Menopause Live (23 April, 2012) from IMS

Two papers from the Women’s Health Initiative (WHI), an updated assessment of the risk of breast cancer in relation to the use of unopposed estrogen therapy (ET) [1] and a review of breast cancer risk in relation to the use of hormone replacement therapy (HRT) [2], have recently been published. 

The WHI randomized, ‘double-blind’ trial of ET versus placebo [3], conducted among hysterectomized women aged 50–79 years, was terminated after an average of 6.8 years of follow-up because of an increased risk of stroke (not considered here). ‘Unblinding’ of the study medication occurred among less than 2% of the women and, overall, 53.8% discontinued their assigned treatments. In an ‘intention-to-treat’ (ITT) analysis, the hazard ratio (HR) for invasive breast cancer was 0.77 (95% confidence interval (CI) 0.59–1.01). In a second analysis [4], after 7.1 years of ‘blinded’ follow-up, the ITT HR was 0.67 (95% CI 0.47–0.97; p = 0.03). In a third ‘as treated’ analysis [5], based on 7.1 years of ‘blinded’ follow-up plus an additional 3.6 years of ‘unblinded’ follow-up (total 10.7 years), the HR was 0.68 (95% CI 0.49–0.95); the risk reduction remained evident when the data were stratified by decade of age. The authors concluded that ‘the decreased risk of breast cancer persisted’. In the most recent publication [1], the median duration of follow-up was 11.8 years, and for invasive breast cancer the ITT HR for the use of ET for a median duration of 5.9 years was 0.77 (95% CI 0.62–0.95; p = 0.02); in an ‘as treated’ analysis, it was 0.68 (95% CI 0.49–0.95; p =0.02). For fatal breast cancer, the ITT HR was 0.37 (95% CI 0.13–0.91; p = 0.03).

In the recent review of the risk of breast cancer in relation to the use of HRT [2], the authors stated that ‘recent results from large prospective cohort studies and the [WHI] randomized placebo-controlled trials have substantially changed concepts regarding how estrogen alone and estrogen plus progestogen (E+P) influence breast cancer’. Whereas ‘the preponderance of the observational studies’ suggested that both ET and E+P increase the risk of breast cancer, in the WHI trials ‘E+P statistically significantly increased the risk of breast cancer’, but the use of unopposed ET ‘statistically significantly decreased the risk’. ‘Differential mammography usage patterns may explain differences between observational and randomized trial results’, since ‘in the WHI randomized trials, mammogram frequency was protocol mandated and balanced between comparison groups’.

Comment
Before the WHI clinical trial of ET was terminated, there was little or no detection bias among women who adhered to treatment. For that reason, the trial data represent the best evidence to date concerning breast cancer risk among ET-exposed women. All trial participants were hysterectomized; since vaginal bleeding did not occur, the ‘unblinding’ rate was less than 2%, and bias or confounding due to knowledge of exposure status was largely, if not entirely, avoided. 

In striking contrast, in the WHI trial of E+P, in which naturally menopausal women were included, the respective ‘unblinding’ rates among exposed women and placebo recipients were 44.4% and 6.85 [6], principally because of vaginal bleeding, making it mandatory to rule out endometrial cancer. That necessity, in turn, selectively alerted ‘unblinded’ E+P users to the possibility that exposure may cause breast cancer, and detection bias was inevitable. 

The WHI investigators claimed that detection bias was largely averted in the E+P and ET clinical trials, since the mammography rates were the same in the treated women and in the placebo recipients [1,6]. As has been pointed out [6], however, that claim was incorrect: annual ‘protocol mandated’ mammograms were only performed among women who continued to adhere to their assigned treatments; ‘protocol mandated’ mammograms were not carried out after the women stopped. Before the E+P trial was terminated, 42% and 38% of the E+P and placebo recipients, respectively, had stopped their treatments. Thereafter, women assigned during the trial to E+P would more commonly have undergone mammography on their own initiative than women assigned to placebo. Contrary to what has been claimed by the WHI investigators, detection bias was not eliminated in the E+P trial.

In the ET trial, 53.8% of the participants stopped their assigned treatments, and the discontinuation rates were similar among the ET and placebo recipients [3]. Among them, mammography rates could possibly have been higher thereafter among women initially assigned to ET than among women assigned to placebo – although this was unlikely, since the ‘unblinding’ rate was less than 2% and, following ‘unblinding’, knowledge of exposure status would not selectively have resulted in more common use of mammography by ET-exposed women. In addition, even if ET users who stopped may more commonly have undergone mammography thereafter, the magnitude of the apparent reduction in the risk of breast cancer would, if anything, have been underestimated.

