200.8 (EPA 1994b) to analyze the samples for manganese. Method 200.8 uses inductively coupled plasma mass spectrometry (ICP-MS) for the analysis of total manganese in drinking water. The method reporting limit (MRL) was $10\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$.

Manganese results were also available for some Minnesota systems through the EPA’s UCMR4 monitoring program starting in 2018 (378 Minnesota UCMR sample results through 28 October 2020). Eurofins Eaton Analytical, Inc. was approved by the EPA to perform the analysis using EPA Method 200.8 (EPA 1994b), with an MRL of $0.4\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$. A total of 2,726 samples were available from both MDH and UCMR monitoring efforts.

If a CPWS had both raw and finished water samples, only the finished water samples were retained. If more recent finished water results from UCMR4 monitoring were available, they were selected over earlier General Water Chemistry Project results. Any General Water Chemistry Project results from wells that had since been sealed or that had a current classification of “emergency use” were excluded from the analysis. For systems that purchase water, the seller concentration results were substituted for the purchasing system in cases where the system did not have its own sampling results. For systems with only raw water samples, we considered each system’s treatment technologies to determine if raw water concentrations likely represent the finished water concentrations. Dissolved manganese, the predominant species present in groundwater, can be removed by oxidation/physical separation, adsorption/oxidation, biological filtration, and precipitative softening (Health Canada 2019). Within-system raw or finished water results were averaged to create one result per system. Results below the MRL were assigned a value of one-half MRL. Buyer/seller information, treatment types, and population-served information are regularly reported to MDH by CPWSs and updated in Minnesota’s Drinking Water Information System, which served as the source of these data.

Infant formula concentrations.

MDH purchased a variety of powdered formulas online from Target Corporation and Walgreens Company in January 2018. A total of 27 powdered formulas were tested (Table S1), representing:

• The top 10 powdered formulas sold in the United States in 2016 (https://www.statista.com/statistics/186157/top-powdered-baby-formula-brands-in-the-us/)

• A variety of suppliers of formula sold in the United States, including major suppliers (Enfamil®, Gerber®, and Similac®), a store brand (Target Corporation’s Up&Up™), and two organic brands (Earth’s Best® and The Honest Company®)

• Formulas with different bases (cow’s milk, soy, and amino acid)

On two separate occasions, powdered formula was mixed according to directions on the formula labels with deionized (ultrapure) water provided by the MDH PHL. Prepared samples were transferred to PHL personnel for analytical testing.

The prepared samples provided to the PHL were fixed according to EPA Method 3050B (EPA 1996). Briefly, samples were equilibrated to room temperature. The samples were shaken by hand until thoroughly mixed, and a $5.0\text{-mL}$ aliquot pipetted into a $50\text{-mL}$ DigiTUBE digestion vessel (SCP Science) along with $10\hspace{0.17em}\text{mL}$ ultrapure water. Prior to heating, $0.5\hspace{0.17em}\text{mL}$ of VWR BDH® Aristar® Plus nitric acid (67–70%) (catalog no. 87003-261; lot no. 1,117,120; VWR International, LLC) was added to each sample vessel and quality control vessel and placed in the HotBlock Pro graphite digestion system (Environmental Express, Inc.). The vessels were capped with disposable ribbed reflux caps (Environmental Express) and refluxed at 95°C for 2 h. After refluxing, the vessels were removed and allowed to cool. The addition of nitric acid and the refluxing process were repeated two additional times.

After the third reflux, $0.25\hspace{0.17em}\text{mL}$ of Aristar® Plus 30–32% hydrogen peroxide (catalog no. 4508A60; lot no. BDH7690-3; VWR International, LLC) was slowly added to each vessel. The vessels were capped with the disposable ribbed reflux caps (Environmental Express) and placed in the HotBlock Pro (Environmental Express) for heating until the effervescence subsided. Once the effervescence subsided, the vessels were removed and allowed to cool. This hydrogen peroxide process was repeated once more. After the second peroxide addition, the cooled vessels were diluted to $50\hspace{0.17em}\text{mL}$ with ultrapure water. The vessels were capped, shaken by hand until mixed thoroughly, and allowed to settle. Once settled, the samples were filtered with FilterMate™ (Environmental Express) (Catalog no. SC0409, $0.45\hspace{0.17em}\mu \mathrm{m}$ PVDF filter with a PTFE Prefilter), and the filtrate transferred to a new digestion vessel (SCP Science) for analysis.

Samples, including quality control samples, were analyzed on a Thermo Fisher Scientific Inc. iCAP™ Q ICP-MS for ⁵⁵Mn according to EPA Method 200.8 (EPA 1994b). (Instructions were followed for samples that were prepared as total recoverable elements, and internal standardization was performed according to Method B.) The ICP-MS was calibrated with 0-, 1-, 10-, and $100\text{-}\mu \mathrm{g}/\mathrm{L}$ standards. Manganese stock standards were from Inorganic Ventures (catalog no. IV-21). Manganese second-source standards were from SPEX CertiPrep (catalog no. CL-CQ-21). Manganese digestion standards were purchased from Inorganic Ventures (catalog no. MNDH-CAL-2). iCAP™ Q ICP-MS software prepared a calibration curve for ⁵⁵Mn by plotting instrument intensity against standard concentration. Concentrations for infant formula samples were calculated from the regression equation generated by the software. All results were reported in microgram per liter to three significant figures after accounting for all dilutions.

Concentrations of manganese on infant formula labels appear as micrograms per 100 calories (or microgram per numeric fluid ounce when prepared as directed). Because label instructions for mixing the powdered formula with water were followed, the manganese concentration reported on each label was divided by the number of fluid ounces identified on each label. This resulted in units of microgram per fluid ounce, which was converted to micrograms per liter ($1\hspace{0.17em}\text{fl}\hspace{0.17em}\text{oz}/0.030\hspace{0.17em}\mathrm{L}$) (21 CFR 101) (FDA 2020b).

