In a new study, published in JAMA, scientists report that among postmenopausal women, hormone therapy was not associated with risk of all-cause, cardiovascular, or cancer mortality during a cumulative follow-up of 18 years.
A roundup accompanied this analysis.
Title, Date of Publication & Journal
Menopausal hormone therapy and long-term all-cause and cause-specific mortality: The Women’s Health Initiative randomized trials’ by Manson et al.
Published: Tuesday 12 September 2017
Study’s main claims – and are they supported by the data
The paper strongly supports the claim that hormone therapy with combined equine oestrogens plus a type of synthetic progesterone (medroxyprogesterone acetate), or combined equine oestrogens alone is not associated with risk of all-cause, cardiovascular, or cancer mortality in postmenopausal women.
The study is a follow-up of two very large randomized controlled trials of a long duration. The conclusions in both the paper and the press release are supported by the evidence and the recommendations for clinical decision making are generally justified.
The trials are randomized, meaning that results are not sensitive to inherent differences between treatment and control groups that might lead to confounding.
The trials have a large sample size and long duration, particularly relative to other randomized controlled trials.
Authors use a concrete objective outcome (all-cause mortality) as well as mortality by a variety of different important causes. All outcomes are predefined, preventing the selecting of significant findings.
Non-parametric statistical tests were used, meaning that fewer assumptions about the data were required for the results of the tests to be valid.
Results from two trials are reported, the results from which are generally very similar.
Authors conducted a sensitivity analysis including only women who took more than 80% of their medication during the first 2 years. This did not lead to any changes in results, though the authors acknowledge there were fewer participants in this analysis to be able to assess the results with the same degree of precision.
There is no explanation for why so few women (4%) continued to use hormone therapy after the trial was completed, and it is unclear how this figure was obtained. This could have been for a variety of reasons. For example, it may mean that there were either limited beneficial effects to continuing hormone therapy beyond 5 to 7 years or there were other unpleasant side effects that were difficult to tolerate (unrelated to mortality). Or perhaps the same hormonal preparations are not easily available in practice
Participants only took hormone therapy for the duration of the study. Whilst this was a long time (5 to 7 years), in practice it may be that women may typically be prescribed hormone therapy for longer periods of time, in which case the results may not be fully generalisable.
As participants were not closely monitored after the completion of the trials, it is unclear if there were any other differences unrelated to previous hormone therapy that may have arisen between treatment and control groups that could have affected mortality after 12 to 18 years follow-up.
A number of limitations were acknowledged by the authors:
Non-parametric statistical test: A type of analysis that does not require assuming that data follow a particular distribution. This therefore makes results from the tests more robust, but often requires a larger sample size to have enough precision to test for statistical differences.
Any specific expertise relevant to studied paper (beyond statistical)?
The reviewer has worked on the National Institute for Heath and Care Excellence guideline for menopause, which included extensive reviews of the evidence for hormone therapy in menopausal women.
* ‘Menopausal Hormone Therapy and Long-term All-Cause and Cause-Specific Mortality: The Women’s Health Initiative Randomized Trials’ by Manson et al. published in JAMA on Tuesday 12 September.
Before The Headlines is a service provided to the SMC by volunteer statisticians: members of the Royal Statistical Society (RSS), Statisticians in the Pharmaceutical Industry (PSI) and experienced statisticians in academia and research. A list of contributors, including affiliations, is available here.