Associations represent the odds ratios (ORs) and 95% CIs of preterm birth per interquartile range increase in mean pregnancy urinary phthalate metabolite concentration in the Pooled Phthalate and Preterm Birth Study (N = 6045). The interquartile range (ng/mL) of each metabolite is as follows: monoethyl phthalate (MEP), 168.2; mono-n-butyl phthalate (MBP), 21.4; mono-isobutyl phthalate (MiBP), 8.6; monobenzyl phthalate (MBzP), 11.0; mono(2-ethylhexyl) phthalate (MEHP), 5.0; mono(2-ethyl-5-hydroxyhexyl) phthalate (MEHHP), 17.3; mono(2-ethyl-5-carboxypentyl) phthalate (MECPP), 26.8; mono(2-ethyl-5-oxohexyl) phthalate (MEOHP), 12.4; mono(3-carboxypropyl) phthalate (MCPP), 2.5; monocarboxy-isooctyl phthalate (MCOP), 18.5; and monocarboxy-isononyl phthalate (MCNP), 2.2 (eTable 7 in the Supplement). Single metabolite results were estimated by multivariable logistic regression models and mixture results were produced by quantile g-computation models. Unadjusted models adjusted for study as a fixed effect. Adjusted models were adjusted for study, maternal age, race and ethnicity, education, and prepregnancy body mass index. Missing covariate values were multiply imputed for all models. The metabolites MCOP and MCNP were excluded from the mixtures analysis owing to limited sample size across cohorts.
Estimates represent the difference in mean probability of preterm birth following a series of hypothetical interventions to proportionally reduce concentrations of 9 phthalate metabolites in the pooled study population (n = 5471), including monoethyl phthalate, mono-n-butyl phthalate, mono-isobutyl phthalate, monobenzyl phthalate, mono(2-ethylhexyl) phthalate, mono(2-ethyl-5-carboxypentyl) phthalate, mono(2-ethyl-5-oxohexyl) phthalate, mono(2-ethyl-5-hydroxyhexyl) phthalate, and mono(3-carboxypropyl) phthalate. G-computation was implemented to estimate probabilities from a multivariable logistic regression model, which adjusted for study, maternal age, race and ethnicity, education, and prepregnancy body mass index. Differences were multiplied by 1000 to estimate the rate per 1000 live births. The 95% CIs were estimated using quantiles of the nonparametric bootstrap distribution across 2000 iterations. Estimations were performed on a single randomly chosen imputed data set.
eFigure 1. Flow diagram of study participant selection and exclusion in the Pooled Phthalate and Preterm Birth Study
eFigure 2. Spearman correlations between pregnancy-averaged concentrations of urinary phthalate metabolites
eFigure 3. Distributions of pregnancy-averaged phthalate metabolite concentrations (a-k) in the Pooled Phthalate and Preterm Birth Study (overall) and by study
eFigure 4. Comparison of main effects (odds ratios) when excluding individual studies
eTable 1. Additional study design elements of cohorts included in the Pooled Phthalate and Preterm Birth Study population
eTable 2. Description of participant exclusions and final sample size in the Pooled Phthalate and Preterm Birth Study population
eTable 3. Participant characteristics (n [%] or mean [SD]) by study (a-p)
eTable 4. Urinary metabolites of phthalate and phthalate alternative compounds measured in the Pooled Phthalate and Preterm Birth study
eTable 5. Limits of detection (LOD) for phthalate metabolites and distribution of samples with concentrations above and below LOD
eTable 6. Sample size for each urinary phthalate metabolite across studies
eTable 7. Distribution of pregnancy-averaged urinary phthalate metabolite concentrations (ng/mL)
eTable 8. Heterogeneity by study in main effects using fixed effect, random effect, and interaction models
eTable 9. Effect estimates and Wald tests for tests of heterogeneity in confounding by study
eTable 10. Comparison of odds ratio (OR) estimates for preterm birth with additional adjustment for year of delivery, maternal smoking, and parity
eTable 11. Odds ratio (OR) for preterm birth in the overall study population and stratified by fetal sex
eTable 12. Urinary phthalate metabolite specified using non-linear term
eTable 13. Estimated change (β) in length of gestation (weeks) per IQR increase in urinary phthalate biomarkers
Customize your JAMA Network experience by selecting one or more topics from the list below.
Welch BM, Keil AP, Buckley JP, et al. Associations Between Prenatal Urinary Biomarkers of Phthalate Exposure and Preterm Birth: A Pooled Study of 16 US Cohorts. JAMA Pediatr. 2022;176(9):895–905. doi:10.1001/jamapediatrics.2022.2252
Is phthalate exposure during pregnancy associated with preterm birth?
In this pooled analysis of 16 studies in the US including 6045 pregnant individuals, phthalate metabolites were quantified in urine samples collected during pregnancy. Higher urinary metabolite concentrations for several prevalent phthalates were associated with greater odds of delivering preterm, and hypothetical interventions to reduce phthalate exposure levels were associated with fewer preterm births.
