Early life organophosphate ester exposures and bone health at age 12 years: The Health Outcomes and Measures of the Environment (HOME) Study

Abstract

Background:

No human studies have evaluated early life organophosphate ester (OPE) exposures with bone health outcomes, despite evidence of osteotoxicity.

Objectives:

We assessed associations of urinary OPE metabolites measured across early life with areal bone mineral density (aBMD) and bone mineral content (BMC) at age 12 years.

Methods:

Among 223 mother-child dyads enrolled in the Health Outcomes and Measures of the Environment (HOME) Study, we quantified concentrations of bis-2-chloroethyl phosphate (BCEP), bis-(1,3-dichloro-2-propyl) (BDCIPP), di-n-butyl phosphate (DnBP), and diphenyl phosphate (DPHP) in urine collected from mothers during pregnancy and children at ages 1, 2, 3, 5, and 8 years. At age 12 years, we performed dual energy x-ray absorptiometry and calculated aBMD and BMC z-scores at six skeletal sites. We estimated overall and sex-stratified BMD/BMC z-score differences per interquartile range (IQR) increase in OPE concentrations at multiple exposure timepoints: gestation (average) and 1–3 (average), 5, and 8 years.

Results:

In adjusted models, overall associations of BCEP and BDCIPP with total hip and 1/3^rd distal radius aBMD and BMC varied significantly by exposure timepoint, as did BDCIPP with whole body aBMD. For example, differences (95% CI) in total hip aBMD z-score per IQR increase in BDCIPP were 0.33 (0.01, 0.64), −0.10 (−0.34, 0.14), −0.18 (−0.40, 0.05), and 0.14 (−0.09, 0.38) for concentrations during gestation and at 1–3, 5, and 8 years, respectively. Overall DnBP and DPHP associations were generally null at all timepoints. We observed sex-specific associations for some timepoints and skeletal sites. For example, an IQR increase in 8-year DPHP was associated with a 0.21 (0.05, 0.38) greater total hip aBMD z-score among females but −0.19 (−0.43, 0.05) lower z-score among males.

Discussion:

Early life OPE exposures may be associated with sex- and exposure period-dependent alterations in early adolescent bone mineral accrual and strength.

1. Introduction

Low bone mineral density (BMD), a measure of bone strength, is associated with increased risk of fractures in children (Clark et al., 2006; Kalkwarf et al., 2011) with about half of boys and a third of girls fracturing at least one bone by age 18 years (Jones et al., 2002; Landin, 1983; Mayranpaa et al., 2010). Lower bone mass accrual in adolescence is associated with low BMD and osteoporosis in adulthood (Weaver et al., 2016), which collectively affects more than 50 million adults over age 50 in the U.S. (Wright et al., 2014). Over 2 million fractures occur each year among individuals with osteoporosis, resulting in substantial morbidity and mortality (Haentjens et al., 2010; Singer et al., 2015) and costing as much as $20 billion in annual direct medical costs (Becker et al., 2010; Burge et al., 2007). Acquisition of bone commences during gestation (Harrast and Kalkwarf, 1998) and bone mineral accretion rates peak around age 12.0–12.5 years for females and 14 years for males (Bailey et al., 1999; Jones et al., 2002; McCormack et al., 2017). Since 95% of adult total body bone mass is achieved within a few years of puberty (Baxter-Jones et al., 2011), identifying modifiable early life risk factors is critically important for fostering optimal skeletal health throughout life.

Organophosphate esters (OPEs) are synthetic flame retardants and plasticizers used in items such as clothing, furniture foam, electronics, baby products, and nail polish (Hoffman et al., 2015; Hoffman et al., 2017; Mendelsohn et al., 2016; Romano et al., 2017). OPEs were introduced to the market as a replacement for polybrominated diphenyl ethers, which were voluntarily phased out by manufacturers in the mid-2000s due to concerns about toxicity and persistence in the environment (Linares et al., 2015). While OPEs are rapidly metabolized and excreted in urine, human biomonitoring studies report widespread exposure to several OPEs among the general U.S. population (Ospina et al., 2018). Several studies have additionally reported OPE concentrations in the urine of pregnant women, breastmilk, and placental tissue (Butt et al., 2014; Ding et al., 2016; Hoffman et al., 2017; Kim et al., 2014; Kuiper et al., 2020; Percy et al., 2020; Romano et al., 2017; Sundkvist et al., 2010). Findings from animal models implicate certain OPEs as neurotoxic, carcinogenic, and reproductive toxicants, as well as endocrine disruptors (Liu et al., 2012; Preston et al., 2017; van der Veen and de Boer, 2012; Xu et al., 2017).

Emerging evidence suggests that endocrine disrupting chemicals, including OPEs, may exhibit osteotoxic effects resulting in altered bone morphogenesis, mineralization, or remodeling (Fernandez et al., 2018). An in vitro study of triphenyl phosphate (TPHP) and Firemaster® 550, a chemical formulation that includes TPHP, reported that both compounds activate peroxisome proliferator-activated receptor gamma (PPARγ) and guided differentiation of bone marrow multipotent mesenchymal stem cells (BMSCs) away from osteogenesis (Pillai et al., 2014). Several other studies have reported that OPEs alter PPARs (Belcher et al., 2014; Fang et al., 2015; Hu et al., 2017; Kojima et al., 2013; Pillai et al., 2014; Tung et al., 2017b) and interfere with sex steroid and thyroid hormones (Farhat et al., 2013; Kim et al., 2015; Krivoshiev et al., 2016; Liu et al., 2012; Meeker and Stapleton, 2010; Percy et al., 2020; Preston et al., 2017; Schang et al., 2016; Wang et al., 2015; Zhang et al., 2016), which may in turn alter bone homeostasis.

In the only animal study of OPEs and frank bone density outcomes, rats were exposed during pregnancy and lactation to 1000 μg/day of Firemaster® 550, and offspring were studied using microCT of the femur at 24 weeks (Macari et al., 2020). Compared with offspring of control rats, femurs of exposed offspring had lower trabecular BMD, lower trabecular thickness, lower cortical thickness, and greater medullary area. Notably, all effects were observed only in male offspring (Macari et al., 2020). There are no prior human studies of OPEs and skeletal outcomes.

We aimed to estimate associations of maternal urinary OPE metabolite concentrations during pregnancy, as well as child OPE metabolite concentrations from early to middle childhood, with BMC and BMD at six skeletal sites among 12-year-old children. We also assessed effect measure modification by child’s sex.

