REJOINDER Authors’ Response, Part I: Observational Studies Analyzed Like Randomized Experiments Best of Both Worlds

W e thank the 3 discussants for their contributions. The Women’ Health Initiative (WHI) investigators found a greater CHD risk in the estrogen plus progestin (ie, combined hormone) therapy arm than in the placebo arm of the trial (hazard ratio HR : 1.24, 95% confidence interval CI : 1.00–1.54). In contrast, the Nurses’ Health Study (NHS) investigators found a lower CHD risk in current users of combined hormone therapy than in never users (HR: 0.68, 95% CI: 0.55–0.83, in their most recent publication). We investigated possible reasons for this discrepancy by reanalyzing the NHS data; we used a novel approach that conceptualizes a follow-up observational study as a sequence of “trials.” The discussants disagree sharply in their assessments of the value of this analytic strategy. Prentice and Hoover are positive, whereas Stampfer finds that our approach combines the limitations of both observational studies and randomized trials, gives biased adherence-adjusted HR estimates, and “adds no new insights on the relation of hormone therapy to CHD.” He criticizes our approach for its complexity, its need for additional assumptions, and its “black box” nature, and argues for the continued use of the conventional methods routinely employed in NHS publications, owing to their transparency and validity. We now address each of these criticisms.