We thank Dr Hartzema for his interest in our study. His concerns seem to arise from several misunderstandings, which we will try to clarify. Our combined cohort consists of 4 cohorts initially assembled in the 1970s and followed up for varying time periods. Each cohort is comprised of an exposed group (with medical record documentation) and an appropriate unexposed comparison group.

Systematic information on outcomes and covariates were assessed in the early 1990s and in subsequent follow-ups using identical questionnaires administered to all study subjects along with validation by medical records and pathology review. Ours is probably one of the more robust efforts within the current push toward combining cohorts to address important questions with greater power than is possible with small individual cohorts.

Dr Hartzema also noted some specific concerns. One was the substantially stronger association between diethylstilbestrol (DES) and cervical intraepithelial neoplasia (CIN)-2+ in the Horne cohort compared with the others. This could easily be due to chance (test for overall interaction, P = .13). Also, as noted in the Comment section, this could be related to the younger age and the earlier gestational week at first use in the Horne cohort. Eliminating the Horne cohort from the analysis, the covariate adjusted hazard ratio (HR) for DES and CIN2+ remained elevated (1.68; 95% confidence interval [CI], 1.10–2.58).

Another concern was the potential role of confounding because Dr Hartzema claims that we adjusted only for study cohort and birth year. As stated in the Materials and Methods and Results sections, the HR for DES and CIN2+ with adjustment for cohort and birth year (and age as the time metric) was 1.98 (95% CI, 1.33–2.94); with additional adjustment for frequency of cervical cancer screening, the HR was similar (HR, 1.97; 95% CI, 1.33–2.93), but it rose slightly with further adjustment for education, smoking, age at first sexual intercourse, and number of male sexual partners as a proxy for human papillomavirus infection (HR, 2.10; 95% CI 1.41–3.13).

Finally, Dr Hartzema is concerned about “reanalyzing a data set previously used to draw hypotheses in order to conclude on the veracity of these hypotheses.” First, these cohorts were not the source of the hypothesis of the association of DES exposure with cervical neoplasia. The hypothesis was raised by several clinical studies conducted in the 1970s and early 1980s. Second, the current analysis is not focused on assessing the veracity of the association, which we believe has already been done, but rather on assessing how long the risk persists, particularly as the exposed cohort moves into the postmenopause.

In the context of Dr Hartzema’s concern over hypothesis generation vs testing, he indicates that “the right procedure would be to test this hypothesis using other data sets.” Although, as noted, we believe our combined cohort study is a robust test of a hypothesis raised in other studies, we agree that another such study would be extraordinarily helpful, even more so if it was conducted by investigators independent of prior studies. It would be wonderful if Dr Hartzema or other committed investigators would take this on.