Women 18–45 years of age were recruited from partners seeking evaluation and treatment for infertility at the Massachusetts General Hospital (MGH) Fertility Center in Boston between November 2004 and December 2009. The present analyses were from a larger prospective open-cohort study, the Environment and Reproductive Health (EARTH) Study, which was designed to examine the relationship between environmental chemical exposures and fertility/pregnancy outcomes. The study was approved by the human studies institutional review boards of the MGH, Harvard School of Public Health and the Centers for Disease Control and Prevention (CDC). Subjects signed an informed consent form after the study procedures were explained by a research nurse and all questions were answered.
Women included in this analysis were recruited preconception (herein referred to as pre- or before pregnancy) and followed until delivery. Conception methods included natural conception, ovulation induction with timed intercourse, intrauterine insemination, or in vitro fertilization. Women provided spot urine samples in polypropylene containers at enrollment, on returning for subsequent clinic appointments before pregnancy, and during pregnancy (first, second, or third trimester). Enrollment urine samples were generally collected on entry into the study and before fertility treatment. Before storing samples at –80°C, urine was aliquoted and specific gravity (SG) was measured using a handheld refractometer calibrated with deionized water before each use (National Instrument Company Inc, Baltimore, MD). Samples were shipped on dry ice to the CDC for analysis.
An intrauterine pregnancy was confirmed by presence of a fetal heartbeat detected by transvaginal ultrasound. We used one of three methods to estimate the date of conception: oocyte retrieval date, which was abstracted from medical records; crown–rump length, which was measured during a fetal ultrasound between 6 and 8 weeks of gestation; or woman’s report of last menstrual period. When more than one dating method was available, priority was given to retrieval date > ultrasound > last menstrual period.
To examine variability of urinary phthalate metabolite concentration before and during pregnancy, we restricted our analyses to women who delivered a liveborn infant and provided two or more pregnancy urine samples and two or more urine samples before that pregnancy. We excluded women with fewer than two urine samples during either or both time periods.
We measured the concentration of BPA and eight phthalate metabolites including MBP, MiBP, MBzP, MEP, and the following four DEHP metabolites: mono(2-ethyl-5-carboxypentyl) phthalate (MECPP), mono(2-ethyl-5-hydroxyhexyl) phthalate (MEHHP), mono(2-ethyl-5-oxohexyl) phthalate (MEOHP), and mono(2-ethylhexyl) phthalate (MEHP), using previously described analytical chemistry methods and quality control procedures. (Silva et al. 2007
; Ye et al. 2008
) We limited our statistical analyses to DEHP metabolites, MBP, MiBP, MBzP, and MEP because of their high frequency of detection in the U.S. population (Silva et al. 2004
; Woodruff et al. 2011
) The limits of detection (LOD) for the target phthalate metabolites were in the low microgram per liter range (~ 0.1 to ~ 1 μg/L) and 0.4 μg/L for BPA. Values less than the LOD were given a value of the LOD/√–
2 (Hornung and Reed 1990
). We applied correction factors of 0.66 and 0.72 to the MEP and MBzP concentrations, respectively, because the analytic standards used were of inadequate purity (Calafat A, personal communication).
Because DEHP is metabolized primarily into MEHP, MEHHP, MECPP, and MEOHP, we used two summary measurements: a) total molar sum of all four DEHP metabolites and b) molar sum of three oxidative DEHP metabolites (MECPP, MEHHP, and MEOHP). We calculated the molar sum of DEHP metabolites by dividing each metabolite concentration by its molar mass and then summing the individual metabolite concentrations. We also present results separately for MEHP concentrations to facilitate comparisons with prior studies.
We accounted for urine dilution by standardizing urinary phthalate metabolite and BPA concentrations using SG. Urine dilution was adjusted using a modified and previously described formula in all of our analyses of BPA and phthalate metabolites (Duty et al. 2005
; Meeker et al. 2009
). We excluded samples with SG values > 1.04 (Boeniger et al. 1993
). All statistical analyses were conducted using SG-standardized biomarker concentrations unless otherwise noted.
Statistical analyses. Descriptive analyses. We first examined the sociodemographic characteristics of participating women (means and proportions). We computed the median and 25th and 75th percentiles of SG-adjusted phthalate metabolites and BPA concentrations from the first (at enrollment) and last urine samples provided before women became pregnant and from samples provided during each trimester of pregnancy. We calculated univariate characteristics of and correlation between the within-woman geometric mean (GM) urinary phthalate metabolite and BPA concentrations for all prepregnancy and pregnancy urine samples. We also compared the difference in phthalate metabolite and BPA concentrations before and during pregnancy using a linear mixed model with log10-transformed phthalate metabolite or BPA concentrations as the outcome. We included an indicator variable to designate samples as being before or during pregnancy. We estimated the percent difference in pregnancy concentrations relative to prepregnancy concentrations.
