• Radical prostatectomy RCT

    Perhaps you’ve heard about the results of the latest randomized controlled trial (RCT) of radical prostatectomy for treatment of prostate cancer. It’s written up in the New England Journal of Medicine along with an editorial. Both are ungated. The editorial highlights the limitations of the study and is short. The full research article takes more time to read, and, my guess, is that many interested in this topic won’t.

    However, Tara Parker-Pope did a good job in the New York Times laying out the results in a more accessible, summarized form and putting them in context. Her whole piece is worth reading, but here’s a portion:

    The latest research, called the Prostate Cancer Intervention Versus Observation Trial, or Pivot, focuses on what happens after a man receives a diagnosis of early-stage cancer, often as a result of a P.S.A. test. The men in the study were randomly assigned either to surgical removal of the prostate or to an observation group, where a man’s cancer was monitored but not treated unless it showed signs of progressing. Although the study originally set out to recruit 2,000 men, the researchers could not reach that goal and revised the research plan to include 731 men.

    By the end of the 15-year study, 354 men had died, but most of them had died of from a cause other than prostate cancer. There was no statistical difference in overall mortality rates in the surgical group, which had 171 deaths, compared with the observation group, which had 183 deaths.

    During the study, only 52 men, or about 7 percent of the study subjects, died of prostate cancer, but again there was no statistical difference in the prostate cancer mortality rate between the groups.

    But in a secondary analysis, the researchers did find an important difference between the groups, suggesting that surgery may benefit men with early-stage disease who have a high P.S.A. score. Among men with a P.S.A. value higher than 10 nanograms per milliliter of blood, surgery lowered a man’s risk of dying by 33 percent compared with the observation group. Overall, among men with high P.S.A. scores, there were 13 percent fewer deaths in the surgery group compared with the observation group. No difference was seen among men with a P.S.A. value of 10 or less.

    The study was funded, in part by the Agency for Healthcare Research and Quality (AHRQ). The House is preparing to vote on terminating that agency.


    • When I read Pope’s piece this AM, I was not enthralled with the very excerpt you chose above, specifically, the reference to the secondary analysis and PSA outcome at level >10.

      Without context, this is a lay person invitation to obtain a PSA, and “why shouldnt they,” as they might be the one with the aggressive disease.

      Absolute and relative risk, along with prevalence must be in these pieces–in digestible language. These types of conveyances take us down the overuse ratholes we are trying to avoid in the first place.

      • If you see a nice, simple presentation (perhaps graphical) of these study results that you think best convey the absolute and relative risk, along with prevalence, please let me know. It’d also be nice to see the other risks and effects of treatment conveyed as well. Some of this was covered in Pope’s piece, and I do encourage readers to read the whole thing.

      • I don’t think there is a study on this… but I’d say based off some assumptions from other work (including that discussed above) if a patient walks into a clinic and says “I’d like a PSA to know if mine is above ten and I have high risk disease,” that’s pretty appropriate….

    • I was not as impressed with the NY Times article and the actual medical article as you were.

      First, the media coverage and the original article essentially are written as if a finding that a given treatment does not have effects that are “statistically significantly different from zero” is the same as a finding that the effect is zero.

      In the overall sample, the point estimate is that as of 12 years after random assignment, the surgery group has a 3% absolute reduction in both prostate cancer mortality and overall mortality. Unfortunately, the confidence intervals are wide enough that we don’t have much precision. An effect of zero is within the confidence intervals. For prostate cancer mortality, so is an effect of 6%; for overall mortality, 10%.

      Unless you think a 3% plus reduction in mortality is of minor significance, the estimates don’t tell us that much.

      Now, if you get to the subgroups, it is true that effects are smaller for lower risk patients. But here the confidence intervals really explode.

      The problem is that with only a little over 700 participants in the study, it’s hard to get the precision we would like to have with the estimates.

      Second, the NY Times article does not emphasize that the study sample is somewhat unusual in that the average age is 67 and 90% are 60 or older. So, the study may have implications for those men in that age range, but it is unclear to what extent the results can be generalized to a man age 50 or 55 with prostate cancer. The results for older men are important, as many men diagnosed with prostate cancer are in that age range. But the study is limited in what it means for younger men.

      Third, the NY Times article and the original article do not get into the notion of selection bias. To get the around 700 participants, they had to ask around 5000 men diagnosed with prostate cancer to participate. So only around 15% of those recruited for the experiment agreed to participate. How might those who chose to participate differ from those who turned the experiment down? For example, might the participants be less likely than non-participants to include men who had relatives who died of prostate cancer?

      I think this study is suggestive in its implications. I do think it is fair to say that the results suggest that lower risk men in their 60s and older might want to opt for active surveillance. However, the limited sample size means that the results are too imprecise to be more than suggestive. And I don’t think the study tells us much about optimal choices for younger men.