Formula-fed infant risk assessment.

We considered both a central tendency scenario using the mean formula manganese concentration from this study and a mean water intake rate and an upper-end scenario using the 95th percentile formula manganese concentration and 95th percentile water intake rate for formula-fed infants 1 to $<3\hspace{0.17em}\text{months}$ of age. Intake rates (IRs) were from the EPA Exposure Factors Handbook (EPA 2019). Total manganese intake was calculated at both the median and 95th percentile manganese concentrations from Minnesota community water systems and compared with MDH’s RfD of $83\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$. The RfD was based on a LOAEL of $25\hspace{0.17em}\text{mg}/\text{kg}/\mathrm{d}$ for neurological effects observed following short-term exposure in neonatal rats (Kern et al. 2010). A total uncertainty factor of 300 was applied [10 interspecies extrapolation, 10 for intraspecies variability, and 3 for LOAEL to no observable adverse effect level (NOAEL) extrapolation] resulting in an RfD of $83\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$ manganese. An RSC of 0.5 was used with the RfD in Equation 1 to derive the MDH HBGV (MDH 2018).

$$\frac{\text{RfD} \left(\frac{\mu \mathrm{g}}{\text{kg}\cdot \mathrm{d}}\right)\times \text{RSC}}{\text{IR}\left(\frac{\mathrm{L}}{\text{kg}\cdot \mathrm{d}}\right)}=\frac{\left(\frac{83 \mu \mathrm{g}}{\text{kg}\cdot \mathrm{d}}\right)\times 0.5}{\left(\frac{0.285 \mathrm{L}}{\text{kg}\cdot \mathrm{d}}\right)}=145.6,\hspace{0.17em}\text{rounded} \text{to}\hspace{0.17em}100 \frac{\mu \mathrm{g}}{\mathrm{L}}.$$ (1)

Household perceptions of water quality and water use.

In 2015, MDH partnered with Dakota County, Minnesota, to conduct a private well water quality study (Wells and Increased Infant Sensitivity and Exposure study) in a suburban Twin Cities community with variable levels of manganese in groundwater (Scher and Demuth 2017). Eight hundred private well households were randomly selected and invited to participate in a study to test several well water quality parameters. Manganese was not mentioned as a contaminant of specific interest in participant communications. The MDH institutional review board (IRB) determined that the study was exempt from IRB review, as private well water quality results are not classified as private or nonpublic under state law, and the survey topics were deemed nonsensitive. Potential participants were advised in the introductory letter that well location and study results were not legally considered private and could be subject to disclosure. A total of 274 well owners responded to the mailing by the deadline and were enrolled. No incentives were offered beyond the free water testing. County staff collected a well water sample of unfiltered, untreated water from an outside spigot at each participating home. An adult member of the household completed an online survey (SurveyMonkey®) about any water treatment devices in the home, the primary source of drinking water, and concerns about well water quality ($n=258$). Participants completed the survey prior to receiving any water quality results. In the second phase of the study, households with an outdoor spigot result above $100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ were invited to participate in free follow-up sampling of manganese from an inside tap after the water had gone through any treatment devices used in the home. An MDH-accredited laboratory (Minnesota Valley Testing Lab) measured manganese in both the exterior spigot and inside tap water samples using EPA method 200.7 (EPA 1994a) for determination of metals and trace elements in water by inductively coupled plasma atomic emission spectrometry (ICP-AES). The method detection limit was $1\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$.

Statistical Analysis

Descriptive statistics are presented for CPWSs and reconstituted formula manganese concentrations. For CPWSs, a sensitivity analysis was conducted to assess whether central tendency and upper percentile manganese concentrations changed when including or excluding a small number of systems that only had raw water results available but treatment known to remove manganese. The Kruskal-Wallis test was used to assess differences in manganese concentration by CPWS size category. As described above, two independent rounds of formula sample preparation were conducted according to label instructions, with each round performed by a different person. A paired-samples t-test compared manganese concentrations from the first and second rounds of sample preparation. The average percent difference in concentration between the two formula sampling rounds was calculated as $\%\hspace{0.17em}\text{diff}=[(\Delta \text{rounds})/(\Sigma \text{rounds}/2)\times 100]$. In the Wells and Increased Infant Sensitivity and Exposure study, chi-square tests were used to assess differences in drinking water esthetic concerns and mitigation measures by ordinal manganese concentration categories. Trends in mitigation actions across manganese concentration categories were assessed using the Cochran-Armitage test. Statistical analyses were performed in SAS (version 9.4; SAS Institute Inc.). All tests were two sided with $\alpha =0.05$.

Results

Public Drinking Water

Manganese results were available for 919 of 965 Minnesota CPWSs (95%). The 46 systems without manganese data serve an estimated 9,874 people and were mainly small housing developments constructed after the General Water Chemistry Project took place or systems near the state border that purchase water from a CPWS outside the state. Of the systems with data, 546 (59%) had finished (i.e., entry point) water concentration results, and 373 systems (41%) only had raw (i.e., source water) results (Table 1). Of systems with raw water results, 145 (39%) had an average manganese concentration above $100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, the MDH HBGV. After reviewing each system’s treatment technologies, 131 of these 145 systems (90%), serving 114,437 people, did not have treatment that would remove manganese. For systems with finished water results, 105 of 546 systems (19%) had an average manganese concentration above $100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$. These systems serve 223,864 people. In sum, nearly all CPWSs in Minnesota had manganese results available, of which 250 systems (27%) exceeded the MDH HBGV based on the averaged sample results. These systems were estimated to serve 345,094 people in Minnesota. Twenty-one systems with raw water samples only (including 14 systems $>100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$) had treatment expected to remove manganese. Lack of finished water results for these systems adds uncertainty to concentration summary statistics. Assuming raw water results reflect finished water for the 21 systems with treatment expected to reduce manganese, the 25th, 50th, 75th, 90th, and 95th percentile concentrations for Minnesota CPWSs would be 5.0, 29.0, 111.0, 313.5, and $562.0\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, respectively. Removing these 21 systems did not substantially impact the 50th and 95th percentiles (28.0 and $562.0\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, respectively), which were used in the risk assessment.

