key: cord-0910122-okanr96o
authors: Reddy, Rohin K.; Charles, Walton N.; Sklavounos, Alexandros; Seed, Paul T.; Khajuria, Ankur
title: Impact of smoking on COVID‐19 severity.
date: 2020-08-25
journal: J Med Virol
DOI: 10.1002/jmv.26456
sha: 2a3b832f1e701ed57d44cbc547a8698b05bef7f3
doc_id: 910122
cord_uid: okanr96o

nan

the external validity of our study is undermined by selection bias, a form of collider bias, due to exclusion of non-hospitalised patients. However, restricting analyses to solely hospitalised patients was due to the objective clinical stratification criteria for severity we used being predicated on parameters often only measured in secondary care, such as radiologic assessment and PaO2/FiO2 ratio, amongst others. Other objective measures we planned to investigate were in-hospital outcomes, such as ICU admission, mechanical ventilation requirement and mortality. It was unlikely that detailed parameters allowing assessment of severity and in-hospital outcomes would be available in community-based studies. Analysing hospitalised patients was therefore a conscious choice predefined in our a priori protocol. 8 Therefore while not generalisable to the entire population, our review is certainly applicable to hospitalised patients with COVID-19, which adds value when considering that This article is protected by copyright. All rights reserved.

approximately 20% of COVID-19 patients have severe or critical disease and the case fatality rate for critically ill patients is 49%. 12 Thus, our work may allow early identification of smokers as a patient population vulnerable to worse in-hospital outcomes, allowing timely triage and initiation of supportive measures. Nevertheless, future studies characterising the role of smoking in susceptibility to initial COVID-19 infection, and outcomes in community settings are warranted, such that efforts to protect smokers may be co-ordinated across the full spectrum of clinical care.

Interestingly, the author proposes that because prevalence of Chinese smokers in our study is lower than Chinese population prevalence, our sample is not representative of population smoking habits. However, caution is advised when applying this logic, as most included studies did not adjust the effect of smoking for baseline covariates and therefore comparing prevalence of smoking in hospitalised COVID-19 patients with overall population estimates is inappropriate, as the populations may be inherently different with regards to demographic factors. 2 Furthermore, this comparison is susceptible to collider bias, as due to sampling dependent on hospitalisation, anything that influences hospitalisation (i.e. smoking 13 ), will become negatively associated with COVID-19 infection, thus appearing protective. 11 This may account for 'protective' effects of nicotine, alluded to by the author 1 and reported across preprints and lay media. 14 Potential reasons for the lower reported prevalence of smoking in our study were outlined, namely misclassification bias, reverse causality and survivorship bias. 2 We wholeheartedly agree with the eloquent description of pitfalls involved with inferring causation from poor-quality observational research. 1 However, in our metaanalysis, the majority of studies (60%) were good-or fair-quality. Additionally, while Accepted Article we hope future works will explore these new chapters, rather than attempting to close the book.

Finally, the author's concern with claiming a meta-analysis based on poorly conducted observational data is 'definitive' is valid. However, we did not state our analysis was definitive and rather, wrote that we "aimed" to definitively quantify the effects of smoking, which we considered a worthy ambition. Indeed, systematic reviews and meta-analyses are key tools in the evidence-based-healthcare armamentarium, offering comprehensive summaries of the best available evidence on a given topic. Considering that almost half of all published systematic reviews now include non-randomised studies of intervention effects 10 , it is crucial that they are conducted and reported conscientiously, due to the aforementioned implicit biases in addition to measured and unmeasured confounders/colliders that abound in observational research. For this purpose, the AMSTAR 2 tool 10 was specifically developed to ensure quality in studies including non-randomised studies.

All previously published reviews investigating smoking and COVID-19 severity range in quality from critically poor to moderate. 2 Our meta-analysis is the first to be deemed 'high-quality', alongside being the largest by considerable distance. Therefore, we believe it is fair to conclude that whilst our work is by no means 'definitive' on this topic, it is certainly the most definitive currently available. In the era of COVID-19, pragmatism should reign supreme and biologically plausible effects that are clinically relevant, caused by exposures amenable to modification, must be recognised by healthcare providers, governments and policymakers to protect vulnerable patient populations and maintain public health.

Deriving 'definitive' results in observational COVID-19 research: a problematic endeavour

The effect of smoking on COVID-19 severity: a systematic review and meta-analysis

Waste in covid-19 research

Active smoking is not associated with severity of coronavirus disease 2019 (COVID-19)

No double-edged sword and no doubt about the relation between smoking and COVID-19 severity

Active smoking is associated with severity of coronavirus disease 2019 (COVID-19): An update of a meta-analysis

Active smoking and severity of coronavirus disease 2019 (COVID-19): The use of significance testing leads to an erroneous conclusion

The effects of smoking on COVID-19 severity: a systematic review and meta-analysis

Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement

AMSTAR 2: A Critical Appraisal Tool for Systematic Reviews That Include Randomised or Non-Randomised Studies of Healthcare Interventions, or Both

Collider bias undermines our understanding of COVID-19 disease risk and severity

Characteristics of and Important Lessons From the Coronavirus Disease 2019 (COVID-19) Outbreak in China: Summary of a Report of 72 314 Cases From the Chinese Center for Disease Control and Prevention

Substance Abuse and Mental Health Services Administration (US)

US Department of Health and Human Services

Smokers are actually at a higher risk of dying from covid-19

Mortality in relation to smoking: 50 years' observations on male British doctors

Reducing Bias in Observational Studies Using Subclassification on the Propensity Score