National Post (National Edition)
Unlikely COVID defence: smoking
Maybe it’s time to talk about cigarettes and COVID-19. As a smoker I did not really want to be at the forefront of this conversation. When early data from China, where half of adult males smoke, showed that current cigarette smokers were bizarrely underrepresented among severe COVID-19 cases, I was happy to let Vice have its countercultural fun with the story. When the same thing seemed to be happening in France, the United States and elsewhere, I let the revered Economist handle it.
Things got a little more serious on Thursday, when researchers based in England, with the renowned Ben Goldacre as joint principal investigator, released a population-based study of COVID-19 risk factors based on National Health Service electronic health records. The “smoker’s paradox,” up until that moment a sort of odd smell that was motivating small-scale French research into nicotine, fell out of the data and lay wriggling like a boated flounder on the floor.
This is prospectively collected data, albeit in rushed “preprint” form, from a sample of 17 million people. It turned out current smokers had a significantly lower risk of COVID-19 death than both never-smokers and former smokers when you correct for other variables, including many of the health problems that smoking intensifies. (Note that former smokers will still have these, but should still, as any doctor will tell you till your ears plop off, be better off than those who have kept the habit.)
I had better make it clear that in a simple model that corrects only for age and sex, and not for particular health and socioeconomic conditions, smokers (both current and former) are still worse off than a nonsmoking member of the general public. When you start feeding other variables into the model, however, the sign of the effect of being a smoker changes. Smoking appears to be a net plus for people with chronic respiratory disease (that they probably got from smoking) and for persons of inferior socioeconomic status (inferred, as is usual in such studies, from the postal code of the patient’s neighbourhood).
“Other individual adjustments,” the authors tell us, “did not remove the positive association between current smoking and outcome.” Which is a hint that they were naturally vexed about the finding and tried extra hard to account for it. Adjusting for ethnicity, as a special extra step in the analysis, pushed the effect in the direction of non-significance. But this involved throwing a quarter of the patients with missing ethnicity data out of the original sample, so that’s not especially surprising.
Meanwhile, on May 5, a totally different study from the U.K. Biobank — which is a 21st-century gene-conscious take on classic longitudinal research efforts like the Framingham Heart Study — had already reported that “regular smokers” seemingly have much lower odds of contracting COVID-19 in the first place. (They did better than “occasional” smokers, another small sign of a potentially real phenomenon — i.e., possible evidence of a dose-response.)
There were only 580 COVID-19 cases among the Biobank participants at the time of this snapshot, and it will be updated continuously as the virus works its way further through suffering Britain, but the difference in the odds ratio was significantly in favour of smokers anyway. Note that the researchers seem to have been pursuing the good old Vitamin D chimera: the poor devils went looking in the fridge for milk and found a pack of Marlboro reds.
For smokers to do better in any health measure taken from any sample in any situation is astonishing. (If you’re studying, say, maxillofacial injuries from korfball accidents, the smokers will almost certainly have worse outcomes than the non-smokers.) One result of this kind would be a surprise, but here, in the past week, we have the cigarette effect turning up in two enormous datasets gathered prospectively, adding to weird factoids about apparent disease prevalence that had already been noticed. Remarkably, the effect is positive vis-a-vis two different endpoints: one paper was measuring infection, the other death.
And, by the same rules of inference that tell us smoking is super terrible for you, we do have to consider it a “cigarette effect” for now. The different endpoints suggest that nicotine might be lowering viral load, or something of that nature, and nicotine is known to have some funky immunological effects. But the benefit, if it exists, can’t be attributed to nicotine until that’s established independently.
What if it’s the actually harmful material in cigarettes, the tar and other combustion byproducts, that is protecting smokers? What if just coughing for a few minutes when you get out of bed is what provides the benefit? Given that doctors in intensive care units are furiously trying to adjust patients’ body positions to improve COVID-19 outcomes, everything is in play.
What is a smoker to make of all this? It is, sadly, information that is useless for any purpose other than influencing COVID-19 research programs, which it has already been doing. This is itself impressive, given the deep suspicion with which most physicians and public health officers regard tobacco. (It is not really surprising that the French ran with that particular football first.) No one, including me, will advise a non-smoker to start smoking to ward off the novel coronavirus.
But, uh … if I had pre-existing plans to quit I might personally have delayed them. Hey, your first crack at quitting probably won’t take anyhow. It is probably all right, if only on the principle of Matthew 6:34, to hold off and take a more serious runup in 2021.