Variability analyses. We conducted three analyses to characterize the variability and change in urinary phthalate metabolite and BPA concentrations before and during pregnancy. These analyses used log10
-transformed urinary SG-adjusted phthalate metabolite and BPA concentrations because of their right-skewed distribution. First, we estimated the variability of urinary phthalate metabolite and BPA concentrations before or during pregnancy by calculating the intraclass correlation coefficient (ICC) using a random intercept-only linear mixed model. The ICC is a measure of reproducibility, calculated by dividing the between-subject variability by the sum of the between- and within-subject variability. Values range from 0, indicating no reproducibility, to 1, indicating perfect reproducibility (Rosner 2000
). Next we estimated the percent change in urinary phthalate metabolite and BPA concentrations over time during the prepregnancy or pregnancy sampling frame using linear mixed models with subject-specific intercepts. For prepregnancy samples we calculated the number of weeks since enrollment by subtracting the date of enrollment from each subsequent collection date. The enrollment urine sample was set to a time of 0. For pregnancy samples, we calculated the number of weeks since conception for each urine sample by subtracting the date of conception from each urine sample collection date. We estimated the percent change in phthalate metabolite and BPA concentration with each 4-week change in time before and during pregnancy. Finally, using spaghetti plots, we graphed a random sample of urinary phthalate metabolite and BPA concentrations in 50 women before and during pregnancy as a function of time since enrollment or conception.
We evaluated the pattern and variability of urine dilution before and during pregnancy by conducting the above analyses using untransformed SG values as the outcome, because changes in urine dilution (i.e., SG) during pregnancy may partially account for changes in urinary phthalate concentrations.
Surrogate category analyses. We conducted three additional analyses to examine the rank-ordering, predictive ability, and consistency of a single urinary phthalate metabolite and BPA concentration during pregnancy among women with all three urine samples (i.e., one sample for each trimester). First, we conducted a classification analysis (Hauser et al. 2004
; Mahalingaiah et al. 2008
). Using the GM of the three individual trimester phthalate metabolite or BPA concentrations, we calculated tertiles of average gestational exposure of the women. We then classified women into surrogate tertiles of phthalate metabolite/BPA concentrations using their trimester-specific urine sample concentration. We categorized the women as being either in the top or bottom two tertiles of either of these measures. We then calculated the sensitivity, specificity, and positive predictive value (PPV) of the top tertile of the surrogate measure with the top tertile of the average gestational measure. The PPV is the probability of being classified as having high average gestational concentrations given a high surrogate measure. We examined trimester-specific surrogate categories to determine if the timing of sample collection influenced the predictive ability of a single spot urine sample.
Second, we examined whether surrogate categories of trimester-specific urine samples were associated with average gestational urinary phthalate metabolite or BPA concentrations of women (Meeker et al. 2005
; Teitelbaum et al. 2008
). Similar to the first analysis, we calculated surrogate tertiles of trimester-specific urinary phthalate metabolite/BPA concentrations (i.e., surrogate categories of low, medium, and high). We then used box plots to examine the distribution of the average gestational exposure of women (e.g., GM of all three urinary biomarker concentrations) within each of these surrogate categories. For example, we calculated the average gestational urinary BPA concentration of each woman using the values from all three trimesters. Based on the first-trimester urinary BPA concentrations for all the women, we then categorized each woman’s first-trimester urinary BPA concentrations into surrogate tertiles. We then plotted the distribution of the average gestational urinary BPA concentration of women for the first-trimester surrogate tertile variable. We then completed the same process for the second and third trimesters. If the surrogate tertile variable provides reasonable rank ordering, then we would expect to see increasing average gestational BPA concentrations across the surrogate trimester categories (i.e., increasing average gestational BPA concentration as one goes from first to second- to third-trimester surrogate categories).
Finally, we examined whether the women remained in the same tertile of exposure over the course of their pregnancy by counting the number of times (one, two, or three) her urine sample concentrations were in the same tertile for each phthalate metabolite or BPA. For instance, if all three urine samples for a woman were in the same tertile, she was assigned a count of three. We calculated separate tertiles for each trimester. All analyses were conducted with SAS version 9.2 (SAS Institute Inc., Cary, NC).