Study participants and data collection. The present analysis of 137 women is part of a larger prospective cohort study designed to investigate the impact of environmental chemicals on fertility and pregnancy outcomes among couples seeking fertility treatment. All women in the larger study who underwent at least one IVF cycle and had urinary BPA analyzed were included in the present analysis. Study participants were female partners of couples seeking infertility evaluation and treatment at the Massachusetts General Hospital (MGH) Fertility Center, Boston, Massachusetts. Couples were recruited between November 2004 and April 2010. Women 18–45 years of age who used their own oocytes (eggs) for IVF were eligible. The women were followed from study entry through each of their IVF cycles until they had a live birth or discontinued treatment at the MGH Fertility Center. The study was approved by the Institutional Review Boards of the Massachusetts General Hospital, Harvard School of Public Health (HSPH), and the Centers for Disease Control and Prevention (CDC). All participants provided an informed consent after the study procedures were explained by a research nurse and all questions were answered.
At recruitment, a brief, nurse-administered questionnaire was used to collect data on demographics, medical history, and lifestyle. Women also completed a detailed take-home questionnaire with additional questions on lifestyle factors, occupation, and medical history (completed by > 90% of participants). Clinical information was obtained from the electronic medical record, and infertility diagnoses were classified according to the Society for Assisted Reproductive Technology (SART) definitions.
Treatment protocols and clinical IVF measures
. All women were initially treated with a cycle of oral contraceptive pills (OCP) to suppress ovulation unless it was contraindicated. The day of OCP-induced menses was referred to as cycle day 1 of the treatment cycle, the day after was cycle day 2 and so on. The patient was then monitored at cycle day 3 at the clinic to ensure ovarian suppression before beginning controlled ovarian stimulation with follicle-stimulating hormone (FSH) and gonadotropin-releasing hormone (GnRH) agonists or antagonists. Patients were monitored as needed during gonadotropin stimulation for serum estradiol (E2), follicle size measurements, follicle count, and endometrial thickness through 2 days before the egg-retrieval procedure. Human chorionic gonadotropin (hCG), a hormone similar to luteinizing hormone (LH), was administered (hCG trigger) approximately 36 hr before the scheduled egg-retrieval procedure to induce ovulation. Measurements of serum FSH and E2, and details of egg retrieval have been previously described (Mok-Lin et al. 2010
). Patients underwent one of three IVF treatment protocols: a
) luteal-phase GnRH agonist protocol using low-, regular-, or high-dose leuprolide (Lupron) with pituitary desensitization begun in the luteal phase; b
) follicular-phase GnRH-agonist/Flare protocol, in which Lupron was begun on day 2 of the follicular phase at 20 units and decreased to the standard dose of 5 units on day 5; and c
) GnRH-antagonist protocol, in which GnRH-antagonist was begun when the lead follicle reached 14 mm in size and/or E2 levels were ≥ 1,000 pg/mL. The antagonist and flare protocols were primarily for poor responders. The flare protocol is indicated for women > 40 years of age with very poor ovarian response (i.e., inadequate follicle recruitment after controlled ovarian stimulation with gonadotropins and low peak E2 level at time of hCG trigger), whereas the antagonist protocol is used in women < 40 years of age with diminished ovarian reserve and poor ovarian response.
Endometrial thickness (millimeters) was measured by transvaginal ultrasound scan (7.5 MHz frequency; GE Logiq 3/5; GE Healthcare, Waukesha, WI) before the administration of hCG, which corresponded to 36 hr before the egg-retrieval procedure. Women with endometrial thickness < 7 mm typically did not undergo a transfer and their embryos were frozen. Following the day of egg retrieval, patients were prescribed progesterone (P), usually by intramuscular administration (50 mg/day). Nine days after egg retrieval, patients began transdermal E2 (Vivelle patches; Novartis Pharmaceuticals Corp., East Hanover, NJ) at a dose of 0.2 mg every other day. Both P and E2 were administered to hormonally prime the endometrium for embryo transfer.
Embryos were evaluated by an embryologist and selected for uterine transfer on day 2, 3, or 5 of embryo maturation in culture. A day-2 transfer was performed when the patient had only one or two embryos for transfer. Because there were few day-2 transfers (n = 13 IVF cycles) and they represent patients with poorer expected outcomes, they were excluded from these analyses and are not included in the 180 cycles. Implantation failure was defined as a serum β-hCG level < 6 mIU/mL typically measured 17 days (97% measured by day 17, range 15–20 days) after egg retrieval.
