Mortality Associated with Hormone Replacement Therapy


  • I have no conflict of interest. I work in private practice as a physician and medical gynecologist. My PubMed publications are under grant ec. I have been publishing papers on progesterones and estrogens since 1962.

To the Editor:—Salpeter et al.'s meta-analysis included 30 randomized controlled trials of hormone replacement therapy (HRT) with at least 1 death and claimed a reduction in total mortality in younger women.1 Not a single one of these studies had found a statistically significant effect on total mortality. Of the 17 trials listed for women with a mean age under 60, the largest double-blind study, PEPI 1995, had 701 HRT-treated women and a mortality odds ratio of 1.75 (0.09 to 34.01), but was allocated only 2.3% of the weight in the meta-analysis. An overall beneficial result was achieved by allocating 4 trials, containing 685 treated women, 72.2% of the total weighting, because they had most of the deaths. These studies were either open-label design or contained only hospitalized or ovarian cancer patients. Guidozzi's 1999 study involved 62 HRT-treated women with ovarian cancer. More than half of both HRT-treated and control women died during the follow-up period; therefore, this small study was allocated 41% of the weighting. Several studies have found that estrogen-HRT increases the risk of fatal ovarian cancer.

If only randomized double-blind trials containing more than 100 previously healthy, HRT-treated women are included in the meta-analyses, a different result can be obtained. The 5 trials of younger women which fulfill these criteria are PEPI 1995, Angerer 2000, Arrenbrecht 2002, Giske 2002, and Watts 2000, totaling 1,450 HRT-treated women and 479 controls. No deaths occurred in the controls but 7 deaths occurred in the HRT-treated women.

Most studies have serious problems because few women have never taken hormones and few studies give the numbers for this only valid control group.2,3 Most control volunteers have used progesterones or estrogens at some time, either for contraception or menopause, and the exact duration of exogenous hormone exposures, either past or current, is also often missing. Follow-up duration is not the same as current use duration. Thickened endometrial arterioles and dilated sinusoids match generalized vascular side effects, and clotting factors alter, during therapy. Therefore, vascular deaths would be expected to be increased in current users. However, extra cancer deaths would still be occurring for decades because breast cancer is still the main cause of death in women up to 20 years after it has been diagnosed.