The American Journal of Medicine
Volume 122, Issue 11 , Pages 1016-1022.e1, November 2009

Bayesian Meta-analysis of Hormone Therapy and Mortality in Younger Postmenopausal Women

  • Shelley R. Salpeter, MD

      Affiliations

    • Santa Clara Valley Medical Center, San Jose, Calif
    • Stanford University, Palo Alto, Calif
    • Corresponding Author InformationRequests for reprints should be addressed to Shelley R. Salpeter, MD, Santa Clara Valley Medical Center, 751 S. Bascom Ave, San Jose, CA 95128
    • ,
  • Ji Cheng, MSc

      Affiliations

    • McMaster University, Hamilton, Ontario, Canada
    • ,
  • Lehana Thabane, PhD

      Affiliations

    • McMaster University, Hamilton, Ontario, Canada
    • ,
  • Nicholas S. Buckley

      Affiliations

    • California Institute of Technology, Pasadena
    • ,
  • Edwin E. Salpeter, PhD (Posthumous)

      Affiliations

    • Cornell University, Ithaca, NY

Article Outline

Abstract 

Background

There is uncertainty over the risks and benefits of hormone therapy. We performed a Bayesian meta-analysis to evaluate the effect of hormone therapy on total mortality in younger postmenopausal women. This analysis synthesizes evidence from different sources, taking into account varying views on the issue.

Methods

A comprehensive search from 1966 through January 2008 identified randomized controlled trials of at least 6 month's duration that evaluated hormone therapy in women with mean age <60 years and reported at least one death, and prospective observational cohort studies that evaluated the relative risk of mortality associated with hormone therapy after adjustment for confounding variables.

Results

The results were synthesized using a hierarchical random-effects Bayesian meta-analysis. The pooled results from 19 randomized trials, with 16,000 women (mean age 55 years) followed for 83,000 patient-years, showed a mortality relative risk of 0.73 (95% credible interval 0.52-0.96). When data from 8 observational studies were added to the analysis, the resultant relative risk was 0.72 (credible interval 0.62-0.82). The posterior probability that hormone therapy reduces total mortality in younger women is almost 1.

Conclusions

The synthesis of data using Bayesian meta-analysis indicates a reduction in mortality in younger postmenopausal women taking hormone therapy compared with no treatment. This finding should be interpreted taking into account the potential benefits and harms of hormone therapy.

Keywords: Bayesian meta-analysis, Hormone therapy, Menopause, Mortality

 

Hormone therapy to treat menopausal estrogen deficiency has been in widespread use for over 60 years.1 Long-term treatment was assumed to prevent the atherosclerosis, osteoporosis, and increase in mortality seen following the menopausal transition.1, 2, 3 Since 1983, several observational studies consistently showed that hormone users, many of whom started treatment shortly after menopause, had a significant reduction in total mortality compared with nonusers, even after adjustment for confounding factors.4, 5, 6, 7, 8 The available evidence supported the routine use of hormone therapy to increase longevity in postmenopausal women.9

Clinical Significance

 


•The synthesis of pooled data indicates that hormone therapy reduces total mortality by 25% in younger postmenopausal women.

•A similar reduction in mortality is seen with randomized trials and observational studies.

•The probability that hormone therapy reduces total mortality in younger women is almost 1.

•This finding of reduced mortality should be interpreted taking into account the potential benefits and harms of hormone therapy.

The publication of the Women's Health Initiative (WHI) in 2002 seemed to contradict these assumptions.10 For women of mean age 63 years, estrogen-progestin treatment increased the risk for composite outcomes, termed the global index, by 13% compared with placebo, without increasing mortality. In the confusion that ensued after this publication, it was possible to make 2 erroneous assumptions: that hormone therapy had similar effects in younger and older women, and that the increased global index translated into an increased risk for death. In 2004, a meta-analysis of randomized trials showed that hormone therapy reduced total mortality by 40% in trials of younger women but not in older women.11 In the wake of the WHI trial, this reduction in mortality was considered to be “implausible” and “difficult to reconcile.”12 In 2006, another meta-analysis of randomized trials found a 32% reduction in coronary heart disease events in younger women.13 It was not until 2007 that age-specific mortality data from both arms of the WHI trial were provided, which showed a 30% mortality reduction in women under 60 years of age.14 A recent cost-effectiveness analysis found that hormone therapy given to younger postmenopausal women for 5-30 years results in a small increase in life expectancy and a substantial increase in quality-adjusted life-years.15

To make sense of the accumulated evidence, we performed a Bayesian meta-analysis of hormone therapy and mortality in younger postmenopausal women, which serves to update the previous meta-analysis11 by including age-specific data from the WHI trials. We focused on mortality because it is the clinical outcome that balances the benefits and harms of treatment, and because conflicting conclusions have emerged about its risk. We believe that a Bayesian meta-analysis is appropriate because it allows us to directly incorporate previous assessments from observational studies into the pooled trial data, and can determine the probability of the hypothesis that hormone therapy reduces mortality in younger women. In addition, Bayesian analysis allows us to quantify the effect that prior beliefs—such as those coming from incorrect assumptions made after the WHI trial—can have on our understanding of the evidence.