2. Methods

2.1. Study Design and Population

We used data from mother-child dyads within the Health Outcomes and Measures of the Environment (HOME) Study, a prospective birth and pregnancy cohort, which enrolled pregnant women from March 2003 to February 2006 from the Greater Cincinnati metropolitan area (Ohio, US). The HOME study was initially designed as a randomized trial of lead and injury hazard controls to reduce children’s blood lead levels; detailed methods have been described previously (Braun et al., 2020; Braun et al., 2017). Briefly, 1,263 pregnant women were recruited from nine prenatal clinics via clinical records and phone interviews who met the following eligibility requirements: 1) maternal age of at least 18 years; 2) a pregnancy that was approximately 16±3 weeks of gestation; 3) residence in a house that was built prior to 1978; 4) an intention of continuing with prenatal care and delivering at one of the collaborating obstetric practices; 5) an HIV negative status; and 6) the absence of medication usage for seizures, thyroid disorders, or chemotherapy/radiation. A total of 468 women enrolled during pregnancy and 398 delivered liveborn singleton infants. We prospectively followed children at study visits up to age 12 years (Braun et al., 2020). For the present study, we included 223 mother-child dyads who had available urinary concentrations of OPEs quantified [bis(1,3-dichloro-2-propyl) phosphate (BDCIPP), bis-2-chloroethyl phosphate (BCEP), diphenyl phosphate (DPHP), and di-n-butyl phosphate (DnBP)] during at least one timepoint as well as bone parameters measured during adolescence at age 12 years. At the age 12 study visit, caregivers provided written informed consent and children provided written informed assent. The HOME Study protocol and this present study were approved by the Institutional Review Board at the Cincinnati Children’s Hospital Medical Center (CCHMC). The Centers for Disease Control and Prevention (CDC) deferred to CCHMC IRB as the IRB of record.

2.2. OPE Metabolite Quantification

HOME Study mothers provided spot urine samples in polypropylene specimen cups at approximately 16 weeks and 26 weeks of gestation during scheduled visits. We quantified OPE metabolites in urine samples collected from children during clinic visits at up to five timepoints during follow-up: ages 1-, 2-, 3-, 5-, and 8 years. We collected children’s urine samples from diaper inserts (Kendall abdominal pads) for non-toilet trained children; inserts lining a training toilet, for toilet-training children; or directly into specimen cups, for toilet-trained children (Stacy et al., 2017). We stored and refrigerated all samples in specimen cups for up to 24 hours before urine aliquots were made. Then, samples were frozen (−20 °C), stored, and shipped overnight on dry ice to the National Center for Environmental Health Laboratory at the CDC. Prior to analysis, the CDC stored samples from HOME Study mothers at −70 °C. We quantified four OPE metabolites including BDCIPP, BCEP, DPHP, and DnBP, using 200 μL of urine. Parent compounds of these metabolites, and selected uses, are presented in Supplemental Material Table S1. Target metabolites underwent enzymatic hydrolysis of metabolite conjugates followed by automated off-line solid-phase extraction, separation by reversed-phase high-performance liquid chromatography, and detection with isotope dilution-electrospray ionization tandem mass spectrometry (Jayatilaka et al., 2019; Jayatilaka et al., 2017). The limit of detection (LOD) was 0.10 μg/L for all four metabolites, with a 98–108% accuracy and an intra- and inter-day imprecision that was less than 10% (Percy et al., 2020). We quantified urinary creatinine at CDC using a modified Jaffe-kinetic method.

2.3. Bone Mineral Content and Density Measures

At age 12-year study visit, trained research staff measured children’s standing height to the nearest 0.1 cm, in triplicate, with an Ayrton Stadiometer Model S100 (Prior Lake, MN). Technicians performed dual energy x-ray absorptiometry (DXA) scans of the whole-body, lumbar spine, hip, and forearm with a Hologic Horizon densitometer (Marlborough, MA). Scans were analyzed using Apex 5.5 software to estimate BMC (g) and areal BMD (aBMD= BMC/ projected area, g/cm²) for the whole body (excluding head), total hip, femoral neck, 1/3^rd distal radius, ultradistal radius, and lumbar spine. We excluded the head from whole body measures since the proportion of BMC contained in the skull inconsistently varies relative to a child’s body size (Taylor et al., 1997). Further, we estimated a measure of spine volumetric BMD (g/cm³) not affected by height via calculation of spine bone mineral apparent density (BMAD, g/cm³): BMC/bone area^1.5 (Kindler et al., 2019). We assessed the long-term calibration stability of our DXA scanner by daily scanning of the Hologic anthropomorphic spine phantom. The precision error expressed as the coefficients of variation (CV) ranges from 0.75% (total hip aBMD) to 1.85% (spine BMC) in children 10–13 years of age (Shepherd et al., 2011).

Given DXA is a 2-dimensional technology, we aimed to mitigate bias resulting from short or tall stature by calculating height-for-age adjusted age-, sex-, and race-specific whole body (excluding head), total hip, femoral neck, 1/3^rd distal radius, and lumbar spine BMC and aBMD z-scores. Additionally, we calculated age-, sex-, and race-specific ultradistal radius aBMD and spine BMAD z-scores. We calculated z-scores using reference ranges from the Bone Mineral Density in Childhood Study (Kindler et al., 2020; Kindler et al., 2019; Zemel et al., 2011).

2.4. Covariates

We a priori selected covariates using a directed acyclic graph (Supplemental Material Figure S1). Using computer-assisted interviews and medical chart reviews, trained research assistants collected information on mothers and children during maternal, perinatal, and/or postnatal study visits. Collected information included maternal sociodemographic characteristics, such as maternal age at delivery, maternal race/ethnicity (as a proxy for exposure to structural inequities and/or racism), household income at each visit, height, mid-pregnancy weight, baseline parity, prenatal vitamin use during the second or third trimester of pregnancy, and whether they ever breastfed their enrolled child during infancy and for how long; and child characteristics, such as sex and age-, weight-, and height at each follow-up visit. We calculated maternal mid-pregnancy body mass index (BMI) as well as child BMI (at each follow-up visit) and converted these to BMI z-scores. For maternal mid-pregnancy BMI we internally standardized to calculate z-scores, whereas for child measures we externally standardized using either World Health Organization (WHO) reference standards (for children < age 24 months) or CDC reference standards (≥ age 24 months) to calculate age- and sex-specific z-scores (Kuczmarski, 2000; WHO, 2006). Additionally, we collected whole blood from mothers at up to three timepoints (16- and 26 weeks gestation and delivery) and stored samples at −80 °C until shipment on dry ice to the CDC to measure lead concentrations using inductively coupled plasma mass spectrometry (Jones et al., 2017). We used the average of log₂-transformed maternal blood lead concentrations from the available timepoints as a measure of average lead exposure during pregnancy. At the age 12 study visit, children self-staged their pubic hair development using visual aids depicting Tanner stages of development (Braun et al., 2020; Chavarro et al., 2017; Morris and Udry, 1980; Taylor et al., 2001; Yayah Jones et al., 2021).

