The dramatic decision to halt the Women's Health Initiative (WHI) study 1,2 has once again put epidemiologic evidence under scrutiny. Observational studies had concluded that a regimen of estrogen and progestin increased the risk of breast cancer but reduced the risk of cardiovascular disease. In the randomized clinical trial, both risks were increased.
It is premature to rush to utter certainty on the relation of hormone replacement therapy (HRT) and heart disease. Still, the discrepancy between the results of the trial and the observational studies must be taken seriously. What can we learn?
Clearly, observational studies of treated and untreated groups are subject to self-selection or indication for treatment 3 that can generate uncontrolled differences in risk. Are there signs to alert us to such problems? For example, bias from the “healthy worker effect” is known to attenuate over time. Were there clues in the observational studies of HRT that the effect of a “healthy hormone user”4 attenuated over time? Cardiovascular disease (with its prodromes) is more susceptible than cancer (with none) to bias from the “healthy worker effect.” Might the same be true for bias attributable to treatment effects that are confounded?
We have invited three experts to provide their perspectives. Karin Michels 5 offers the viewpoint of a researcher firmly established in the world of observational studies. Steven Piantadosi 6 looks over the fence from the other side, as a clinical trialist. Alice Whittemore 7 writes as a person with broad methodologic interests. These commentaries raise excellent and thoughtful points.
Along the way, it is good to be reminded that the conclusions of clinical trials can themselves be frustratingly limited. Strictly speaking, the results apply only to the specific group and regimen under study. This can create daunting conundrums for clinical trialists, as Zena Stein and her co-authors 8 discuss in this issue. The tools of epidemiologic thinking become relevent when the perfectionist demands of clinical trials crash against the shoals of real-world conditions.
Epidemiologic thinking might even contribute to interpretation of the WHI. For example, WHI enrolled women who had used HRT previously, but the analysis considers HRT only from the time of randomization. 1 An epidemiologic perspective would suggest incorporating lifetime HRT use in analyses of the clinical trial. Although these analyses would not have all the benefits of randomization, they could provide additional insights on the impact of short-term HRT for control of menopausal symptoms.
Observational studies are here to stay. They will be conducted by necessity, if not always by preference. We may be tempted to defend observational studies, but such protectiveness is not in the best interests of science or public health. As editors, we encourage manuscripts that address the new issues raised by clinical trials. These cannot help but sharpen the peculiar and rigorous qualities of thinking demanded by observational data.
1. Writing Group for the Woman's Health Initiative Investigators. Risks and benefits of estrogen plus progestin in healthy postmenopausal women. Principal results from the Women's Health Initiative randomized controlled trial. JAMA 2002; 288: 321–333.
2. National Institutes of Health, National Heart Lung and Blood Institute. NHLBI stops trial of estrogen plus progestin due to increased breast cancer risk, lack of overall benefit. NIH News Release
9 July 2002. Available at: http://www.nhlbi.nih.gov/new/press/02–07–09.htm
3. Walker AM. Confounding by indication. Epidemiology 1996; 7: 335–336.
4. Sturgeon SR, Schairer C, Brinton LA, Pearson T, Hoover RN. Evidence of a healthy estrogen user survivor effect. Epidemiology 1995; 6: 227–231.
5. Michels KB. Hormone replacement therapy in epidemiologic studies and randomized clinical trials—are we checkmate? Epidemiology 2003; 14: 3–5.
6. Piantadosi S. Larger lessons from the Woman's Health Initiative. Epidemiology 2003; 14: 6–7.
7. Whittemore AS, McGuire V. Observational studies and randomized trials of hormone replacement therapy: what can we learn from them? Epidemiology 2003; 14: 8–10.
8. Stein Z, Myer L, Susser M. The design of prophylactic clinical trials: the case of microbicides for HIV. Epidemiology 2003; 14: 80–83.