This weeks NEJM:
The PIVOT trial is published in the NEJM, a randomized comparison of observation vs radical prostatectomy for localized prostate cancer. The headline findings: no difference in overall survival. However, this one is worth looking a little more closely at.
The first item that strikes me is the low numbers of this trial. This is in theory a non-inferiority trial of observation, however, one can rarely design these with any relevant power with less than a couple of thousand. And in fact when reviewing the statistical design, the investigators downgraded their definition of non-inferiority due to poor accrual - to at 25% reduction in mortality. This is a pretty low bar for non-inferiority. The trial was initially designed to accrue 2000.
Moreover, in the observation arm 20% of the enrollee's entered into a curative treatment, 10% prostatectomy, and 10% radiation (mix of EBRT and brachy). This further dilutes the potential power of this study. 15% of the prostatectomy arm also underwent observation.
And in fact, when one looks at the forest plots in figure 3, the hazard ratios for benefit are pretty consistently in the favoring prostatectomy side, with only the confidence intervals overlapping one. Only in the low risk subgroup did the HR cross to favoring observation.
I feel the investigators should have qualified their conclusions a little more in this trial. When a study is underpowered one expected not to see a significant result: however, in non-inferiority trials, when a positive result hinges on there being no difference in results between two arms - these issues of power, accrual and crossover become much more troublesome. All these factors work to deteriorate the quality of a trial, and make a non-statisically significant result more likely.
Radical Prostatectomy versus Observation for Localized Prostate Cancer: New England Journal of Medicine, Volume 367, Issue 3, Page 203-213, July 2012.