Clinical vs statistical significance

Abstract

In 1991 a prospective observational study of 48 500 women from the Nurses’ Health Study was published.¹ It documented nearly 340 000 person-years of patient follow-up, and showed that oestrogen supplementation in postmenopausal women was associated with an adjusted relative risk (RR) of 0.56 (95% confidence interval [CI] 0.4–0.8) for major coronary disease and similar risk reduction for fatal cardiovascular disease. The following year a quasi-meta-analysis of 35 epidemiological studies showed a reduction in coronary heart disease, stroke and hip fracture, but a significant increase in breast cancer.² This scare led to a randomised controlled trial (RCT) where patients were randomised to oestrogen and progesterone or placebo, with outcomes of coronary artery disease and invasive breast cancer.³ This trial showed that hormone replacement therapy was associated with an increase in coronary heart disease events, strokes and breast cancer and a reduction in hip fractures. The recommendation was that hormone replacement therapy should not be used as a primary prevention.³ In a 10-year period, there was a turnaround in findings from protection to harm with hormonal replacement therapy in postmenopausal women. How did this happen?