Endometriosis - Fertility and Sterility

2 downloads 0 Views 144KB Size Report
Mar 14, 2012 - c Department of the Environment and Primary Prevention, Unit of Toxicological Chemistry, Italian ... 299-00161 Rome, Italy (E-mail: [email protected]). .... pollutants were determined by ion-trap mass spectrometry. ..... GSTP1 and total PCB (OR ¼ 8.5; 95% CI: 1.6–44.6). In ..... 10–19 cigarettes/day. 9.
Glutathione transferase polymorphisms and risk of endometriosis associated with polychlorinated biphenyls exposure in Italian women: a gene–environment interaction Susanna Vichi, Ph.D.,a Emanuela Medda, B.Sc.,b Anna Maria Ingelido, Ph.D.,c Annamaria Ferro, M.D.,d Serena Resta, M.D.,d  , M.D., Ph.D.,b Elena De Felip, Ph.D.,c Maria Grazia Porpora, M.D.,d Annalisa Abballe, Ph.D.,c Lorenza Nistico Simonetta Gemma, Ph.D.,a and Emanuela Testai, Ph.D.a a Department of the Environment and Primary Prevention, Unit of Mechanisms of Toxicity, Italian National Institute for Health; b National Centre for Epidemiology, Surveillance and Health Promotion, Unit of Genetic Epidemiology, Italian National Institute for Health; c Department of the Environment and Primary Prevention, Unit of Toxicological Chemistry, Italian National Institute for Health; and d Department of Gynaecology, Obstetrics and Urology, ‘‘Sapienza’’ University of Rome, Rome, Italy

Objective: To investigate the occurrence of a gene–environment interaction between glutathione transferase (GST) gene polymorphisms (GSTM1, GSTT1, GSTP1, and GSTA1) and serum polychlorinated biphenyls (PCBs) levels. This is suggested as possible risk factors for endometriosis, a multifactorial gynecological disease. Design: Case-control study conducted from 2002 to 2005. Setting: Policlinico Umberto I, ‘‘Sapienza’’ University of Rome and Italian National Institute for Health, Rome. Patient(s): Italian women (N ¼ 343), with laparoscopic diagnosis and histologic confirmation of the presence (cases, N ¼ 181) or the absence (controls, N ¼ 162) of endometriosis. Intervention(s): Genomic DNA extraction, multiplex polymerase chain reaction (PCR), and restriction fragment length polymorphism analysis. Determination of serum concentrations of selected PCBs by ion-trap mass spectrometry (subgroup, 63 cases and 63 controls). Main Outcome Measure(s): Endometriosis diagnosis by laparoscopy, GST genotypes, serum PCB levels. Result(s): The genotype distributions of GSTM1, GSTA1, and GSTP1 did not show any statistically significant difference between cases and controls. The GSTT1 null genotype was negatively associated with the disease. The GSTP1 wild-type genotype in the presence of medium-high blood levels of PCB153, total PCBs, or of high levels of PCB180 significantly increased the risk of endometriosis, suggesting a multiplicative interaction. Conclusion(s): The GSTs polymorphisms per se do not increase the risk of developing endometriosis. However, a gene–environment interaction was observed for GSTP1Ile/Ile and GSTM1 null genotypes, modulating the effect of PCB153, PCB180, and of total PCBs on disease risk. (Fertil SterilÒ 2012;97:1143–51. Ó2012 by American Society for Reproductive Medicine.) Key Words: Endometriosis, GST polymorphism, PCBs, biomonitoring, gene–environment interaction

E

ndometriosis is a multifactorial gynecologic disease defined as the presence of hormonally

dependent endometrial glandular and stromal cells outside uterus. Because of its negative impact on the quality

Received August 2, 2011; revised and accepted February 21, 2012; published online March 14, 2012. S.V. has nothing to disclose. E.M. has nothing to disclose. A.M.I. has nothing to disclose. A.F. has nothing to disclose. S.R. has nothing to disclose. M.G.P. has nothing to disclose. A.A. has nothing to disclose. L.N. has nothing to disclose. E.D.F. has nothing to disclose. S.G. has nothing to disclose. E.T. has nothing to disclose. Study carried out within the framework of the Research Project ‘‘Exposure to Organohalogenated Compounds as a Risk Factor for Women's Reproductive Health’’ funded by the Italian National Institute for Health and the Ministry of Health. Reprint requests: Simonetta Gemma, Ph.D., Department of the Environment and Primary Prevention, Unit of Mechanisms of Toxicity, Italian National Institute for Health, Viale Regina Elena, 299-00161 Rome, Italy (E-mail: [email protected]). Fertility and Sterility® Vol. 97, No. 5, May 2012 0015-0282/$36.00 Copyright ©2012 American Society for Reproductive Medicine, Published by Elsevier Inc. doi:10.1016/j.fertnstert.2012.02.027 VOL. 97 NO. 5 / MAY 2012