Table 1.

Manganese mean drinking water concentrations for 919 Minnesota community public water systems.

Concentration category ($\mu \mathrm{g}/\mathrm{L}$)	Systems with finished water samples ($n=546$)			Systems with raw water samples only ($n=373$)
	$n$ (%)	Number with Mn removal treatment	Population served	$n$ (%)	Number with Mn removal treatment	Population served (with no Mn treatment)
$<50$	380 (70)	228	3,698,912	168 (45)	3	134,487 (133,024)
50–100	61 (11)	39	208,724	60 (16)	4	37,635 (31,593)
$>100–200$	47 (9)	32	109,059	54 (15)	5	29,601 (28,522)
$>200–300$	16 (3)	8	82,750	37 (10)	4	49,711 (46,179)
$>300–400$	9 (2)	5	3,983	16 (4)	0	17,618 (17,618)
$>400–500$	7 (1)	5	2,017	8 (2)	2	1,499 (1,359)
$>500–1,000$	17 (3)	7	20,478	26 (7)	3	19,499 (17,457)
$>1,000$	9 (2)	1	5,577	4 (1)	0	3,302 (3,302)

Note: Raw water samples are collected from a well or surface water intake before the water has undergone any treatment (e.g., filtration, chemical addition). Finished water samples are collected at the entry point to the distribution system after the water has gone through any treatment.

Comparing CPWS results to the EPA HAL, 42 systems (8%) with finished water samples had manganese concentrations above $300\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$. Of systems with raw water results only, 49 (13%) were above the EPA HAL and had no treatment in place to remove manganese. Overall, 96 Minnesota CPWSs (10%) with available manganese results had average concentrations above the EPA HAL. Surprisingly, many finished water (i.e., posttreatment) results from systems that have technologies to remove manganese were still above the MDH HBGV and EPA HAL (Table 1). We also found that the manganese concentration increased as the system size decreased (Table 2), with the smaller systems having significantly higher manganese levels [Kruskal-Wallis test, ${\chi }^2=94.4$, degrees of freedom $\left(\text{df}\right)=3$, $p<0.0001$].

Table 2.

Manganese drinking water concentration ($\mu \mathrm{g}/\mathrm{L}$) by system size.

Number of people served	Number of systems	25th percentile	50th percentile	75th percentile	95th percentile
500 or less	425	16.7	47.1	146.0	666.0
501–3,300	295	5.0	26.3	85.3	518.0
3,301–10,000	89	5.0	5.0	35.9	276.5
$>\mathrm{10,000}$	89	0.91	6.7	27.1	202.3

Note: Because the finished water concentration is unknown, Table 2 excludes the 21 systems that have results based on raw water samples and treatment technology expected to reduce manganese.

Infant Formula

Two independent rounds of sample preparation according to label instructions were conducted, each round performed by a different person. A statistically significant difference in manganese concentration was found between Round 1 (mean $231\pm 147\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$) and Round 2 (mean $216\pm 147\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$) in the paired-samples t-test; $p=0.0016$. Manganese concentrations from Round 1 tended to be slightly and consistently higher (Figure S2). The average percent difference in concentration between the two sampling rounds was 9% and ranged from $<1\%$ to 26%. Because some level of difference between individuals and preparation events is expected, the average formula concentration was used in all further data analyses. Average manganese concentrations from the 27 prepared formulas ranged between 69.8 and $741\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, with a mean of $224\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ and a median of $191\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ (Table 3). Soy formulas had higher concentrations than cow’s milk formulas. Although only one amino acid formula was tested, it had the highest manganese concentration ($741\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$).

Table 3.

Manganese concentration in 27 reconstituted infant formulas ($\mu \mathrm{g}/\mathrm{L}$).

Formula type	$n$	$\text{Mean}\pm \text{SD}$	Min	25th percentile	50th percentile	75th percentile	95th percentile	Max
All	27	$224\pm 145$	69.8	133	191	273	531	741
Cow’s milk	22	$170\pm 61$	69.8	132	165	214	271	291
Soy	4	$388\pm 101$	306	325	359	422	509	531
Amino acid	1	$\text{Concentration}=741$

Note: Numerical data can be found in Table S2. Max, maximum; min, minimum; SD, standard deviation.

The formula concentrations were compared with a concentration equivalent to the MDH’s neonatal-based RfD (Figure 2). To calculate the equivalent concentration, the RfD ($83\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$) was divided by the mean ($0.136\hspace{0.17em}\mathrm{L}/\text{kg}/\mathrm{d}$) or 95th percentile ($0.290\hspace{0.17em}\mathrm{L}/\text{kg}/\mathrm{d}$) formula-fed infant water consumption rate (EPA 2019; also see Figure S1). Only the amino acid formula exceeded the mean consumption-based RfD equivalent concentration ($610\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}=83\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}÷0.136\hspace{0.17em}\mathrm{L}/\text{kg}/\mathrm{d}$). Six of the 27 formulas exceeded the 95th percentile consumption-based RfD equivalent concentration ($286\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}=83\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}÷0.290\hspace{0.17em}\mathrm{L}/\text{kg}/\mathrm{d}$). These six included all of the amino acid and soy formulas as well as one cow’s milk formula.

Figure 2.

