A big, new study recently concluded just one alcoholic drink a day may boost risk for certain cancers. Of course, this politically correct finding made a big splash in the press. Maybe you even saw the headlines.
But hopefully, you ignored them. Because the study is complete bunk
Actually, it’s a perfect example of how a large, expensive, high-profile study with a politically correct agenda–no matter how poor the design and how weak the findings–still garners attention.
The study followed more than 150,000 healthcare professionals in the U.S. The researchers say they found light-to-moderate drinking of alcohol (intake of 15 grams per day in women, and 30 grams per day in men) was associated with a “small, non-significant” increase in cancer risk.
Hold on right there.
The whole reason to conduct a large statistical analysis in the first place is to try to account for the possibility that an observed association is due to chance alone.
When statistical calculations do reach “significance,” it means the association is probably not due just to chance alone. But if statistical calculations show “not significant,” especially in a large study–as this large study did–it means any association is probably just due to chance. Therefore, there is no real connection.
Plus, any mainstream or medical editor who looked at the results of a study on nutritional supplementation that showed a “small, non-significant” benefit would laugh that study out of the editorial room. It would never see the light of day.
But if your results are politically correct (even if not statistically correct)–as this study’s were–well, there’s your new headline.
The second problem with this study?
In a typical, smaller study, if you find a small association that is “not significant,”
it often indicates that the small study did not have the statistical power to prove a significant association. In other words, I might accept a smaller association in a smaller study. You just don’t see it better because the study was too small to have the possibility for stronger statistical findings
But a larger study like this one should yield higher statistical significance, assuming there actually is any effect for what you are studying.
And this new study was about as big as it gets. Researchers followed 150,000 people, for goodness sake. If there was any real association, there is no excuse for seeing only a “small, non-significant” effect–unless there is really no significant effect. Or whatever the effect, it’s just not “significant” enough that reasonable people should take real notice.
But let’s keep going, anyway, as they did.
When the basic study didn’t show real results, they started breaking it up into different subsets to see if they could cherry-pick any significant associations among different sub-groups.
But such subset analyses are notoriously sketchy. They allow statisticians to play ever more games with data.
As a graduate student, my professors repeatedly warned us about the hazards of doing subset analysis. But these warnings came during my Ph.D. studies. Not a word was said in medical school for the future doctors who will actually read these kinds of reports.
Ultimately, they brought it all down to a single, weak, subset connection between alcohol and breast cancer in women. And we have heard about this connection before.
But the relative risk of drinking alcohol for women is nowhere near the large reproductive risk factors for breast and other cancers I’ve told you about before.
So–this flimsy statistical story about alcohol and cancer is based on mostly small, non-significant associations, despite have huge numbers of people to study and supposedly high statistical power to detect a real significant effect.
Yet, in the traditions of places like the former National Socialist Germany, or Soviet Union in Russia, this flimsy stuff served the politically correct agenda enough to inspire one researcher, Jürgen Rhem, Ph.D. to make the following chilling statement in an accompanying editorial
“Even when we consider all-cause mortality attributable to alcohol, drinking more than 10 grams of pure alcohol a day for women, or 20 g for men over a lifetime can lead to a magnitude of risk not considered acceptable for voluntary behavior in modern societies. “
Maybe Dr. Jürgen Rhem sounds better in German. (Or maybe not!)
Of course, Dr. Rehm also said we “definitely” still need more research. However, he says let’s just go ahead anyway, even without more research and start “acting” on these flimsy findings. Dr. Rehm advises people to reduce “their alcohol to below the recommended limits, or even abstain altogether…”
I vonder if zhey haff vays of making us comply?
Mark down another victory for the behavioral control freaks hiding behind flimsy, politically correct “findings.” And this is the kind of stuff now found in our modern medical science literature!
I would advise you not to suddenly start following this new advice. In fact, have a drink or two tonight. It might make things a little less alarming when they finally come knocking on your door, smashing your wine and beer bottles in the street.
And tune back in tomorrow for Oktoberfest. In Friday’s Daily Dispatch, I’ll tell you about several legitimate studies with findings that show the health benefits of drinking beer.
- “Light to moderate intake of alcohol, drinking patterns, and risk of cancer: results from two prospective US cohort studies,” British Medical Journal, 8/18/2015