Back to Article Outline

Methods 

Study Design 

The MEDLINE, EMBASE, CINAHL, and Cochrane databases were searched comprehensively to identify studies in any language published between 1966 and January 2008 that evaluated hormone therapy, defined as the oral or transdermal use of estrogen or estrogen-progestin therapy, in postmenopausal women. Women were considered to be postmenopausal after 12 consecutive months of amenorrhea. The search was augmented by scanning selected journals and references of identified articles.

Two investigators independently evaluated studies for inclusion. Observational studies were included if they were prospective cohort studies of postmenopausal women that evaluated the relative risk of mortality associated with hormone use and used multivariate analysis to adjust for standard confounding factors. Randomized trials were included if they evaluated women with mean age of <60 years at start of trial, compared hormone therapy with placebo or no treatment for at least 6 month's duration, and reported at least one death. Attempts were made to contact the investigators to obtain information concerning deaths during the trial.

Two independent reviewers extracted data, reconciling differences by consensus. For observational studies, the outcome measured was adjusted relative risk or hazard ratio for total mortality. For studies in which more than one multivariate analysis was reported, the estimate for long-term use was chosen, if available. For randomized trials, the outcome measured was total deaths in the treatment and control groups.

Statistical Analysis 

To capture variability from all sources, to make the probability statement for the treatment effect and to incorporate the prior beliefs and external information, we synthesized the results from the randomized controlled trials using a hierarchical Bayesian random-effects model.16, 17, 18, 19, 20 Details of the Bayesian analysis can be found in the Appendix (available online).

To synthesize trial data in conjunction with observational data, we applied 3 different prior distributions to the model: a “non-informative” prior distribution,21 an “informative” prior distribution generated from the observational studies,22 and a “sceptical” prior distribution using observational data.17, 23, 24 In addition, to assess the effect of incorrect assumptions, we generated an “erroneous” prior distribution using the global index from the WHI trial to approximate an increased relative risk of death in younger women when, in fact, no increase in mortality was seen.10

For the “non-informative” prior distribution, we set a relative risk of 1.0 with a large variance, so that the pooled trial data dominated the posterior distribution. This result is similar to that obtained from a traditional non-Bayesian meta-analysis. For the “informative” prior distribution, we applied the same random-effects model to the pooled observational studies; randomized trial data were then added via the Bayes rule to produce posterior distributions. The “sceptical” prior distribution used the assumption that most clinically important interventions reduce the relative risk of major outcomes by 10%-20%.24 We assigned a highly skeptical prior distribution to our model that allowed only a 5% chance to observe a large benefit, such as the 30% risk reduction taken from observational studies.

The treatment effects were obtained from the posterior distributions of the Bayesian analysis and are reported as relative risk with 95% associated credible interval (CrI), which is a Bayesian analog of the 95% confidence interval from traditional meta-analyses. The results are presented in a forest plot, which displays the relative risk for both the individual trials and also the pooled results. The posterior probability of a mortality benefit with hormone therapy, that is, a relative risk of <1, also was reported. The analysis was performed using the software WinBUGS 1.4 (MRC Biostatistics Unit, Cambridge, UK), which generates inferences using the Gibbs sampler (Appendix [available online]).

Role of Funding Source 

Funding for this analysis came from salary support for S. Salpeter, J. Cheng, and L. Thabane and from a Podell Emeriti Award for E. Salpeter. The institutions had no role in the design, conduct, or reporting of the study. No sponsorship from the institutions or the pharmaceutical industry was provided to conduct this analysis.

