CRD summary

The review found that among younger postmenopausal women, mortality appears to be lower in those taking hormone therapy. The authors’ conclusions appear to reflect the randomised evidence presented, but may require some caution in interpretation in view of limited reporting of study characteristics.

Authors' objectives

To assess the effect of hormone therapy on mortality in younger postmenopausal women.

Searching

MEDLINE, EMBASE, CINAHL and The Cochrane Library were searched without language restrictions for studies published from 1966 to January 2008. Selected journals and reference lists of retrieved articles were checked.

Study selection

Randomised controlled trials (RCTs) and prospective cohort studies of hormone therapy (oral or transdermal oestrogen or oestrogen/progestin) in postmenopausal women (defined as 12 consecutive months of amenorrhoea) were eligible for inclusion. RCTs were required to enrol women with a mean age of under 60 years, use placebo or no treatment for controls, to be of at least six months’ duration and report at least one death. Cohort studies were required to report mortality in postmenopausal women and to adjust for standard confounding factors, using multivariate analysis.

Mean age of women in the included RCTs was 54.5 years. Participants received various forms of hormone therapy (such as conjugated equine oestrogen, oral esterified oestrogen, transdermal oestrogen; with or without progestin; continuous or cyclic). Most RCTs did not report mortality as a primary or secondary outcome. Mean duration of study follow-up was 5.1 years (range one to 6.8 years) in RCTs and 13.8 years (range six to 22 years) in cohort studies. Two reviewers independently selected studies for inclusion.

Assessment of study quality

The authors reported on dropout rates and blinding in the RCTs. They did not state that they assessed other aspects of study validity.

Data extraction

For the RCTs, relative risks (RRs) and 95% credible intervals (CrIs) were calculated from the number of events in each group. For cohort studies, adjusted relative risks or hazard ratios were extracted using the longest-term data available.

Two reviewers independently extracted the data. Disagreements were resolved by consensus. Further information was sought from authors of primary studies.

Methods of synthesis

Studies were combined using a hierarchical Bayesian random-effects model using four different prior distributions (informative, non-informative, sceptical and erroneous (highly sceptical)) to generate pooled relative risks and 95% credible intervals. Analysis was stratified by study design (RCT or cohort study). Posterior probability of a mortality benefit with hormone therapy was calculated for each type of distribution. Methods used to assess statistical heterogeneity were not reported. Funnel plots were used to assess for publication bias.

Results of the review

Nineteen RCTs (n=16,283, range 26 to 8,832) and eight cohort studies (n=212,171) were included in the review. Twelve RCTs were double-blinded, one single-blinded and six open label; dropout rates ranged from 4% to 41% across groups.

Using a non-informative prior distribution, risk of mortality in the RCTs was significantly lower in the intervention group compared to placebo (RR 0.73, 95% CrI 0.52 to 0.96); posterior probability of a mortality benefit from hormone therapy was 0.985.

When the cohort data were pooled and combined with the RCT data, resultant relative risk using an informative prior distribution was 0.72 (95% CrI 0.62, 0.82); posterior probability of a mortality benefit from hormone therapy was 1.00.

There was wide variation between studies (for example in trial size, medicines used, methods of administration and populations), but little inter-study heterogeneity was observed in the results. Funnel plots did not show evidence of publication bias.

Results for analyses based on different prior distributions were reported in the review.

Authors' conclusions

Among younger postmenopausal women, mortality appeared to be lower in those who took hormone therapy.

CRD commentary

Objectives and inclusion criteria of the review were clear in most respects; it was unclear whether the restriction to younger participants applied to RCTs also applied to observational studies. Relevant sources were searched for studies without language restrictions. Search terms were not reported. The restriction by publication status meant that the review was open to publication bias. However, formal testing showed no evidence of publication bias. Steps were taken to minimise risks of reviewer bias and error by having more than one reviewer undertake study selection and data extraction. It was unclear whether study validity was systematically assessed. Few details were provided about the included RCTs (such as randomisation methods, treatment compliance, clinical characteristics of participants). No information was provided about characteristics of observational studies (such as participant age, selection procedure, sample size). This made it difficult to assess reliability and applicability of the review findings. Statistical methods used to combine the RCTs appeared suitable. The authors reported that although there was significant variability in study methods, there was little heterogeneity seen in the results. It was unclear whether addition of data from cohort studies was appropriate in view of the high susceptibility to bias of observational data and lack of evidence that these data were applicable to younger women. Thus the somewhat imprecise effect estimate derived from the RCTs appears more reliable than estimates that included observational data. The authors’ conclusions appear to reflect the randomised evidence presented, but may require some caution in interpretation in view of limited reporting of study characteristics.

Implications of the review for practice and research

Practice: The authors stated that the reduction in mortality associated with hormone therapy among younger women should be interpreted taking into account potential benefits and harms of such therapy.

Research: The authors did not state any implications for research.

Funding

No research, institutional or pharmaceutical funding. Funding for authors was from salary support and Podell Emeriti award.

Bibliographic details

Salpeter SR, Cheng J, Thabane L, Buckley NS, Salpeter EE. Bayesian meta-analysis of hormone therapy and mortality in younger postmenopausal women. American Journal of Medicine 2009; 122(11): 1016-1022.e1. [PubMed: 19854329]

Original Paper URL

http://www.amjmed.com/article/S0002-9343(09)00666-4/abstract

Indexing Status

Subject indexing assigned by NLM

MeSH

Age Factors; Bayes Theorem; Cause of Death /trends; Female; Hormone Replacement Therapy /methods; Humans; Middle Aged; Postmenopause /drug effects; Survival Rate /trends; United States /epidemiology

AccessionNumber

12009110060

Database entry date

12/05/2010

Record Status

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.