Between March 2007 and May 2008 the Center for Androgen-related Research and Discovery at Cedars-Sinai Medical Center (CSMC) in Los Angeles, CA recruited fifty-two women with PCOS and fifty controls through media advertisements and subspecialty clinics. Most (>90%) of PCOS patients agreed to store blood and urine samples for future studies. As part of the CSMC study, controls volunteered in response to advertising in the surrounding community and university for “healthy” women aged 25–45 years. All women were aged 18–45 years. Both PCOS patients and controls provided CSMC with information about their age, race, ethnicity, virilization and body mass index which are associated with case status. Participants enrolled in the CSMC study provided written informed consent to allow CSMC to provide Centers for Disease Control and Prevention (CDC) with anonymized data and samples for our nested pilot study. This study was approved by the Internal Review Board of the Centers for Disease Control and Prevention (CDC protocol #4918).

Exclusion criteria included concurrent pregnancy, the use of hormonal (including oral contraceptives) or other medications for the prior three months, diabetes, menopause, and inability to provide written consent.

We defined case-patients based on the strict National Institutes of Health 1990 criteria for PCOS, specifically including anovulation or oligo-ovulation; clinical (i.e., a modified Ferriman-Gallwey [mFG] hirsutism score >6) or laboratory evidence of hyperandrogenism; and the exclusion of related disorders (i.e. thyroid dysfunction, hyperprolactinemia, non-classic adrenal hyperplasia, androgen-secreting tumors, etc.), as previously described [16]. We defined controls as healthy women with long-term regular and predictable menstrual cycles, without hirsutism (i.e. mFG scores of ≤2), and without clinical and laboratory evidence of hyperandrogenism or other hormonal dysfunction.

Recruited participants provided single spot urine and blood samples at CSMC. To avoid environmental contamination, participants were asked to open the urine containers when ready to provide a sample and to avoid touching the inside of the container. Serum samples where obtained by venipuncture performed using standard protocol and precautions, and the samples drawn into one 10 ml red top Vacutainer^® tube per participant. Tubes were placed upright and blood allowed to clot at room temperature for 60 min, after which they were centrifuged at 3,000 rpm (~1000 g) for 10 min. Using a disposable pipet, the serum was transferred to CDC-provided containers (10 mL amber glass bottle with Teflon-lined screw cap [Wheaton, Millville, NJ] Serum vial for serum PBDE content and 2 mL Nalgene^® [Rochester, NY] cryovials for serum PFCs and lipids content). The serum PBDE vials had been previously cleaned in a laboratory dish washer and baked at 300°C overnight and the screw caps had been rinsed with methanol [17]. After recapping, each container was frozen upright in a CDC-provided cardboard storage box at -70°C until shipping.

First morning void urine samples were collected in individually wrapped collection cups provided by the CDC. At the time of collection participants were asked to wash their hands with soap and water, to collect at least 10–15 ml of urine in the cup, to not remove the cap from the cup until ready to void, and to not touch the inside of the cup or cap. Using a sterile disposable plastic pipet, urine samples were transferred into prelabeled 5 ml Nalgene^® cryovials (for urinary Phthalates/BPA content) and 2 ml Nalgene^® cryovials (for urinary creatinine content). The urinary samples were then placed in CDC-provided storage boxes and frozen at -20°C until shipping.

All sample collection vials were labeled with preprinted bar-coded labels provided by the CDC, and the date and time of collection was added using a permanent marker. Operators used disposable powder-free nitrile gloves and surgical masks during sample collection and processing, and all collection and processing supplies were provided by the CDC to maximize the integrity of the samples. All sample processing was performed in a fume hood lined with aluminum foil to minimize contamination by ambient dust, as PBDEs are well known contaminants in indoor dust. The laboratory space was further thoroughly cleaned before any laboratory work was undertaken to minimize any dust or particulate matter that could contaminate the samples. All samples were shipped to the CDC on dry ice and upon arrival stored at -70°C until the time of analysis [17].

Blood samples were analyzed for concentration of 36 PCBs, 9 OCPs, 11 PBDEs and 8 PFCs. Laboratory results of blood-agent concentration included both whole serum and lipid-adjusted concentration values of PCBs, OCPs and BDEs and serum only for PFCs.

Analytical determination of PCBs, OCPs and PBDEs in serum were performed by gas chromatography isotope dilution high resolution mass spectrometry, after solid phase extraction (SPE) and co-extracted lipid removal techniques [17]. Total cholesterol and triglycerides were measured using a Roche Hitachi Mod P Chemistry Analyzer (Roche Hitachi, Basel, Switzerland), using single-point, forced-through-zero calibration curves and Roche colorimetric methods as described in the product applications #11491458216 V15 (total cholesterol) and #04843673003 V13 (total triglycerides).