In this large observational study, urinary biomarkers of common phthalates used in consumer products were a risk factor for preterm birth.
Phthalate exposure is widespread among pregnant women and may be a risk factor for preterm birth.
To investigate the prospective association between urinary biomarkers of phthalates in pregnancy and preterm birth among individuals living in the US.
Design, Setting, and Participants
Individual-level data were pooled from 16 preconception and pregnancy studies conducted in the US. Pregnant individuals who delivered between 1983 and 2018 and provided 1 or more urine samples during pregnancy were included.
Urinary phthalate metabolites were quantified as biomarkers of phthalate exposure. Concentrations of 11 phthalate metabolites were standardized for urine dilution and mean repeated measurements across pregnancy were calculated.
Main Outcomes and Measures
Logistic regression models were used to examine the association between each phthalate metabolite with the odds of preterm birth, defined as less than 37 weeks of gestation at delivery (n = 539). Models pooled data using fixed effects and adjusted for maternal age, race and ethnicity, education, and prepregnancy body mass index. The association between the overall mixture of phthalate metabolites and preterm birth was also examined with logistic regression. G-computation, which requires certain assumptions to be considered causal, was used to estimate the association with hypothetical interventions to reduce the mixture concentrations on preterm birth.
The final analytic sample included 6045 participants (mean [SD] age, 29.1 [6.1] years). Overall, 802 individuals (13.3%) were Black, 2323 (38.4%) were Hispanic/Latina, 2576 (42.6%) were White, and 328 (5.4%) had other race and ethnicity (including American Indian/Alaskan Native, Native Hawaiian, >1 racial identity, or reported as other). Most phthalate metabolites were detected in more than 96% of participants. Higher odds of preterm birth, ranging from 12% to 16%, were observed in association with an interquartile range increase in urinary concentrations of mono-n-butyl phthalate (odds ratio [OR], 1.12 [95% CI, 0.98-1.27]), mono-isobutyl phthalate (OR, 1.16 [95% CI, 1.00-1.34]), mono(2-ethyl-5-carboxypentyl) phthalate (OR, 1.16 [95% CI, 1.00-1.34]), and mono(3-carboxypropyl) phthalate (OR, 1.14 [95% CI, 1.01-1.29]). Among approximately 90 preterm births per 1000 live births in this study population, hypothetical interventions to reduce the mixture of phthalate metabolite levels by 10%, 30%, and 50% were estimated to prevent 1.8 (95% CI, 0.5-3.1), 5.9 (95% CI, 1.7-9.9), and 11.1 (95% CI, 3.6-18.3) preterm births, respectively.
Conclusions and Relevance
Results from this large US study population suggest that phthalate exposure during pregnancy may be a preventable risk factor for preterm delivery.
Preterm birth is a leading cause of neonatal mortality and morbidity.1 The societal burden of preterm birth is particularly high in the US,2 with approximately 10% of pregnancies delivered preterm annually.3 While the underlying risk factors for most preterm births are unknown, exposure to environmental chemicals like phthalates may play a role.
Phthalates are synthetic chemicals used in everyday consumer products such as personal care items and food processing or packaging.4 Exposure can occur through many sources, including household dust, diet, and personal care products like cosmetics.5 Consequently, phthalate exposure is ubiquitous among pregnant individuals.6,7 Human and animal studies suggest that prenatal phthalate exposure is associated with adverse effects on children’s neurodevelopment and male reproductive tract development.8,9 While several studies have found positive associations between prenatal biomarkers of phthalate exposure and preterm birth,10-16 others have shown null17-20 or inverse21-23 associations. This may be partly due to the limited number of preterm births included, differences in exposure assessment methods, and variation in the baseline risk of preterm birth and phthalate exposure.
The purpose of this analysis was to pool individual-level data from 16 prospective studies conducted in the US11,12,14,17,21-32 and examine associations between prenatal urinary biomarkers of phthalate exposure and preterm birth. We also considered the potential influence of exposure to an overall phthalate mixture and evaluated how hypothetical interventions to reduce this exposure could impact preterm birth.
In May 2019, we systematically reviewed the literature to identify epidemiologic studies conducted in the US with data on urinary phthalate metabolites quantified during pregnancy and gestational age at delivery (eMethods 1 and 2 in the Supplement). We focused on US studies to facilitate generalizability of results to the US general population, which experiences relatively high levels of phthalate exposure33 and high rates of preterm birth.34 Of 21 unique studies, 17 had sufficient sample size (N > 50) and 16 corresponding authors agreed to collaborate (eFigure 1 in the Supplement). Participating studies received ethics approval from the institutional review board or human research ethics committees from their respective institutions. Participants provided written or verbal informed consent. Analysis of anonymized data sets sent to the National Institute of Environmental Health Sciences was deemed to not be human subjects research by the National Institute of Environmental Health Sciences institutional review board. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.
