Holmberg and Anderson recently published a research letter in the journal Lancet regarding early cessation of the HABITS study,1 a Scandinavian trial designed to evaluate the risk of breast cancer recurrence in women with previously treated breast cancers who were given hormone therapy (HT) for relief of menopausal symptoms. The primary objective of the study was to investigate the safety of administering HT over a 2-year period for the treatment of menopause-related symptoms in women with a history of breast cancer; secondary goals included evaluation of quality-of-life issues and associated breast cancer mortality. The trial was directed only at women with climacteric symptoms; both perimenopausal and postmenopausal patients were eligible for enrollment. In terms of follow-up, study participants were to be evaluated at least twice yearly for the first 3 years, with a total of 5 years observation in all.
The study was terminated early because the results of an interim safety analysis indicated an unacceptable risk of breast cancer recurrence in women randomized to HT, compared to those randomized to “best symptomatic treatment without hormones.” After a median of 2.1 years, 26 women (14%) in the HT group and seven (4%) in the non-HT group had a new breast cancer event (local or distant recurrence or disease in the contralateral breast). These findings disagree with the results of previous studies that have found either a decreased or no increased risk of recurrence associated with HT use in breast cancer survivors.2-6 The HABITS investigators indicated that the hazard ratio of 3.5 (95% CI, 1.5-8.1) also differed significantly (p=0.02) from the unpublished, interim results of a similar trial among breast cancer survivors in Stockholm, Sweden, which found no increased risk of recurrence among HT users (RH=0.82, 95% CI 0.35-1.9)
As one of relatively few prospective randomized trials investigating the safety of HT use among breast cancer survivors, the early cessation of the HABITS study is important to note. Equally important to consider, however, are a number of limitations in the trial’s design. Primarily, the study is not double blinded or placebo controlled. Randomization occurred centrally, but the only categories used for data stratification were the participating research center, current use of tamoxifen, and history of HT use before cancer diagnosis (patients could have a history of HT use, but must not have been on HT for 3 months prior to randomization). In addition, while patients were randomized to either HT or “best symptomatic treatment without hormones,” the protocol was highly flexible, and decisions about the specific type of treatment were made by local practitioners. As a result, women in the HT group received varying types and doses of estrogen and progestin, and we do not know the impact of differing dose or preparation on the study results. The protocol recommends administration of “medium potency estrogen replacement (with or without progestins) therapy as it is commonly given in the environment where the patient lives and the clinician works” for those in the HT arm, although the term “medium potency”is not defined. Local vaginal hormone treatment was allowed in the control arm. Furthermore, it is difficult to assess the degree of adherence to other aspects of the study protocol; for example, some patients in the HT group received tibolone even though the investigators noted in their research letter that this therapy was not allowed. Similarly, the protocol recommended recent mammograms (within 3 months) for all patients before inclusion in the study, although no documentation is provided to confirm that this was done.
The brevity of the published report from the safety analyses on which termination of the trial was based leaves many additional questions unanswered:
Patients with a history of previously treated in situ, stage 1 or stage 2 breast cancer were eligible for the trial, but we are not provided with the numbers of patients in each stage in either HT or non-HT arms of the study. Patients whose tumors were estrogen receptor positive, receptor negative, or receptor status unknown were enrolled in the trial, although we do not know how many women in each of these groups were included in the HT and non-HT arms. An individual patient could also be included if she had four or fewer positive nodes, if nodal status was known. There is no documentation regarding the number of patients in whom nodal involvement was investigated, or the percentage of patients in whom data concerning the presence of positive nodes was lacking. At baseline, the HT population included 5% more patients with positive nodes than those randomized to the non-HT group. From this incomplete information, we cannot accurately assess if the arms of the study are equivalent for comparison, as there is no correction for the most important risk factors, stage and node status. Clearly, the patient with four positive nodes carries a worse prognosis than the patient with only one positive node.
The purpose of the trial was to investigate whether treatment of menopausal symptoms with HT was safe in women with a previously treated breast cancer, but the degree to which patients were symptomatic is not specified. We do not know the number of patients who were menopausal versus postmenopausal, or the length of time that elapsed since last menses. The study includes women up to 75 years of age-how many of these older women had menopausal symptoms?
It was anticipated that analysis would require enrollment of 250 to 300 patients per year, for a total population of 1300 participants. Only 381 patients were enrolled in the trial from its initiation in May 1997 until May 2002. The HABITS protocol was then amended to include collaboration with the Stockholm trial to pool data for safety and final analyses. Only 434 patients were enrolled by September 2003, and the study was discontinued in December 2003. Interestingly, only 79% of enrolled patients were included in the data analysis, as the other 21% did not have at least one follow-up examination. Information regarding compliance with follow-up is not given. The investigators recommended clinical mammograms every 12 to 24 months or participation in screening, although no information about adherence to these guidelines is provided. Thus, we cannot tell if there is a difference in compliance with imaging follow-up between the two arms.
The impact of duration of treatment and compliance with assigned therapy is unknown as well. The authors indicate that 18% of the patients assigned to the non-HT group received HT. In addition, we do not know the duration of therapy in those randomized to the HT group. Although the authors state that all participants diagnosed with recurrences in the HT group took hormones, 19% were not currently receiving HT when the cancer was discovered. It is also unknown if the patients treated with tamoxifen were compliant with its use, and if tamoxifen use affected the recurrence rate.
The follow-up period of 2.1 years is very short and information regarding the timing of breast cancer recurrence is not provided. Certainly, we know that when breast cancer is diagnosed it has been in the breast for up to 10 years or longer; thus, it is likely that the study patients with cancer recurrence or contralateral breast cancer had unrecognized disease at the time of initial randomization. Long-term follow-up of recurrence incidence in the non-HT group will be required before conclusions can be made regarding the potential role of HT as a promoter.