2.5. Statistical Analyses

2.5.1. Urinary dilution and longitudinal OPE exposure measures

OPE concentrations <LOD were replaced by the $\text{LOD/}\sqrt{2}$ (Hornung and Reed, 1990). All metabolites had detection frequencies ≥80% at all timepoints, except for DnBP at 26 weeks of gestation (77%) and the 8-year visit (66%) (Supplemental Material Table S2). We adjusted OPE metabolite concentrations for urinary dilution using urine creatinine as the proxy of hydration status and a covariate-adjusted standardization approach (O’Brien et al., 2016). Briefly, covariate-adjusted standardization accounts for individual differences in hydration status by multiplying the observed metabolite concentration by the ratio of predicted to observed creatinine concentration. We used separate linear regression models to estimate predicted urinary log₂-transformed creatinine concentrations at each timepoint. For maternal creatinine at 16- and 26-weeks of gestation, we included as covariates: maternal age, maternal mid-pregnancy BMI, maternal height, maternal race/ethnicity, household income, hypertension status, and gestational diabetes status as covariates. For child postnatal creatinine, we included as covariates: sex, age, BMI z-score, height, race/ethnicity (White non-Hispanic, Black/African American non-Hispanic, and other/multiple race or Hispanic ethnicity), and household income. We then multiplied OPE metabolite concentrations by the ratio of predicted to observed creatinine concentrations, resulting in creatinine-standardized metabolite concentrations. We log₂-transformed all creatinine-standardized metabolite concentrations given their log-normal distributions. Finally, we created averages of available log₂-transformed creatinine-standardized OPE metabolite concentrations during gestation (maternal urinary concentrations at 16- and 26- week) and ages 1–3 years (child urinary concentrations at 1, 2, and 3 years) to reduce the number of hypothesis tests. For DnBP, we used only the 3-year timepoint concentrations due to the relatively high proportion of DnBP results that could not be reported because of the presence of analytical interferences during metabolite quantification of 1- and 2-year samples.

2.5.2. Descriptive statistics

For each timepoint, we calculated the detection frequency of each urinary OPE metabolite as well as the geometric mean (GM), geometric standard deviation (SD), and percentiles of metabolite concentrations. We also calculated the proportion of participants within categories of discrete variables and mean (standard deviation [SD]) of continuous variables. To assess where there was evidence of selection bias due to intermittent loss to follow-up, we compared maternal characteristics of our study participants with those of singleton mother-child dyads in the full HOME Study cohort and also compared maternal and child characteristics of participants at each OPE timepoint (gestation, 1–3 year, 5 year, and 8 year).

2.5.3. Multiple informants regression models

We used a multiple informants approach to evaluate the prospective, unadjusted and covariate-adjusted associations of gestational and childhood urinary creatinine-standardized OPE metabolite concentrations at each timepoint (gestation, 1–3 year, 5 year, and 8 year) with BMC and BMD measures at age 12 years. Specifically, we fit separate regression models for each timepoint, then used a seemingly unrelated regression equations approach (with cluster-based sandwich estimation of standard errors) to perform Wald F-tests for global differences in effect estimates across timepoints (Zellner, 1962), with a p-value ≤ 0.2 suggestive of a difference across timepoints. While there is no set rule for what alpha value is appropriate in the context of a test of differences across timepoints, we used this less conservative cutoff given that relatively small sample sizes within each timepoint limits our power to detect potential underlying differences. To facilitate comparisons among OPE effect estimates, we standardized all log₂-transformed OPE metabolite concentrations by their interquartile range (IQR) derived from all timepoints, prior to model fitting. The overall IQRs were: 2.09 (BCEP), 2.29 (BDCIPP), 1.39 (DnBP), and 1.40 (DPHP). While we evaluated for non-linearity of associations of OPE metabolite concentrations with aBMD and BMC z-score outcomes using quadratic and cubic terms, we did not observe consistent evidence of non-linearity, and therefore assumed linear, monotonic relations for all analyses. We used this parameterization of the multiple informants model over the more common generalized estimating equations approach (Buckley et al., 2019; Sánchez et al., 2011) as it allows for different covariate adjustment sets for each timepoint. Covariates for all models included maternal age at delivery, maternal race/ethnicity (White non-Hispanic; all other races/ethnicities), average maternal blood lead concentration during pregnancy, prenatal vitamin use during the second or third trimester (ever, never), baseline parity (nulliparous, primiparous, multiparous), child age at the 12-year visit, and child sex. Since < 6% of participants had a mother who identified as a race other than White or Black/African American or identified as Hispanic ethnicity, we collapsed the non-Hispanic Black/African American category together with all other races, multiple races, and/or Hispanic ethnicity. Additionally, the gestational timepoint models included maternal mid-pregnancy BMI z-score and baseline household income (midpoint of reported category), while childhood timepoint models included household income (at visit), BMI z-score (at visit), and duration of breastfeeding (weeks up to the 3-year visit). In addition to overall models, we examined effect measure modification via sex-stratified models. For all models, we used multiple imputation by chained equations (50 imputed datasets) to account for missing covariate information (< 10% at all timepoints). Finally, given that BMD is the primary clinical indicator of bone diseases, such as osteoporosis, we present associations with aBMD as the primary results and BMC as secondary.

As sensitivity analyses, to assess the separate impacts of diet and physical activity on adjusted associations, we fitted models in the overall study sample that were additionally adjusted for maternal fruit and vegetable consumption during pregnancy (< 4 days per week, ≥ 4 days per week) as well as the child’s average 2010 Healthy Eating Index score (average of scores from up to three available 24-hour recalls) at the 12-year visit (Gajjar et al., 2022; Guenther et al., 2013), or additionally adjusted for the child’s composite Physical Activity Questionnaire score at the 12-year visit (Kowalski et al., 2004).

