CYP17 genotype and ovarian cancer: a null case-control study.

Introduction Long-term differences in steroid hormone levels between women likely contribute to the interindividual variation in ovarian cancer risk, although the precise biological mechanisms are unclear (1). The P450c17a (CYP17) gene codes for 17ahydroxylase and 17,20 lyase which catalyze the rate-limiting step in androgen biosynthesis, cleaving the C21 steroids to the C19 steroids, androstenedione and dihydroepiandrosterone. The 59 untranslated region has a single bp polymorphism (T3C transition) that creates an Sp-1 type (CCACC box) promoter site (2). The presence of the variant A2 allele has been associated with elevated transcription of progesterone and estradiol in premenopausal women (3), which might modulate the release of pituitary gonadotropin and increase the risk of ovarian cancer. The A2 allele has also been associated with polycystic ovarian syndrome, a condition resulting from high androgen levels (4). Polycystic ovarian syndrome was a significant risk factor for ovarian cancer in the Cancer and Steroid Hormone Study (5), and preand postmenopausal ovarian cancer cases had significantly higher prediagnostic levels of androstenedione and dehydroepiandrosterone than nested controls in the Washington County cohort (6). The association of the CYP17 polymorphism with the risk of breast cancer, another hormoneassociated cancer, was shown in a multiethnic cohort study conducted in Hawaii and Los Angeles (2), although results from other studies have been inconsistent (7, 8). A recent report from the Nurses’ Health Cohort did not find an association of breast cancer with the CYP17 gene (7), but the analysis did find an inverse association of late age at menarche with breast cancer that was absent among women with the A2 allele, suggesting that CYP17 influences early ovulatory events and perhaps the risk of ovarian cancer. Blood samples from a case-control study in Hawaii were used to test the hypothesis that women with the CYP17 variant A2 allele are at increased risk of ovarian cancer. Materials and Methods Eligible cases for this population-based, case-control study in Hawaii comprised all patients with histologically confirmed, primary, epithelial ovarian cancer diagnosed between July 1, 1993, and June 30, 1999, in any of the major hospital centers on Oahu. Eligible women were 18 to 84 years of age and were residents of Oahu. Interview information was obtained from 218 (75%) of 291 ovarian cancer cases eligible for participation in the study. The control pool consisted of population-based lists of female Oahu residents who were interviewed by the Health Surveillance Program of the Hawaii Department of Health. This source was supplemented with women 65 years of age and older who were Health Care Financing Administration participants on Oahu. Potential controls were randomly selected from the pool so that the ethnic (e.g., Chinese) and 5-year age-group distribution would match that of the case group with a 1:1 ratio. Four hundred and sixteen women meeting these eligibility criteria were contacted to participate in the study. Interviews were obtained for 284 (68%) of these women, with 132 (32%) eligible women refusing to participate. We were able to draw blood from 146 (67%) of the interviewed cases and 192 (68%) of the interviewed controls. We selected 129 cases with complete questionnaire (e.g., menstrual regularity) and tumor registry (e.g., histology) information to be included in the genotyping analysis. One hundred and forty-four controls were also selected to match the age and ethnicity of the cases. Laboratory personnel were blinded to the case-control status of the subjects. DNA was purified from peripheral blood leukocytes by SDS/proteinase K treatment and phenol/chloroform extraction. Genotyping for the CYP17 A2 polymorphism was evaluated as described by Fiegelson et al. (2). Unconditional multiple logistic regression models were used to estimate the association (ORs and 95% CIs) of each genotype of interest with case-control status by creating binary indicator variables representing the levels of the exposure. Adjustment variables included age (as a continuous variable), ethnicity by indicator variables (Caucasian, Asian, other), education (,13 years, 13–14 years, $15 years), pregnancy history (ever versus never), oral contraceptive pill use (ever versus never), and history of tubal ligation (yes versus no). Gene dosage effects were modeled by assigning the value 1, 2, or 3 to a genotype trend variable according to the subject’s number of variant alleles (zero, one, or two variant alleles, respectively). Logistic regression was used to explore gene-environment interactions by modeling each level of interaction between the pairs of variables using subjects who had A1/A1 genotypes and who were “unexposed” as the reference category. The likelihood ratio test was used to compare this interaction model with one containing main-effect terms only.