Study acronyms and design characteristics are provided in Table 1, and eligibility criteria are described in eTable 1 in the Supplement. All studies prospectively enrolled participants during prepregnancy (North Carolina Early Pregnancy Study [EPS]14 and Environment and Reproductive Health Study [EARTH]27) or pregnancy and all participants had live births between 1983 and 2018. The only case-control study included was LIFECODES,11 a study of preterm birth nested within a prospective cohort. Studies provided gestational age at delivery (defined by last menstrual period, early pregnancy ultrasonography, date of conception in pregnancies using assisted reproductive technologies, or some combination thereof). We defined preterm birth as delivery prior to 37 weeks’ gestation. Our final analytic sample included 6045 participants after excluding 1136 of 7181 participants in the total pooled sample (eFigure 1 and eTable 2 in the Supplement).
Participants provided urine samples during pregnancy for quantification of phthalate monoester metabolites. Urinary phthalate metabolites are the preferred biomarker of phthalate exposure35 and are highly stable in urine samples stored at ≤20 °C, as they were for all cohorts.36,37 All studies collected spot urine samples, except for EPS14 and Markers of Autism Risk in Babies-Learning Early Sign (MARBLES)29 that pooled multiple samples prior to measurement (eTable 1 in the Supplement). Phthalate metabolite measurements were performed separately by cohort. Most studies measured at the US Centers for Disease Control and Prevention (CDC) or using CDC-developed methods, and targeted the same metabolites as the CDC biomonitoring program. Briefly, after enzymatic hydrolysis of phthalate metabolite conjugates, phthalate metabolites were extracted from urine using online solid phase extraction, separated by high-performance liquid chromatography, and detected by isotope dilution tandem mass spectrometry. The analysis of deidentified specimens at the CDC was determined not to constitute engagement in human subjects research. We included 11 metabolites based on availability in at least 50% of participants (eTable 4 in the Supplement): monoethyl phthalate, mono-n-butyl phthalate (MBP), mono-isobutyl phthalate, monobenzyl phthalate, mono(2-ethylhexyl) phthalate, mono(2-ethyl-5-hydroxyhexyl) phthalate, mono(2-ethyl-5-carboxypentyl) phthalate (MECPP), mono(2-ethyl-5-oxohexyl) phthalate, mono(3-carboxypropyl) phthalate (MCPP), monocarboxy-isooctyl phthalate, and monocarboxy-isononyl phthalate.
Using multiple imputation by chained equations, we simultaneously imputed (1) phthalate biomarker concentrations below the limit of detection without instrument-read values (eMethods 3 in the Supplement) and (2) missing covariates (eTable 5 in the Supplement). We performed all subsequent analyses on the imputed data sets and pooled results using Rubin’s rules.38 Studies measured urinary specific gravity or creatinine to account for urine dilution (eTable 1 in the Supplement). We used covariate-adjusted standardization to correct phthalate metabolite concentrations for urine dilution (eMethods 4 in the Supplement).39,40 Most studies (9 of 16) quantified phthalate metabolites in multiple (range, 2-10) urine samples (eTable 1 in the Supplement). After dilution standardization, we calculated the within-participant geometric mean of phthalate metabolite concentrations across pregnancy. Subsequently, we natural-log–transformed concentrations and standardized concentrations by dividing by the interquartile range (IQR) to facilitate interpretability.
We used multivariable logistic regression to examine associations of mean pregnancy phthalate metabolites with odds of preterm birth. Odds ratios and 95% CIs were interpreted as the change in log-odds of preterm birth per 1-IQR increase in mean phthalate metabolite concentration. Crude models adjusted for study (via fixed effects for each study) and adjusted models included additional covariates that were measured across all 16 studies. We selected primary confounders a priori from the literature, including self-reported maternal race and ethnicity (categorical),18,41,42 education (categorical),12,17,18,28,41 maternal age at enrollment (years),12,18,28,41 and prepregnancy body mass index.17,18,28,41 Race and ethnicity was used as a confounder based on the consistent disparities in preterm birth43 and environmental exposures41 experienced by minoritized racial and ethnic populations in the US, which is driven by social determinants including racism and discrimination.44 We defined race and ethnicity by combining several self-identified categories to maximize sample size and consistency across pooled studies, including non-Hispanic Black, Hispanic/Latina, non-Hispanic White, and other (including American Indian/Alaskan Native, Native Hawaiian, >1 racial identity, or reported as other).
We used 2 complementary methods, quantile g-computation and standard g-computation, to examine the association of an overall mixture of phthalate metabolites and preterm birth. The mixture included all metabolites except monocarboxy-isooctyl phthalate and monocarboxy-isononyl phthalate, which were excluded a priori because fewer participants (n = 3758) and studies (10 total) quantified these biomarkers. This provided 5471 participants (14 studies) for the mixture analyses (eTable 6 in the Supplement). We used quantile g-computation to examine the odds of preterm birth per IQR increase in all phthalate metabolites in the mixture.45 We used standard g-computation to estimate the probability of preterm birth following several hypothetical interventions to reduce concentrations of the phthalate metabolite mixture,46 which provides potentially more interpretable results than model coefficients.47,48 Hypothetical interventions reduced each metabolite in the mixture by 10% to 90% in 10% increments. The 95% CIs were estimated using nonparametric bootstrapping (2.5th and 97.5th percentiles across 2000 iterations).46 We transformed results to be interpreted as the estimated number of preterm births prevented per 1000 live births by contrasting each hypothetical intervention with no intervention.