Manganese concentrations in 27 infant formulas reconstituted with purified water compared with MDH’s neonatal-based reference dose (RfD). Note: Two independent rounds of sample preparation were conducted. The average concentration from the two rounds of sample preparation is presented in this figure. Numeric data are presented in Table S2. Formula concentration corresponding to 100% of the $\text{RfD}=\text{RfD}\hspace{0.17em}(83\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d})÷\text{mean}\hspace{0.17em}(0.136\hspace{0.17em}\mathrm{L}/\text{kg}/\mathrm{d})$ or 95th percentile ($0.290\hspace{0.17em}\mathrm{L}/\text{kg}/\mathrm{d}$) formula-fed infant water consumption rates (EPA 2019; Table 3–5).

The calculated label-based concentration of manganese in $\mu \mathrm{g}/\mathrm{L}$ was also compared with the concentration of manganese measured by the PHL in the prepared infant formula. Label-based manganese concentrations were consistently lower than the measured concentrations in all cases (Figure 3). The mean ratio of measured to labeled concentration was 2:3, with a range of 1:3 to 4:8.

Figure 3.

Measured vs. labeled manganese concentration in 27 infant formulas reconstituted with purified water. Note: Numeric data are provided in Table S3.

Formula-Fed Infant Risk Assessment

Estimated risks from formula alone, and formula reconstituted with drinking water, are shown in Table 4 (central tendency exposure scenario) and Table 5 (upper-end exposure scenario). The central tendency exposure estimates are based on a mean water consumption rate and formula concentration, combined with median or 95th percentile CPWS water concentration. Manganese intake from cow’s milk and soy formula alone, without any contribution from water, constituted 28% and 64% of the MDH’s infant-based RfD, respectively. Cow’s milk and soy formula reconstituted with water at the median CPWS concentration ($29.0\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$) marginally increased intake to 33% and 68% of the RfD, respectively. Combined formula plus water intake at the 95th percentile CPWS concentration ($562\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$) increased manganese intake to 120% and 156% of the RfD. Based on this analysis, central tendency exposure estimates from cow’s milk and soy formulas would exceed the MDH’s infant-based RfD when reconstituted with water at manganese concentrations greater than 440 and $220\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, respectively. The central tendency exposure estimate for manganese intake from the amino acid formula alone constituted 122% of the MDH’s infant-based RfD and increased to 127% and 213% when reconstituted with water at the median and 95th percentile CPWS concentrations, respectively.

Table 4.

Central tendency estimate of manganese intake in infants 1–3 months old from formula reconstituted with CPWS drinking water at the median ($29\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$) or the 95th ($562\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$) percentile concentrations.

	Formula contribution only			Formula reconstituted with median CPWS water concentration		Formula reconstituted with 95th percentile CPWS water concentration
	Measured concentration ($\mu \mathrm{g}/\mathrm{L}$)	Intake ($\mu \mathrm{g}/\text{kg}/\mathrm{d}$)a	Intake as percent of MDH RfD ($83\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$)	$\text{Formula\hspace{0.17em}}+\text{\hspace{0.17em}water intake}$b ($\mu \mathrm{g}/\text{kg}/\mathrm{d}$)	$\text{Formula\hspace{0.17em}}+\text{\hspace{0.17em}water intake}$ as % of MDH RfD ($83\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$)	$\text{Formula\hspace{0.17em}}+\text{\hspace{0.17em}water intake}$b ($\mu \mathrm{g}/\text{kg}/\mathrm{d}$)	$\text{Formula\hspace{0.17em}}+\text{\hspace{0.17em}water intake}$ as % of MDH RfD ($83\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$)
Cow’s milk formula	170	23.1	28	27.1	33	99.6	120%
Soy formula	388	52.8	64	56.8	68	129	156%
Amino acid formulac	741	101	122	105	127	177	213%

Note: No adjustments for absorption differences from neonatal rats have been made.

^aFormula $\text{intake}=\text{mean}$ formula manganese $\text{concentration}\times \text{mean}$ total water intake rate of $0.136\hspace{0.17em}\mathrm{L}/\text{kg}/\mathrm{d}$ for 1- to $<3\text{-month}$-old formula-fed infant (EPA 2019; Table 3–5).

^bWater intake at median CPWS $\text{concentration}=29.0\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}\times 0.136\hspace{0.17em}\mathrm{L}/\text{kg}/\mathrm{d}=3.94\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$. Water intake at 95th percentile CPWS $\text{concentration}=562\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}\times 0.136\hspace{0.17em}\mathrm{L}/\text{kg}/\mathrm{d}=76.4\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$.

^cMean and 95th percentile calculations both use formula concentration of $741\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$.

Table 5.

Upper-end estimate of potential manganese intake in infants 1–3 months old from formula reconstituted with CPWS drinking water at the median ($29\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$) or the 95th ($562\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$) percentile concentrations.

	Formula contribution only			Formula reconstituted with median CPWS water concentration		Formula reconstituted with 95th percentile CPWS water concentration
	Measured concentration ($\mu \mathrm{g}/\mathrm{L}$)	Intake ($\mu \mathrm{g}/\text{kg}/\mathrm{d}$a)	Intake as % of MDH RfD ($83\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$)	$\text{Formula\hspace{0.17em}}+\text{\hspace{0.17em}water intake}$b ($\mu \mathrm{g}/\text{kg}/\mathrm{d}$)	$\text{Formula\hspace{0.17em}}+\text{\hspace{0.17em}water intake}$ as % of MDH RfD ($83\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$)	$\text{Formula\hspace{0.17em}}+\text{\hspace{0.17em}water} \text{intake}$b ($\mu \mathrm{g}/\text{kg}/\mathrm{d}$)	$\text{Formula\hspace{0.17em}}+\text{\hspace{0.17em}water intake}$ as % of MDH RfD ($83\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$)
Cow’s milk formula	271	78.6	95	87.0	105	242	292
Soy formula	509	148	178	156	188	311	375
Amino acid formulac	741	215	259	223	269	378	455

Note: No adjustments for absorption have been made.