Back to Article Outline

Results 

The database search identified approximately 4000 articles, of which 398 were potentially relevant studies. Of these, 8 observational studies5, 6, 7, 8, 25, 26, 27, 28 and 19 randomized controlled trials10, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46 met the inclusion criteria (Table 1 [available online], Figure 1). A previous meta-analysis provided information on 17 of the randomized trials.11 A subsequent publication14 provided age-specific mortality data for the 2 WHI trials.10, 46 In the observational cohorts, a total of 212,171 women were followed for a mean of 13.8 years (range, 6-22 years). In the randomized trials, there were 8689 women in the treatment groups and 7594 women in the control groups, with a mean age of 54.5 years followed for 5.1 years (range, 1-6.8 years). Interventions included conjugated equine estrogen, oral esterified estrogen, or transdermal estrogen, alone or in combination with a progestin, given on a continuous or cyclic basis.

Table 1. Characteristics of Included Randomized Trials
Study⁎, Year, ReferenceDesign, DurationNumber (n) in Treatment ControlDropout in Treatment ControlMean Age (Years) in Treatment ControlInterventionOutcomesComments
Angerer, 200029Double blind, 1 year21528%59.0Estradiol plus gestodone vs placeboCarotid artery dispensability
6629%59.5
Arrenbrecht, 200230Double blind, 2 years10824%50.5TD estradiol vs placeboBone mineral density
5324%50.5
Giske, 200231Double blind, 2 years12311.4%49.1Estradiol vs placeboBone mineral density
4330.2%49.6
Guidozzi, 199932Open label, 4 years624%51CEE plus MPA vs placeboSurvival
684%51
Hall, 199433Open label, 2 years10037%55.8TD estradiol vs placeboBone mineral density
10016%56.1
Hall, 199834Single blind, 1 year4020%58.6TD estradiol plus MPA vs placeboAngina
2030%61.3
Komulainen, 199935Open label, 5 years1155.2%52.9Estradiol plus cyproterone acetate vs placeboBone mineral densityVitamin D also studied
1155.2%52.6
Kyllonen, 199836Double blind, 2 years5222%52.6Estradiol plus cyclic MPA vs placeboLumbar spine mobility
2622%52.6
Lindsay, 197637Double blind, 5 years636.3%44-50Mestranol vs placeboBone mineral content
575.3%44-50
MacDonald, 199438Double blind, 1 year4022.5%53TD estradiol with or without cyclic norethisterone vs placeboBone mineral density and rheumatoid arthritis disease activity
2240.9%55
Mijatovic, 199839Double blind, 2 years135%55.7CEE vs placeboHomocysteine levelsRaloxifene also studied
136.7%54.9
Mosekilde, 200240Open label, 5 years5029.6%49.5Estradiol plus cyclic norethisterone vs placeboBone mineral density and forearm fracture
5049.1%50.0
Nachtigall, 197941Double blind, 10 years843.6%55.3CEE plus MPA vs placeboMedical illness or death
848.3%54.9
PEPI trial Writing Group, 199542Double blind, 3 years70116%56.1CEE plus MPA or cyclic MPA or cyclic micronized progesterone vs placeboLipids, fibrinogen, blood pressure and insulin
17432.8%56.1
Perez-Jaraiz, 199643Open-label, 1 year265.8%48TD estradiol plus cyclic MPA vs calciumBone mineral densityCalcitonin also studied
525.8%50
Ravn, 199944Open label, 4 years11025.5%55CEE plus MPA or estradiol plus cyclic norethisterone vs placeboBone mineral densityAlendronate also studied
10926.7%55
Watts, 200045Double blind, 2 years30310%51.8CEE vs placeboBone mineral density
10310%51.3
WHI Writing Group, 200210Double blind, 5.2 years44766.3%50-59CEE plus MPA vs placeboDisease events and mortalityData for 50-59 years14
43566.1%50-59
WHI Writing Group, 200446Double blind, 6.8 years16374.9%50-59CEE vs placeboDisease events and mortalityData for 50-59 years14
6735.5%50-59

CEE=conjugated equine estrogen; MPA=medroxyprogesterone acetate; TD=transdermal; WHI=Women's Health Initiative.

⁎Most studies are identified by the first author's surname.

Using a “non-informative” prior distribution to pool data from 19 randomized trials, the mortality relative risk (RR) was 0.73 (95% CrI, 0.52-0.96), and the posterior probability of a mortality benefit was 0.985 (Table 2, Figure 2). There were a total of 156 deaths in 8689 participants in the hormone therapy group (1.80%) and 211 deaths in 7594 participants in the control group (2.64%) over the course of 5.1 years, indicating an absolute risk reduction of 0.84%.

