The distribution of period of entry into the study and age at entry for the 54
362 women in the cohort has been described previously.28
The median year for entry was 1989, the median age at first evaluation of infertility was 30 years, and the median age at the end of follow-up was 47 years. The women were followed up for a median of 16.0 years (range 0.0-42.6 years), with 25% followed for more than 23 years. The 54
362 women contributed 957
454 person years of observation. The median time from entry into the cohort and diagnosis of cancer was 14.5 years (range 0.02-34.2 years), and the age of the women at the time of diagnosis ranged from 25 to 68 years (median 46 years). Fertility drugs were used by 77 of the 156 (49%) women with ovarian cancer and 615 of the 1241 (50%) subcohort members. Clomifene was the most commonly used fertility drug, taken by 58 cases (37%) and 415 subcohort members (33%), followed by human chorionic gonadotrophins (31% and 33%), gonadotrophins (17% and 15%), and gonadotrophin releasing hormone (10% and 9%).
Maternal age at birth of first and last child did not significantly affect the overall risk of ovarian cancer. In contrast, parous women had a notably lower risk than nulliparous women (rate ratio 0.45, 95% confidence interval 0.32 to 0.63), which decreased with number of births (rate ratio 0.68, 0.47 to 0.99; table 1). Furthermore, the overall risk of ovarian cancer among women in the cohort was not significantly affected by either use of oral contraceptives (compared with ever use, the rate ratio for never use 0.46, 0.20 to 1.05) or causes of infertility (compared with anovulation, the rate ratio for tubal disease was 1.47, 0.80 to 2.68, for endometriosis was 0.92, 0.26 to 3.31, for uterine disorder was 0.65, 0.14 to 2.93, and for male factor was 0.57, 0.27 to 1.24).
Table 1 Rate ratios for ovarian cancer according to reproductive factors
After adjustment for parity and number of additional births, the overall risk of ovarian cancer was not significantly affected by use of any fertility drug (rate ratio 1.03, 0.73 to 1.47; data not shown), gonadotrophins (0.83, 0.50 to 1.37), clomifene (1.14, 0.79 to 1.64), human chorionic gonadotrophins (0.89, 0.62 to 1.29), or gonadotrophin releasing hormone (0.80, 0.42 to 1.51; table 2). For all groups of fertility drugs the risk of ovarian cancer was not substantially different according to number of cycles of use or years since first use (table 2). For all four types of fertility drug, the overall risk of ovarian cancer was not noticeably affected by parity, and none of the interaction terms were statistically significant (table 3).
Table 2 Rate ratios for ovarian cancer according to use of fertility drugs
Table 3 Rate ratios for ovarian cancer in nulliparous and parous women according to use of fertility drugs
As a substantial proportion of the women in the cohort had used more than one type of infertility drug, the main effect of each type of drug was assessed. The general results were not changed when the risk of ovarian cancer was assessed only in those women who had used one type of fertility drug compared with never users (data not shown). When the rate ratios for the most prevalent combinations of fertility drugs were calculated, none of the five most commonly used combinations significantly affected risk: clomifene plus human chorionic gonadotrophins (rate ratio 1.13, 95% confidence interval 0.71 to 1.80), clomifene plus gonadotrophins plus human chorionic gonadotrophins (0.60, 0.25 to 1.42), gonadotrophins plus human chorionic gonadotrophins plus gonadotrophin releasing hormone (1.02, 0.34 to 3.03), clomifene plus gonadotrophins plus human chorionic gonadotrophins plus gonadotrophin releasing hormone (0.98, 0.44 to 2.18), and clomifene plus gonadotrophins (2.20, 0.49 to 9.86).
Table 4 presents the risks of different histological types of ovarian cancer. Of the 156 epithelial tumours, serous tumours constituted the majority (58%), followed by endometrioid tumours (15%), other tumours (15%), mucinous tumours (8%), and clear cell tumours (5%). The risk of serous cancers was significantly increased after use of clomifene (1.67, 1.07 to 2.61). The increase in risk was found mainly among women followed for 15 years or more after first use of clomifene when compared with never use (<5 years 1.22, 0.45 to 3.34; 5-9 years 1.76, 0.87 to 3.59, 10-14 years 1.23, 0.58 to 2.59, ≥15 years 2.17, 1.15 to 4.08). The risk of serous cancer was not, however, related to the number of cycles using clomifene, and the risk did not differ according to parity (data not shown). A significantly decreased risk for other types of ovarian cancers was observed after use of human chorionic gonadotrophins (0.24, 0.08 to 0.79), whereas the risks for mucinous, endometrioid, and clear cell tumours were not significantly affected by use of any of the four groups of fertility drug.
Table 4 Rate ratios for histological subgroups of ovarian cancer according to use of fertility drugs
To allow a minimum latency period between first use of fertility drugs and development of ovarian cancer, the few women with a diagnosis of ovarian cancer (n=2) within one year after first use of fertility drugs were excluded from the analyses. Virtually no changes in the risk estimates were observed (data not shown), however, and therefore the results were not noticeably biased by a potential fertility drug induced acceleration of the growth of an existent ovarian tumour.
Finally, to evaluate the effects of various causes of infertility (anovulation, tubal disease, endometriosis, uterine disorder, cervical disorder, and male factor) and any use of oral contraceptives as potential confounders, all the analyses were adjusted for these two variables in the subsets of women for whom information was available: causes of infertility, 45% of cases and 51% of subcohort members, and any use of oral contraceptives, 33% of cases and 55% of subcohort members. None of the risk estimates changed noticeably after adjustment for these potential confounders (data not shown).