Chris Lawson emails me about screening for prostate cancer:
I hope you don't mind me expressing myself at length here. I apologise in advance for doing so, but epidemiology and statistics is one of my areas of interest and prostate screening in particular gets my goat because of the rampant misinformation about it in the mass media. I won't go into great detail about the pros and cons of prostate screening - you've obviously heard those arguments, but I would like to point out a few problems with the Guardian's article.
1. The Guardian quotes a 20 per cent reduction in prostate cancer deaths - but this is not the full story. In the screened group there were 261 deaths out of 737,397 men, as opposed to 363 out of 878,547 in the control group. This means that the actual number of preventable cancers is very small - as the paper points out you would need to screen 1,400 men for 10 years to prevent one cancer death. This is why epidemiologists prefer absolute numbers to relative numbers.
2. Even that's not the whole story. If you look at the numbers, total deaths were the same in the control and the treated group. So what this study really showed was that the reported cause of death changed. Screening didn't save anyone's life, it just saved them from the label of 'prostate cancer' on their death certificate.
3. The study itself is on very shaky ground. The screening group was recruited from many different countries with many different screening protocols. Also, some of the people in the control group had screening done outside the study. In other words, it was rather messy and full of biases and confounders. This is, to be fair to the researchers, because the subject is difficult to study. But given that the authors of the paper were open about most of the shortcomings, I find myself grumbling at the Guardian for not mentioning them.
4. The paper did not measure adverse outcomes such as post-operative incontinence or impotence - very common outcomes of prostate surgery. The study referred to earlier studies using different populations - but that is not sufficient. A randomised control trial should measure primary outcomes in both benefits and risks, not simply measure the benefits and state that the risks have been previously demonstrated. (There is another critical problem with this paper that I won't mention unless you let me know you want to know more about lag-time bias. Suffice to say, though, that even the small reported effect was possibly a design blip.)
5. This paper was published in the very same edition of the New England Journal as another study, which had its own flaws but was more reliably designed. It showed no difference in prostate cancer deaths (and if anything the trend was to a higher rate in the screened group). To quote the conclusion of the second, better study: '[W]e now know that prostate-cancer screening provided no reduction in death rates at 7 years and that no indication of a benefit appeared with 67% of the subjects having completed 10 years of follow-up. Thus, our results support the validity of the recent recommendations of the U.S. Preventive Services Task Force, especially against screening all men over the age of 75 years.' (That Task Force recommended against screening, as you may have gathered, pending further evidence.)
6. The second study also did not report extensively on complication rates, but at least the researchers say they collected the data and will be publishing the results in the near future.
7. Why is it, do you think, that the Guardian would so positively trumpet the findings of one trial, ignore its shortcomings, and not even report on the other (better) trial in the exact same issue of the exact same journal? In contrast, the NYT reported the studies with the headline 'Prostate Test Found to Save Few Lives'. This NYT piece, by the way, is an excellent piece of health journalism that covers both papers, interviews some of the researchers involved as well as other key opinions in the field, and does a great job of explaining the issues.
Some other links: the European study suggesting a benefit from screening; the US study suggesting no benefit from screening; the NEJM editorial making it quite clear that the question has not been resolved by either study; the US Preventive Services Task Force recommendations for prostate screening.