We conducted several sensitivity analyses. (1) To assess heterogeneity in effect estimates by study, we qualitatively compared estimates from fixed-effect models to mixed models in which we specified study indicator as a random intercept49; used Wald tests of goodness of fit for an interaction term between study and metabolite in the primary model49; and examined differences in effect estimates after we fit models that drop participants from single cohorts. This leave-1-out analysis provides a way to examine how overall results may have been influenced by individual cohorts. (2) We used Wald tests to assess potential differences in confounding across studies by fitting a series of models that additionally included interaction terms between study and each of the following covariates: maternal age, prepregnancy body mass index, race and ethnicity, and education. (3) We fit models additionally adjusted for precision variables associated with phthalate exposure or preterm delivery, including delivery year, smoking, or parity. (4) We assessed potential effect measure modification by fetal sex using model stratification and a nonstratified model with an interaction term between phthalate metabolite and sex.24 (5) We examined nonlinearity in associations by fitting quadratic terms. (6) We examined metabolite associations with gestational age at delivery (continuous) using multivariable linear regression using the same covariates but applied inverse probability of sampling weights to account for the LIFECODES study design.50 We chose not to conduct sensitivity analyses for other pregnancy complications (eg, preeclampsia) because evidence suggests such conditions are potentially on the causal pathway between phthalate exposure and preterm birth.51-53 We considered Wald tests or interactions statistically significant if 2-sided P values were less than .05. We performed analyses using R version 4.0.3 (R Foundation).
The overall study population consisted of 6045 pregnant individuals (mean [SD] age, 29.1 [6.1] years), of whom 539 (9%) delivered preterm (eFigure 1 in the Supplement). Overall participant characteristics are presented in Table 2 and characteristics by study are shown in eTable 3 in the Supplement. A total of 802 individuals (13.3%) were Black, 2323 (38.4%) were Hispanic/Latina, 2576 (42.6%) were White, and 328 (5.4%) had other race and ethnicity (including American Indian/Alaskan Native, Native Hawaiian, >1 racial identity, or reported as other). Participant characteristics were similar between individuals who delivered term vs preterm (Table 2). Concentrations of urinary phthalate metabolites included for analysis were detectable in 96% or more of urine samples, except for mono(2-ethylhexyl) phthalate (83%) and MCPP (90%) (eTable 5 in the Supplement) and were highest for monoethyl phthalate, MBP, and MECPP (eTable 7 in the Supplement). Correlations were highest between metabolites with shared parent chemicals (eFigure 2 in the Supplement). Overall, there was substantial overlap in the distributions of phthalate metabolite concentrations across studies (eFigure 3 in the Supplement). However, concentrations for several metabolites (eg, monobenzyl phthalate, MCPP) were higher for EPS,14 which was the only study to collect samples in the 1980s.
Regression analyses showed that higher concentrations of most phthalate metabolites were associated with slightly higher odds of preterm birth (Figure 1). After covariate adjustment, there was a 12% to 16% higher odds of preterm birth associated with an IQR increase in urinary concentrations of MBP (OR, 1.12 [95% CI, 0.98-1.27]), mono-isobutyl phthalate (OR, 1.16 [95% CI, 1.00-1.34]), MECPP (OR, 1.16 [95% CI, 1.00-1.34]), and MCPP (OR, 1.14 [95% CI, 1.01-1.29]). Other phthalate metabolites also displayed positive but nonsignificant associations. An IQR increase in the mixture of 9 phthalate metabolites was associated with 25% higher odds of preterm birth (OR, 1.25 [95% CI, 0.88-1.77]), although the confidence interval included the null. Based on results from g-computation, hypothetical interventions to reduce the phthalate metabolite mixture were estimated to prevent a mean of 2 to 32 preterm births per 1000 live births (Figure 2). For example, reducing the mixture of phthalate metabolite concentrations by 10%, 30%, or 50% was estimated to prevent 1.8 (95% CI, 0.5-3.1), 5.9 (95% CI, 1.7-9.9), and 11.1 (95% CI, 3.6-18.3) preterm births per 1000 live births, respectively.
Fixed-effects and random-effects models produced nearly equivalent estimates and metabolite by study interactions were not statistically significant (eTable 8 in the Supplement), indicating minimal heterogeneity by study. Magnitudes of associations were similar after excluding participants from individual study populations (eFigure 4 in the Supplement). However, associations were attenuated for MBP, MECPP, and MCPP after exclusion of LIFECODES participants.11 Heterogeneity in confounding was not detected (eTable 9 in the Supplement). We did not observe differences in associations when models were additionally adjusted for precision variables (delivery year, smoking, or parity) (eTable 10 in the Supplement) or evidence of effect measure modification by fetal sex (eTable 11 in the Supplement). We did not find evidence of nonlinear associations (eTable 12 in the Supplement). Importantly, direction of associations was consistent when gestational age at delivery was evaluated continuously (eTable 13 in the Supplement).