of women's life in terms of pain and infertility and the economic burden for diagnosis, pharmacologic and surgical treatments, as well as for assisted fertilization practice, endometriosis has been recognized as a health priority. In industrialized countries endometriosis occurs in 6%–10% of the general female population; in women with pain, infertility, or both, the frequency is 35%–50% (1). Nevertheless these numbers are likely an underestimation, as among asymptomatic fertile women undergoing tubal sterilization procedures incidence ranges from 4%–43% (2). 1143

ORIGINAL ARTICLE: ENVIRONMENT AND EPIDEMIOLOGY The etiopathogenesis of this disease remains still unclear. The hypothesis of Sampson that retrograde menstruation triggers endometriosis is not supported by the high prevalence of this event (90%), indicating the involvement of additional risk factors. Endometriosis is characterized by a general inflammatory response in the peritoneal cavity with production of reactive oxygen species (ROS), which might act by increasing growth and adhesion of endometrial cells in the peritoneal cavity, with progression of endometriosis and infertility (3). Contrasting results on the association between inflammationinduced oxidative stress and endometriosis have been reported (4). Hormonal imbalance, genetic predisposition, failure of host immune response, and environmental factors have been suggested to concur to its onset and progression (5), but the association still remains controversial and no single theory seems to cover all the aspects of this disease. The hypothesis for a role of persistent environmental contaminants, such as polychlorinated dibenzo-p-dioxins (PCDDs) and polychlorinated biphenyls (PCBs), as risk factors in endometriosis, has been initially based on data on primates (6). Some criticisms (7) were related to bias and methodological deficiencies of the study. However, PCDDs and PCBs can interfere with synthesis, function, storage, or metabolism of hormones, bind to estrogen or androgen receptors. Many studies on the possible association between PCB serum levels and endometriosis, although controversial, indicated a slight increase of risk associated to PCB exposure (8–14). A genetic contribution to endometriosis etiopathogenesis was suggested by a 6% increased risk for close relatives, although no simple Mendelian inheritance was shown. Genetic susceptibility was explored also by studying mutations in genes responsible for detoxication, such as glutathione transferase (GST), as a possible risk factor to endometriosis (15–22). The GSTs, phase II enzymes, are critical in controlling free radical-associated injuries, making the association between the endometriosis and any alterations in their enzymatic activity biologically plausible. Many functional polymorphic variants at the four major loci have been described, including deletions for GSTM1 and GSTT1 and single-nucleotide polymorphisms in exon 5 of GSTP1 gene (an A/G transition causing a Ile/Val substitution in position 105 of the protein) (23) and in the proximal promoter of GSTA1 (C-69T) (24). The GSTM1 and GSTT1 gene deletion results in the absence of protein expression in 42%–60% and 13%–26%, respectively, of the Caucasians (25), lacking the related enzymatic activity. The GSTP1Ile/Val and GSTP1Val/Val enzymes (present in 30%–50% and 4%–16% of the Caucasians) show substrate-dependent alteration of its catalytic activity (23). The GSTA1 (C-69T) T allele (present in homozygosis in 14%–18% and in heterozygosis in 44%–49% of the Caucasians) significantly lowers enzyme hepatic expression (24, 26). Because endometriosis has the characteristics of a polygenic and multifactorial disease, it is likely that both environmental and genetic determinants interact. Therefore the concurrent contributions of both factors need to be investigated, as well as the presence of a gene–environment interaction accounting for multiplicative joint effects. To this aim 1144

a case-control study was carried out on Italian women aged 18–45 years, to examine whether functionally relevant polymorphisms in GSTA1, GSTT1, GSTM1, and GSTP1 genes are associated with the risk of endometriosis per se and in association with exposure to PCBs.

