“Truth” in screening … How many men to screen and treat to “save a life”?
Posted Feb 13 2011 12:00am
One of the statistical games ongoing since publication of the results of the PLCO and the ERSPC screening trials in the New England Journal of Medicine (nearly 2 years ago) has been the estimation of how many men need to undergo regular PSA-based screening, and how many of those men need immediate treatment, to “save a life” (i.e., to avoid a single prostate-cancer specific death).
The number of men that needs to be screened (NNS) and the number of men that needs to be treated (NNT) to save a life at 9 years of follow-up in the ERSCP trial were “only” 1,254 and 43, respectively.
At 10 and 12 years of follow-up NNS would decrease to 837 and 503, respectively.
At 10 and 12 years of follow-up NNT would decrease to 29 and 18, respectively.
With additional follow-up, if the mortality difference continues to grow, the NNT to save a life with PSA screening will decrease.
We have absolutely no doubt whatsoever that the model used by Loeb et al. really does suggest such numbers. We also have absolutely no doubt that if we ever get 20-year data from long-term follow-up of the ERSCP study, it will include a lot more men who died of prostate cancer than the 214 prostate-cancer deaths among the 72,952 men originally assigned to the screening group and the 326 in the 89,435 assigned to the control group as initially reported. However, whether the mortality difference will continue to grow is a quite different question.
There is (understandably) a lot of confusion about how to interpret these data accurately (“the truth”). We should therefore take a step back and acknowledge that at this point in time there is no “truth” about any of these secondary data. All we have are estimates based on statistical models, all of which require the modelers to make assumptions which all come with greater or less degrees of accuracy and sophistication. Nothing can be proven at this time and people tend to create models that will support their own beliefs. It’s human nature.
So in taking our “step back,” we thought it might be helpful to focus on some really simple facts:
In the PLCO screening trial, as originally reported by Andriole et al. , there was no prostate cancer-specific and no overall survival benefit associated with PSA screening whatsoever. In other words, it didn’t matter how many men you screened or treated because no lives were saved. (However, it is also true that the data from this trial were probably revealed at least 5 years too early.)
In the ERSPC screening trial, as originally reported by Schröder et al. , there was a 20 percent reduction in the risk of prostate cancer-specific mortality among men in the screening arm compared to men in the unscreened arm of the trial.
Schröder et al. also estimated on the basis of their data that it would require 1,410 men to be screened (NNS) and 48 to be treated (NNT) to save a single case of prostate cancer-specific death at the 9-year time point.
In a follow-up study, some of the ERSPC investigators, after “improving” their model to take account of a whole variety of complications (largely associated with the messy nature of the trial design) suggested that the NNT might be as low as 24 to avoid a single case of prostate cancer-specific death.
At this point in time, The “New” Prostate Cancer InfoLink is comfortable drawing just three conclusions from all these data:
One can not apply currently available data from screening of European populations to draw conclusions about the effects of screening in the USA (or any other well-defined region for that matter); the baseline criteria simply are not the same.
The number of men who need to be screened to save one man from prostate cancer-specific mortality is somewhere between about 12 (in the Göteborg trial) and about 38,000 (in the PLCO trial), depending on the past history of screening in the population being studied.
There is no evidence at this time that prostate cancer screening has any impact at all on overall survival of male populations.
There are a lot of people who would like to be able to use the currently available data to justify annual screening of all men for prostate cancer. The “New” Prostate Cancer InfoLink does not believe that the available data can be used to justify such national policies. However, we do believe that there is sufficient information to suggest that baseline PSA and other kallikrein data, examined at 40-50 years of age, and combined with information about family history, may allow us to accurately identify a subset of men men with a significantly greater than average risk for a diagnosis of prostate cancer. Those men may well benefit from regular monitoring of their risk, although the optimal frequency of such monitoring is still not well defined. In addition, the ability to determine which of those men actually need invasive treatment at the time of initial diagnosis (if they meet low- or very low-risk criteria) is also still not well defined.