Efficacy Endpoints: Mortality, breast cancer mortality, breast cancer diagnoses
Harm Endpoints: Death, due to breast cancer, diagnosis of breast cancer
Narrative: Hormone replacement therapy (HRT) remains a controversial intervention with complex effects. The study targeted here for appraisal and translation is a follow-up analysis examining participants in the Women"s Health Initiative study originally published in 2002.1 The original study was the largest and best of its kind and surprised many with results suggesting that coronary heart disease was not reduced among those taking HRT, but indeed may have been slightly increased. This was a stark contrast to both the conventional wisdom and the results of many observational (cohort) studies examining the issue. In addition, the study was stopped early due to a secondary finding of increased breast cancer diagnoses among those taking HRT. In this analysis the authors extend the analysis of breast cancer outcomes by an additional 7 years and report on not just breast cancers, but also deaths due to these breast cancers among women who had been enrolled in the original study.
The result suggests an increase in the chances of death due to breast cancer, although the differences are small for individuals. Taken at face value the report indicates that more than 99.9% of women were unaffected in terms of breast cancer by HRT for each year of therapy, and 99% were unaffected after 11 years cumulatively. Perhaps most importantly, one in every 5000 women taking HRT per year lost her life to breast cancer because of the HRT. These numbers may be contextualized for women and physicians attempting to make a decision: First, based on this study, roughly 1 in every 7 additional breast cancers due to HRT would result in death. Second, based on the earlier report, it is also true that fractures related to osteoporosis and colorectal cancers are favorably affected. For every 8 cases of breast cancer caused by HRT there would also be 6 colorectal cancers prevented, 5 hip fractures prevented, 6 back fractures prevented, and hot flashes would be reduced by approximately 75%.
Caveats: The topic is complex and difficult to study even in randomized trials, and an individual"s baseline risk of cancer (due to genetics, lifestyle, etc.) are clearly much more important determinants than any pill. Therefore the numbers reported here and from any other study on the topic, should be tailored whenever possible to an individual"s baseline risk. This would offer a more individualized estimate of true risk.
An important additional note is that this analysis is a secondary analysis of a trial for which breast cancer was a secondary endpoint. Analyzing subjects years after they have left their original study groups means that in most cases they are no longer taking the original treatment or placebo and are now subject to risks and confounders that make cause-effect conclusions unstable and prone to error. In addition, as the authors point out, the original study was designed to detect coronary events (heart attacks, cardiovascular deaths, etc.) and therefore the study was designed to answer a different question than the question asked here. This limits, though does not invalidate, the importance of these findings. An ideal study would be designed to seek and report breast cancer outcomes primarily rather than secondarily. Despite these caveats this is certainly the best available data on the topic and should be considered a robust and reasonable best estimate of the impact of HRT on breast cancer.
Author: David Newman, MD
Published/Updated: October 20, 2010
The title bar is color-coded with our overall recommendation.
If you have suggestions, requests, or questions about a particular NNT review, please send us a message and we’ll try to address it as soon as possible.