-
Do cancer screening exams really extend patients’ lives?
It is time for the “non-science” to stop (actually it is long overdue). It is unclear how papers such as this pass peer review. The Randomized Controlled Trials (RCT) of breast cancer screening have proven that early detection saves lives. There were statistically significantly fewer women ages 40-74 who died from breast cancer in these trials. The present paper apparently tried to evaluate the effect that fewer breast cancer deaths had on overall, “all cause”, mortality deaths from all causes. This clearly ignores the fact that, although it is the most common non-cutaneous cancer among women, breast cancer still accounts for only 3% of [b]all[/b] deaths each year. If you reduce breast cancer deaths by 30% this means that you would reduce all-cause mortality by 1%. Tabar showed years ago that it would take an RCT of more than 2 million women to show that reducing breast cancer deaths by 30%, statistically, significantly, reduces all-cause mortality (Tabar L, et al. All-cause mortality among breast cancer patients in a screening trial: support for breast cancer mortality as an end point. J Med Screen. 2002;9(4):159-62.). You would need more than 3 times as many women than have participated in all the trials put together. This is simple statistics. It is unclear why, scientifically unsupportable, analyses continue to be published, especially by those who should know better.
Tabar did show in the same paper that if you look at women diagnosed with breast cancer, screening, significantly, reduced “all-cause” mortality among these women.
-
3 Replies
-
Quote from Dr Dan Kopans
If you reduce breast cancer deaths by 30% this means that you would reduce all-cause mortality by 1%.
No it doesn’t. If you die from other things in any case, you aren’t reducing all cause mortality. That’s why the end point is called all cause mortality. I’m not promoting anything with this point except to point out that you called something “non-science” and you don’t even understand all cause mortality.
What else doesn’t reduce all cause mortality? Statin drugs in people who don’t have Metabolic Syndrome. Note the commonality usually between things, while possibly just an association, is probably not if you know anything about human beings and incentives ($).-
I have to admit I do not understand your point. I think we agree that all-cause mortality means deaths from all causes??
You write If you die from other things in any case, you arent reducing all cause mortality
That is the point. This paper is arguing that screening for breast cancer did not reduce deaths from all causes (“other things” including breast cancer), or, essentially, the total number of deaths. The question is Does reducing deaths from breast cancer reduce the absolute number of deaths from all-causes?
In order to show that screening reduces deaths from all causes you would need to see fewer [b]total[/b] deaths among women who participated in screening compared to the total for the women in the control groups, and the difference between the two groups would need to be statistically significant. The point is that the trials were not large enough to have the statistical power to show that a 30% decrease in breast cancer deaths, which would reduce all deaths (all-cause) by 1%, could be shown to decrease deaths from all causes. It is a specious argument.
-
-
“It is unclear why, scientifically unsupportable, analyses continue to be published, especially by those who should know better.”
A moments thought yields the answer, which you and I know instinctively:1) journals need papers to publish – and the more surprising the better
2) authors need to publish papers to keep their jobs and burnish their image
The worst effect of this is that people with no particular statistical expertise, but with a cause they are pushing, will quote such papers with appended adjective of “Peer-reviewed” which to the even less sophisticated audience will seem to be a shining medal labeled “truth”
