Urinary BPA Levels During Pregnancy and Preterm Birth Risk
Urinary BPA Levels During Pregnancy and Preterm Birth Risk
Study demographic characteristics have been previously described (Ferguson et al. 2014). Briefly, the study population was predominantly white (58.5%) and highly educated (85.9% had some college-level education) and were nonsmokers (92.3%). Approximately 10% of the women used some form of ART, and 47.3% of the population was either overweight (25 to ≤ 30 kg/m BMI) or obese (> 30 kg/m BMI). These population characteristics did not differ significantly between PTB cases and controls (Ferguson et al. 2014). Overall, BPA concentrations were detected in 1,350 (79.6%) of the samples analyzed, with the breakdown by visit and case–control status detailed in Table 1. There were 327 (92.9%) controls and 114 (87.7%) cases that had three or more repeated samples available for analysis during pregnancy. Visit 4 urine samples were collected on 66 (50.8%) cases and 314 (89.2%) controls. The mean (range) of the final gestational age for those cases who provided a visit 4 urine sample was 36.3 (35.0–36.9) weeks.
As expected, BPA was log-normally distributed and levels were ln-transformed for statistical analysis. SG-corrected urinary BPA geometric means and standard deviations by individual visit are presented in Table 1. Concentrations were relatively constant across gestation, with no significant BPA concentration differences at individual visits by Wilcoxon rank-sum test between PTB cases and controls. Urinary specific gravity was the highest early in pregnancy (SG = 1.017) and had a significantly decreased trend across pregnancy (p < 0.01). Spearman correlations between study visits for BPA were low and ranged from 0.17 to 0.26 (all p-values < 0.01). ICC for SG-corrected BPA indicated low temporal reliability (ICC = 0.21; 95% CI: 0.16, 0.27), and ICCs were slightly higher in cases of PTB compared with controls (Table 2). SG concentrations had low to moderate reproducibility across pregnancy (ICC = 0.38; 95% CI: 0.33, 0.43).
Crude and adjusted ORs (95% CIs) of overall preterm birth in association with averaged and cross-sectional (i.e., by visit) urinary BPA levels are presented in Table 3. Adjusted models controlled for maternal age, maternal race/ethnicity, educational level attained, medical insurance, parity, prior history of preterm birth, and BMI. A third set of models additionally included subject-specific averages for the ΣDEHP metabolites.
Generally, in both crude and adjusted models no significant relationships were observed between overall preterm birth and either averaged (OR = 1.21; 95% CI: 0.79, 1.85) or individual-visit urinary BPA levels, although, except for visit 3, ORs tended to be > 1.0 (Table 3). In our joint models using the random slopes and intercepts as predictors in the adjusted logistic regression models, we found no significant associations (see Supplemental Material, Table S1 http://ehp.niehs.nih.gov/wp-content/uploads/123/9/ehp.1408126.s001.acco.pdf). Fitting generalized additive mixed models indicated that the association between gestational age at sample collection and urinary BPA concentrations was linear, and there was no interaction with PTB (data not shown). Follow-up analysis using linear mixed-effects models confirmed no significant interactions between preterm birth and gestational age at urine sampling (interaction term: B = –0.00198; SE = 0.005; p = 0.67). Table 4 presents the secondary analysis in which PTB was further classified into either spontaneous or placental. We observed no significantly elevated odds for placental PTB in relation to averaged or visit-specific BPA concentrations. It is worth noting, though, that the ORs for the placental subtype PTB are consistently positive, with a suggestive increase in odds in relation to visit 2 BPA concentrations, and should be followed up in a larger study. Except for visit 4, there were no significant associations between BPA levels and spontaneous PTB. At this time point, a ln-unit increase in BPA concentrations was associated with 2.46 (95% CI: 1.42, 4.25) times greater odds of delivering a spontaneous PTB (n = 25). This relationship persisted even after additional adjustment for ΣDEHP metabolites. No significant relationships were observed between BPA concentrations and the medical protocol–driven PTBs (data not shown).
After stratifying by sex, we observed significantly elevated odds of overall PTB for female infants in relation to averaged BPA concentrations (OR = 1.80; 95% CI: 1.02, 3.13), though statistical significance did not persist after additional adjustment for phthalate concentrations (Table 5). We observed no significant associations between average or visit-specific BPA exposure and PTB for male infants (OR = 0.89; 95% CI: 0.47, 1.70).