Table 2. Table of Results
Source of AssumptionsPrior DistributionRelative Risk (95% Credible Interval)Probability of Relative Risk <1Between-study Variance on Log-relative Risk Scale

External information:

Synthesis of randomized trial data with observational studies

Non-informative prior distribution0.73(0.52-0.96)0.9850.067
Informative prior distribution⁎0.72(0.62-0.82)1.0000.024
Sceptical prior distribution†0.83(0.68-1.00)0.9750.026

Prior beliefs:

Erroneous assumptions

Erroneous prior distribution‡1.09(0.97-1.23)0.0770.308

⁎The Bayesian estimates obtained from 8 observational studies are: log-relative risk=−0.35 (−0.53, −0.16); SD of log-relative risk=0.092, therefore variance=0.008 and precision=125; between-study SD=0.17.

†Mean=0; variance=(0.008*sqrt[8]) such that the prior probability that the true relative risk is >0 is 0.05; between-study SD is adopted from the Bayesian estimate from observational studies.

‡Mean=0.13; variance=(0.06*0.06); precision=278; between-study SD=0.4. These estimates were derived using the Bayesian random-effects model, and were based on an erroneous interpretation of the Women's Health Initiative (probabilities of global index: treatment group=571/8506, placebo=623/8102).

Using an “informative” prior distribution, the 8 observational studies were pooled to obtain a RR of 0.78 (CrI, 0.69-0.90). When trial data were added, the RR was 0.72 (CrI, 0.62-0.82), with the posterior probability of a mortality benefit of 1.0 (Table 2, Figure 2).

Using a “sceptical” prior distribution, the RR was 0.83 (CrI, 0.68-1.00), with the posterior probability of a mortality benefit of 0.975 (Table 2, Figure 2). When an “erroneous” prior distribution was used, based on a belief that hormone therapy increased mortality, the resultant RR was 1.09 (CrI, 0.97-1.23). Using this assumption, the posterior probability that hormone therapy reduced mortality was 0.077 (Table 2).

Back to Article Outline

Discussion 

This Bayesian meta-analysis synthesized the available data in order to understand the effect of hormone therapy on mortality in younger postmenopausal women. We focused our investigation on mortality because it is an important clinical outcome, one that balances the benefits and harms of treatment. In addition, mortality is an area where the evidence seems uncertain, leading to the current controversies about hormone therapy. We used a Bayesian approach, which pools data from different sources, taking into account varying views on the issue.

Using pooled data from 8 observational studies, which followed 200,000 women for an average of 14 years, the mortality relative risk was 0.78. Pooled data from 19 randomized controlled trials, which followed younger postmenopausal women for over 83,000 patient-years, showed a similar mortality relative risk of 0.73. These results are similar to that found in the WHI trials, which showed a mortality hazard ratio of 0.7 in this younger age group.14 In this analysis, when randomized trial data were added to observational studies using an “informative” prior distribution, the resultant mortality relative risk was 0.72. The mortality reduction was 20%-30% over the course of 5 years' treatment, with an absolute risk reduction of close to 1%.

Bayesian analysis has the ability to determine the level of certainty about our conclusions. When randomized trial data are pooled using a “non-informative” prior distribution so the treatment effect is similar to that obtained using traditional meta-analysis, the posterior probability that hormone therapy provides a mortality benefit in younger women is 0.985. When an “informative” observational prior distribution is used, the posterior probability of a mortality benefit is essentially 1. When greater uncertainty is placed on the results from observational studies with a “sceptical” prior distribution, the results still remain robust, with a posterior probability of a mortality reduction of 0.975. Overall, our analysis shows that there is convergence of evidence from different sources.

From this synthesis of the data, we can see that the results from observational studies and randomized trials are remarkably similar, both showing reductions in total mortality of approximately 25%. We also can see that the age-specific results from the WHI are similar to those of other trials in younger women. A meta-analysis of randomized trials that was published before the availability of WHI data found a 39% reduction in total mortality in younger women,11 whereas the WHI trials found a 30% reduction in this age group.14 It is important to note that there still is substantial uncertainty in prevailing beliefs about a mortality benefit in early postmenopausal women.47, 48 This might be because age-specific mortality data for the WHI were not available until 2007,14 5 years after the original publication.10 The authors initially reported that mortality was not increased for all ages combined (hazard ratio 0.98), but that the “global index” of composite outcomes was increased by 13%.10 Now it seems plausible that misconceptions may have occurred in the interpretation of the available data, with a tendency of over-interpreting the results toward greater harm than benefit and assuming that hormone therapy had similar effects in younger and older women.47, 48 Quite possibly, many concluded after the first WHI publication that observational data had been misleading because there was not adequate adjustment for socioeconomic status, and that hormone therapy in fact increases mortality in all ages.