Urine sample collection and urinary bisphenol A concentrations. The 137 enrolled women provided ≤ 2 spot urine samples per IVF cycle, the first collected between cycle day 3 and day 9 of the treatment cycle, and the second on the day of the egg-retrieval procedure, typically before the procedure. Urine was collected in a sterile clean polypropylene specimen cup. Specific gravity (SG) was measured at room temperature using a handheld refractometer (National Instrument Co. Inc., Baltimore, MD) calibrated with deionized water before each measurement. The urine was divided into aliquots and frozen and stored at –80°C. Samples were shipped on dry ice overnight to the CDC where they were stored at ≤ –40°C until analysis.
The urinary concentration of free and conjugated BPA species (total BPA) was measured using online solid phase extraction (SPE) coupled to isotope dilution–high performance liquid chromatography (HPLC)–tandem mass spectrometry (MS/MS) as described before (Ye et al. 2005
). First, 100 µL of urine was treated with β-glucuronidase/sulfatase (Helix pomatia
, H1; Sigma–Aldrich, St. Louis, MO) to hydrolyze the BPA-conjugated species. BPA was then retained and concentrated on a C18 reversed-phase-size-exclusion SPE column (Merck KGaA, Darmstadt, Germany), separated from other urine matrix components using a pair of monolithic HPLC columns (Merck KGaA), and detected by negative ion–atmospheric pressure chemical ionization–MS/MS. The limit of detection (LOD) for BPA was 0.4 µg/L. In addition to study samples, each analytical run included low-concentration and high-concentration quality control materials, prepared with spiked pooled human urine, and reagent blanks to assure the accuracy and reliability of the data (Ye et al. 2005
). BPA concentrations < LOD were assigned a value equal to the LOD divided by √–
2 (Hornung and Reed 1990
) before adjustment for urine dilution by SG, as described previously (Meeker et al. 2010
Statistical analysis. Characteristics of the women and IVF cycles were summarized using means, standard deviations, and percentages, as appropriate. The geometric mean of the SG-adjusted BPA concentrations from two spot–urine samples collected during each IVF cycle was used as a measure of cycle-specific urinary BPA concentration. The distribution of these geometric mean BPA concentrations was summarized using percentiles.
Multivariate generalized estimating equation (GEE) models for repeated measures were used to evaluate the association between cycle specific urinary SG-adjusted BPA concentrations and potential risk factors for implantation failure. We used an autoregressive correlation structure to account for correlation between outcomes across repeated IVF cycles within the same woman. We modeled SG-adjusted urinary BPA concentrations in quartiles. Age (≥ 37 years or < 37 years), and day of embryo transfer (day 5 vs. day 3) were retained in the final model because of their biological and clinical relevance as indicated in previous studies (Giorgetti et al. 1995
; Rehman et al. 2007
Other variables considered as potential confounders included the number of embryos transferred (single vs. multiple); IVF protocol type (flare/antagonist vs. regular luteal phase protocol); day-3 serum FSH level (international units per liter, a measure of ovarian reserve); peak serum E2 (this corresponded to the last serum E2 measurement, 2 days before egg retrieval); endometrium thickness (< 9 mm vs. ≥ 9 mm); smoking (ever vs. never smoker); and body mass index (BMI) [overweight/obese (BMI ≥ 25 kg/m2) vs. normal/underweight (BMI < 25 kg/m2)]. Covariates that predicted implantation failure with p < 0.2 in univariate models and present in ≥ 5% of the cohort were evaluated for inclusion in the multivariate model using backward selection, and they were retained in the final model if their p was ≥ 0.10 or if the effect estimate changed by > 10% when the covariates were removed. A test for trend was performed to determine if there was a linear dose–response relationship between quartiles of urinary BPA and implantation failure. The trend test was performed by modeling the BPA quartiles as an ordinal categorical variable, using the integer values 0, 1, 2, and 3 for BPA quartiles 1, 2, 3, and 4, respectively.
As an alternative to the GEE model approach based on a logistic link, log binomial models for the relative risk were also fit. Because these models did not always converge, but yielded similar results when they did, they are not presented here. In addition, analyses restricted to first cycles only were also performed.
Because the choice of IVF treatment protocol used in a given cycle is determined by the patient’s anticipated probability of success based in part on past implantation failures and because past urinary BPA concentrations may be associated with past implantation failures and correlated with current BPA concentrations, we considered protocol type as a potential intermediate on the causal pathway between urinary BPA concentration and implantation failure. We therefore present model results both adjusted and unadjusted for treatment protocol. We also conducted a stratified analysis by type of treatment protocol. All data analyses were performed using SAS version 9.2 (SAS Institute Inc., Cary, NC).