In this pooled analysis of more than 6000 pregnancies from 16 prospective studies in the US, we observed that higher maternal pregnancy concentrations of several urinary phthalate metabolites, particularly MBP, mono-isobutyl phthalate, MECPP, and MCPP, were associated with higher odds of preterm birth. While ORs were seemingly small in magnitude, g-computation estimates suggested that joint reductions in phthalate metabolites could produce significant population-level reductions in preterm births. Our findings suggest that exposure to multiple phthalates is associated with an increased risk of preterm birth.
At the population-level, modest effect sizes can be important when exposures are widespread and the outcome is prevalent.54 The imprecision of our estimates, as reflected by our confidence intervals, may be related to inconsistencies of methods used across pooled studies. Several studies quantified phthalates using spot urine samples collected at single time points in different periods of pregnancy,17,21,22,26,30-32 and such isolated measures are not ideal estimators of long-term exposure to be attributable to short half-life.55 Further, we did not have the data to subdivide preterm births into those that were spontaneous vs indicated, which may be important for assessing risk.11,12
Our results are important to consider in the context of the literature. As in our study, urinary metabolites of di-n-butyl phthalate, di-isobutyl phthalate, and di(2-ethylhexyl) phthalate have been associated with reduced gestational age at delivery or increased likelihood of preterm birth in several prospective US studies included here11-14,26,31,32 as well as studies from China16 and Mexico.15 Although null17-20,56 or contradictory14,21-23 associations have also been observed, associations between metabolites of these parent chemicals and preterm birth appear to be more consistent than other phthalate metabolites. Variation across studies with respect to magnitudes of association and statistical significance is expected owing to differences in (1) sample size and preterm birth prevalence, (2) metabolite distributions, (3) exposure assessment approaches, (4) gestational age at exposure assessment, and (5) geographic location, where some populations may have different underlying susceptibilities or patterns of exposure.33,34 While pooling data cannot address all systematic biases, our study directly addressed several limitations by achieving larger sample size and examining associations across wide distributions of phthalate biomarkers.
The mechanistic pathway between phthalate exposure and preterm birth is unclear, but several lines of evidence provide biologic plausibility for a relationship. Associations of phthalate metabolites with preterm birth may be mediated by oxidative stress and inflammation at the maternal-fetal interface.57,58 Additional mechanisms may include dysregulated trophoblast differentiation and endocrine disruption, as phthalate biomarkers have been associated with downregulated expression of placental genes responsible for these processes.59
Our findings provide additional evidence of the need to reduce phthalate exposures among pregnant individuals, which could take the form of behavioral interventions or regulations. Although phthalate exposure can occur through many sources and environments,4,7,60,61 there has been a long-standing scientific effort to accurately determine whether a single source drives the majority of human exposure.62 The US Consumer Product Safety Commission attempted to estimate exposure by source and found food and medications, not children’s toys, were the primary sources of exposure.63 Unfortunately, there is still substantial uncertainty in the primary source of exposure. In the US, phthalate exposure varies widely by sociodemographic factors,64 including whether a person is pregnant,65 at a disadvantaged socioeconomic status,64,66 or is of a particular marginalized race or ethnicity.66
Targeted interventions may help modify consumer behaviors that lead to phthalate exposures, such as altering the type of personal care products purchased.67,68 However, behavioral approaches are difficult to implement on a population scale because of the vast number of available consumer products containing phthalates and the limited ability of US consumers to access accurate ingredient lists.69 For example, the US Food and Drug Administration does not require phthalates to be listed as ingredients when designated as part of the fragrance. Alternatively, interventions to reduce exposures through diet have had mixed results.68,70 Compounding these difficulties, economic disparities may make access to phthalate-free products and diet more difficult for certain populations.28,41 Past public health efforts have successfully led to federally mandated restrictions on the use of certain phthalates in consumer products intended for children,4,71 but few restrictions exist for products intended for people who are pregnant. The US Food and Drug Administration also has the power to regulate phthalates in food, but 28 phthalates are currently allowed as food additives or in food contact materials.72 Given this reality, Project TENDR (Targeting Environmental Neuro-Development Risks) recommends a multipronged approach to reducing human exposure to multiple phthalates, including regulations at the federal and state levels, as well as voluntary action on the part of retailers and manufacturers.8
Our analysis of hypothetical interventions to reduce exposure to the phthalate mixture, regardless of whether reductions occur via behavioral or regulatory mechanisms, helps to highlight the potential magnitude of effect that population-level phthalate exposure may have on preterm birth, meanwhile addressing the fact that realistic interventions will change exposure to multiple phthalates simultaneously, rather than one at a time. Based on the rate of about 90 preterm births per 1000 live births birth in the pooled study population, hypothetical interventions of 10% to 50% would correspond with an estimated mean of 2% to 12% reduction in preterm births. Given that most individuals are exposed to multiple phthalates, regulatory approaches to mitigate population-level health effects from phthalates would be most effective when considering phthalates as a class, rather than as individual chemicals.8 We took an approach used by previous studies48,73,74 and evaluated a range of possible decrements in exposure. This approach allowed us to evaluate whether any reductions, large or small, in phthalate exposure would be worth pursuing based on the potential to result in fewer preterm births in community settings. Our results are consistent with the hypothesis that modest, but potentially feasible, reductions in phthalate exposure could reduce rates of preterm birth. However, our results should be interpreted cautiously in light of the assumptions required for causality (eMethods 5 in the Supplement).48 Although g-computation is often used to facilitate causal inference,75 it is still a statistical model and thus we opt for associational rather than causal language. Regardless, “preterm births prevented” uses causal language because there is not useful associational language for this statistic.