In this Bayesian analysis, we see that if one assumed that hormone therapy had similar effects in younger and older women and that the global index from the WHI translated into an increased risk for death, there would now be a great deal of uncertainty in interpreting the available age-specific mortality data. In fact, when using an “erroneous” prior distribution—which assumed an increased mortality in younger women—the addition of randomized trial data resulted in a posterior probability of 0.077 that hormone therapy reduces mortality in younger women. Thus, the “erroneous” prior distribution dominated the empirical evidence from field trials. This may explain why there is such difficulty at present accepting the conclusion that hormone therapy has a mortality benefit in this age group.

This meta-analysis has several potential limitations. Standard meta-analytic results are uncertain when the number of events per study is small, as is the case with mortality. In addition, there was wide variability in the 19 trials, such as the size of the trial, medications used, populations of women studied, and method of administration. We included both blinded and open-label trials with a wide range of methodological quality. Random-effects Bayesian meta-analysis is a useful approach when pooling heterogeneous studies, as it accounts for the possibility of between-trial variations. Despite these differences, little inter-study heterogeneity was seen in the results. Another limitation is that the 8 observational studies used varying methods in the adjustment for confounding variables, such as age, comorbid illnesses, and socioeconomic status. However, the results for trial data were remarkably similar to those of observational data, which indicate that most sources of bias were accounted for.

Estimates for relative risks were pooled from data on estrogen therapy and estrogen-progestin therapy, despite the fact that there are differences between these 2 types of treatment. However, evidence from the WHI trial indicates that similar mortality reductions are seen in younger women with single or combined treatment.14 This analysis was based only on published data, and therefore may be subject to publication bias. However, funnel plots of effect size versus standard error did not show evidence of bias. Furthermore, most trials did not report mortality as a primary or secondary outcome, so it is unlikely that deaths in those trials would have affected the decision to publish.

There are advantages and disadvantages of using Bayesian meta-analysis. Bayesian analysis allows us to compile all of the available data together, synthesize them into a coherent summary, and calculate the posterior probability of a given hypothesis. It can incorporate external information such as observational studies into pooled trial data to obtain a more precise estimate of treatment effect. In addition, it can illustrate how previously held assumptions—quantified in a form of prior distribution—can affect our understanding of the results. A disadvantage of this technique is that the results of the meta-analysis are dependent on the choice of prior distributions used. We performed the analysis using “informative,” “non-informative,” and “sceptical” prior distributions to assess how sensitive the results were to the prior assumptions used, and the results remained robust, showing strong beneficial mortality effect in each analysis.

It is clear that these findings need to be interpreted in the light of potential benefits and harms of hormone therapy. The available evidence indicates that hormone therapy in younger postmenopausal women increases the risk of breast cancer and pulmonary embolism, and reduces the risk of cardiovascular events, colon cancer, and hip fracture.13, 14, 49, 50, 51, 52, 53 The cardiovascular benefit is a result of a small absolute increase in stroke and a greater reduction in coronary heart disease events.14, 51, 54 The total mortality benefit for younger women seen in the randomized trials and observational studies indicates that the reduction in deaths from coronary heart disease, fracture, and colon cancer outweighed the increase in deaths from breast cancer, stroke, and pulmonary embolism. In addition to this mortality benefit, hormone therapy in younger women provides an improvement in quality-of-life measures, at least in the first few years of treatment.55, 56, 57, 58, 59

The past 6 years have seen marked fluctuations in our understanding of the effects of hormone therapy in young postmenopausal women. Guidelines published after the first WHI report concluded that hormone therapy had greater harms than benefits in women of all ages and should be used only for short durations in women with severe menopausal symptoms.60, 61, 62 After age-specific data from the WHI surfaced, position statements from 2007 concluded that the initiation of hormone therapy in younger women may in fact reduce cardiovascular morbidity and mortality, but no mention was made of an overall reduction in mortality.63, 64 It is our hope that this Bayesian meta-analysis will help to clarify the role of hormone therapy in younger women. The aim is to optimize the synthesis of evidence from different sources while taking into account varying views on the subject. These issues have bearing on development of guidelines to inform policy and practice.

Back to Article Outline

Supplementary data 

Download file

download text

Supplementary data.