We used Stata v15.1 (Stata Statistical Software: Release 15. College Station, TX: StataCorp LLC) for all analyses.

3. Results

3.1. Sample characteristics

Of the 223 mother-child dyads contributing information to this study, the average maternal age at delivery was 29 years and 57% of women identified as White non-Hispanic (Table 1). Maternal characteristics of women included in this study were similar to those in the full HOME Study (Table 1). Time-fixed covariates were also similar across participants at the four OPE measure timepoints (Supplemental Material Table S3). The average age of children at the age 12 visit was 149 months (12.4 years) with 25% of males and 44% of females identifying as Tanner stages 4 or 5 (Table 2). BDCIPP and DPHP were the most frequently detected OPE metabolites in urine at all timepoints, with BDCIPP detected in 90% (26 weeks gestation) to 100% (8 years) of samples and DPHP detected in 98% (26 weeks gestation) to 100% (1-, 2-, 5-, and 8 years) of samples (Supplemental Material Table S2). Median concentrations of all OPEs were generally higher at all childhood timepoints compared to gestation (Supplemental Material Table S2). Within timepoint correlations across OPEs metabolite were weak to moderate, ranging from 0.16–0.44 (gestation), −0.02–0.53 (1–3 year), 0.25–0.51 (5 year), and 0.17–0.31 (8 year), and strongest between BCEP and BDCIPP at 1–3 years (ρ = 0.53) (Supplemental Material Figure S2). Within OPE correlations across timepoints were generally lower, and ranged from 0.00–0.31 (BCEP), −0.07–0.38 (BDCIPP), −0.08–0.23 (DnBP), and 0.01–0.27 (DPHP) (Supplemental Material Figure S2).

Table 1.

Baseline characteristics of HOME Study participants included in study sample and in full sample. Statistics are mean (SD) unless otherwise stated.

Characteristics	Study sample (n = 223)	Full sample (n = 398)
Maternal age at delivery, years	29 (5.9)	29 (5.8)
Maternal mid-pregnancy body mass index (kg/m2)	28 (6.8)	27 (6.7)
Missing, n	15	29
Household baseline income, $USD, n (%)
< $20,000	57 (26)	88 (22)
$20,000 – < $40,000	35 (16)	67 (17)
$40,000 – < $80,000	77 (35)	140 (35)
≥ $80,000	54 (24)	103 (26)
Maternal average pregnancy blood lead concentration (μg/dL)	0.68 (1.5)	0.70 (1.4)
Maternal race/ethnicity, n (%)
White non-Hispanic	127 (57)	241 (61)
Black non-Hispanic	84 (38)	129 (33)
Other race/ethnicity	12 (5)	25 (6)
Missing	0	3
Parity, n (%)
Nulliparous	90 (42)	171 (44)
Primiparous	72 (33)	124 (32)
Multiparous	54 (25)	92 (28)
Missing	7	11
Gestational vitamin use, n (%)
Never used	27 (13)	46 (12)
Ever used	188 (87)	340 (88)
Missing	8	12
Breastfeeding duration (weeks)	24 (24.0)	23 (24.6)
Missing, n	12	25

Abbreviations: kg = kilogram; m = meter; mL = milliliter; ng = nanogram; USD = United States Dollars

Table 2.

Characteristics of HOME Study children at the 12-year visit (n = 223). Statistics are mean (SD) unless otherwise stated.

Characteristics	Overall (n = 223)	Males (n = 99)	Females (n = 124)
Age, months	149 (8.4)	149 (8.8)	148 (8.0)
Weight (kg)	52 (17)	49 (15)	55 (19)
Height (cm)	156 (9.4)	155 (11)	156 (8.3)
Body mass index (kg/m2)	21 (5.6)	20 (4.8)	22 (6.0)
Tanner stage, n (%)
1	23 (10)	17 (17)	6 (5)
2	56 (25)	31 (31)	25 (20)
3	65 (29)	26 (26)	39 (32)
4	45 (20)	16 (16)	29 (24)
5	33 (15)	9 (9)	24 (20)
Missing	1	0	1
BMC (g)
Whole body, excluding head	1229 (325)	1174 (328)	1272 (318)
Total hip	25.2 (6.7)	25.6 (7.6)	24.8 (5.9)
Femoral neck	3.69 (0.78)	3.68 (0.77)	3.70 (0.80)
Distal radius	1.45 (0.28)	1.42 (0.27)	1.47 (0.28)
Spine	38.9 (11.8)	35.7 (11.1)	41.5 (11.8)
BMC z-scorea
Whole body, excluding head	−0.05 (0.82)	−0.16 (0.78)	0.04 (0.85)
Total hip	−0.08 (0.92)	−0.12 (0.94)	−0.06 (0.90)
Femoral neck	0.23 (0.97)	0.13 (0.93)	0.31 (0.99)
Distal radius	0.10 (0.98)	−0.02 (0.93)	0.19 (1.0)
Spine	0.09 (0.86)	0.06 (0.87)	0.12 (0.85)
aBMD (g/cm2)
Whole body, excluding head	0.826 (0.094)	0.809 (0.092)	0.841 (0.093)
Total hip	0.853 (0.137)	0.838 (0.130)	0.866 (0.142)
Femoral neck	0.798 (0.136)	0.782 (0.122)	0.810 (0.145)
Distal radius	0.603 (0.064)	0.592 (0.062)	0.612 (0.064)
Ultradistal radius	0.350 (0.056)	0.347 (0.047)	0.352 (0.063)
Spine BMAD (g/cm3)	0.228 (0.036)	0.209 (0.026)	0.244 (0.036)
aBMD z-scorea
Whole body, excluding head	−0.38 (0.90)	−0.43 (0.89)	−0.34 (0.92)
Total hip	−0.02 (1.1)	−0.06 (1.1)	0.02 (1.1)
Femoral neck	−0.02 (1.1)	−0.10 (1.1)	0.05 (1.0)
Distal radius	0.23 (0.96)	0.22 (0.92)	0.23 (1.0)
Ultradistal radius	0.28 (1.2)	0.18 (1.0)	0.36 (1.3)
Spine BMAD	0.33 (1.0)	0.32 (1.0)	0.34 (1.0)

Abbreviations: aBMD = areal bone mineral density; BMAD = bone mineral apparent density; cm = centimeter; g = grams; kg = kilograms; m = meter

