Today’s headline was modified from the Telegraph, one of the least lurid and sensationalistic of those generated by the peer-reviewed paper “Sexual partners, sexually transmitted infections, and prostate cancer risk” by Andrea Spence, Marie-Claude Rousseau, and Marie-Élise Parent in Cancer Epidemiology.
The Telegraph reported, “The University of Montreal has found that men who had sex with more than 20 women lower their prostate cancer risk.” Twenty wouldn’t do it; neither 19. Had to be more than 20. But what, you ask, about sodomy?
In contrast, men who slept with 20 men doubled their risk of developing prostate cancer compared with men who have never had sex with another man.
That level of specificity can only arise via models, neighbors and friends, statistical models. Let’s see what happened.
Some 3,208 Canadian men, half with and half without cancer, were asked in-person personal questions. Now since these were extremely intimate questions about highly sensitive topics (sexual “orientation”, STDs, etc.), it is assured, academically speaking, that everybody told the truth. Right? Right?
Turns out that men who reported having had only 1 sexual partner over their lifetime, 25% of men in each group, half had cancer. The same thing was found for people who had 2-3 partners (about 16% of men; cancer rate about 50%). Those who reported 4-7, 19% of men with cancer and 16% of men without, 54% had cancer. Oho! Another similarity for those reporting 8-20 partners (about 20% of men, cancer rate about 50%). Only 25 men claimed celibacy, 15 with cancer and 10 without: too few to say anything about (though this didn’t stop anybody).
Say, how did we jump from just 1, then 2-3, then 4-7, then 8-20? Hey! That was peer reviewed. So don’t ask. These divisions may seem arbitrary to us, but not to scientists.
Finally, those men who reported having greater than 21 partners, 14% versus 16% of men, the cancer rate was 47%. A wee p-value confirmed that this was “statistically significant”, meaning…what, exactly? We’ll come to that. First, it’s interesting that of those who refused to tell how many “partners” they had, 5% of the men with cancer and 7% without, the cancer rate was only 40%!
That’s just “partners” so far, mind. Not man on woman, but man on anything. So the clever authors split the data to count just those who admitted having had female “partners”—which was all but 10 of the people surveyed. Since this was nearly everybody, the overall results are scarcely different. Yet they are reported as being a whole new separate part of the study, especially in the media.
Stay with me.
The authors then split the cancer cases into those with high and low Gleason scores. A Gleason score is a subjective rating (2 to 10) about how ugly the prostate cancer tissue looks under a microscope. The authors chose 7 as the demarcation point because why not?
According to the hypothesis tests, no “significant” differences between the groups in any “partner” category except the greater than 21 “partners”, and only for those with Gleason scores less than 7. Say, that’s a lot of testing! Any adjustment of p-values? Nope.
The authors then broke the data out by those who admitted having committed sodomy. This was 141 men. Of these, 55% had cancer. 22% of those 141 who admitted having 21+ “partners” had cancer and 11% did not, for a rate of 68%.
Question: was it sexual activity, or its lack, that caused these differences in prostrate cancer rates? Wasn’t that the point of the study? To show what caused observed differences? If activity is causal, then having sex with 4-7 women causes more cancers and having sex with 21+ causes fewer, at the same rates. (This may be why men notch belts: so that they don’t lose count. We already knew sodomy is destructive to health.)
Parent told the Telegraph, “It is possible that having many female sexual partners results in a higher frequency of ejaculations”, so she certainly thought the statistics demonstrate causality. Sort of.
But when asked whether public health authorities should recommend men to sleep with many women in their lives Dr Parent added: “We’re not there yet.”
However, the authors in the paper say, “ours is the first study to report a protective effect of having several female sexual partners over the lifetime.” That’s cause-and-effect language, sisters and brothers. Except for a guessed-at more frequent emptying of seminal fluids (any why can’t that happen inside a marriage?), they know not how this cause works. But they’re sure it’s there.
Yet the authors had no awareness that people could have lied—misremembered, maybe, but not lied. The arbitrary cuts were never acknowledged as arbitrary. The huge number of tests hunting for wee p-values were not mentioned. They did admit that only 63% of those without cancer agreed to participate, but 86% of those with cancer did. And since we’re talking about differences between these groups, this large difference might be meaningful.
And they never realized that statistical models aren’t causal.
Thanks to reader and contributer Bob Kurland for bringing this paper to our attention.