Back to Article Outline

References 

  1. Wallach S, Henneman PH. Prolonged estrogen therapy in postmenopausal women. J Am Med Assoc. 1959;171:1637–1642
  2. Robinson RW, Higano N, Cohen WD. Increased incidence of coronary heart disease in women castrated prior to the menopause. Arch Intern Med. 1959;104:908–913
  3. Wilson RA, Brevetti RE, Wilson TA. Specific procedures for the elimination of the menopause. West J Surg Obstet Gynecol. 1963;71:110–121
  4. Bush TL, Cowan LD, Barrett-Connor E, et al. Estrogen use and all-cause mortality (Preliminary results from the Lipid Research Clinics Program Follow-Up Study). JAMA. 1983;249:903–906
  5. Petitti DB, Perlman JA, Sidney S. Noncontraceptive estrogens and mortality: long-term follow-up of women in the Walnut Creek Study. Obstet Gynecol. 1987;70(3 Pt 1):289–293
  6. Cauley JA, Seeley DG, Browner WS, et al. Estrogen replacement therapy and mortality among older women (The study of osteoporotic fractures). Arch Intern Med. 1997;157:2181–2187
  7. Grodstein F, Stampfer MJ, Colditz GA, et al. Postmenopausal hormone therapy and mortality. N Engl J Med. 1997;336:1769–1775
  8. Paganini-Hill A, Corrada MM, Kawas CH. Increased longevity in older users of postmenopausal estrogen therapy: the Leisure World Cohort Study. Menopause. 2006;13:12–18
  9. Mishell DR. Estrogen replacement therapy: an overview. Am J Obstet Gynecol. 1989;161(6 Pt 2):1825–1827
  10. Rossouw JE, Anderson GL, Prentice RL, et al. Risks and benefits of estrogen plus progestin in healthy postmenopausal women: principal results From the Women's Health Initiative randomized controlled trial. JAMA. 2002;288:321–333
  11. Salpeter SR, Walsh JM, Greyber E, et al. Mortality associated with hormone replacement therapy in younger and older women: a meta-analysis. J Gen Intern Med. 2004;19:791–804
  12. Furburg CD, Psaty BM. Review: hormone replacement therapy may reduce the risk for death in younger but not older postmenopausal women. ACP J Club. 2005;142:1
  13. Salpeter SR, Walsh JM, Greyber E, Salpeter EE. Brief report: coronary heart disease events associated with hormone therapy in younger and older women (A meta-analysis). J Gen Intern Med. 2006;21:363–366
  14. Rossouw JE, Prentice RL, Manson JE, et al. Postmenopausal hormone therapy and risk of cardiovascular disease by age and years since menopause. JAMA. 2007;297:1465–1477
  15. Salpeter SR, Buckley NS, Liu H, Salpeter EE. The cost-effectiveness of hormone therapy in younger and older postmenopausal women. Am J Med. 2009;122:42–52
  16. Whitehead A. Meta-analysis of Controlled Clinical Trials. Chichester, England: John Wiley & Sons; 2002;
  17. Spiegelhalter DJ, Abrams KR, Myles JP. Bayesian Approaches to Clinical Trials and Health-care Evaluation. Chichester, England: John Wiley & Sons; 2004;
  18. Sutton AJ, Abrams KR. Bayesian methods in meta-analysis and evidence synthesis. Stat Methods Med Res. 2001;10:277–303
  19. Sung L, Hayden J, Greenberg ML, et al. Seven items were identified for inclusion when reporting a Bayesian analysis of a clinical study. J Clin Epidemiol. 2005;58:261–268
  20. Warn DE, Thompson SG, Spiegelhalter DJ. Bayesian random effects meta-analysis of trials with binary outcomes: methods for the absolute risk difference and relative risk scales. Stat Med. 2002;21:1601–1623
  21. Lambert PC, Sutton AJ, Burton PR, et al. How vague is vague? (A simulation study of the impact of the use of vague prior distributions in MCMC using WinBUGS). Stat Med. 2005;24:2401–2428
  22. Higgins JP, Whitehead A. Borrowing strength from external trials in a meta-analysis. Stat Med. 1996;15(24):2733–2749
  23. Higgins JP, Spiegelhalter DJ. Being sceptical about meta-analyses: a Bayesian perspective on magnesium trials in myocardial infarction. Int J Epidemiol. 2002;31:96–104
  24. Pogue J, Yusuf S. Overcoming the limitations of current meta-analysis of randomised controlled trials. Lancet. 1998;351(9095):47–52
  25. Bush TL, Barrett-Connor E, Cowan LD, et al. Cardiovascular mortality and noncontraceptive use of estrogen in women: results from the Lipid Research Clinics Program Follow-up Study. Circulation. 1987;75:1102–1109