^aHeight-for-age adjusted and age-, sex-, and race-standardized

3.2. Overall OPE Metabolite Regression Models

In the overall study sample, we observed significant differences by exposure timepoint in adjusted associations of BCEP and BDCIPP (but not DnBP or DPHP) with age 12 year aBMD at multiple skeletal sites (Table 3). Specifically, differences in associations across timepoints were significant for both metabolites with total hip and 1/3^rd distal radius aBMD, and additionally for BDCIPP with whole body aBMD (Table 3). Greater BCEP concentrations at 1–3 years and 5 years were generally associated with lower BMD z-scores at most sites, and strongest for total hip aBMD (Figure 1A and Table 3). An IQR increase in BCEP concentration at 1–3 years and 5 years was associated with a −0.15 (95% CI: −0.36, 0.06) and −0.17 (95% CI: −0.34, 0.01) lower total hip aBMD z-score, respectively (Figure 1A and Table 3). An IQR increase in BDCIPP concentration during gestation was associated with higher BMD z-scores at all sites and was strongest for the total hip (difference: 0.33; 95% CI: 0.01, 0.64) (Figure 1B and Table 3). Conversely, greater BDCIPP at 5 years was associated with lower total hip aBMD (difference: −0.18; 95% CI: −0.40, 0.05) (Figure 1B and Table 3). Associations with DnBP were null and did not significantly differ across timepoints (Figure 1C and Table 3). While there was no evidence of a statistical difference across timepoints, an IQR increase in DPHP at 8 years was associated with greater BMD z-scores at several skeletal sites (Figure 1D and Table 3) and was strongest for whole body aBMD (difference: 0.12; 95% CI: 0.01, 0.24) (Table 3).

Table 3.

Adjusted associations of an IQR increase in log₂-urinary organophosphate ester biomarker concentration with bone mineral density z-score^a measures at age 12 years, overall sample (n = 223)

Biomarker/timepoint	Sample size	Whole body (excluding head)	Total hip	Femoral neck	1/3rd Distal radius	Ultradistal radius	Spine BMAD
BCEP
Gestationb	223	0.00 (−0.15, 0.15)	−0.05 (−0.24, 0.13)	−0.09 (−0.27, 0.09)	0.13 (−0.05, 0.31)	0.00 (−0.21, 0.22)	−0.09 (−0.28, 0.09)
1–3 yearc	193	−0.06 (−0.24, 0.11)	−0.15 (−0.36, 0.06)	−0.01 (−0.20, 0.18)	−0.08 (−0.24, 0.07)	−0.01 (−0.21, 0.19)	0.09 (−0.10, 0.29)
5 year	157	−0.13 (−0.29, 0.03)	−0.17 (−0.34, 0.01)	−0.10 (−0.27, 0.07)	−0.11 (−0.30, 0.08)	−0.03 (−0.24, 0.17)	−0.12 (−0.32, 0.09)
8 year	184	0.04 (−0.10, 0.18)	0.11 (−0.05, 0.27)	0.10 (−0.05, 0.25)	0.05 (−0.08, 0.18)	0.00 (−0.16, 0.16)	−0.07 (−0.22, 0.08)
p-valued		0.4	0.06	0.2	0.1	1	0.4
BDCIPP
Gestationb	223	0.29 (0.02, 0.56)	0.33 (0.01, 0.64)	0.23 (−0.09, 0.56)	0.23 (0.01, 0.46)	0.10 (−0.24, 0.45)	0.30 (0.03, 0.58)
1–3 yearc	193	−0.03 (−0.23, 0.18)	−0.10 (−0.34, 0.14)	0.01 (−0.21, 0.23)	−0.10 (−0.32, 0.11)	0.01 (−0.22, 0.24)	0.07 (−0.20, 0.34)
5 year	155	−0.17 (−0.39, 0.05)	−0.18 (−0.40, 0.05)	−0.11 (−0.32, 0.10)	−0.07 (−0.31, 0.17)	0.00 (−0.25, 0.24)	0.03 (−0.18, 0.23)
8 year	187	0.09 (−0.10, 0.28)	0.14 (−0.09, 0.38)	0.12 (−0.12, 0.36)	0.17 (−0.01, 0.36)	0.07 (−0.21, 0.34)	0.03 (−0.21, 0.27)
p-valued		0.06	0.04	0.3	0.07	1	0.4
DnBP
Gestationb	223	0.08 (−0.12, 0.27)	0.09 (−0.16, 0.35)	0.08 (−0.16, 0.32)	0.08 (−0.11, 0.27)	0.02 (−0.21, 0.25)	0.04 (−0.19, 0.28)
3 yearc	122	−0.05 (−0.18, 0.09)	−0.04 (−0.22, 0.14)	−0.08 (−0.25, 0.10)	−0.01 (−0.15, 0.13)	−0.10 (−0.26, 0.07)	0.04 (−0.08, 0.17)
5 year	155	−0.05 (−0.22, 0.11)	−0.02 (−0.20, 0.16)	−0.01 (−0.18, 0.17)	−0.11 (−0.29, 0.07)	−0.06 (−0.26, 0.13)	0.01 (−0.17, 0.18)
8 year	187	0.05 (−0.09, 0.19)	0.02 (−0.14, 0.19)	0.03 (−0.14, 0.21)	−0.03 (−0.18, 0.11)	−0.11 (−0.28, 0.05)	0.05 (−0.11, 0.22)
p-valued		0.6	0.8	0.7	0.5	0.8	1
DPHP
Gestationb	223	0.01 (−0.16, 0.19)	−0.04 (−0.25, 0.17)	−0.05 (−0.25, 0.16)	0.06 (−0.16, 0.28)	0.00 (−0.21, 0.21)	−0.11 (−0.31, 0.09)
1–3 yearc	193	−0.03 (−0.24, 0.17)	−0.08 (−0.31, 0.15)	−0.01 (−0.23, 0.21)	−0.02 (−0.19, 0.16)	0.06 (−0.21, 0.32)	0.00 (−0.21, 0.22)
5 year	158	0.04 (−0.11, 0.18)	0.01 (−0.13, 0.15)	0.02 (−0.12, 0.17)	0.04 (−0.12, 0.20)	−0.01 (−0.20, 0.18)	0.00 (−0.15, 0.15)
8 year	187	0.12 (0.01, 0.24)	0.10 (−0.04, 0.24)	0.09 (−0.07, 0.25)	0.06 (−0.06, 0.19)	0.06 (−0.10, 0.21)	0.11 (−0.01, 0.24)
p-valueb		0.3	0.4	0.6	0.9	0.9	0.2

