In studies where people are observed until the event occurs (or not), it is insufficient merely to record the proportion of the population that becomes ill. It is also necessary to take into account the time during which the participants were at risk of developing the disease. By measuring the length of time contributed by all persons during the period they were in the population and events were counted, we can calculate incidence rates. In our cohort, 106,074 person-years were associated with HRT use, during which 372 cases of invasive breast cancer occurred; 209,012 person-years were associated with no HRT use, during which 576 cases occurred. This leads to a crude incidence rate ratio of 1.27 [i.e., (372/106,074)/(576/209,012)].
Regarding confounding, we performed careful adjustments (a further stratification on year of birth did not alter our results). Of course residual confounding may still remain, but the high consistency between the randomised controlled data and the observational data with regard to the relation between HRT and breast cancer is reassuring and would suggest that there is little confounding in these observational studies.
The BMI reported in Table I is derived from the questionnaire returned the closest to the date of menopause (i.e., measured at a mean age of 52 years). A validation study of self-reported values for anthropometric measurements was undertaken on a small sample of the E3N participants, showing a very high correlation coefficient between self- and external BMI measurements.1 Self-reported measurements for BMI may thus be considered accurate. Women in our cohort all belong to a national health insurance plan that covered mostly teachers, have a relatively high level of education and are health conscious, which could explain why they are particularly lean.
Regarding progestins, indications for their use before menopause were not specifically recorded in our study; however, compared to women who did not use progestins during premenopause or used them for less than 5 years, women who used progestins for more than 5 years were more likely to have suffered from benign breast, ovarian or uterine diseases. Of concern is the possibility that women who have been prescribed progestins during a long premenopausal period may be more likely to develop breast cancer than others (because of benign breast diseases or other hormonal imbalance) and are also more likely to subsequently use HRT (perhaps particularly those containing synthetic progestins). However, the resulting confounding bias would have been minimised because our models were adjusted for premenopausal progestins use. We also performed a sensitivity analysis in which we excluded all these women who had used progestins before menopause. This analysis, although less powered, led to the same conclusions as those reached through our main paper.
Finally, nomegestrel (in France acetate de nomegestrol) has been marketed for a long time and is widely used both alone in premenopausal women and combined with estrogens after menopause.