Our study represents the largest prospective investigation of phthalate exposure in pregnancy and preterm birth, to date and to our knowledge, and includes individual-level data from almost all US studies that have quantified phthalate metabolites in pregnancy. Thus, we were not restricted to studies that only published on associations with preterm birth or gestational age at birth25,27,30 and avoided publication bias. Pooled participant characteristics (eg, exposure distributions, geographic locations, education, and race and ethnicities) were more diverse than any single prior study, which provided better representation of the US population. Further, our mixtures approach helped reflect the reality that pregnant individuals are exposed to a variety of phthalates in their environments, which should be a central consideration for any future policies intended to reduce phthalate exposures.8
Several limitations in our study are important to acknowledge. First, there was variation in exposure assessment methods across studies. This may have produced measurement error of metabolites, which could have contributed to observed exposure differences and could not be disentangled from true differences in exposure levels across the study populations. However, there was large overlap in distributions across studies, and we adjusted for known confounders. Although calculating mean values across multiple spot urine samples can improve characterization of exposure,76 single spot urine samples may provide lower accuracy.77 Second, ORs from our statistical approach will tend to overestimate risk ratios, which are arguably more interpretable. We selected a logistic model to ensure that the model predictions remain within logical bounds without placing constraints on the phthalate distribution, and we use g-computation to allow easier interpretation of results. Third, we were also unable to examine potentially important confounders, such as diet.78 Concentrations of certain phthalate biomarkers are higher in individuals who have diets high in ultraprocessed food, fast food, or meat and dairy.60,61,79 Because some parameterizations of poor diet that include these foods are also associated with increased risk of preterm birth,80 residual confounding may exist in our analysis. However, phthalate exposure can come from many dietary pathways,70 so the role of diet in this relationship is uncertain.
In this pooled analysis of 16 prospective US studies, higher concentrations of several urinary phthalate metabolites in pregnancy were associated with preterm birth. These findings highlight the need for public health and policy measures to reduce phthalate exposures among pregnant individuals.
Accepted for Publication: May 4, 2022.
Published Online: July 11, 2022. doi:10.1001/jamapediatrics.2022.2252
Corresponding Author: Kelly K. Ferguson, PhD, Epidemiology Branch, Division of Intramural Research, National Institute of Environmental Health Sciences, 111 TW Alexander Dr, Research Triangle Park, NC 27709 (email@example.com).
The Pooled Phthalate Exposure and Preterm Birth Study Group Authors: Akram N. Alshawabkeh, PhD; José F. Cordero, MD; John D. Meeker, ScD; Emily S. Barrett, PhD; Nicole R. Bush, PhD; Ruby H. N. Nguyen, PhD; Sheela Sathyanarayana, MD; Shanna H Swan, PhD; David E. Cantonwine, PhD; Thomas F. McElrath, MD, PhD; Jenny Aalborg, MPH; Dana Dabelea, MD, PhD; Anne P. Starling, PhD; Russ Hauser, MD, ScD; Carmen Messerlian, PhD; Yu Zhang, BA; Asa Bradman, PhD; Brenda Eskenazi, PhD; Kim G. Harley, PhD; Nina Holland, PhD; Michael S. Bloom, PhD; Roger B. Newman, MD; Abby G. Wenzel, PhD; Joseph M. Braun, PhD; Bruce P. Lanphear, MD; Kimberly Yolton, PhD; Pam Factor-Litvak, PhD; Julie B. Herbstman, PhD; Virginia A. Rauh, ScD; Erma Z. Drobnis, PhD; Amy E. Sparks, PhD; J. Bruce Redmon, MD; Christina Wang, MD; Alexandra M. Binder, ScD; Karin B. Michels, ScD, PhD; Donna D. Baird, PhD; Anne Marie Z. Jukic, PhD; Clarice R. Weinberg, PhD; Allen J. Wilcox, MD, PhD; David Q. Rich, ScD; Barry Weinberger, MD; Vasantha Padmanabhan, PhD; Deborah J. Watkins, PhD; Irva Hertz-Picciotto, PhD; Rebecca J. Schmidt, PhD.