Abbreviations: BCEP = bis-2-chloroethyl phosphate; BDCIPP = bis-(1,3-dichloro-2-propyl) phosphate; DnBP = di-n-butyl phosphate; DPHP = diphenyl phosphate

Note: Regression models adjusted for maternal age at delivery, maternal race/ethnicity, household income (time-varying), BMI z-score (time-varying), gestational vitamin use, parity, average maternal blood lead concentration during pregnancy, duration of breastfeeding (except gestation model), child age and child sex

^aHeight-for-age adjusted and age-, sex-, and race-standardized

^bAverage of available log₂ transformed creatinine-standardized maternal OPE biomarker concentrations at 16- and 26- weeks of gestation

^cAverage of available log₂ transformed creatinine-standardized child OPE biomarker concentrations at age 1, 2, and 3 years. For DnBP, concentrations in 1- and 2-year samples are excluded because of the presence of analytical interferences

^dGlobal F-test of differences in effect estimates across timepoints

Figure 1.

Adjusted total hip aBMD z-score difference (95% CI) per IQR increase in creatinine-standardized OPE concentration for (A) BCEP, (B) BDCIPP, (C) DnBP, and (D) DPhP. X-axis key: G = gestation, 1–3Y = 1–3 year average, 5Y = 5 years, 8Y = 8 years; circle = overall, square = males, triangle = females. Regression models adjusted for maternal age at delivery, maternal race/ethnicity, household income (time-varying), BMI z-score (time-varying), gestational vitamin use, parity, average maternal blood lead concentration during pregnancy, child age and (for overall models) child sex. Abbreviations: BCEP = bis-2-chloroethyl phosphate, BDCIPP = bis-(1,3-dichloro-2-propyl) phosphate, aBMD = bone mineral density, CI = confidence interval, DnBP = di-n-butyl phosphate, DPHP = diphenyl phosphate, IQR = interquartile range, OPE = organophosphate ester.

We observed comparable patterns of associations with BMC, though they were frequently more attenuated (Supplemental Material Table S4). For both BMD and BMC, unadjusted associations were either comparable or greater in magnitude, and more precise, compared to adjusted associations (Supplemental Material Table S5 and Table S6).

In our sensitivity analyses, additional adjustment for diet-related variables yielded comparable associations to those estimated from our primary models (Supplemental Material Table S7 and Table S8), as were associations from models additionally adjusted for physical activity (Supplemental Material Table S9 and Table S10).

3.3. Sex-Stratified OPE Metabolite Regression Models

For both males and females, there were several notable differences in patterns and magnitudes of OPE associations with aBMD measured at age 12 years, across timepoints (Figure 1, Figure 2, Supplemental Material Table S11, and Supplemental Material Table S12). Among males, BCEP associations differed across timepoints for all sites except for the 1/3^rd distal- and ultradistal radius, and for BDCIPP at all sites (Supplemental Material Table S11). Higher BCEP and BDCIPP concentrations at 1–3 years and 5 years were associated with lower BMD z-scores at several skeletal sites (Supplemental Material Table S11). For example, an IQR increase in BCEP concentration at 1–3 years and at 5 years was associated with a −0.38 (95% CI: −0.74, −0.03) and −0.30 (95% CI: −0.62, 0.03) lower total hip aBMD z-score, respectively (Figure 1A). Similarly, an IQR increase in BDCIPP at 1–3 years and at 5 years was associated with a −0.28 (95% CI: −0.64, 0.07) and −0.47 (95% CI: −0.83, −0.11) lower total hip aBMD z-score, respectively (Figure 1B). In contrast, an IQR increase in BDCIPP during gestation was associated with higher BMD z-scores at most sites and was strongest for total hip aBMD (difference: 0.46; 95% CI: −0.04, 0.95) (Figure 1B and Supplemental Material Table S11).

Figure 2.

Adjusted spine BMAD z-score difference (95% CI) per IQR increase in creatinine-standardized OPE concentration for (A) BCEP, (B) BDCIPP, (C) DnBP, and (D) DPhP. X-axis key: G = gestation, 1–3Y = 1–3 year average, 5Y = 5 years, 8Y = 8 years; circle = overall, square = males, triangle = females. Regression models adjusted for maternal age at delivery, maternal race/ethnicity, household income (time-varying), BMI z-score (time-varying), gestational vitamin use, parity, average maternal blood lead concentration during pregnancy, child age and (for overall models) child sex. Abbreviations: BCEP = bis-2-chloroethyl phosphate, BDCIPP = bis-(1,3-dichloro-2-propyl) phosphate, BMAD = bone mineral apparent density, CI = confidence interval, DnBP = di-n-butyl phosphate, DPHP = diphenyl phosphate, IQR = interquartile range, OPE = organophosphate ester.

Among females, associations for DnBP with femoral neck and ultradistal radius aBMD differed across timepoints (Supplemental Material Table S12). Similarly, associations for DPHP differed across timepoints with most sites, except the 1/3^rd distal- and ultradistal radius (Supplemental Material Table S10). An IQR increase in DnBP at 3 years was associated with a −0.27 (95% CI: −0.47, −0.06) lower femoral neck aBMD z-score and a −0.37 (95% CI: −0.57, −0.16) lower ultradistal radius aBMD z-score (Supplemental Material Table S12). For most skeletal sites, magnitudes of DPHP associations ordinally increased across timepoints (Supplemental Material Table S12). As examples, an IQR increase in gestational DPHP concentration was associated with a −0.20 (95% CI: −0.40, 0.00) lower total hip aBMD z-score (Figure 1D) and −0.19 (95% CI: −0.41, 0.03) lower spine BMAD z-score (Figure 2D); however, at 8 years, an IQR increase in DPHP was associated with a 0.21 (95% CI: 0.05, 0.38) higher total hip aBMD z-score (Figure 1D) and 0.15 (95% CI: 0.00, 0.30) higher spine BMAD z-score (Figure 2D).

Patterns with BMC z-scores were similar to those we observed for aBMD for both males and females, though magnitudes of associations were generally attenuated (Supplemental Material Tables S13 and S14).

