Interpretation of PIVOT findings

The results of the long awaited Prostate Cancer Intervention versus Observation Trial (PIVOT) have been recently published in the New England Journal of Medicine (1). The primary outcome of the trial was overall survival in men treated with radical prostatectomy (RP) or observation for clinically localized prostate cancer detected with PSA screening. Seven hundred thirty one US veterans were randomized between 1994 and 2002 to RP or observation. Mean age of the cohort was 67 years, median PSA was 7.8 ng/mL, approximately half of patients had non-palpable disease, and same proportion had Gleason score ≤7. Overall survival was 2.9% higher and cancer specific survival was 2.6% higher in surgically treated patients at the10-year median survival mark. The most prominent overall survival benefit of 12.6% was observed in patients with intermediate-risk tumors by D'Amico criteria. The likelihood of bone metastases was twice lower in the RP arm of the trial. The authors concluded that RP does not reduce significantly overall and cancer specific survival in men diagnosed with PSA screening. Does it mean that all men with clinically localized prostate cancer should not be offered RP or should PSA screened be abandoned if active treatment does not matter? Probably not. First, prostate cancer, even detected with PSA screening and clinically localized, is a heterogeneous disease. There is plenty of evidence that low-grade small tumors may be treated expectantly while high proportion of aggressive tumors may present as pathologically extra-prostatic disease destined to recur despite surgery, not so for the middle of the spectrum. Indeed, in PIVOT study RP was mostly beneficial in the intermediate risk tumors. Unfortunately, due to low recruitment rates the study was underpowered to detect survival differences in the clinically relevant subgroups. In the retrospect, it is apparent that at the time of study design almost two decades ago the need to focus on particular risk group may not have been that obvious. Furthermore, the control of recruitment rate is beyond researches ability. Second, by the randomized nature of the trial not all participants complied with the intended treatment. In the RP arm 15% chose observation, 7% chose radiation and in 2% surgery was abandoned, while in the observation arm 20% received RP. The inconsistent adherence to the assigned therapy results essentially in comparing '77% surgery' to '80% observation'. It may make sense epidemiologically to avoid bias by allowing the 'intention to treat' analysis, however may not apply …