Results
Study demographic characteristics have been previously described (Ferguson et al. 2014). Briefly, the study population was predominantly white (58.5%) and highly educated (85.9% had some college-level education) and were nonsmokers (92.3%). Approximately 10% of the women used some form of ART, and 47.3% of the population was either overweight (25 to ≤ 30 kg/m BMI) or obese (> 30 kg/m BMI). These population characteristics did not differ significantly between PTB cases and controls (Ferguson et al. 2014). Overall, BPA concentrations were detected in 1,350 (79.6%) of the samples analyzed, with the breakdown by visit and case–control status detailed in Table 1. There were 327 (92.9%) controls and 114 (87.7%) cases that had three or more repeated samples available for analysis during pregnancy. Visit 4 urine samples were collected on 66 (50.8%) cases and 314 (89.2%) controls. The mean (range) of the final gestational age for those cases who provided a visit 4 urine sample was 36.3 (35.0–36.9) weeks.
As expected, BPA was log-normally distributed and levels were ln-transformed for statistical analysis. SG-corrected urinary BPA geometric means and standard deviations by individual visit are presented in Table 1. Concentrations were relatively constant across gestation, with no significant BPA concentration differences at individual visits by Wilcoxon rank-sum test between PTB cases and controls. Urinary specific gravity was the highest early in pregnancy (SG = 1.017) and had a significantly decreased trend across pregnancy (p < 0.01). Spearman correlations between study visits for BPA were low and ranged from 0.17 to 0.26 (all p-values < 0.01). ICC for SG-corrected BPA indicated low temporal reliability (ICC = 0.21; 95% CI: 0.16, 0.27), and ICCs were slightly higher in cases of PTB compared with controls (Table 2). SG concentrations had low to moderate reproducibility across pregnancy (ICC = 0.38; 95% CI: 0.33, 0.43).
Crude and adjusted ORs (95% CIs) of overall preterm birth in association with averaged and cross-sectional (i.e., by visit) urinary BPA levels are presented in Table 3. Adjusted models controlled for maternal age, maternal race/ethnicity, educational level attained, medical insurance, parity, prior history of preterm birth, and BMI. A third set of models additionally included subject-specific averages for the ΣDEHP metabolites.
Generally, in both crude and adjusted models no significant relationships were observed between overall preterm birth and either averaged (OR = 1.21; 95% CI: 0.79, 1.85) or individual-visit urinary BPA levels, although, except for visit 3, ORs tended to be > 1.0 (Table 3). In our joint models using the random slopes and intercepts as predictors in the adjusted logistic regression models, we found no significant associations (see Supplemental Material, Table S1 http://ehp.niehs.nih.gov/wp-content/uploads/123/9/ehp.1408126.s001.acco.pdf). Fitting generalized additive mixed models indicated that the association between gestational age at sample collection and urinary BPA concentrations was linear, and there was no interaction with PTB (data not shown). Follow-up analysis using linear mixed-effects models confirmed no significant interactions between preterm birth and gestational age at urine sampling (interaction term: B = –0.00198; SE = 0.005; p = 0.67). Table 4 presents the secondary analysis in which PTB was further classified into either spontaneous or placental. We observed no significantly elevated odds for placental PTB in relation to averaged or visit-specific BPA concentrations. It is worth noting, though, that the ORs for the placental subtype PTB are consistently positive, with a suggestive increase in odds in relation to visit 2 BPA concentrations, and should be followed up in a larger study. Except for visit 4, there were no significant associations between BPA levels and spontaneous PTB. At this time point, a ln-unit increase in BPA concentrations was associated with 2.46 (95% CI: 1.42, 4.25) times greater odds of delivering a spontaneous PTB (n = 25). This relationship persisted even after additional adjustment for ΣDEHP metabolites. No significant relationships were observed between BPA concentrations and the medical protocol–driven PTBs (data not shown).
After stratifying by sex, we observed significantly elevated odds of overall PTB for female infants in relation to averaged BPA concentrations (OR = 1.80; 95% CI: 1.02, 3.13), though statistical significance did not persist after additional adjustment for phthalate concentrations (Table 5). We observed no significant associations between average or visit-specific BPA exposure and PTB for male infants (OR = 0.89; 95% CI: 0.47, 1.70).
Source...