What do we actually know about smoking and COVID-19?
Almost as soon as China released the first COVID-19 statistics, scientists noticed some unexpected demographic characteristics among hospitalized cases. For example, early reports showed that men were over represented, which some suggested was likely due to gender imbalance in China’s smoking rates (more men than women tend to smoke in China). This supposition wasn’t particularly surprising. Because COVID-19 is a respiratory disease, and smoking causes heart and lung damage, many health experts expected smokers to be at higher risk of severe health outcomes from the novel coronavirus.
However, when more comprehensive data was later released, it became clear that smokers were actually underrepresented among hospitalized cases.
And, as COVID-19 became a pandemic and other countries released additional data on smoking prevalence among COVID-19 patients, the debate about smoking’s impact on disease progression intensified, as estimates of smoking’s association with severe COVID-19 could support several possible relationships. Recently, however, the OpenSAFELY Collaborative—a group of esteemed scientists from top British universities and institutions—published a study detailing the factors associated with COVID-19-related hospital death among 17 million adults. With such an enormous sample size, this study offers strong evidence that can help disentangle the association between smoking and COVID-19.
To conduct their retrospective cohort study, the Collaborative linked electronic medical records to national records of inpatient hospital deaths and COVID-19 diagnosis. The resulting dataset is equivalent to about 40 percent of the English population, which provides sufficient power to detect even small differences in associations between the covariates and death from COVID-19. The authors modeled the direct (minimally adjusted) associations between covariates and created several models that adjust for the effects of all covariates. Due to some quirks of the dataset, the authors also conduct sensitivity analyses to validate the results and ensure that the necessary model assumptions do not change the results.
What the researchers discovered about risk of death from COVID-19 among smokers is that in the minimally adjusted model, which only accounted for age and sex, smokers and former smokers were at 25 percent and 80 percent, respectively, higher risk of dying from COVID-19 compared to nonsmokers. However, when the researchers adjusted their model for all potentially confounding covariates, they found that former smokers were only at 25 percent higher risk of dying from COVID-19 than nonsmokers, and current smokers had a slightly lower risk of dying from COVID-19 than nonsmokers.
Any time an association switches from positive to negative after adjusting for related confounding factors, it should pique a researcher’s interest—especially when the association is counter to what the academic literature suggests. Upon further analysis, the researchers found that chronic respiratory disease and socioeconomic status (called ‘deprivation’ by the authors) accounted for much of the apparent protective effect of smoking in the fully adjusted model.
Perhaps what is more interesting than the small potentially protective effect of smoking is the absence of a large negative effect. Medical literature offers few examples of smoking having beneficial effects on health — and only marginally more examples of neutral effects. This is especially true of respiratory conditions. This leads one to ponder why COVID-19 is the exception and as the authors of the study show, the protective effect of smoking found in some studies may be a relic of residual confounding.
A number of factors could also explain the small protective effect of smoking on death from COVID-19. One possibility is that smoking’s strong association with chronic respiratory disease confounds the relationship between COVID-19 death and smoking. Alternatively, there could be effect moderation between smoking and chronic respiratory diseases that accounts for the observed protective effect. The authors’ analyses also suggest that at least one confounding variable explains the observed protective effect in the sample.
Additionally, when ethnicity was included in the full model, the association between current smoking and death from COVID-19 became insignificant, which indicates that not adjusting the model for ethnicity confounded the observed relationship between smoking and COVID-19 death. A similar result was observed when the authors added deprivation and chronic respiratory disease, individually, to the minimally adjusted model—only adjusted for sex and age—and found that the association between smoking and COVID-19 death was also no longer significant. One final oddity is that former smokers had a higher risk of death from COVID-19 than current smokers in both models. In most studies of smoking’s association with death, the opposite is observed. Given that the researchers found what may be only a small protective effect from smoking, they rightfully state that, “even if smoking does have a small protective effects [sic] against COVID-19, this would still be massively outweighed by the well-established adverse health effects of smoking.”
Some researchers have already keyed in on the well-established fact that nicotine is not responsible for the grave health effects of smoking, and have begun experimenting with nicotine as a treatment and prevention tactic for COVID-19. This has left some tobacco harm reductionists considering how these findings will affect the tobacco control debate, and the future of reduced-risk products that provide nicotine without the many toxic chemicals found in combustible products. While robust and critically important, the OpenSAFELY study is just the beginning of our understanding of the relationship between smoking, nicotine and COVID-19.