4. Discussion

In this prospective study of OPE exposures from gestation through age 8 years with measures of BMC and BMD at age 12 years, associations were skeletal site-, sex-, and exposure period-specific. We found that urinary BCEP and BDCIPP concentrations during gestation and at 8 years were generally positively associated with bone z-scores while concentrations at 1–3 and 5 years were inversely associated. These patterns were stronger among males than females. For DPHP, we observed strong sex and timepoint differences with gestational concentrations being inversely associated and later childhood concentrations being positively associated with bone z-scores, and these patterns were more pronounced among females. Associations of bone outcomes with DnBP were generally weaker than other OPEs. Across OPEs, magnitudes of association were generally strongest at the total hip while sex differences were most apparent at the spine.

In our study, OPE metabolite concentrations during pregnancy were generally comparable to those from other pregnancy studies with more contemporary samples (Kuiper et al., 2020; Romano et al., 2017). Similarly, concentrations measured during pregnancy in our study were comparable to those measured among adults (ages 20–59 years) in the 2013–2014 cycle of the Nutrition and Health Examination Survey (NHANES) (Ospina et al., 2018). Additionally, 5- and 8-year BCEP and DnBP concentrations in our study were comparable to those for participants 6–11 years of age in the 2013–2014 cycle of NHANES, whereas concentrations for BDCIPP and DPHP were ~1.7x higher in our study (Ospina et al., 2018). In the only animal study to evaluate the relation of OPE exposure with BMD outcomes, Wistar rats were exposed to 1000 μg/day of Firemaster© 550 from gestational day 8 to postnatal day 21, and femurs were extracted at age six months for microCT and histomorphometric analyses (Macari et al., 2020). Male rats showed evidence of trabecular bone loss, reduced overall BMD, reduced osteoblast numbers, and increased yellow bone marrow (the portion of bone marrow predominately comprised of adipocytes) compared to controls (Macari et al., 2020). Though no differences in body weight at euthanasia were observed among males, exposed female rats had significantly greater body weight at euthanasia despite no differences in bone microarchitecture or composition (Macari et al., 2020). It is difficult to directly compare these findings to those in our study, given that Firemaster © 550 is a mixture of multiple OPEs and brominated chemicals (Phillips et al., 2016), rats were exposed up through postnatal day 21 (thus constituting cumulative exposure from gestation to early postnatal life), and differences in age at bone measures (i.e., post-pubescent adult rats vs. early adolescent children). However, in our study, BCEP at gestation was associated with lower spine BMAD (a skeletal site comprised primarily of trabecular bone) among males, whereas we observed comparable associations for DPHP with spine BMAD among females. Taken together, these findings suggest that some OPEs may have sexually dimorphic effects on formation of trabecular bone (i.e., the spongy, inner osseous) as opposed to cortical bone (i.e., hard, outer osseous), particularly when exposure occurs in utero.

The exact mechanisms by which OPEs may exert effects on bone homeostatic processes are not well established; however, endocrine disruption is one possibility given that bone homeostasis is a hormonally-regulated process and is particularly dependent on sex steroids (e.g., estrogen and testosterone) (Khosla et al., 2012; Mohamad et al., 2016; Snyder et al., 2017; Streicher et al., 2017) as well as thyroid hormones (Hill et al., 2018; Liu et al., 2019; Meeker et al., 2013; Meeker and Stapleton, 2010; Preston et al., 2017; Ren et al., 2016; Wang et al., 2020; Xu et al., 2015; Yao et al., 2021). This is evidenced by animal studies which have shown that greater gestational or early postnatal exposure to estrogen or estrogenic chemicals significantly inhibits osteoclasts (bone cells responsible for bone resorption), and is associated with higher peak bone mass in adulthood (Migliaccio et al., 1996; Migliaccio et al., 1995; Migliaccio et al., 2000). Further, the rate of bone formation and mineralization of in vitro murine bone cells treated with estrogen (in females) and testosterone (in males) was increased in a dose-dependent manner, while bone resorption was unaffected (Schwartz et al., 1991). Considering the endocrine disrupting potentials of OPEs, several in vitro studies have shown that TPHP (one of the parent compounds of DPHP), tris(2-chloroethyl) phosphate (TCEP, the primary parent compound of BCEP), tris(1,3-dichloro-2-propyl) phosphate (TDCPP, the parent compound of BDCIPP), and tris(butyl) phosphate (TnBP, the primary parent compound of DnBP) have anti-estrogenic (Kojima et al., 2013; Krivoshiev et al., 2016) as well as estrogenic properties (Krivoshiev et al., 2016; Liu et al., 2012), with TCEP, TDCPP, and TPHP also capable of acting as androgen antagonists (Kojima et al., 2013; Liu et al., 2012). Notably, the seemingly contradictory actions for some of these OPEs has been speculated to be from receptor-independent and -dependent mechanisms (Krivoshiev et al., 2016). In addition to sex steroids, it is recognized that thyroid hormones are critically important for adequate bone development and attainment of peak bone mass as the skeleton is a triiodothyronine (T3)-target organ and T3 is partially responsible for regulating endochondral ossification, long bone formation, and linear growth (Bassett and Williams, 2016; Williams, 2013). Further, human and animal studies have shown that OPEs can modulate thyroid hormones, providing another mechanism by which they may exert osteotoxic effects (Hill et al., 2018; Liu et al., 2019; Meeker et al., 2013; Meeker and Stapleton, 2010; Preston et al., 2017; Ren et al., 2016; Wang et al., 2020; Xu et al., 2015; Yao et al., 2021). As such, future studies should evaluate the potential mediating effects of thyroid hormones on associations of OPEs with bone health outcomes.

