Unspun: Alcohol and Cancer Risk
Let's set aside the sin-shaming, and look at the data
People are judgy about drinking.
Plus, health authorities commonly try to influence behaviors by scaring the bejesus out of the public. Last month the New York Times asked “Should Alcoholic Beverages Have Cancer Warning Labels?”, noting governments in Ireland and some other countries are considering the step. This comes on the heels of the World Health Organization admonishing “No safe amount of alcohol consumption for cancers can be established.”
But conventional wisdom and public messaging on topics like alcohol and its relationship to cancer get muddy, and even political.
So let’s unspin it.
There’s one big caveat for research translation in this arena: the studies aren’t good enough for certainty. And they never will be, since no study is going to randomly assign 50,000 people to get soused every night for a decade, and another 50,000 to be teetotalers. (I’d sign up, though I might quit early based on group assignment).
Which means observational research is all we have, instead of trials. But no research actually ‘observes’ alcohol consumption, it’s self-reported—and notoriously under-reported. In one study researchers found Americans purchase three times more alcohol than they claim to drink. Apparently there’s a lot of beer just sitting in the fridge.
With that grain of salt, the question is: Does light or moderate alcohol consumption increase the risk of cancer?
Why ‘light or moderate’? Because seriously heavy drinking can definitely lead to cirrhosis, liver cancer, and liver failure (among other health and personal disasters). This is undisputed. But the people agonizing over alcohol consumption aren’t severe alcoholics. It’s the light and moderate drinkers, usually at 1-2 drinks a day or less, who consider modifying their habits, and are targets for public health messaging.
The CDC, the National Cancer Institute, and the WHO, however, wag their fingers furiously at these amateur drinkers. The NCI: “Even those who have no more than one drink per day and binge drinkers… have a modestly increased risk of some cancers.”
Curious word choices, ‘modestly’ and ‘some’.
The research the NCI cites for this waffle is worth digging into. For example, one of the five studies represents 130,000 participants. In this cohort light and moderate drinkers had a risk of cancer 2-4% higher than non-drinkers for women, and 3-6% for men.
To be blunt, these differences are meaningless. Those who never drink and those who regularly drink are very different people, with yawning gaps in their cultural, genetic, and behavioral traits. The notion that in groups of people this different a piddly 2% discrepancy in cancers can be attributed to alcohol is preposterous. And with hundreds of thousands of participants, ‘statistically significant’ but utterly fatuous differences will constantly pop up (and usually at magnitudes much greater than 2%).
To put the 2%-3% differences in perspective, a smoker’s risk of lung cancer is between 2,000% and 4,000% higher than a non-smoker’s. Yup. That’s what a cause-and-effect relationship looks like, mathematically.
A second paper cited by the NCI meta-analyzed 572 studies and found a 13% increase among light drinkers for oral cancer, 4% for breast and 4% for prostate. Again, tiny. But in that same paper there was another finding, not mentioned by the NCI or any other health authority: for thyroid, kidney, lung, gastric, and lymph node cancers, the associations were negative. In others words, by the NCI’s logic, light to moderate alcohol consumption prevents thyroid, kidney, lung, gastric, and lymph node cancers.1
Bottoms up!
But of course these are not cause-and-effect relationships, in either direction. That is a conceit of public health authorities pressing agendas. As noted, differences this small are far more likely to be due to genetics, geography, culture, and a host of other lurking variables that poison observational data. Moreover, their pitiful magnitude (2%??) effectively rules out a causal relationship, because we know what causation looks like. And it’s not this.
Drunk driving kills, alcoholism leads to liver and other cancers, and drunken behaviors can be awful. And maybe, light drinking can lead to heavy drinking. We should be wary and attentive to these widespread problems. But instead of sin-shaming, we should be honest about the data, because the results are consistent: Light to moderate alcohol intake is simply not a meaningful risk factor for cancer.
So pour yourself a drink, and go argue about what it means.
Cheers!
FOOTNOTE: This negative association with cancers is not a novel finding, and while for obvious reasons I’m quoting and digging into studies cited by the NCI, there are lots more. For instance, note a 2009 study of 1.3 million women in the UK, in whom overall cancer rates were exactly the same in nondrinkers as light to moderate drinkers. And again, kidney cancer, thyroid cancer, and lymphomas were less common in light drinkers than nondrinkers.
Based on this, we could reasonably modify the WHO statement to say: “In regard to alcohol and cancer, there is no level of abstention from alcohol that can be considered safe.” This would be just as evidence-based as claiming there’s no safe level of consumption.
Going back to the 2009 study, the highest associations between cancer and drinking amounted to 2% increases for the <1 drink/day group, and 5% for the 2 drinks/day groups—but only after the researchers shifted their reference group to those drinking <2 drinks/week, instead of nondrinkers. When nondrinkers were the comparison group there was no increase with drinking until levels hit more than 2 drinks a day (even then it was 15%, trifling). Their explanation for shifting the reference group is dubious, but the point is nondrinkers and light-to-moderate drinkers had either identical risks, or within a few percentage points, no matter how you look at the data.
Larger point: I’m showing a small slice of the research (though the meta-analysis of 572 studies covers most of it), but the findings I’m highlighting are perfectly consistent with the rest of the world’s literature on this topic. And, to be explicit about something I alluded to earlier, it is highly likely that the people reporting 1-2 drinks per day (the alcohol consumption levels for which I am pulling data) are consuming more than 1-2 drinks per day, which by definition renders all of the associations even weaker (i.e. these tiny associations were associated with even more alcohol consumption than the studies claim).