The measurements of PCBs, OCPs, and PBDEs were made in batches of twenty-four unknowns, three quality controls and three method blanks. The method blanks were used to track any analytical background during the sample preparation and the final analytical results were blank subtracted. The limits of detection (LODs) were calculated as standard deviations (SDs) of the method blanks analyzed in parallel with the unknowns after subtracting the average blank concentration or as the instrumental LOD in the absence of a detectable blank level (Additional file 1). The concentration ratio between 2,2′,4,4′,5-pentabromodipehnyl ether (BDE-99) and 2,2′,4,4′-tetrabromodipehnyl ether (BDE-47) can be used as an indicator of contamination during sample collection. The median ratio (BDE-99/BDE-47) in this study was 0.19 (range 0.10 – 0.36) which is consistent with a metabolized pattern and hence did not indicate contamination of indoor dust which is expected to have a ratio close to 1 (i.e. the concentration ratio between BDE-99 and BDE-47 in the commercial pentaBDE product used as a flame retardant).

We measured PFCs in serum using a modification of the on-line SPE coupled to high-performance liquid chromatography (HPLC)-isotope dilution tandem mass spectrometry (MS/MS) approach previously described in detail [18].

Urine samples were analyzed for 11 phthalate metabolites and for BPA (total concentrations) using a modification of the on-line SPE-high performance liquid chromatography-isotope dilution tandem mass spectrometry approaches described before [19, 20]. Urinary creatinine, used to adjust for the dilution of the urine, was measured using an enzymatic reaction on a Roche Hitachi 912 chemistry analyzer (Roche Hitachi, Basel, Switzerland).

We used the serum concentrations or urine concentrations of the chemicals to categorize exposures to the chemicals or their parent compounds. We split observations into tertiles based on chemical concentrations then performed multivariable logistic regression analyses to explore associations between the concentrations of the chemicals (in μg/L for PFCs; lipid-adjusted for PCBs, OCPs and PBDEs and creatinine-adjusted for phthalate metabolites and BPA) and PCOS case status after adjusting for potential confounders (i.e. age [divided into 5 categories], body mass index [BMI], and White race [compared to all other races]). This multivariable logistic regression model is represented as PCOS (0/1) = α + β₁(chemical concentration tertile) + β₂(age) + β₃(BMI) + β₄(race). Hispanic ethnicity was not significantly related to PCOS when tested in preliminary analyses by logistic regression alone or when controlling for race, BMI, or age; therefore we did not statistically control for Hispanic ethnicity in the final models to conserve power. Due to the high ratio of variables in the model to study participants we calculated exact adjusted odds ratios, exact p values, and exact confidence intervals using the network method [22]. All statistical analyses were conducted using SAS v9.3 (Cary, NC). More than 50% of the samples had serum or urine concentrations < LOD for 22 PCBs, 3OCPs, 2 phthalate metabolites, and 2 PFCs; thus, no further analyses were conducted for these compounds (Additional file 2).

PCOS case-patients had perfluorooctanoate (PFOA) and perfluorooctane sulfonate (PFOS) geometric mean serum-concentration higher than controls (4.1 μg/L vs. 2.3 μg/L, p = 0.001; and 8.2 μg/L vs. 4.9 μg/L, p = 0.01, respectively) (Table 3). Participants were 6.9-fold more likely to have PCOS if they had PFOA concentrations in the highest tertile (geometric mean = 5.5 μg/L, range = 4.1–13.4 μg/L) when compared to those in the lowest tertile (geometric mean = 1.6 μg/L, range = 0.2–2.6 μg/L). Participants were also 5.8-fold more likely to have PCOS if they had PFOS concentrations in the highest tertile (geometric mean = 12.5 μg/L, range = 8.6–27.9 μg/L) (Table 4).Table 3

Serum and lipid-adjusted concentrations of brominated diphenyl ethers, polychlorinated biphenyls, and persistent organic pesticides, and perfluorinated compounds in polycystic ovary syndrome (PCOS) case-patients and controls

Participants were 5.79-7.5 times more likely to have PCOS if they had PCB 153, 170, 180, 183, or 196 and 203 whole weight (pg/mL serum) concentrations in the middle tertile (but not the highest tertile), when compared to those in the lowest tertile (Table 4). A similar pattern also occurred among lipid-adjusted measurements of PCB 170, 180, and 196 and 203 concentrations. PCOS case-patients had PCB180 geometric mean lipid-adjusted concentrations (4.3 ng/g) lower than controls (5.3 ng/g) (p = 0.04) (Table 3).