^aFormula $\text{intake}=95\text{th}\hspace{0.17em}\text{percentile}$ formula manganese $\text{concentration}\times 95\text{th}\hspace{0.17em}\text{percentile}$ total water intake rate of $0.290\hspace{0.17em}\mathrm{L}/\text{kg}/\mathrm{d}$ for 1- to $<3\text{-month}$-old formula-fed infants (EPA 2019; Table 3–5).

^bWater intake at median CPWS $\text{concentration}=29.0\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}\times 0.290\hspace{0.17em}\mathrm{L}/\text{kg}/\mathrm{d}=8.41\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$. Water Intake at 95th percentile CPWS $\text{concentration}=562\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}\times 0.290\hspace{0.17em}\mathrm{L}/\text{kg}/\mathrm{d}=163.0\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$.

^cMean and 95th percentile calculations both use formula concentrations of $741\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$.

The upper-end exposure estimates are based on a 95th percentile water consumption rate and formula concentration combined with median or 95th percentile CPWS water concentrations. Manganese intake from cow’s milk, soy, and amino acid formula alone constituted 95%, 178%, and 259% of the MDH’s infant-based RfD, respectively, essentially leaving no room for manganese in drinking water (Table 5). Upper-end intake estimates for formula reconstituted with water at the median CPWS concentration ($29.0\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$) increased intake to 105%, 188%, and 269% of the RfD, and using a water concentration at the CPWS 95th percentile ($562\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$) increased intake to 292%, 375%, and 455% of the RfD. These estimates are concerning, as the upper-end intake from water alone at the CPWS 95th percentile is $163\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$, which is nearly twice the MDH’s infant-based RfD of $83\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$, leaving no room for contribution from formula.

Central tendency and upper-end exposure assessments were also conducted using the MDH’s HBGV of $100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ and EPA’s HAL of $300\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ (Tables S4 and S5). The resulting estimates also highlight concerns for infants with upper-end water consumption rates. Of particular concern is the upper-end intake estimate from water alone at EPA’s HAL, which exceeded the MDH’s neonatal-based RfD.

It should be noted that the comparisons to the MDH’s RfD presented in Tables 4 and 5 assume infant absorption rates are similar to neonatal rats. Unfortunately, there are very limited data regarding manganese absorption in young infants. Dörner et al. (1989) evaluated manganese intake and retention from human milk and cow’s milk formula in full-term and preterm infants. The two cow’s milk formulas contained 77 and $99\hspace{0.17em}\mu \mathrm{g}$ manganese/L. Absorption in full-term infants was estimated to be 16–21% for cow’s milk formula, compared with 37% for breastmilk. Keen et al. (1986) evaluated manganese absorption from human milk, cow’s milk formula, and soy formula in neonatal rats. Estimated absorption in 14-d-old rats was similar for human milk and cow’s milk formula (82% and 77%, respectively) but lower for soy formulas ($\sim 65\%$). These limited data suggest that absorption rates may be lower in infants than neonatal rats. However, this comparison is based on a single human infant study conducted only on cow’s milk formula, which contained relatively low manganese concentrations.

The relative absorption rate in human infants compared with the absorption in neonatal rats, which would result in the combined upper-end exposure estimates exceeding the MDH’s RfD ($83\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$) when formula was reconstituted with water at the median or 95th percentile CPWS manganese concentration, ranged from 37% to 95% or 22 to 34%, respectively (Table 6). Based on absorption rates reported in human infants (Dörner et al. 1989) and neonatal rats (Keen et al. 1986), it is possible that absorbed intake from upper-end cow’s milk formula reconstituted with water at or below the 95th percentile CPWS concentration will not exceed the MDH’s RfD, but this is not certain. For soy formula, the relative absorption rate would need to be 27% or lower. Absorption rates for soy formula have not been reported in infants; however, $\sim 65\%$ absorption has been reported in neonatal rats (Keen et al. 1986), indicating a possible health risk. No human or animal data exist to compare to the maximum allowable absorption of amino acid formula.

Table 6.

Percent absorption from reconstituted formula relative to absorption in neonatal rats required to maintain intake at or below the MDH’s manganese reference dose (RfD) of $83\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$.

	$\text{Formula\hspace{0.17em}}+\text{\hspace{0.17em}water intake}$ at CPWS, 50th percentile ($\mu \mathrm{g}/\text{kg}/\mathrm{d}$)	Max allowable absorption (%)	$\text{Formula\hspace{0.17em}}+\text{\hspace{0.17em}water} \text{intake}$ at CPWS, 95th percentile ($\mu \mathrm{g}/\text{kg}/\mathrm{d}$)	Max allowable absorption (%)
Central tendency intake estimate
Cow’s milk formula	27.1	100	99.6	83
Soy formula	56.8	100	129	64
Amino acid formula	105	79	177	47
Upper-end intake estimate
Cow’s milk formula	87.0	95	242	34
Soy formula	156	53	311	27
Amino acid formula	223	37	378	22

Note: Max, maximum.

