ABSTRACT
We illustrate how observational studies have inherent biases that can be eliminated with randomized controlled trials (RCTs). This lesson is driven home through the hormone replacement therapy controversy. Observational studies suggested that hormones protect women against cardiovascular events, and that loss of hormones after menopause removes that protection. A very large, well-designed observational study (The Nurse's Health Study) supported that conclusion, but a large randomized trial showed that women randomized to hormone replacement had a higher risk of cardiovascular events. We discuss what could account for the difference in conclusions of the two studies. We also discuss and present an example of the more common scenario in which results of observational studies and randomized controlled trials are consistent. In that example, observational studies and the large Strategic Timing of Antiretroviral Therapy (START) trial showed that patients with HIV should begin antiretroviral treatment well before their CD4 T-cell count drops to low levels.