MATERIALS AND METHODS Study Population and Data Collection Three hundred forty-three (181 cases and 162 controls) Italian women, undergoing laparoscopy for suspected endometriosis or other benign gynecologic conditions, were enrolled in a case-control study from 2002 to 2005 at the Department of Gynaecology and Obstetrics, of the Policlinico Umberto I, ‘‘Sapienza’’ University of Rome. All patients met the inclusion criteria: 18–45 years of age, residence in Rome in the past 5 years, nulliparity/no breastfeeding history, absence of immunologic, hormonal disorders, or chronic diseases, and no occupational exposure to PCBs or pesticides. All enrolled women signed an informed consent form; a physician unaware of the indications to laparoscopy administered a questionnaire before surgery that documented age, education, job, medical/gynecologic/obstetric history, height and weight, and smoking habits. Among the 386 women characterized by these features, 343 underwent (incidental) laparoscopy and fully met the inclusion criteria. Before laparoscopy, a fasting blood specimen (1 mL) was collected from the cubital vein in vacutainer tubes to be used for DNA extraction. In a subset of women giving their consent (63 cases and 63 controls) the amount of blood collected was 30 mL to analyze genotype and serum PCB levels. For the subset, the body mass index (BMI) was calculated and a more detailed questionnaire was used to obtain information on potential confounders relevant for the exposure part (i.e., metabolic diseases, gravidity, parity, dietary, and weight changes in the past years) (Supplemental Table 1, available online). A detailed medical and gynecologic history was taken, and all patients underwent clinical and ultrasound examinations. More detailed information on questionnaire design, and diagnostic methodologies were previously described (8). A 10-mm laparoscopy was performed under general anesthesia. All women undergoing laparoscopy also had endometrial implants excised for histology to confirm the presence of endometriosis. The disease was staged according to the revised American Society of Reproductive Medicine classification. The control group consisted of women without complaints of infertility or pelvic pain, undergoing laparoscopy for benign gynecologic conditions (benign adnexal masses [61.7%], fallopian tubes abnormalities/diseases [24.7%], uterine myomas [13.6%]), that are not expected to share etiology with endometriosis and had no visual evidence of histologic features of endometriosis in random peritoneal biopsies.

DNA Extraction Blood samples from cases and controls were collected in ethylenediaminetetraacetic acid (EDTA) vacutainer tubes and VOL. 97 NO. 5 / MAY 2012

Fertility and Sterility® stored at 20 C until subjected to analysis. Genomic DNA extraction from whole blood was obtained using the QIAamp DNA blood mini kit (Qiagen) according to the manufacturer's instructions. Genotyping of the four loci was performed on the entire study population sample (181 cases and 162 controls) by personnel blinded to case-control status. To validate genotyping procedures 5% of samples were typed twice.

GSTM1 and GSTT1 Genotyping The detection of GSTM1 and GSTT1 homozygote null genotype (-/-) was performed simultaneously in a single assay by a multiplex polymerase chain reaction (PCR) reaction as previously described (27). The absence of GSTM1- or GSTT1specific PCR products indicated that GSTM1 or GSTT1 genes were deleted on both alleles (null). In contrast, PCR products indicated the presence of at least one copy of GST gene (positive: -/þ or þ/þ).

GSTP1 Genotyping The GSTP1 polymorphism (rs1695) was analyzed by restriction fragment length polymorphism following the method of Harries et al. (28) with minor modifications. The homozygous wild-type (Ile/Ile), homozygous mutant (Val/Val), and the heterozygous (Ile/Val) genotypes were characterized by the presence of one band of 176 bp, two bands of 81 and 95 bp, and three bands of 176, 81 and 95 bp, respectively.

GSTA1 Genotyping The GSTA1 C-69T polymorphism (rs3957357) was determined by restriction fragments length polymorphism following the method by Coles et al. (24) with minor modifications. The point mutation C-69T introduces a restriction site for EarI. Individuals homozygous for the wild-type allele (base C) showed one 480-bp band, homozygous for the mutant allele (base T) showed two bands of 100 and 380 bp, heterozygous were characterized by the presence of three bands of 480, 380, and 100 bp.