PCOS case-patients had monobenzyl phthalate (mBzP) geometric mean creatinine-adjusted urine concentrations (7.5 μg/g creatinine) lower than controls (11.7 μg/g creatinine) (p = 0.02) (Table 5). After controlling for BMI, age, and race, participants were less likely to have PCOS if they had creatinine-adjusted urine concentrations of mBzP or mono-n-butyl phthalate (mBP) in the middle or highest tertile compared to the lowest and mono-2-ethylhexyl phthalate (mEHP), and monoethyl phthalate (mEP) in the middle (but not the highest) tertile compared to the lowest (Table 6).Table 5

Urinary concentrations of phthalate metabolites and bisphenol A in polycystic ovary syndrome (PCOS) case-patients and controls

PCOS case-patients and controls had similar geometric mean lipid-adjusted PBDEs, and OCPs serum concentrations (Table 3), and urinary BPA concentrations (Table 5), with the exception of BB153 of which controls had higher concentrations than case-patients (Table 3). After controlling for BMI, age, and ethnicity, no increase in likelihood for PCOS was associated with any of the measured PBDEs, OCPs, or BPA, with one exception (Tables 4 &6).

Our findings suggest that increasing odds of PCOS case-status was associated with higher serum concentrations of two PFCs (PFOA and PFOS) and with several PCB congeners. PFOS, PFOA, PCBs and phthalates have all been implicated as endocrine disruptors and have been linked to other reproductive outcomes. Previous epidemiologic investigations suggest that exposure to PFCs may increase women’s risk of early menopause [23], thyroid disease [24], and delayed pregnancy and subfecundity [25–27]. Exposure to certain PCBs has been associated with other reproductive health effects such as delayed pregnancy [28]. PCB exposure has also been linked to increased breast cancer [29], diabetes and thyroid disease [30] risk. This is the first documented evidence of an association between PCOS and serum concentrations of PFCs or PCBs.

PCOS case-patients had urinary mBzP concentration lower than controls, and lower concentrations of mBzP, mBP, mEHP and mEP were associated with an increased likelihood of PCOS. Previous research suggests that some phthalates, including di-2-ethylhexyl phthalate (DEHP), the precursor of mEHP, have anti-androgenic effects in animals [12]. Also, epidemiologic studies have linked mEHP and a metabolite of di-isononyl phthalate (DiNP) with decreased testosterone production in men [31] and mBP, mono-isobutyl phthalate, mBzP, and the sum of metabolites of DEHP and of DiNP with delayed pubarche in women [32]. PCOS is characterized by hyperandrogenemia; therefore, our results are consistent with previous evidence of the anti-androgenic effects of certain phthalates.

Unlike previous studies [33–35], we found no association between PCOS and BPA. Previous investigators quantified serum BPA concentrations using a commercially available enzyme-linked immunosorbent assay (ELISA) kit, whereas we measured urinary concentrations of this toxicant using the gold-standard detection technique, isotope dilution mass spectrometry [20]. ELISA lacks adequate analytical selectivity and specificity, and because matrix effects may induce performance anomalies, ELISA is not adequate for the quantitative determination of BPA in clinical specimens [34, 36].

This study is subject to several limitations. First, our sample size of 52 PCOS case-patients and 50 controls may not be large enough to generate enough statistical power to detect a difference in some of the toxicant concentrations among PCOS case-patients and controls. Similarly, the small sample size limits the possibility of considering non-monotonic dose–response relationships and some significant associations can be chance findings due to the large number of statistical tests.

We obtained a single spot serum and urine specimen. Therefore, the concentrations of each toxicant represent a snapshot of each woman’s exposure at a given time. A single measurement design may be ineffective in detecting an association between PCOS and pollutants that metabolize quickly. For example, some of the measured analytes (such as PCBs, PBDEs, OCPs, and PFCs) persist in the body for years, while phthalates and BPA metabolize quickly and are eliminated from urine within a few hours after exposure.

It is unclear whether the observed associations suggest that PFOA, PFOS, and the implicated PCB congeners increase risk for PCOS or whether the endocrine milieu of the disorder alters the storage and clearance of these chemicals, leading to increased serum measurements in PCOS patients. Oligomenorrhea and amenorrhea are common symptoms of PCOS. Thus, PCOS cases, who menstruate less frequently compared to controls, may have similar exposures to women without PCOS, but higher blood-toxicant concentrations [23]. PCBs, however, are stored mostly in adipose tissue making amenorrhea an unlikely explanation for differences in measured exposure.

Finally, a larger proportion of case-patients were of White race compared to controls, and PCOS patients were younger and had significantly higher BMI than controls, although we attempted to control for these parameters known to be associated with case-status in the logistic regression models.