So many awkward studies about smoking and COVID-19

If you check out my list of COVID-19/smoking studies you'll find more than 70 epidemiological studies looking at how smokers fare when faced with the virus. In short, the evidence overwhelmingly shows that they are much less likely to get infected. 

Almost the only exceptions are a handful of Mendelian Randomisation studies that can't distinguish between smokers and nonsmokers and instead assume that people with genes that are associated with a propensity to smoke are smokers. For various reasons, that is not a sound assumption after decades of anti-smoking education and legislation. When it comes to lifestyle risk factors, the blunt tools of MR are only any good if you don't want to find an association.

The largest epidemiological study of its kind was published in Nicotine and Tobacco Research a few days ago. It involved 2.4 million people in California and it found that smokers were not only less likely to get infected but were less likely to be hospitalised, admitted to intensive care and die as well. 

 

Current smoking was associated with lower adjusted rates of SARS-CoV-2 infection (aHR=0.64 95%CI:0.61-0.67), COVID-19-related hospitalization (aHR=0.48 95%CI:0.40-0.58), ICU admission (aHR=0.62 95%CI:0.42-0.87), and death (aHR=0.52 95%CI:0.27-0.89) than never-smoking.

 

Coming from a state that has taken anti-smoking regulation to almost fascistic lengths, this must have broken the authors' hearts. To be fair, they seem to take it quite well and acknowledge that the 'smoking paradox' hypothesis should probably now be considered a fact.

 

In what has been termed the “smoker’s paradox,” studies across the globe have generally found lower risk of SARS-CoV-2 infection for current versus never smoking, an inverse association between smoking prevalence and the prevalence of SARS-CoV-2 infection, and a lower than expected prevalence of current smoking among patients hospitalized with COVID-19.

... A national matched case-control study from Korea found that current (OR=0.33, 95%CI:0.28-0.38) and former smoking (OR=0.81, 95%CI:0.72-0.91) was associated with a lower odds of SARS- CoV-2 infection than never-smoking. Data from 38 European countries found that after covariate adjustment, smoking prevalence was inversely related to SARS-CoV-2 infection. Further evidence comes from a cohort study of an aircraft carrier crew exposed to SARS-CoV-2 while at sea. Current smoking was associated with a lower odds of SARS-CoV-2 infection (OR=0.64, 95%CI:0.49-0.84), with even lower odds for those smoking more heavily...

That was a particularly amusing finding and an interesting study.

Many researchers who find that smokers are at less risk of Covid are keen to downplay their findings. They sometimes ignore them in the text altogether But the authors of the Californian study seem more interested in testing the various explanations that have been put forward and their study does this rather well.

Our study has important strengths. First, it is now recognized that non-representative sampling (e.g., hospitalized patients, people tested for active infection, voluntary participants) in many observational studies of risk factors for COVID-19 can lead to collider bias distorting true associations between risk factors and outcomes. A unique strength of our study is the inclusion of a large defined cohort of patients at-risk for COVID-19 within a closed healthcare system followed from testing and infection to death. 

Since all patients were insured, results are unlikely due to variations in access to care. 

Our retrospective cohort study design properly estimates risk over time, making it more rigorous than convenience sample studies. 

Further, the semi-parametric Cox proportional hazards model flexibly allows the underlying baseline risk to vary over the study period, accounting for changes in risk/exposure as the pandemic unfolded. By assessing smoking status during standard care pre-pandemic, our smoking data do not reflect short-term changes resulting from infection (e.g., if smokers with severe COVID-19 consequently quit smoking and report former smoking status). The small percentage missing smoking status was excluded rather than included with never-smoking, reducing the likelihood of misclassification. 

 

As for the idea that smokers are more likely to get tested, we know from mass seroprevalence studies (in which everyone gets tested) that this doesn't explain the association, but the authors offer further evidence.

 

Prior studies have speculated that people who smoke may be more likely to get tested for COVID-19 when asymptomatic (e.g., due to Centers for Disease Control and Prevention [CDC] guidance characterizing them as at-risk) or due to smoking-related symptoms mimicking COVID-19 symptomatology (e.g., cough), increasing their percentage of negative tests. While we are unable to directly test this, it is reassuring that in our study, COVID testing prevalence was comparable by patient smoking status (24.7% current, 28.1% former, and 24.6% never- smoking) and with a similar number of tests, on average. 

This must be the first worthwhile study about smoking produced in California for about thirty years. Does it matter? It's not as if governments are going to start telling people to start smoking, and everyone's going to get COVID-19 sooner or later anyway.

I think it does matter and so the authors. Firstly, because the truth is important no matter how politically incorrect it may, but also because this information could turn out to be really useful.

Understanding whether smoking is associated with risk of SARS-CoV-2 infection and COVID-19 severity is critical for informing public health strategies to mitigate risk during future outbreaks and prioritize at-risk groups for vaccination outreach, boosters, and treatments as they become available.

It will also never stop being funny.