Organochlorinated Persistent Pollutants Analysis In the subset of women (63 controls and 63 cases) enrolled in the study, serum levels of 11 organochlorinated persistent pollutants were determined by ion-trap mass spectrometry. The PCB congeners were selected for the analysis on the basis of their abundance in human tissues, and of toxicological relevance. They were the six non-dioxin-like congeners (NDLPCBs) currently referred to as indicator (29), namely PCBs 28, 52, 101, 138, 153, and 180, known to account for 50%– 80% total PCBs in human tissues (30), plus NDL-PCB170, and the dioxin-like congeners most abundant in human tissues, that is PCBs 105, 118, 156, and 167. Total PCB serum concentration was calculated as the sum of all of the 11 congeners analyzed. Serum samples (10 mL) were added with a 13C mixture of labeled PCBs (28, 52, 101, 118, 138, 153, 156, 180) and allowed to rest overnight. Samples were added with 15 mL of VOL. 97 NO. 5 / MAY 2012

formic acid:2-propanol (4:1, vol/vol), sonicated, and extracted by manual shaking with N-hexane. After centrifugation, the organic phase was removed and collected. The extraction process was repeated four times, and the N-hexane aliquots were pooled in a centrifuge tube and concentrated. After acidic treatment with concentrated H2SO4, the organic phase was reduced in volume and transferred into a 1-mL vial to undergo instrumental analysis. Instrumental analysis was carried out by ion-trap mass spectrometry (Thermofinnigan Polaris Q) coupled to high resolution gas chromatography and used in the MS–MS mode. The isotope dilution technique was applied throughout. The PCBs are highly lipophilic compounds, known to be strongly associated to serum lipids, therefore concentrations of total cholesterol, phospholipids, and triglycerides were determined by enzymatic methods, as previously described (8). The PCB concentrations were expressed in per gram of serum fat. This choice is in line with the approach established by World Health Organization for the studies of Persistent Organic Pollutants in biological samples, and specifically targeted to monitor possible differences of exposure between groups of individuals.

Statistical Analysis The c2 test was applied to verify that observed and expected genotype frequencies of GSTP1 and GSTA1 loci were in Hardy-Weinberg equilibrium. For GSTM1 and GSTT1, frequency of two genotypes (-/þ and þ/þ) could not be observed, thus deviations from Hardy-Weinberg proportions were not tested. Genotypic distributions among cases and controls were compared by c2 test. A P value %.05 was considered as statistically significant. Crude odds ratio (OR) with 95% confidence intervals (CI) were also estimated for all GST polymorphisms. Based on published data on enzymatic activity or protein expression, GSTM1, GSTT1, and GSTA1 were grouped in high risk and low risk genotypes (GSTM1 -/þ and þ/þ; GSTT1 -/þ and þ/þ; GSTA1 C/T and C/C). Conflicting data on GSTP1 Ile105Val did not allow us to choose a reference genotype a priori. In the sample of case and control women in which exposure was estimated (n ¼ 126), the distribution of each PCB found to be associated with the disease (PCB118, PCB153, PCB138, PCB170, PCB180, and total PCBs) was divided into tertiles. After evaluation of the risk conferred by each stratum, women were grouped at low or medium-high exposure for PCB118, PCB153, PCB138, PCB170, and total PCBs. For exposure to PCB180 women were grouped in low-medium and high serum level. Crude ORs and 95% CIs were estimated for the second and third tertile of each analyte (8). The homogeneity of the association between PCB levels and the disease across GST genotype strata was verified using the heterogeneity test and by introducing a first-order multiplicative term in logistic regression model. Unconditional logistic regression models were used to estimate the magnitude of the association (OR) between the risk factors and the outcome controlling for the effects of all other variables in the model. Multiplicative term 1145