  26. Criqui MH, Suarez L, Barrett-Connor E, et al. Postmenopausal estrogen use and mortality (Results from a prospective study in a defined, homogeneous community). Am J Epidemiol. 1988;128:606–614
  27. Folsom AR, Mink PJ, Sellers TA, et al. Hormonal replacement therapy and morbidity and mortality in a prospective study of postmenopausal women. Am J Public Health. 1995;85(8 Pt 1):1128–1132
  28. Stahlberg C, Lynge E, Andersen ZJ, et al. Breast cancer incidence, case-fatality and breast cancer mortality in Danish women using hormone replacement therapy—a prospective observational study. Int J Epidemiol. 2005;34:931–935
  29. Angerer P, Kothny W, Stork S, von Schacky C. Hormone replacement therapy and distensibility of carotid arteries in postmenopausal women: a randomized, controlled trial. J Am Coll Cardiol. 2000;36:1789–1796
  30. Arrenbrecht S, Boermans AJ. Effects of transdermal estradiol delivered by a matrix patch on bone density in hysterectomized, postmenopausal women: a 2-year placebo-controlled trial. Osteoporos Int. 2002;13:176–183
  31. Giske LE, Hall G, Rud T, Landgren BM. The effect of 17beta-estradiol at doses of 0.5, 1 and 2 mg compared with placebo on early postmenopausal bone loss in hysterectomized women. Osteoporos Int. 2002;13:309–316
  32. Guidozzi F, Daponte A. Estrogen replacement therapy for ovarian carcinoma survivors: a randomized controlled trial. Cancer. 1999;86:1013–1018
  33. Hall GM, Daniels M, Doyle DV, Spector TD. Effect of hormone replacement therapy on bone mass in rheumatoid arthritis patients treated with and without steroids. Arthritis Rheum. 1994;37:1499–1505
  34. Hall G, Pripp U, Schenck-Gustafsson K, Landgren BM. Long-term effects of hormone replacement therapy on symptoms of angina pectoris, quality of life and compliance in women with coronary artery disease. Maturitas. 1998;28:235–242
  35. Komulainen M, Kroger H, Tuppurainen MT, et al. Prevention of femoral and lumbar bone loss with hormone replacement therapy and vitamin D3 in early postmenopausal women: a population-based 5-year randomized trial. J Clin Endocrinol Metab. 1999;84:546–552
  36. Kyllonen ES, Heikkinen JE, Vaananen HK, et al. Influence of estrogen-progestin replacement therapy and exercise on lumbar spine mobility and low back symptoms in a healthy early postmenopausal female population: a 2-year randomized controlled trial. Eur Spine J. 1998;7:381–386
  37. Lindsay R, Hart DM, Aitken JM, et al. Long-term prevention of postmenopausal osteoporosis by oestrogen (Evidence for an increased bone mass after delayed onset of oestrogen treatment). Lancet. 1976;1(7968):1038–1041
  38. MacDonald AG, Murphy EA, Capell HA, et al. Effects of hormone replacement therapy in rheumatoid arthritis: a double blind placebo-controlled study. Ann Rheum Dis. 1994;53:54–57
  39. Mijatovic V, Netelenbos C, van der Mooren MJ, et al. Randomized, double-blind, placebo-controlled study of the effects of raloxifene and conjugated equine estrogen on plasma homocysteine levels in healthy postmenopausal women. Fertil Steril. 1998;70:1085–1089
  40. Mosekilde L, Beck-Nielsen H, Sorensen OH, et al. Hormonal replacement therapy reduces forearm fracture incidence in recent postmenopausal women - results of the Danish Osteoporosis Prevention Study. Maturitas. 2000;36:181–193
  41. Nachtigall LE, Nachtigall RH, Nachtigall RD, Beckman EM. Estrogen replacement therapy II: a prospective study in the relationship to carcinoma and cardiovascular and metabolic problems. Obstet Gynecol. 1979;54:74–79
  42. PEPI Trial Writing Group. Effects of estrogen or estrogen/progestin regimens on heart disease risk factors in postmenopausal women (The Postmenopausal Estrogen/Progestin Interventions (PEPI) Trial. The Writing Group for the PEPI Trial). JAMA. 1995;273:199–208
  43. Perez-Jaraiz MD, Revilla M, Alvarez de los Heros JI, et al. Prophylaxis of osteoporosis with calcium, estrogens and/or eelcatonin: comparative longitudinal study of bone mass. Maturitas. 1996;23:327–332
  44. Ravn P, Bidstrup M, Wasnich RD, et al. Alendronate and estrogen-progestin in the long-term prevention of bone loss: four-year results from the early postmenopausal intervention cohort study (A randomized, controlled trial). Ann Intern Med. 1999;131:935–942