Household Perceptions of Water and Water Use

In the Wells and Increased Infant Sensitivity and Exposure Study, 71% of untreated outdoor faucet samples exceeded the MDH’s HBGV of $100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, and 56% of samples exceeded the EPA HAL of $300\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$. The MDH and Dakota County first assessed whether the manganese well water concentration, represented by the outside spigot sample, was associated with well users’ concerns about esthetic issues (taste, odor, or color) or their water use and treatment behaviors. Household concern about the taste, odor, or color of their water was compared across three manganese concentration categories ($n=254$ homes). We found that the higher the manganese concentration, the higher the level of concern regarding taste, odor, or color (Figure 4; Table S6). For example, 16% of those with manganese below $100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ said they were “somewhat” or “very” concerned about the taste, odor, or color of their water compared with 27% in the next higher bin and 46% in the highest concentration bin. These differences were statistically significant (chi-square test; ${\chi }^2=35.72$, $\text{df}=6$, $p<0.0001$). However, 54% of respondents with well water concentrations over $300\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ still reported being “not at all” or “not very” concerned about the taste, odor, and color of their water. Iron can also impart noticeable taste and color to water, and manganese and iron concentrations were positively correlated in this study (Scher and Demuth 2017). Therefore, we do not know to what extent these esthetic concerns can be attributed solely to manganese. We also compared households that reported “mostly or always” drinking softened or filtered water or purchasing bottled water to those that do not “mostly or always” take these mitigative actions by manganese concentration category ($n=258$) (Table S6). Households with manganese concentrations above $300\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ were significantly more likely to report treating or softening their water or using bottled water (85%) compared with those with an outside tap concentration between 100 and $300\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ (76%) or $<100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ (71%) (${\chi }^2=6.72$, $\text{df}=2$, $p=0.0347$). Although there was a significant trend across concentration groups ($p=0.0099$), there was no statistically significant difference in mitigative action when limiting the comparison to the two lowest manganese concentration bins ($<100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ vs. $100-300\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$; ${\chi }^2=0.3545$, $\text{df}=1$, $p=0.5516$).

Figure 4.

Esthetic concerns about household drinking water by outdoor tap manganese concentration category ($n=254$). Note: Numeric data are provided in Table S6.

A subset of households with outside spigot manganese concentrations $>100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ participated in follow-up sampling from the inside tap after the water had gone through any treatment devices in the home ($n=99$). Of the 37 households in which the inside tap concentration was still over $100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, only 8 households (22%) practiced water avoidance (reported “mostly or always” drinking bottled water or getting drinking water from another source). Of the 26 households with an inside tap result over $300\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, only 4 households (15%) practiced water avoidance.

Discussion

For nearly 50 y, researchers have identified early life as a period of high exposure and biological vulnerability to manganese toxicity (Miller et al. 1975). Over a quarter-century ago, the National Academy of Science recommended changes to regulatory practice to ensure proper characterization of risks to infants and children (NRC 1993). In 1995, the EPA issued its policy on evaluating risks to children, which stated that the EPA would consistently and explicitly consider risks to infants and children as part of risk assessment and standard setting (EPA 1995), reaffirmed in 2013 (EPA 2013). The WHO, the Organization for Economic Co-operation and Development (OECD), and the European Food Safety Authority (EFSA) have also acknowledged the need to identify the most vulnerable subpopulations, based on windows of sensitivity or high exposure, and to incorporate this information into risk assessment (Cohen Hubal et al. 2014; EFSA Scientific Committee et al. 2017; OECD 2019). Although there appears to be a clear consensus regarding the importance of characterizing risks to infants and children, actions to characterize these risks have not been routinely implemented. In the EPA’s most recent manganese risk assessment, the EPA did not develop a separate child-specific health reference level (EPA 2003b). The EPA did consider the vulnerability of infants in establishing the HAL of $300\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ for manganese by stating that infants should not be given drinking water above the lifetime HAL for more than 10 d (EPA 2004). However, the EPA did not identify a drinking water level that would be safe for infants to consume for longer than 10 d. The FDA has not assessed the need to establish a maximum concentration of manganese in infant formula despite long-standing warnings about the potential for overexposure in the published literature (Keen et al. 1986; Davidsson 1989; Dörner et al. 1989; Lönnerdal 1997; Ljung et al. 2011) and from science advisory committees (Raiten et al. 1998; IOM 2001; Koletzko et al. 2005). Manganese falls into a particularly complicated regulatory gap in which formula manufacturers are not required to consider the contribution of manganese in drinking water when developing their powdered formulas.

Drinking Water Concentrations

The existence of a Minnesota study that measured manganese in nearly all current CPWSs in the state allowed us to determine that manganese exceeded the MDH HBGV and the EPA HAL in 27% and 10% of Minnesota systems, respectively. Seventeen percent of systems (serving 214,460 people) exceeded $192\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, the benchmark concentration associated with a 5% drop in IQ performance for girls in a recent benchmark dose modeling analysis (Kullar et al. 2019). Elevated manganese concentrations tended to occur in smaller systems, raising environmental health equity concerns. Water systems obtain most of their revenue from user charges (ASCE 2017), making small systems at a greater disadvantage in providing safe drinking water because of budget constraints. Ninety-seven percent of water systems in the United States are considered small to medium systems, defined by the EPA as serving 10,000 or fewer people (https://www.epa.gov/water-research/small-drinking-water-systems-research).

There are uncertainties in the drinking water analysis. Although naturally occurring manganese levels in groundwater are expected to be stable over time, some CPWS monitoring results may not reflect current manganese concentrations if changes had been made to system operations since monitoring took place. To address this issue, any tested wells that had been sealed or that had a current classification of “emergency use” were excluded from the analysis. We were unable to account for changes in well status for finished water samples because these samples may be a blend of multiple water sources. It is also possible that some systems already manage esthetic issues associated with manganese by limiting the pumping of higher-manganese wells. Not all system wells/intakes or entry points may have been sampled, and average results may not account for concentrations at the consumer tap because of pressure zones and mixing within a system. We lacked data for 46 small systems; systems of this size were more likely to have elevated levels. Concentrations at the tap may also be underestimated over short periods because of the potential for intermittent release of accumulated manganese from biofilms and scale deposits in distribution system pipes, which has been previously documented (Health Canada 2019).

