Cyclin E Overexpression Relates to Ovarian Cancer Histology but not to Risk Factors

In Response: We appreciate the opportunity to respond to the concerns raised by Dr. Risch related to our recent publication about the relationship between cyclin E overexpression and epidemiologic risk factors in ovarian cancer (1). The first concern was with the statistical significance of the difference in the number of lifetime ovulatory cycles among the cyclin E–positive and cyclin E–negative ovarian cancer cases. Dr. Risch suggested that we use a 2-degree-of-freedom test to evaluate the significance of this relationship. The second concern related to the inclusion of mucinous cases, which Dr. Risch believes are not associated with reproductive risk factors such as ovulation and oral contraceptive use that predispose to other histologic types of ovarian cancer.

In response to these concerns, we provide Table 1 below, which shows the case-case analysis of lifetime ovulatory cycle tertiles previously reported in Table 4 of the article, but now without mucinous cases. We have also included the P values from the 2-degree-of-freedom test. Removing the 60 (7 cyclin E positive and 53 cyclin E negative) mucinous cases results in the point estimates becoming somewhat stronger, and the P values show strong statistical significance. Addressing the second concern, we repeated the analysis in Table 3 of our article omitting mucinous cancers, and the results were virtually identical to our original findings. These results are given in Table 2. The point estimate for the relationship between lifetime ovulatory cycles among cyclin E–positive cases compared with controls remained the same as in the original analysis [odds ratio, 1.18; 95% confidence interval (95% CI), 1.07–1.30], whereas the point estimate for the association between lifetime ovulatory cycles and the cyclin E–negative cases compared with controls changed slightly from 1.05 (95% CI, 0.98–1.14) to 1.06 (95% CI, 0.97-1.15). The odds ratios and 95% CIs for the relationship between the number of years pregnant and years of oral contraceptive use in the analysis of cyclin E–positive cases and cyclin E–negative cases compared with controls are also similar to the original analyses that included the mucinous cases.

We also want to point out that the P value of 0.11 derived by Dr. Risch for the relationship between oral contraceptive use and cyclin E–positive and cyclin E–negative cases clearly cannot be correct because the 95% confidence limits do not even overlap (Table 3). We do not know how Dr. Risch came up with what he calls a “rough estimate” P value, but when we conducted a case-only analysis of this relationship we found a P value of 0.0004. In a similar analysis of the number of lifetime ovulatory cycles in positive cases compared with negative cases, we calculated a P value of 0.003. In doing this reanalysis, we noticed that in the footnote to Table 3 we incorrectly reported that the sample size of cyclin E–negative cases was 166, when in fact the sample size is 299. This may explain the discrepancy between Dr. Risch's rough estimate P value being >0.05 and the fact that the confidence intervals do not overlap.

We appreciate Dr. Risch's thoughtful comments and the opportunity to provide further analyses, which address these issues.