Affiliations of The Pooled Phthalate Exposure and Preterm Birth Study Group Authors: National Institute of Environmental Health Sciences, Research Triangle Park, Durham, North Carolina (Baird, Jukic, Weinberg, Wilcox); University of North Carolina at Chapel Hill, Chapel Hill (Starling); Harvard T.H. Chan School of Public Health, Boston, Massachusetts (Hauser, Messerlian, Zhang); Northeastern University, Boston, Massachusetts (Alshawabkeh); University of Georgia, Athens (Cordero); University of Michigan School of Public Health, Ann Arbor (Meeker, Watkins); Rutgers School of Public Health, Piscataway, New Jersey (Barrett); University of California, San Francisco, San Francisco (Bush); University of Minnesota School of Public Health, Minneapolis (Nguyen); University of Washington and Seattle Children's Research Institute, Seattle (Sathyanarayana); Icahn School of Medicine at Mount Sinai, New York, New York (Swan); Brigham and Women's Hospital, Boston, Massachusetts (Cantonwine, McElrath); University of Colorado Anschutz Medical Campus, Aurora (Aalborg, Dabelea); University of California, Merced, Merced (Bradman); University of California, Berkeley, Berkeley (Eskenazi, Harley, Holland); George Mason University, Fairfax, Virginia (Bloom); Medical University of South Carolina, Charleston (Newman, Wenzel); Brown University, Providence, Rhode Island (Braun); Simon Fraser University, Burnaby, British Columbia, Canada (Lanphear); Cincinnati Children's Hospital Medical Center, Cincinnati, Ohio (Yolton); Mailman School of Public Health, Columbia University, New York, New York (Factor-Litvak, Herbstman, Rauh); University of Missouri, Columbia, Columbia (Drobnis); University of Iowa, Iowa City (Sparks); University of Minnesota Medical School, Minneapolis (Redmon); The Lundquist Institute at Harbor, UCLA Medical Center, West Carson, California (Wang); University of Hawaii Cancer Center, Honolulu (Binder); University of California, Los Angeles, Los Angeles (Michels); University of Rochester Medical Center, Rochester, New York (Rich); Cohen Children's Medical Center of New York, Northwell Health, Queens (Weinberger); University of Michigan Medical School, Ann Arbor (Padmanabhan); University of California, Davis, Davis (Hertz-Picciotto, Schmidt).
Author Contributions: Drs Welch and Ferguson had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Welch, Keil, Engel, O'Brien, James-Todd, Ferguson, Meeker, Swan, McElrath, Factor-Litvak, Rauh, Hertz-Picciotto.
Acquisition, analysis, or interpretation of data: Welch, Buckley, Calafat, Christenbury, Engel, Rosen, James-Todd, Zota, Ferguson, Alshawabkeh, Cordero, Meeker, Barrett, Bush, Nguyen, Sathyanarayana, Swan, Cantonwine, McElrath, Aalborg, Dabelea, Starling, Hauser, Messerlian, Zhang, Bradman, Eskenazi, Harley, Holland, Bloom, Newman, Wenzel, Braun, Lanphear, Yolton, Factor-Litvak, Herbstman, Drobnis, Sparks, Redmon, Wang, Binder, Michels, Baird, Jukic, Weinberg, Wilcox, Rich, Weinberger, Padmanabhan, Watkins, Hertz-Picciotto, Schmidt.
Drafting of the manuscript: Welch, Engel, Rosen, James-Todd, Ferguson, Sathyanarayana, Swan, Weinberg.
Critical revision of the manuscript for important intellectual content: Welch, Keil, Buckley, Calafat, Christenbury, Engel, O'Brien, Rosen, James-Todd, Zota, Ferguson, Alshawabkeh, Cordero, Meeker, Barrett, Bush, Nguyen, Sathyanarayana, Swan, Cantonwine, McElrath, Aalborg, Dabelea, Starling, Hauser, Messerlian, Zhang, Bradman, Eskenazi, Harley, Holland, Bloom, Newman, Wenzel, Braun, Lanphear, Yolton, Factor-Litvak, Herbstman, Rauh, Drobnis, Sparks, Redmon, Wang, Binder, Michels, Baird, Jukic, Wilcox, Rich, Weinberger, Padmanabhan, Watkins, Hertz-Picciotto, Schmidt.
Statistical analysis: Welch, Keil, Engel, O'Brien, Rosen, James-Todd, Ferguson, Cantonwine, McElrath.
Obtained funding: Engel, Ferguson, Alshawabkeh, Cordero, Meeker, Barrett, Bush, McElrath, Dabelea, Hauser, Messerlian, Bradman, Holland, Braun, Factor-Litvak, Herbstman, Rauh, Padmanabhan, Schmidt.
Administrative, technical, or material support: Welch, Calafat, Christenbury, Alshawabkeh, Cordero, Meeker, Barrett, Bush, Nguyen, Sathyanarayana, McElrath, Aalborg, Messerlian, Zhang, Bradman, Yolton, Herbstman, Drobnis, Sparks, Wang, Binder, Michels, Baird, Wilcox, Weinberger, Watkins, Hertz-Picciotto, Schmidt.