UCMR4 sampling for manganese took place through 2020 at all large CPWSs (serving more than 10,000 people) and at 1.5% of small and medium systems (serving $\le 10,000$ people) to provide a nationally representative sample of CPWSs (EPA 2016b). However, the EPA’s reliance on national-level occurrence data and large system results dilutes the exposure picture for areas of the country that are known to have higher groundwater concentrations and/or more heavily rely on groundwater sources for drinking water (e.g., 75% of Minnesota residents rely on groundwater for drinking water). Therefore, it is possible that the EPA may again determine that regulating manganese will not present a meaningful opportunity for health risk reduction. Although Minnesota, like many states, has never formally promulgated a regulatory standard for contaminants in drinking water lower than a federal standard, it is already taking actions to reduce manganese drinking water exposure. The MDH has begun retesting all CPWSs with General Water Chemistry Project results above the HBGV of $100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ that were not recently tested under the UCMR4 program and will develop response and communications plans to address exceedances. Because results show that some systems with treatment to remove manganese still had concentrations above $100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, the MDH is also developing tools to help operators optimize filter performance.

This assessment does not include private well drinking water concentrations upon which 20% of people in Minnesota rely (MDH n.d.). Private well households in Minnesota are disproportionately impacted by elevated manganese in drinking water, with approximately one-half and one-quarter of private wells estimated to exceed the MDH HBGV and EPA HAL, respectively (MGWA 2015). This is not a Minnesota-specific issue, as 18% of all wells and 10% of drinking water wells were found to exceed the EPA HAL in the larger U.S. glacial aquifer system covering 26 northern states (Groschen et al. 2008). Seven states in the United States have a higher percent of HAL groundwater exceedances than Minnesota (McMahon et al. 2019), suggesting private well impacts may be far-reaching. More outreach to private well households is needed to promote manganese testing prior to giving the water to an infant. The MDH’s “Well Water and Your Baby” brochure recommends testing for manganese and is a communications tool that can be adapted for use by other states and tribes (MDH 2019).

Formula Concentrations

Levels of manganese in prepared formula in this study were generally higher than several previous reports in which cow’s milk and soy formula contained $30-100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ and $\sim 100-340\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ manganese, respectively (Lönnerdal 1997; Cockell et al. 2004). In comparison, we found concentrations of $69.8-291\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ and $306-531\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ for cow’s milk and soy formula, respectively. A contemporaneous study (Frisbie et al. 2019), however, has reported even higher concentrations, with $230-430\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ and $420-790\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ in U.S. cow’s milk and soy infant formulas, respectively. Forty-seven percent of formulas in Frisbie et al. (2019) were above the IEG’s recommended maximum of $338\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ without water contribution.

In deriving the 2012 HBGV of $100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ for manganese in drinking water, the MDH used an RSC of 50% for infant exposure (MDH 2008). This apportionment was intended to acknowledge formula as a significant known source of nonwater exposure and was based, in part, on upper range concentrations reported for cow’s milk ($50\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$) and soy ($300\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$) formula (Aschner and Aschner 2005). The 95th percentile manganese concentrations in cow’s milk and soy formula empirically measured by the MDH’s PHL were 5- and 1.7-fold higher, respectively, than the formula concentrations considered in deriving the MDH’s guidance value. These results suggest that formula may contribute a much greater fraction of exposure. Nearly 90% of the formulas tested contained concentrations greater than the upper range concentrations considered in the HBGV derivation. Because formula and water can each be a significant source of manganese, there is a need to address both sources of exposure in order to address potential risks to infants.

While fewer soy formulas were tested, concentrations in these formulas were roughly two times higher than cow’s milk formula. Soy protein formula naturally contains higher concentrations of manganese than cow’s milk (Lönnerdal et al. 1981, 1983; Stastny et al. 1984; Lönnerdal 1994, 1997; Cockell et al. 2004), but the lack of studies on formula-based absorption differences in young infants limits interpretation. Only one published study, in neonatal rats (Keen et al. 1986), has evaluated differences in absorption by formula type. This study did not find a large difference in absorption (77% for cow’s milk vs. 65% for soy formula).

The highest formula concentration found in this study was in the amino acid formula ($741\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$). This is particularly concerning because these formulas are used not only for infants with cow’s milk protein allergy but may also be recommended for infants that are otherwise medically vulnerable (e.g., premature infants with feeding intolerance) (Meyer et al. 2018). To our knowledge, only Frisbie et al. (2019) have tested amino acid infant formulas. Concentrations in the two tested formulas were markedly different (830 and $210\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$) and demonstrate that not all formulas of this type may be high in manganese. Roughly a third of U.S. infants are not fed cow’s milk formula (Rossen et al. 2016), which is generally lowest in manganese (Cockell et al. 2004; Frisbie et al. 2019; current study). This fact further emphasizes the need for additional testing of manganese formula types and absorption differences.

The consistently higher prepared vs. labeled manganese concentrations (2.3-fold higher on average) are concerning, as the difference is greater than what one may expect because of analytical variability between laboratories. It may be in the manufacturers’ interest to ensure that levels are well above the legal minimum concentration, as well as labeled levels, to avoid products being found misbranded or adulterated by the FDA, which can lead to product recall (21 CFR 107.200–107.210) (FDA 2020c). The results indicate that a consumer may not be able to depend on labeled information regarding the concentration of manganese in prepared formula. Concentrations of manganese above health-based guidelines from formula alone, and the differences found between prepared and labeled concentrations, suggests the need for the FDA to establish maximum levels of manganese in infant formula, as previously recommended by the IOM and the Codex IEG. Given existing concentrations of manganese in water, unnecessarily high formula concentrations translate into a greater percentage of infants exposed above health-based guidance and at increased risk. Therefore, the EPA and FDA must work together to jointly consider risks to formula-fed infants from infant formula and the water used to reconstitute it.