ORIGINAL ARTICLE: ENVIRONMENT AND EPIDEMIOLOGY represents a combination of exposure and potential effect modifier and was added to the logistic model to allow the assessment of interaction. Confidence limits around this estimate were obtained using the coefficient and its related standard error. Age (years), BMI (kilograms per square meter), evidence of relevant weight modification (>10 kg) in the past 5 years (yes vs. no), and smoking habits (ex-smokers vs. nonsmokers; current smokers vs. nonsmokers) were tested as covariates. Age and BMI were evaluated as continuous variables. In addition, a 2  4 table approach was used (31) according to which, beside the effect of each individual factor, any excess risk due to the gene–environment combination was also calculated. Case-only analysis, a powerful strategy to detect gene– environment interactions, in which controls group is used to validate the independence assumption between gene and environmental exposure, was performed. Case-only ORs for the relevant interactions were estimated by unconditional logistic regression adjusted for age. Our sample size (181 cases and 162 controls) had 80% probability to detect a risk equal to 1.7, 1.9, and 1.9, for GSTM1, GSTT1, and GSTA1 high risk homozygous genotypes, respectively (alpha ¼ 0.05, 1-sided; recessive mode of inheritance). The same power was obtained at the GSTP1 locus for a risk of 1.85 conferred by the homozygous genotype (2-sided, unchanged other parameters). The subsample of 63 cases and 63 controls had 80% power to detect a gene–environment interaction of approximately 7.5, 10, 11.5, and 7 between each of the GSTM1, GSTT1, GSTA1, and GSTP1 high risk genotypes and medium-high levels of each PCB. Statistical analysis was performed with STATA software (version 9) and QUANTO (version 1.2.4; http://hydra.usc.edu/gxe/) was used for power calculation. According to the rules of the Italian National Institute for Health this study has been authorized and approved by the Head of the Department of the Units involved in the study and by the President of the Institute (authorization ID number 36006).

RESULTS All participants (N ¼ 343) were of Italian ethnicity. There was no significant difference in age distribution between cases (mean, 33.0  6.5 years) and controls (mean, 30.7  8.3 years). More than 50% of cases and controls were in the age range of 26–35 years; the rest of the individuals were equally distributed (20% each) in the lower and higher age intervals. Sociodemographic, clinical characteristics, and all the information used to obtain data on potential confounders relevant for the exposure part related to the subset analyzed for serum PCB levels and GST loci (N ¼ 126) are described in Supplemental Tables 1 and 2, available online.

Genetic Analysis The GSTM1, GSTT1, GSTP1, and GSTA1 genotype distributions were studied in 181 women with endometriosis and in 162 controls (Supplemental Table 3, available online). Genotypes of GSTP1 and GSTA1 loci were in Hardy-Weinberg equilibrium in case and control subjects. 1146

Women with GSTM1 null genotype have a nonsignificant increased risk of developing endometriosis compared with those carrying at least one allele. The GSTT1 homozygous null genotype confers a significant reduced risk of disease (Supplemental Table 3). The GSTP1Ile/Ile frequency was higher in cases (48.6%) than in controls (42.0%). Women homozygous for the Ile variant appear to have a higher risk of endometriosis (OR ¼ 1.31; 95% CI: 0.8–2.1) than subjects who carry at least one Val allele (Supplemental Table 3, right side). Genotype frequencies of GSTA1 did not differ between cases and controls, therefore no significant association was found with endometriosis (Supplemental Table 3). Genotype distribution of GSTM1, GSTT1, GSTP1, and GSTA1 were similar between the subgroups of women that were also analyzed for PCBs (N ¼ 126) and those who were not (N ¼ 217). In addition, we did not detect significant differences in the GSTs genotype distribution among the four stages of the disease.

PCB Analysis The PCB exposure has been studied on a subset of the enrolled women (63 patients and 63 unaffected women). The PCB congeners 118, 138, 153, 170, and 180 makes up 70% or more of total amount. Therefore they represent the majority of the PCB exposure. Among the 11 analyzed congeners, only medium-high levels of PCB congeners 118, 138, 153, and 170 and high level of PCB180 are associated with a significant increased risk of endometriosis compared with the low ones (OR range, 2.4–3.7) (Table 1). The risk of endometriosis appears to be also significant for total PCBs, obtained as the sum of all the 11 analyzed congeners (OR ¼ 4.6; 95% CI: 1.9–11.0). The cutoff points of the tertiles are indicated in Table 1 (see Table footnotes).

Interaction Between GST Genotype and PCB Exposure The association between levels of serum organochlorine and risk of endometriosis was evaluated after stratification by GST genotypes for the assessment of gene–environment interaction (Table 2 and 3). Women with medium-high levels of PCB118, PCB153, PCB138, PCB170, total PCBs, and GSTM1 null genotype showed a higher (3.560.4 ng/g fat. h Low %208 ng/g fat, medium 209–305 ng/g fat, high >305 ng/g fat. i Chi-square test. Vichi. GST and PCB interaction in endometriosis. Fertil Steril 2012.

