Women’s Health Initiative View of Estrogen Avoidance and All-Cause Mortality

In a recent article, Sarrel et al.¹ assert that estrogen avoidance since 2002 has caused tens of thousands of premature deaths among posthysterectomy women aged 50 to 59 years in the United States. They fault Women’s Health Initiative (WHI) investigators for inadequate efforts to communicate the benefits of unopposed estrogen and to contrast (unopposed) estrogen findings from those for estrogen plus progestin in reporting on the WHI randomized controlled trials.^2–5

We, as WHI Investigators, disagree with the calculations and oversimplified conclusions of these authors.¹ The claimed mortality toll from estrogen avoidance derives from an all-cause mortality hazard ratio (HR; 95% confidence interval [CI]) of 0.73 (0.53, 1.00),³ based on 65 and 89 deaths in the active and placebo groups among women who were aged 50 to 59 years when enrolled in the WHI unopposed estrogen trial. The upper confidence limit of 1.00 corresponds to no estrogen effect on all-cause mortality and, in the terminology of Sarrel et al.,¹ to an excess mortality of zero from estrogen avoidance. Hence their lower confidence limit of 18 601 projected premature deaths from estrogen avoidance is simply wrong.

Also, the above WHI “nominal” confidence interval would be widened to include projected estrogen avoidance all-cause mortality reductions if adjusted for multiple comparisons. Substantial multiple testing issues attend this confidence interval because of subset analyses for various participant characteristics and various clinical outcomes (all-cause mortality was not a primary outcome in the WHI trials). In fact, the estrogen trial is not entirely convincing concerning any dependence of the all-cause mortality hazard ratio on age: the nominal age-interaction P value was 0.04, rising to 0.13 upon adjustment for participant reconsent characteristics.³ Without age restriction, the all-cause mortality HR (95% CI) was a null 1.02 (0.91, 1.15).

We agree with Sarrel et al.¹ that WHI trial data are complex and require a nuanced presentation. A development more nuanced and comprehensive than taking a specific subset hazard ratio and confidence interval at face value would be needed to project health risks and benefits from any decline in unopposed estrogen prescriptions following the WHI trials. Finally, concerning efforts to communicate trial findings, in addition to more than 110 hormone trial publications to date, a comprehensive analysis of data from the WHI trials with side-by-side comparisons of the benefits and risks of the two preparations (including stroke and venous thromboembolism increases) was recently published.⁶