Formula-Fed Infant Risk Assessment

The risk assessment suggests that central tendency exposure estimates for cow’s milk and soy formulas could pose an unacceptable risk when formula is reconstituted with water containing manganese greater than 440 or $220\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, respectively. Approximately 7% of Minnesota CPWSs (serving 50,680 people) exceeded $440\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$, and 15% of Minnesota CPWSs (serving 166,759 people) exceeded $220\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$. Upper-end exposure estimates for cow’s milk, soy, and amino acid formulas were equal to or exceeded the RfD without any manganese contribution from water. For water to contribute an equal amount as formula, water concentrations would need to approach or exceed $600\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$. If water containing $600\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ was used to reconstitute formula, which approximates the 96th percentile CPWS concentration in Minnesota, the RfD would be exceeded by more than 2-fold. These estimated doses, however, were not adjusted for potential absorption differences between infants and neonatal rats (the basis of the RfD). More research on absorption of manganese in neonates is needed, specifically for specialty formulas such as amino acid formula. Formula manufacturers may have additional information on absorption, such as unpublished studies, and would thus serve as valuable research partners.

The MDH’s RfD is based on changes in behavior and learning ability in male neonatal rats (Kern et al. 2010). Primates may be a more appropriate model than rodents to assess neurologic effects. Golub and colleagues (Golub et al. 2005, 2012) reported changes in sleep, activity, and behavior patterns in neonatal male rhesus monkeys exposed to 106 and $323\hspace{0.17em}\mu \mathrm{g}/\text{kg}/\mathrm{d}$ manganese in soy formula. The upper-end formula plus water exposure estimates reported here fall within these dose levels. Several study design and data reporting limitations precluded use of the Golub study as the basis of the MDH’s RfD. Nevertheless, this study underscores the potential toxicological significance of manganese exposures in formula-fed infants.

There are some uncertainties in the parameters used in the risk assessment:

• Point of departure: The lowest dose used by Kern et al. (2010) was identified as a LOAEL ($25\hspace{0.17em}\text{mg}/\text{kg}/\mathrm{d}$) based on behavior and learning deficits in male neonatal rats. A LOAEL-to-NOAEL uncertainty factor of 3 was applied; however, there may be the risk of neurological effects in human neonates at lower concentrations. The WHO, in their recently released draft guidance, applied a LOAEL-to-NOAEL uncertainty factor of 10 (WHO 2020).

• Formula concentration: The range of manganese concentrations is based on a limited number of samples and a limited number of cans of formula. Actual manganese formula concentrations may differ between lots and from can to can, and therefore, the actual range may be larger than reported here. We also based the risk assessment on the mean, not maximum, manganese concentration from two rounds of sample preparation. As we observed in this study, formula concentration will differ between (and likely within) individuals based on preparation practices, even when the label directions are followed.

• Water concentration: The median and 95th percentile were on a CPWS basis, not on a population basis (i.e., number of infants exposed).

• Water intake rates: EPA-recommended consumption rates were based on survey respondents who reported any ingestion of formula that was reconstituted with community water at least once in the past 2 d and, therefore, likely include the lower water intakes of infants receiving breastmilk supplemented with formula.

• Manganese absorption: Due to inadequate data regarding manganese absorption in infants, the estimated manganese intakes were not quantitatively adjusted for absorption rate. The chemical valence state of manganese affects its absorption. Infant formula contains manganese in the divalent state, which, unlike the complex-associated trivalent state of manganese found in human breastmilk, does not bind to lactoferrin, and therefore, intestinal uptake cannot be regulated by lactoferrin receptors (Health Canada 2019). If absorption in neonatal humans is similar to neonatal rats (the basis of the MDH’s RfD), the upper-end exposure estimates from reconstituted formula are within 100-fold of the LOAEL.

The analysis presented here provides a central and upper-end estimate of exposure and potential risk. This work could be expanded to include probabilistic risk estimates (e.g., Chiu and Slob 2015). Distributional information for many of the exposure parameters (e.g., formula and water concentrations, water consumption rates) is readily available. However, distributional information regarding dose–response [e.g., critical effect(s) metric, identity of sensitive age window(s), and uncertainty factors] relevant to neonatal exposure to manganese would be far more challenging.

Recently, Yoon et al. (2019) published a physiologically-based pharmacokinetic (PBPK) model assessing children’s internal exposure to manganese in drinking water. Modeling results indicated that manganese brain concentrations in formula-fed infants are actually lower than those of breastfeeding infants, up to water concentrations of $\sim 500\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$. However, it should be noted that the PBPK model utilized an extremely low manganese absorption rate of only 0.06% for reconstituted formula (1/1,000th of that used for breastmilk), which is not consistent with limited available literature on absorption in human infants or adults (Davidsson et al. 1989; Dörner et al. 1989) or neonatal rats (Keen et al. 1986).

Household Perceptions of Water Quality and Water Use

We found that elevated levels of manganese in tap water may increase esthetic concerns but were not an absolute consumption deterrent, even at concentrations well above the EPA secondary standard of $50\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$. While the Wells and Increased Infant Sensitivity and Exposure study was limited to one geographic area, the results suggest that consumption of water with elevated levels of manganese could be a common occurrence, as 78% of households with an inside tap result above $100\hspace{0.17em}\mu \mathrm{g}/\mathrm{L}$ reported drinking the water. One limitation of the analysis is the positive correlation found between manganese and iron, as both impart taste and staining issues at elevated levels. In addition, the number of households participating in the inside tap water sampling portion of the study was small and not limited to those with infants. However, the results are consistent with previous studies that find that a consumer acceptability threshold for manganese cannot be assumed to offer protection from concentrations above levels of health concern (Cohen et al. 1960; Sain and Dietrich 2014; Frisbie et al. 2015; Ander et al. 2016).

Taking a broader view, manganese is just one naturally occurring element that may be present in both powdered infant formula and the water used to reconstitute it, with potential health concerns at high levels of exposure (e.g., boron, zinc). Unintended silos across the disciplines of nutrition, toxicology, epidemiology, geology, and pharmacology have likely served to impede awareness and assessment of the combined contribution to total oral exposure from infant formula and water constituents. At the federal regulatory level, coordination between the FDA and the EPA is needed to address these combined exposures and to jointly consider and address risks to formula-fed infants.