To estimate and compare risks conferred by either gene and/or environmental exposures, a 2  4 table was used. Reference category consists of women with the low-risk genotypes and low PCB (or low-medium for PCB180) blood concentrations. As shown in Table 4 (left side), the presence of GSTM1 null genotype did not confer risk per se; medium-high PCB exposure (or high for PCB180) increased the risk of endometriosis in the absence of GSTM1 null genotype (1.410 kg) Yes 8 No 55 Smoking status Nonsmokers 32 Ex-smokers 6 Smokers: 1–9 cigarettes/day 7 10–19 cigarettes/day 9 R20 cigarettes/day 7 Food consumption [times/month (mean  SD)] Milk and dairy products 31.5  16.5 Meat 25.2  10.6 Fish 7.3  5.4 Laparoscopic findings Stage of endometriosis (revised ASRM) I II III IV Ovarian endometrioma Yes No Peritoneal lesions Yes No Type of peritoneal lesion Typical Subtle Both Deep endometriosis Yes No

%

Cases (N [ 63)

%

32.1  6.3

P valuea .09

25.4 54.0 20.6

11 35 17

17.5 55.6 27.0

96.8 3.2 –

59 2 1 21.1  3.0 12.2  1.5

95.2 3.2 1.6

73.0 14.3 12.7

43 12 7

69.4 19.6 11.3

.75

12.7 87.3

7 55

11.3 88.7

.81

52.5 9.8

33 13

52.4 20.6

11.5 14.7 11.5

5 7 5

7.9 11.1 7.9

39.7  24.7 24.3  11.7 7.3  5.4

.48

.60 .04

.48 .03 .66 .42

5 4 34 20

7.9 6.3 54.0 31.8

57 6

90.5 9.5

34 29

54.0 46.0

17 6 9

53.1 18.8 28.1

58 5

92.1 7.9

Note: ASRM ¼ American Society for Reproductive Medicine; BMI ¼ body mass index. a t-test or Chi-square test. Vichi. GST and PCB interaction in endometriosis. Fertil Steril 2012.

1151.e1

VOL. 97 NO. 5 / MAY 2012

Fertility and Sterility®

SUPPLEMENTAL TABLE 2 Serum concentrations (geometric mean [95% CI]; ng/g fat) of PCBs in cases and controls.

PCB118 PCB153 PCB138 PCB180 PCB170 Total PCBs

Controls (N [ 63)

Cases (N [ 63)

P valuea

15.8 (13.5–18.5) 61.9 (50.1–76.4) 32.8 (27.5–39.1) 34.6 (28.1–42.5) 6.7 (5.1–8.8) 209.6 (182.2–241.2)

24.1 (19.5–29.9) 96.3 (82.0–113.0) 48.8 (40.6–58.5) 46.8 (39.3–55.7) 9.3 (7.4–11.6) 301.3 (264.8–342.8)

.002 .001 .002 .03 .07 .0002

Note: CI ¼ confidence interval; PCB ¼ polychlorinated biphenyl. a t-test on log-transformed values. Vichi. GST and PCB interaction in endometriosis. Fertil Steril 2012.

VOL. 97 NO. 5 / MAY 2012

1151.e2

ORIGINAL ARTICLE: ENVIRONMENT AND EPIDEMIOLOGY

SUPPLEMENTAL TABLE 3 Genotypic frequency distributions of glutathione transferase polymorphisms (n [ 343). Gene GSTM1 GSTT1 GSTP1 (rs1695) GSTA1 (rs3957357)

Genotype b

Positive Nullc Positiveb Nullc Ile/Ile Ile/Val Val/Val C/C C/T T/T

Controls % (N [ 162)

Cases % (N [ 181)

OR

47.5 52.5 80.3 19.7 42.0 47.5 10.5 37.0 44.4 18.6

42.5 57.5 89.0 11.0 48.6 40.9 10.5 35.9 49.2 14.9

1 1.22 1 0.50 1 0.74 0.86 1 1.14 0.83

Grouped genotypes

OR

95% CI

.43

Ile/Ile Ile/Val, Val/Val

1.31 1

0.83–2.05

.58

C/C, C/T T/T

1 0.77

0.43–1.36

95% CI

P valuea

0.80–1.88

.35

0.27–0.93

.02

0.47–1.17 0.42–1.79 0.71–1.82 0.44–1.56

Note: CI ¼ confidence interval; OR ¼ odds ratio. a 2 c test. b (/þ and þ/þ). c (/). Vichi. GST and PCB interaction in endometriosis. Fertil Steril 2012.

1151.e3

VOL. 97 NO. 5 / MAY 2012