Last Call For Alcohol-Cancer Papers

Time (sigh) for yet another example of the scrofulous, not to say ubiquitous, Epidemiologist Fallacy. This is peer-reviewed-paper generating technique of declaring X CAUSES Y, but where X is never measured. And where the correlation between Not-X-But-Called-X and Y is called a cause using the Wee P Fallacy.

A two-for-one double impact Fallacy.

You’d think researchers would never get away such a brazen ploy of saying X CAUSES Y, when anybody can read their papers and see X was never measured. Well, just as you’re about to notice the blatant attempt at obfuscation, some hairy-but-balding paunchy researcher jumps in front of you and starts waving his P at you. “Look how small it is!” he demands.

As ludicross as this sounds, it works, and works the great effect.

Our example today (thanks to Dr Anon for alerting us) is the peer-reviewed paper “Global burden of cancer in 2020 attributable to alcohol consumption: a population-based study” by Harriet Rumgay (yes) and others in The Lancet Oncology.

X CAUSES Y: Alcohol Causes Cancer. To even think about proving that, you have to at least measure how much booze individuals drink, and whether or not they have cancer. Which is only the merest whisper of a beginning. Stop with just those direct measures, and nothing else, and you’d only have correlations. And what do we know about correlations?

Did Rumgay—seriously, now: are they really asking us to believe this name is the lead author on an temperance paper?—measure how much wine, spirits, and beer individuals drank, and then measure whether they contracted various forms of cancer?

No, sir, Rumgay did not. Rumgay not only did not measure X, Rumgay did not measure Y! What we have is the Double Epidemiologist Fallacy, when neither X nor Y is measured.

Here is what Rumgay did, with my emphasis: “Country-specific estimates of incident cancer cases were extracted from the GLOBOCAN 2020 database for lip” and other cancers. Next, “alcohol consumption estimates for 2010 were obtained from the Global Information System on Alcohol and Health as adult per capita alcohol consumption in litres of alcohol per year by country disaggregated by age”.

If countries could catch cancer (if cancer can be caught), then maybe Rumgay’s analysis makes sense. But countries can’t. Only individuals can. Yet even if countries could catch cancer, Rumgay’s analysis is absurd, because Rumgay did not even get direct measures of alcohol and cancer, but only estimates!

Those estimates would have wide plus-or-minuses attached, which means any attempt to correlate country-level data would see weak signals, at best, if those plus-and-minuses were properly accounted for.

And that is before insinuating country-level signals apply with equal strength to the individuals within those countries. Which is not possible. Not possible to know, that is.

How did Rumgay, and how do the Rumgays of science, fool themselves so badly? Facility with software, for one reason. Instead of pausing to think through what we just discussed, Rumgay jumped immediately to this:

We calculated the effect of alcohol consumption on the incidence of cancer worldwide in 2020 using a Levin-based population attributable fraction (PAF) method¹⁰ adapted from Shield and colleagues⁵ and based on a theoretical minimum-risk exposure of lifetime abstention from alcohol consumption (appendix pp 2–5).

There follows (what might appear to some) an intimidating equation, with integrals and other mysterious objects. Once you let yourself even see these toys, you are lost. You cannot resist but to play with them. It’s math! And sophisticated! Rumgay used them and succeeded in fooling herself with science.

Is there any benefit in examining Rumgay’s results? None at all. To do so would to be seduced by math and colorful pictures. The likelihood of forgetting the objections we begin with is large, its effects potent. You will find yourself telling stories of the same kind Rumgay surely told herself. “Look at that correlation there! And in that country. Oho!” Many such cases. The anchoring has begun.

The best you can do with data like this is to use it as a rough and crude guide: Let’s go look for cancer and alcohol here and not there. That kind of thing. There is nothing worth writing a paper about, but perhaps, and only perhaps, something useful for planning real studies.

What you cannot do is to make causal connections. That’s not possible.

Rumgay claims cause: “Globally, about 741 000, or 4.1%, of all new cases of cancer in 2020 were attributable to alcohol consumption.” And later (my emphasis) “we found that alcohol use causes a substantial burden of cancer”.

That’s a mighty precise number. Two precise numbers, rather. Too precise.

What about increasing surveillance, the growing voices calling for early testing, the increases in measurement? Might these correlate with cheaper and easier to get booze? And vice versa? Point is, you could substitute in another measure that has similar trends to alcohol purchases, or even not that similar, and you will be able to use Rumgay’s very equations and claim your measurement caused cancer. These things I have tried to teach in the Class over and over again.

I remind you severely that bad science does not suddenly become good because it is in your favor. Stop trusting “research shows” until you investigate every claim thoroughly.

Subscribe or donate to support this site and its wholly independent host using credit card click here. Or use PayPal. Or use the paid subscription at Substack. Cash App: \$WilliamMBriggs. For Zelle, use my email: matt@wmbriggs.com, and please include yours so I know who to thank. BUY ME A COFFEE.