  45. Watts NB, Nolan JC, Brennan JJ, Yang HM. Esterified estrogen therapy in postmenopausal women (Relationships of bone marker changes and plasma estradiol to BMD changes: a two-year study). Menopause. 2000;7:375–382
  46. Anderson GL, Limacher M, Assaf AR, et al. Effects of conjugated equine estrogen in postmenopausal women with hysterectomy: the Women's Health Initiative randomized controlled trial. JAMA. 2004;291:1701–1712
  47. Utian WH. NIH and WHI: time for a mea culpa and steps beyond. Menopause. 2007;14:1056–1059
  48. Pines A, Sturdee DW, Birkhauser MH, et al. HRT in the early menopause: scientific evidence and common perceptions (Summary of the First IMS Global Summit on menopause-related issues). http://www.imsociety.org/pdf_files/comments_and_press_statements/ims_press_statement_13_05_08.pdf?SESSID=qr3rsun20036dp70id40v17m87March 29-30, 2008;Accessed June 15, 2008
  49. Beral V. Breast cancer and hormone-replacement therapy in the Million Women Study. Lancet. 2003;362(9382):419–427
  50. Beral V, Banks E, Reeves G. Evidence from randomised trials on the long-term effects of hormone replacement therapy. Lancet. 2002;360(9337):942–944
  51. Grodstein F, Manson JE, Colditz GA, et al. A prospective, observational study of postmenopausal hormone therapy and primary prevention of cardiovascular disease. Ann Intern Med. 2000;133:933–941
  52. Nelson HD, Humphrey LL, Nygren P, et al. Postmenopausal hormone replacement therapy: scientific review. JAMA. 2002;288:872–881
  53. Manson JE, Allison MA, Rossouw JE, et al. Estrogen therapy and coronary-artery calcification. N Engl J Med. 2007;356:2591–2602
  54. Bath PM, Gray LJ. Association between hormone replacement therapy and subsequent stroke: a meta-analysis. BMJ. 2005;330(7487):342
  55. Zethraeus N, Johannesson M, Henriksson P, Strand RT. The impact of hormone replacement therapy on quality of life and willingness to pay. Br J Obstet Gynaecol. 1997;104:1191–1195
  56. Smith AJ, Hall DR, Grove D. Postmenopausal hormone therapy and quality of life. Int J Gynaecol Obstet. 2006;95:267–271
  57. Dinger J, Zimmermann T, Heinemann LA, Stoehr D. Quality of life and hormone use: new validation results of MRS scale. Health Qual Life Outcomes. 2006;4:32
  58. Ylikangas S, Sintonen H, Heikkinen J. Decade-long use of continuous combined hormone replacement therapy is associated with better health-related quality of life in postmenopausal women, as measured by the generic 15D instrument. J Br Menopause Soc. 2005;11:145–151
  59. Heikkinen J, Vaheri R, Timonen U. A 10-year follow-up of postmenopausal women on long-term continuous combined hormone replacement therapy: update of safety and quality-of-life findings. J Br Menopause Soc. 2006;12:115–125
  60. Postmenopausal hormone replacement therapy for primary prevention of chronic conditions: recommendations and rationale. Ann Intern Med. 2002;137:834–839
  61. North American Menopause Society. Amended report from the NAMS Advisory Panel on Postmenopausal Hormone Therapy. Menopause. 2003;10(1):6–12
  62. Wathen CN, Feig DS, Feightner JW, et al. Hormone replacement therapy for the primary prevention of chronic diseases: recommendation statement from the Canadian Task Force on Preventive Health Care. CMAJ. 2004;170:1535–1537
  63. Pines A, Sturdee DW, Birkhauser MH, et al. IMS updated recommendations on postmenopausal hormone therapy. Climacteric. 2007;10:181–194
  64. North American Menopause Society. Estrogen and progestogen use in peri- and postmenopausal women: March 2007 position statement of The North American Menopause Society. Menopause. 2007;14:168–182

 Funding: No research funding.

 Conflict of Interest: Dr. Salpeter has consulted in the past with law firms representing Wyeth Pharmaceuticals on issues involving hormone therapy, and was paid on an hourly basis. Dr. Thabane consults with GlaxoSmithKline on statistical and methodological issues. The other authors report no potential conflicts of interest.

 Authorship: All authors had access to the data and had a role in writing the manuscript. Edwin Salpeter died after the preparation and approval of the manuscript.

PII: S0002-9343(09)00666-4

doi:10.1016/j.amjmed.2009.05.021

The American Journal of Medicine
Volume 122, Issue 11 , Pages 1016-1022.e1, November 2009

Article Tools

* Image Usage & Resolution