Supervision: Keil, Calafat, Ferguson, Nguyen, Swan, McElrath, Dabelea, Newman, Yolton, Rauh, Redmon, Jukic, Schmidt.
Conflict of Interest Disclosures: Dr Engel reported grants from the National Institutes of Health (NIH)/National Institute of Environmental Health Sciences (NIEHS) and the US Environmental Protection Agency (EPA) during the conduct of the study; honorarium for grant review from the NIH/Center for Scientific Review outside the submitted work; and honorarium for advisory board participation from the University of Montana outside the submitted work. Dr Cordero reported grants from the NIH during the conduct of the study and outside the submitted work and from Medtronic Foundation outside the submitted work. Dr Barrett reported grants from NIH during the conduct of the study. Dr Bush reported grants from the NIH during the conduct of the study. Dr McElrath reported research support to their institution and equity from NxPrenatal Inc; serving on the scientific advisory board of and equity from Mirvie Inc; and serving on the scientific advisory board of and cash payment from Hoffmann-La Roche, Momenta Pharmaceuticals, Comanche Biopharma; and Tectonic Therapeutic. Dr Starling reported grants from the NIH during the conduct of the study. Dr Hauser reported grants from NIEHS during the conduct of the study. Dr Eskenazi reported grants from the NIH and EPA during the conduct of the study. Dr Harley reported grants from the NIEHS during the conduct of the study. Dr Holland reported grants from the NIEHS during the conduct of the study. Dr Bloom reported grants from the NIH during the conduct of the study. Dr Braun reported grants from the NIH during the conduct of the study and served as an expert witness for plaintiffs in litigation related to perfluoroalkyl substances–contaminated drinking water for Morgan & Morgan Law Firm (funds were not paid to Dr Braun directly; all compensation was paid to a discretionary account that cannot be used for salary or fringe) outside the submitted work. Dr Factor-Litvak reported grants from the NIH during the conduct of the study. Dr Jukic reported grants from the NIEHS during the conduct of the study. Dr Weinberg reported salary support from the NIEHS during the conduct of the study. Dr Weinberger reported grants from the NIH and the New Jersey Department of Environmental Protection during the conduct of the study. Dr Watkins reported grants from the NIH and EPA during the conduct of the study. Dr Schmidt reported grants from Autism Science Foundation during the conduct of the study. No other disclosures were reported.
Funding/Support: This research was supported in part by the Intramural Research Program of the National Institutes of Health (NIH)/National Institute of Environmental Health Sciences (NIEHS). The project was also supported by the NIEH (grants P42ES017198 to Dr Alshawabkeh, P30ES005022 to Dr Barrett, R21ES031231 to Dr Bloom, P01ES009605 and R01ES021369 to Drs Bradman, Eskenazi, Harley, and Holland, R01ES024381 to Dr Braun, R01ES030078 to Dr Buckley, R01ES016863 to Dr Bush, P42ES017198 to Dr Cordero, R01ES022934 to Dr Dabelea, P30ES010126 and P01ES09584 to Dr Engel, R01ES013543, R01ES014393, and R01ES08977 to Dr Factor-Litvak, R01ES009718 to Dr Hauser, ES013543 to Dr Herbstman, P30ES023513 to Dr Hertz-Picciotto, Z01ES103333 to Dr Jukic, R01ES031591 and P42ES017198 to Dr Meeker, R01ES031657 to Dr Messerlian, P01ES022844 and R01ES017500 to Dr Padmanabhan, T32ES007018 to Ms Rosen, R01ES0125169-01 to Dr Sathyanarayana, R21ES025551 and R24ES028533 to Dr Schmidt, R01ES016863-04 and R01ES016863-02S4 to Dr Swan, P30ES005022 to Dr Weinberger, P01ES011261 to Dr Yolton), NIH (grants UH3OD023251 to Dr Alshawabkeh, UH3OD023365 to Dr Hertz-Picciotto, P30ES005022 to Dr Rich, UH3OD023342 to Dr Schmidt), US Environmental Protection Agency (grants R82670901 and R827039 to Dr Engel), National Institute of Diabetes and Digestive and Kidney Diseases (grant R01DK076648 to Dr Dabelea), National Cancer Institute (grant R21CA128382 to Dr Michels), National Center for Advancing Translational Sciences (grant UL1TR001881 to Dr Wang), and Eunice Kennedy Shriver National Institute of Child Health and Human Development (grant R21HD058019 to Dr Weinberger).
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The findings and conclusions of this report are those of the authors and do not necessarily represent the official position of the US Centers for Disease Control and Prevention (CDC). Use of trade names is for identification only and does not imply endorsement by the CDC, the Public Health Service, or the US Department of Health and Human Services.
Additional Contributions: We thank Sharon Soucek, PhD (National Institute of Environmental Health Sciences), for oversight of data transfer agreements required for pooling data and Elena Colicino, PhD (Icahn School of Medicine at Mount Sinai), for her feedback on the analysis plan. These individuals were not compensated.