Examining the Evidence
Analyzing the HABITS data: Is HT safe in women with previous breast cancer?
To evaluate the safety of hormone therapy (HT) compared to nonhormone therapy (NHT) in women with previously diagnosed breast cancer who are experiencing menopausal symptoms.
Researchers found an increased risk of breast cancer recurrence among women taking HT, with 14% developing recurrence/contralateral breast cancer compared with 4% of nonusers, and terminated the trial early.
Women were eligible for the HABITS prospective randomized trial if they had a history of in situ, stage I or II breast cancer with up to 4 positive lymph nodes and climacteric symptoms. Both premenopausal and postmenopausal women were accepted into the study, which began accruing participants in May 1997. The trial assigned 434 previous breast cancer patients with menopausal symptoms to either HT or NHT.
The main endpoint was any new breast cancer event (recurrence/contralateral breast cancer), with all analyses done according to intention to treat. Secondary aims were to examine quality of life and risk of death from breast cancer.
After a median follow-up of 2.1 years, 26 women in the HT group and 7 nonusers had a new breast cancer event. All women in the HTgroup and 2 women in the NHT group were exposed to HT; most experienced their event while on treatment.
Researchers determined that HT posed an unacceptable risk and ended the trial on Dec 17, 2003.
Imperfect study design. Although it was a prospective randomized trial, HABITS leaves much to be desired. It was an open study that was neither placebo-controlled nor blinded.
Treatment was not described in either arm of the study. For example, HT was suggested to be estrogen with or without progestin of “median potency” (undefined), similar to that “commonly given in the environment where the patient lives and the clinician works.” In the NHT arm, therapy was the “best symptomatic treatment without hormones” and could include clonidine, beta blockers, psychological support, physical exercise, and acupuncture. Local estrogen could be used but “natural products” could not. In 2002, because of poor accrual, the “Stockholm” study was folded into the HABITS trial.
The HABITS study was an equivalency trial designed to stop if the hazard ratio (HR) surpassed 1.36. The investigators state that the HR for HT was 3.5 (14% of women had a recurrence/contralateral cancer, compared with 4% in the nonhormonal group). The investigators note that the Stockholm trial had an HR of 0.82 and was not included in the research letter published in Lancet. Since most HABITS participants were from Sweden, one wonders why there are such different results from the same apparent population base.
Unanswered questions. Analyzing the HABITS trial is problematic on several fronts. For instance, mammograms and follow-up are suggested but apparently not required (more than 20% of randomized women were not included in the analysis because they had not had at least 1 follow-up visit). Were these items equal in both groups? Did the HT group have better mammography compliance than the nonhormonal group? Compliance to therapy was not detailed.
Roughly 20% of women in the HT group who developed a recurrence were not on HT at the time of recurrence. With such a short follow-up (2.1 years), details on the length of HT and relationship to time of randomization and recurrence are important. Since breast cancer can reside in the breast for 10 years or more before diagnosis, it is reasonable to assume that the recurrences and contralateral breast cancers were present at the time of randomization.
Two of the most important risk factors in breast cancer are stage and lymphnode status. These 2 items were not stratified at the time of randomization. Were they equal in the 2 groups? What was the receptor status in each group? Were the groups equal in this regard?
Tamoxifen was allowed and stratified at randomization. Since tamoxifen can have an impact on menopausal symptoms, was compliance the same in the 2 groups?
Although the authors noted that a Cox proportional hazard model would be used in their analysis, no such data was given. The results could change once such a model is used.
We appreciate that a study like this is difficult, as the Women’s Health Initiative amply proves. This one involved low accrual and lenient guidelines regarding treatment, prognostic variables, and compliance. Still, this preliminary research letter is just that: preliminary. We should await thorough evaluation before giving the findings much credence. Certainly the statement in the commentary accompanying this letter, that this study “can now reasonably guide clinical practice for women with breast cancer,” appears premature and without merit.