Despite these considerations, it is appropriate to interpret the apparent risk reduction in the ET users cautiously, until independently confirmed. Numbers were limited, large proportions of the follow-up data were observational – not randomized or ‘double-blind’ – and the findings could have been due to unidentified sources of bias or confounding. However, it is unlikely that bias or confounding could have explained a 63% reduction (1.00 minus the HR of 0.37 × 100) in the breast cancer death rate after 11.8 years of follow-up. That estimate was based on sparse data, but it is nonetheless impressive.

Turning next to the review of breast cancer risk in relation to the use of HRT [2], as described above, the claim, repeated in the review, that differential mammography was avoided in the WHI trials, was incorrect. The review was also inaccurate: in table 1, the Collaborative Reanalysis [7] was represented as a meta-analysis of 51 case–control studies. In fact, the Collaborative Reanalysis included cohort studies as well as case–control studies, and the findings based on the two approaches were materially different [7]: in the cohort studies, in which selective recall of HRT use by the breast cancer cases (information bias) was unlikely, the relative risk (RR) was 1.09, and not statistically significant; in the hospital- and population-based case–control studies, the respective RRs were 1.27 and 1.15, and significant. Over 80% of the HRT use in the Collaborative Reanalysis was unopposed estrogen, the data in the cohort studies were largely free of information bias and, in the absence of that bias, there was no evidence of an increased risk of breast cancer.

In conclusion, the evidence for a possible reduction in the risk of breast cancer among ET users is tenuous, principally because of limited numbers, and further study is needed. But whether or not ET reduces the risk, the WHI clinical trial has furnished statistically robust evidence to suggest that ET does not increase it. The importance of that finding cannot be overstated: since the clinical trial data for ET were unbiased, while the data for E+P were demonstrably biased, the ET findings raise the possibility that HRT, overall, may not increase the risk of breast cancer and that the observational studies, as well as the WHI clinical trial of E+P (which soon became an observational study [6]) suggested that they did reach incorrect conclusions, largely because they shared much the same biases. Not only do the WHI findings for ET need to be independently confirmed, but the entire topic of whether or not HRT, or certain types of HRT, do or do not increase (or possibly even reduce) the risk of breast cancer now needs to be revisited.

Samuel Shapiro
Department of Public Health and Family Medicine, University of Cape Town, South Africa

References
1. Anderson GL, Chlebowski RT, Aragaki AK, et al. Conjugated equine oestrogen and breast cancer incidence and mortality in postmenopausal women with hysterectomy: extended follow-up of the Women’s Health Initiative randomized placebo-controlled trial. Lancet Oncol 2012 March 6. Epub ahead of print.
http://www.ncbi.nlm.nih.gov/pubmed/22401913
2. Chlebowski RT, Anderson GL. Changing concepts: menopausal hormone therapy and breast cancer. J Natl Cancer Inst 2012;104:517-27.
http://www.ncbi.nlm.nih.gov/pubmed/22427684
3. Anderson GL, Limacher M, Assaf AR, et al.; The Women’s Health Initiative Steering Committee. Effects of conjugated equine estrogen in postmenopausal women with hysterectomy: the Women’s Health Initiative randomized controlled trial. JAMA 2004;291:1701-12.
http://www.ncbi.nlm.nih.gov/pubmed/15082697
4. Stefanick ML, Anderson GL, Margolis KL, et al. Effects of conjugated equine estrogens on breast cancer and mammographic screening in postmenopausal women with hysterectomy. JAMA 2006;295:1647-57.
http://www.ncbi.nlm.nih.gov/pubmed/16609086
5. LaCroix AZ, Chlebowski RT, Stefanick ML, et al. Health outcomes after stopping conjugated equine estrogens among postmenopausal women with prior hysterectomy: a randomized controlled trial. JAMA 2011;305:1305-14.
http://www.ncbi.nlm.nih.gov/pubmed/21467283
6. Shapiro S, Farmer RDT, Mueck AO, Seaman H, Stevenson JC. Does hormone replacement therapy cause breast cancer? An application of causal principles to three studies. 2. The Women’s Health Initiative: estrogen plus progestogen. J Fam Plann Reprod Health Care 2011;37:165-72.
http://www.ncbi.nlm.nih.gov/pubmed/21642264
7. Shapiro S, Farmer RDT, Seaman H, Stevenson JC, Mueck AO. Does hormone replacement therapy cause breast cancer? An application of causal principles to three studies. 1. The Collaborative Reanalysis. J Fam Plann Reprod Health Care 2011;37:103-9.
http://www.ncbi.nlm.nih.gov/pubmed/21454266