Similarities and differences between adjuvant hormonal therapy in breast and prostate cancer

Various reasons exist for the difference in the clinical research data between prostate cancer and breast cancer. First, breast cancer occurs in younger women while prostate cancer occurs in older men.

Obviously, a patient with a 40-year life expectancy is more interested in what happens long term than a patient with a 10-year life expectancy.

Second, the early hormonal treatments for prostate cancer were unpleasant. They consisted of castration and diethylstilbestrol (DES), which was discovered to be seriously cardiotoxic and would actually do more harm than good in terms of life expectancy. As soon as DES was no longer used and alternative means of turning off testicular function were discovered, trials began.

Hormonal therapy for prostate cancer substantially delays progression of the disease and moderately delays death from the disease. The effects of immediate hormonal treatment versus deferred hormonal treatment in a man with prostate cancer are comparable to the effects of five years of adjuvant tamoxifen in a woman with hormone-sensitive breast cancer. Additionally, hormonal therapy prevents a number of complications of metastatic disease, such as spinal metastases, ureteric obstruction and the need for further surgery.

The prostate cancer trials were not as large as the breast cancer trials, so the results were muddled by the deaths from other causes. The curves are similar, but the prostate trials have statistical noise from the large numbers of deaths that are unrelated to prostate cancer or its treatment. When patients are older, deaths from other causes confuse trial results.

Potential impact of early hormonal therapy for prostate cancer on survival

The problem with evaluating hormone therapy for prostate cancer is that only a few thousand men with prostate cancer were being randomly assigned to therapy, compared to tens of thousands of women with breast cancer. That is why the evidence of benefit in breast cancer is so much better.

In breast cancer, we have seen impressive decreases in death rates in middleaged women as a result of early use of tamoxifen and chemotherapy. I believe the effects of earlier treatment with hormonal therapy in prostate cancer over the next five or 10 years will be comparable to that produced by tamoxifen in breast cancer.

Does bicalutamide have an adverse effect on mortality in patients with low-risk disease in the Early Prostate Cancer (EPC) program trials?

No good evidence indicates that bicalutamide treatment affects mortality from causes other than prostate cancer. Currently, the number of deaths from prostate cancer in the EPC trials is so limited that it is difficult to obtain any clear evidence of an effect on prostate cancer mortality.

The question as to the effect on overall mortality is well worth asking, but it needs to be answered by separate analyses of prostate cancer mortality and nonprostate cancer mortality. One should ask, “Is there any evidence of hazard?” No. “Is there any evidence of benefit?” At some point, the answer to that question may well turn out to be “yes.”

No good evidence indicates that bicalutamide increases the overall death rate from causes other than prostate cancer. If you have no overall evidence and you begin looking for subgroups of this and subgroups of that, you’re almost bound to find a subgroup in which the results seem favorable and a subgroup in which the results seem unfavorable, but that is just statistical noise.

Risk of false-negative results from subgroup analyses

Years ago, we published a paper in The Lancet that analyzed astrological birth signs as a subgroup in a trial evaluating aspirin as treatment for acute heart attack. The study proceeded with a controlled randomization of approximately 17,000 heart attack patients. There were 1,000 deaths in the placebo arm compared to 800 deaths in the aspirin-treated arm. That equates to a five standard error difference in mortality, which is an excellent result.

The Lancet agreed to publish the study; however, they insisted on knowing which subpopulation was going to derive benefit — older versus younger, male versus female, those with a previous infarct versus those without. The suggestion was completely ridiculous because if the treatment works that well, it’s going to be of some value for everybody.

Dividing a five standard error difference into subgroups will result in false negatives. So, as an absolute matter of principle, we said “no.” The Lancet refused to publish the data unless we complied with subgroup analyses. Finally, we relented. We classified patients according to astrological birth sign, performed a subgroup analyses and sent the results to The Lancet. The paper was published August 13, 1988 (ISIS 1988). Aspirin didn’t seem to work as treatment for heart attack if you’re born under Libra or Gemini, but it produced halving of risk if you were born under Capricorn. It’s just complete junk. And, actually, a lot of subgroup analyses are junk.

Select publications

Professor Sir Richard Peto is Professor of Medical Statistics and Epidemiology and co-founder and co-director of the Clinical Trial Service Unit at the University of Oxford in the United Kingdom.