OPEs may also exhibit osteotoxicity by affecting BMSC differentiation. In a recent study, ex vivo treatment of mouse embryos (gestational day 13) with 1 μM of TPHP displayed decreased limb differentiation, at 3 μM displayed decreased longitudinal growth of limb buds as well as inhibited differentiation of hypertrophic chondrocytes and osteoblasts, and at 10 μM nearly all limb bud growth and differentiation was halted (Yan and Hales, 2019). Further, 10 μM of TPHP was associated with downregulation of Sox9 mRNA expression as well as Runx2 and Sp7 (Yan and Hales, 2019). Importantly, whereas Runx2 and Sp7 are regulators of osteogenesis, Sox9 is the master regulator of chondrogenesis, a process by which cartilage—the predominant form of the skeleton in early human development—is derived from mesenchymal tissue (Robert et al., 2020). While sex-specific effects were not evaluated in the above studies, in our study, DPHP (a primary metabolite of TPHP) at gestation was associated with lower BMD at several skeletal sites (total hip, femoral neck, and spine) among females, whereas we observed null associations among males. Related to BMSC differentiation, OPEs may disrupt osteogenesis via interaction with peroxisome proliferator-activated receptors (PPARs) (Belcher et al., 2014; Cano-Sancho et al., 2017; Pillai et al., 2014; Tung et al., 2017a; Tung et al., 2017b) or by altering expression of PPAR genes (den Broeder et al., 2017). Specifically, evidence suggests that agonistic activation of PPARγ directs BMSCs towards adipogenesis, suppressing osteoblast formation and consequently reducing BMD (Chen et al., 2016; Kawai, 2013; Moerman et al., 2004; Sadie-Van Gijsen et al., 2013; Tontonoz and Spiegelman, 2008; Zhuang et al., 2016). However, studies evaluating interactions of OPEs with PPARγ have been mixed, with some finding that TPHP and DPHP act as PPARγ agonists (Belcher et al., 2014; Cano-Sancho et al., 2017; Pillai et al., 2014; Tung et al., 2017a; Tung et al., 2017b), and others either finding no evidence of agonism (Kojima et al., 2013; Suzuki et al., 2013) or even evidence of antagonism (Young et al., 2021). Notably, of the studies finding PPARγ agonism, most only evaluated Firemaster© 550 and/or its components or component metabolites (including TPHP and DPHP) (Belcher et al., 2014; Cano-Sancho et al., 2017; Pillai et al., 2014; Tung et al., 2017a; Tung et al., 2017b).

We observed both positive and negative associations of OPEs with aBMD and BMC z-scores. Both higher and lower aBMD and BMC are potentially adverse for long-term bone health. Low peak bone mass accrual in adolescence is related to low aBMD in adulthood, the primary risk factor and clinical indicator of osteoporosis (Weaver et al., 2016). Alternatively, high aBMD in early adolescence may be due to early pubertal onset (Gilsanz et al., 2011), which can also have important implications for long-term health (Golub et al., 2008).

Our study has several notable strengths. First, we estimated effects of OPE metabolite concentrations at several timepoints across early life, including gestation, to inform the developmental periods that may be most susceptible to OPE osteotoxicity. Second, we evaluated associations at multiple skeletal sites, not just the whole body, which allows for comparisons of associations with sites that are comprised predominantly of cortical versus trabecular bone. While both are important for overall biomechanical function, it is useful to distinguish sites by bone composition since it could provide insight into mechanisms of chemical action as well as allow specificity when comparing to animal toxicological studies that measure both trabecular and cortical bone. Third, we adjusted estimates for several potential confounders, including duration of breastfeeding and maternal lead exposure, both of which have been associated with bone development (Foley et al., 2009; Jones et al., 2013; Min et al., 2008; Mølgaard et al., 2011). However, we did not have measures of physical activity until the 12-year visit, and therefore associations of OPE exposures with BMD and BMC outcomes may be residually confounded by a child’s physical activity, particularly for the 5- and 8-year visits. Lastly, we assessed the potential for sex-specific effects.

Still, this study is not without limitations. There was loss to follow-up over time resulting in a moderate sample size at each timepoint, limiting our power to detect associations, particularly in sex-stratified models. We also relied on OPE metabolite concentrations from single spot urine samples as a proxy for average OPE exposure. However, given the likely episodic nature of exposures and the short biological half-lives of OPEs (Ospina et al., 2018), a single urinary OPE measure at each timepoint may not represent typical exposures. This measurement error is likely to be independent and non-differential but may also result in bias. As such, additional studies are necessary to replicate findings. We also limited our analyses to OPE metabolites with the highest detection frequencies which may have excluded other important and possibly osteotoxic metabolites. Additionally, participants at the 12-year visit were at various stages of pubarche, with females in later stages of pubertal development than males. However, we used height-for-age adjusted, sex-, age-, and race-specific BMC and BMD z-scores to account for differences in BMC/BMD values that are due to age, bone size, and sexual maturation. Given our sample size and the differing directions of metabolite associations within skeletal sites, we did not evaluate mixture effects. Relatedly, given that other epidemiologic studies have shown relations of other chemical classes (e.g., per- and polyfluoroalkyl substances (Buckley et al., 2021; Cluett et al., 2019; Jeddy et al., 2018), phthalates (Kuiper et al., 2022; van Zwol-Janssens et al., 2020), phenols (van Zwol-Janssens et al., 2020)) with BMD, future, well-powered studies should evaluate chemical mixtures that include multiple chemical classes. Further, while we focused on general trends and patterns of associations, we conducted many statistical tests given the number of OPE metabolites, skeletal sites, and sex-stratification.

In this cohort, we found evidence of sex- and exposure period-specific associations of certain OPE exposures during pregnancy and early childhood with early adolescent bone mineral accrual and strength. Future studies assessing clinical outcomes, such as fracture and osteoporosis, can help contextualize the potential long-term health implications of early childhood OPE exposures.

Abbreviations

aBMD

areal bone mineral density

BMAD

bone mineral apparent density

BMC

bone mineral content

BMD

bone mineral density

BMSC

bone marrow multipotent mesenchymal stem cells

BCEP

bis-2-chloroethyl phosphate

BDCIPP

bis-(1,3-dichloro-2-propyl)

CI

confidence interval

cm

centimeter

dL

deciliter

DnBP

di-n-butyl phosphate

DPHP

diphenyl phosphate

g

grams

IQR

interquartile range

kg

kilogram

m

meter

OPE

organophosphate ester

PPAR

peroxisome proliferator-activated receptor

T3

triiodothyronine

TCEP

tris(2-chloroethyl) phosphate

TDCIPP

tris(1,3-dichloro-2-propyl) phosphate

TnBP

tri(n-butyl) phosphate

TPHP

triphenyl phosphate

μg

microgram

μL

microliter

μM

micromolar

6. Data sharing

Data from the HOME Study is available upon request. The HOME Study principal investigators have actively engaged in collaborative data-sharing projects. We welcome new collaborations with other investigators. Investigators interested in HOME Study data can explore options at the following location: https://homestudy.research.cchmc.org/ and use the available link to contact the investigators to discuss collaborative opportunities. The Data Sharing Committee meets regularly to review proposed research projects and ensure they do not overlap with extant projects and are an efficient use of scarce resources (e.g., biological samples). Funds to support staff efforts in the assembly of data sets are required.