Tuesday, 6 April 2021

Another awkward study about smoking and COVID-19

A study in pre-print from Germany looks at a particularly deadly outbreak of COVID-19 in the county of Tirschenreuth in spring 2020. The researchers took blood samples from 4,203 residents to see who had antibodies (and therefore who had contracted COVID-19). All the participants filled in a questionnaire with various questions about their lifestyle, e.g. alcohol consumption, how much TV they watch, physical activity, etc. 
Overall, 8.6% of those tested had antibodies. To the surprise of the researchers - but perhaps not to regular readers of this blog - only one lifestyle factor was associated with having antibodies for SARS-CoV-2 - and it was a negative association.

We were also interested whether seropositivity was associated with factors capturing a less healthy lifestyle (low physical activity, current or ex-smoking, drinking alcohol, or increased body-mass-index as a marker for excess calorie intake). We found no association with any lifestyle factor, except for current smoking. Remarkably, the odds of being seropositive was substantially decreased for current smokers compared to never-smokers (OR=0.36, 95%-CI: 0.24-0.53), while there was no association for ex-smokers (Supplemental Table 8). This would correspond to a 2.8-fold increase in the odds of being seropositive among never-smokers or ex-smokers compared to current smokers (Figure 5).

The researchers went to great lengths to find an explanation for smokers having a third of the risk of getting COVID-19, preferably one which didn't involve smoking.

The finding that current smoking was associated with a smaller probability of being seropositive is counter-intuitive under a hypothesis that smoking was associated with a less healthy attitude and more risky behavior. 
We conducted several sensitivity analyses in the search for potential confounding effects: first, we found stable associations across age groups and sex (unadjusted OR of seropositivity for current smokers vs. ex-/never smokers = 0.43, 0.34, and 0.47 for age groups 14-39 (young), 40-59 (mid), 60+ (old), respectively; 0.32 for men and 0.51 for women) (Supplemental Table 9). 
Second, we considered the possibility that the association of current smoking with reduced probability of seropositivity was due to an impaired ability of smokers to raise antibodies based on a previous report [20,21]. However, among the 74 individuals reporting a positive SARS-CoV-2 PCR test, we observed only 5 individuals without antibodies and these included 1 current smoker and 4 never-smokers (20% smoker as in the full sample). 
Furthermore, the association persisted when restricting to individuals having reported a previous PCR test: current smoking was associated with a positive test report versus negative test report (among the n=501 tested: OR=0.35), which is in line with a hypothesis that current smoking was associated with a lower risk of infection (Supplemental Table 10). This finding was not compromised by a higher proportion of current smokers among those tested, as previously reported by others [22]: 21.0% of those tested and 20.2% of those not tested were smokers (OR=1.07 adjusted for age, sex). Of note, we observed more women than men tested (14.4% and 9.4%, respectively), but no difference across age groups (9.7%, 12.2%, 11.5%, for age 14-19, 20-69, 70+, respectively). 
Finally, we also found a significant dose-response between the number of daily smoked cigarettes and seropositivity among all participants (zero cigarettes for never and ex-smokers OR=0.50 per 10 cigs/day, 95%-CI: 0.37-0.65, modelling reported # cigarettes smoked daily in a linear fashion, adjusted for age and sex) and a similar association restricting to current smokers (OR=0.69 per 10cigs/day, 95%-CI: 0.43- 1.07, Supplemental Table 11). Allowing for non-linear dose-response supported this finding (Supplemental Figure 2). 
A reverse epidemiology effect could induce a bias here, when individuals being current smokers at the height of the outbreak included severely ill individuals (e.g. cancer, severe heart disease) that prevented individuals from going outside or primed individuals to be particularly careful avoiding infection. However, the smoking- associated severe diseases would affect more likely older individuals and are less likely among younger individuals; therefore, the stable effect estimates across age groups provide evidence against such bias that would fully explain the strong association across age groups. 
Another potential bias needs to be considered when smokers who were infected were less likely to participate in the study. However, most individuals have not known their seropositivity at the time of questionnaire completion and participation.

Having come up empty-handed, they are forced to concede that smokers really are significantly less likely to be infected, as many other studies have shown. 

Particularly important to discuss is the finding that current smoking compared to never smoking, but not ex-smoking, was associated with a decreased risk of being seropositive and for reported infection. The association is strong, stable across age groups and sex, and exhibited a dose-response-effect. While the sample size of smokers among individuals with reported PCR-test was limited, the stable risk estimates and the few individuals with reported infection not showing antibodies suggests that the finding was not due to a lack of antibody building, but due to lower infection probability.

But here comes the cope...

This does not allow for the conclusion that smoking per se was protective. 

It does though, doesn't it? 

While smoking is generally linked to a less healthy and more risky behavior, a behavior associated with smoking that guards against infection could result in the same observation, e.g. gathering socially more outside or less frequently.

Hmm. It's a stretch.
And finally...
Any conclusion that smoking was preventive for SARS-Cov-2 infection needs to be rigorously challenged and, if substantiated, conclusions made need to weigh in the adverse public health impact due to the severe other implications of smoking like lung cancer.
It goes without saying that all conclusions in science should be "rigorously challenged", but it is difficult to imagine quite so much rigour being called for if the study had found smokers to be at greater risk of infection.

Hat top to Phil for bringing this to my attention.

No comments: