Swedish HABITS Is Cancelled

Abstracts & Commentary

Hormonal replacement therapy after breast cancer—is It Safe? (HABITS) began in May 1997, to compare breast cancer survivors treated for at least 2 years with hormone therapy with treatment other than hormones. By September 2003, a total of 434 women had been randomized and, in December 2003, the steering committee of the HABITS study made the decision to stop the trial because there were 26 women in the treated group and 7 in the non-treated group with new breast cancer diagnoses. Three women died of breast cancer in the treated group and 4 in the nonhormone group.

Women with Stage I and Stage II disease were eligible for thestudy, and tamoxifen treatment (but not tibolone) was allowed. The safety analysis indicated a relative hazard of 1.8 (CI, 1.03-3.1). The accompanying editorial stated that the HABITS trial "will probably be the last word" on the use of hormone therapy in women with breast disease, and that clinical decisions should hinge on the HABITS conclusion that hormone therapy, even short-term, is associated with an unacceptably high risk of breast cancer recurrence. (Holmberg L, Anderson H. Lancet. 2004;363:453-455; Chlebowski RT, Col N. Lancet. 2004;363:410-411.)

Comment by Leon Speroff

Prior to this publication, the information in the literature regarding hormonal treatment in breast cancer survivors has been uniformly reassuring. More than 1000 patients in case series, at least 2 case-control studies, and a preliminary report from the ongoing clinical trial at M.D. Anderson in Texas had all failed to detect an increase in treated women. Therefore the almost automatic reaction to the Swedish trial, as articulated in the accompanying editorial, is that the previous data reflect biases, such as selection of low-risk women for treatment (in my view, a very reasonable criticism), restaging of women at entry that removes women with early recurrences, and selective publication of studies reporting a protective effect (how in the world would we ever know if other studies were performed and not publicized?). But isn’t it time that we apply the same critical approach to randomized trials? Didn’t our experience with the Women’s Health Initiative teach us this? Therefore, let’s consider some criticisms of HABITS.

HABITS was not the only clinical trial organized in Sweden to study hormone therapy in breast cancer survivors. A similar trial was ongoing in Stockholm. At the time of the safety analysis, the relative hazard in Stockholm was 0.82 (CI = 0.35-1.9), compared to 3.3 (CI = 1.5-7.4) in HABITS. The HABITS report pointed out that the difference between the studies represented a statistically significant heterogeneity (the difference between the 2 trials was statistically significant), and it was a puzzle why the trials differed. Unfortunately, because of the decision regarding the HABITS trial, the Stockholm trial was also canceled, because of anticipated difficulties with recruitment and compliance. Indeed, the reason for analyzing both trials together is that a decision was made in 2002 to pool the data because of slow recruitment. Does the difference invalidate the HABITS’ conclusion? That’s a difficult question to answer.

HABITS was not a double-blinded, placebo-controlled trial. Women in the treated group were openly compared to the nonhormone treated group. The treated group (174 women) consisted of estrogen only (21%), sequential estrogen-progestin (26%), and continuous combined estrogen-progestin (46%). Comparison of regimens did not yield significant differences. However, 11% of the treated group used other treatments, including tibolone, even though it was supposed to be excluded. In the comparison group (171 women), 18% of the women were treated with hormone therapy. From the reported analysis, the effect of this variation and crossover cannot be ascertained.

In the treated group, 21 of the 26 women with recurrent disease were diagnosed while on the planned 2-year treatment. Did this rapid appearance of recurrent disease (median follow-up of 2.1 years) reflect an impact of hormone treatment on pre-existing disease, a bias that could have been present because of the failure to re-stage disease on entry to the study?

The HABITS report indicated that the characteristics of the 2 groups did not influence the results with the possible exception of hormone receptor status. The two groups did differ, with the following characteristics more prevalent in the hormone-treated group: positive nodes, positive hormone receptors, unknown hormone receptor status, and fewer mastectomies. A statistical analysis of these differences was not provided. The relative hazard comparing the hormone-treated women and the comparison women in each of the subgroups (grouped according to baseline characteristics) was presented, and a significant increase was present in the following subgroups: positive hormone receptors, no tamoxifen, and hormone therapy before diagnosis. The problem is that the confidence intervals were very wide because of small numbers, and the report acknowledged the very low precision. Is it possible that these differences represented the greater prevalence in the treated group of the previously noted baseline characteristics?

Hopefully the report from HABITS will not cause cancelation of the other ongoing clinical trials investigating the use of hormone therapy in breast cancer survivors. It seems to me that there are too many questions and problems with the HABITS study to consider it definitive. Having said that, the report certainly makes it more difficult for clinicians and breast cancer patients to accept the unknown risks and use hormone therapy. I will continue to support patients with whatever decision they make once they are fully informed regarding the accumulated data.

Leon Speroff, MD, Professor of Obstetrics and Gynecology at Oregon Health Sciences University, Portland and Editor of OB/GYN Clinical Alert.