The results of several pivotal phase iii
trials have been reported or published recently (see ). The studies reporting survival data13,17,18,21
showed no difference in overall and prostate cancer–specific survival between groups. However, those studies were underpowered and contained patients with both metastatic and nonmetastatic disease.
Summary of phase iii clinical trials of intermittent compared with continuous androgen deprivation therapy (adt) in advanced prostate cancer
Two large trials with long-term follow up and homogeneous patient populations—the pr
and Southwest Oncology Group (swog
) 9346 trials20
, intended as companion studies—were designed to address the impact of adt
on cause-specific and overall survival in a definitive fashion. Both were designed as noninferiority studies. The results, which are somewhat contradictory, are fascinating.
The pr.7 trial, in nonmetastatic patients, was led by the ncic and supported by swog and the U.K. Coordinating Committee on Cancer Research. It enrolled 1386 patients with a psa level greater than 3 ng/mL more than 1 year after primary or salvage radiotherapy for localized prostate cancer. Treatment with iad was provided in 8-month cycles, with nontreatment periods determined according to psa level. The primary endpoint was overall survival. Secondary endpoints included quality of life, time to castration-resistant disease, and duration of nontreatment intervals. Median follow-up was 6.9 years. In the intermittent therapy group, full testosterone recovery occurred in 35% of patients, and testosterone recovery to the trial entry threshold occurred in 79%. Patients on the intermittent arm were on treatment only 27% of the time. Benefits for qol with iad were seen in the domains of physical function, fatigue, urinary problems, hot flashes, libido, and erectile function. Deaths numbered 268 in the intermittent therapy group and 256 in the continuous therapy group. Median overall survival was 8.8 years in the intermittent therapy group compared with 9.1 years in the continuous therapy group [hazard ratio (hr): 1.02; 95% confidence interval (ci): 0.86 to 1.21]. The associated 7-year cumulative rates of prostate cancer mortality were 18% and 15% respectively (p = 0.24).
.7 trial is the first to definitively demonstrate equivalence in terms of overall survival and prostate cancer–specific survival in men randomized to iad
compared with continuous androgen deprivation (cad
). Time to castration resistance was longer in the iad
arm, likely reflecting a bias in the trial design. As originally constructed, patients off treatment with a rising psa
had to be placed back on therapy, with psa
being seen to rise again before they were deemed to have castration resistance. In fact, many patients had prolonged androgen suppression during the off-treatment interval, and their rising psa
reflected true castration resistance. That result was not foreseen at the time the trial was initiated in the mid-1990s. Thus, the time to androgen independence was biased in favour of the intermittent arm. The fact that prostate cancer deaths trended higher in the intermittent than in the continuous arm (balanced precisely by an increased number of non–prostate cancer deaths in the continuous arm) reinforces the suggestion that the difference in time to castration resistance was probably an artefact. Time to castration resistance was nonsignificantly different between the groups in several other phase iii
The swog 9346 trial was designed along similar lines to evaluate overall and disease-specific survival for iad compared with cad given to patients with metastatic prostate cancer. The study registered 1535 men with lymph node, visceral, or bone metastases and a psa exceeding 5 ng/mL and placed them on 7 months of goserelin and bicalutamide. If psa was less than 4 ng/mL by month 6, the men were then randomized to either discontinuation of therapy (iad) or cad. Treatment was reinitiated when psa reached a baseline of 20 ng/mL and discontinued again after 7 months if psa was less than 4 ng/mL. The trial was sized to demonstrate noninferiority with a delta of 1.2.
Approximately 90% of the patients have now died. Overall survival showed a nonsignificant trend to improvement in the cad arm (hr: 1.09; 95% ci: 0.95 to 1.24). Median survival was 5.8 years compared with 5.1 years—a 7-month difference. A sub-analysis (not specified a priori) stratified by minimal disease (confined to axial skeleton and pelvis or to lymph nodes) and extensive disease (any or a combination of ribs, long bones, skull, or viscera) showed a benefit in the minimal-disease group (hr: 1.23; 95% ci: 1.02 to 1.49; p = 0.034) and no benefit in the extensive-disease group (hr: 0.95; 95% ci: 0.80 to 1.16). In the minimal-disease group, median overall survival was 5.2 years with iad and 7.1 years with cad. The authors claim that the study demonstrates the inferiority of intermittent therapy in that setting.
The interpretation of swog
9346, which was still unpublished at the time this manuscript was written, is controversial. It reflects a somewhat arcane statistical controversy over the correct interpretation of noninferiority trials. The issue is the significance of a result in which the confidence limits cross both unity and the pre-specified delta—that is, the minimum difference thought to be clinically significant at the time of trial design (for interested readers, I recommend an informative article by Piaggio et al.22
on the interpretation of noninferiority trials). First, the primary endpoint showed no significant difference between the two arms (95% ci
: 0.95 to 1.24). The fact that the ci
crossed the pre-specified delta does not imply inferiority; rather, it means that the results are inconclusive with respect to noninferiority, which cannot be ruled out. Coincidentally, superiority also cannot be ruled out. Importantly, those findings also apply to the results in the minimal metastatic disease group. Because both sides of the 95% confidence limits were not beyond the delta of 1.2, even that result is inconclusive for noninferiority. Second, the stratification analysis was conducted post hoc,
making it hypothesis-generating, not proof. Third, considering these two large well-conducted studies together, pr
.7 shows noninferiority for iad
in nonmetastatic disease, and swog
9346 purports to show both inferiority of iad
in minimal metastatic disease and noninferiority in extensive metastatic disease. It is a struggle to make biologic sense of the supposed inferiority of iad
in minimal metastatic disease, encased between the bookends of the apparent noninferiority of iad
in both nonmetastatic and extensive metastatic disease. A possible explanation is that at the favourable (nonmetastatic) and unfavourable (extensive metastatic) ends of the spectrum, adt
adds little or nothing to patient survival, and therefore whether treatment is intermittent or continuous is irrelevant, and that, for the minimal metastatic disease where adt
does make a difference, continuous is seen to be better than intermittent. In the context of the interpretation of the earlier-mentioned swog
data, the foregoing interpretation is only a hypothesis.
The intermittent arm of pr.7 experienced 9% more prostate cancer deaths, but that difference did not achieve significance. It is, of course, possible that the trial was simply underpowered to show a small difference in cancer-specific survival. Given that the increased mortality from non–prostate cancer death in the continuous arm perfectly balanced that increase in cancer deaths, the question seems moot. One might argue that, if a population of men without cardiovascular morbidity were to be treated, a difference in overall survival might emerge. Again, that suggestion is a hypothesis.
The bottom line is that the substantial qol benefits of intermittent therapy, combined with the apparent protective effects on bone mineral density (bmd), surpass the hypothetical survival benefits of continuous therapy, which failed to emerge in two studies following almost 3000 patients. Further, in clinical practice, patients who are not likely to benefit from intermittent therapy can be identified early (by clinical parameters or by observation during a short off-treatment interval) and switched to continuous therapy.