Associations between COVID-19 infection, tobacco smoking and nicotine use, common respiratory conditions and inhaled corticosteroids: a prospective QResearch-Case Mix Programme data linkage study January-May 2020

Nicola Lindson, Min Gao, Jamie Hartmann‐Boyce, Margaret Smith, Paul Aveyard, Duncan Young, Carol Coupland, Pui San Tan, Ashley Kieran Clift, David A Harrison, Doug W Gould, Ian Pavord, Peter Watkinson, Julia Hippisley–CoxmedRxiv 2020 · DOI

Introduction

Early in the COVID-19 pandemic, a series of epidemiological observations sparked significant controversy: current smokers appeared to be underrepresented among patients hospitalized with the virus. This phenomenon, dubbed the "smoking paradox," led to hypotheses that nicotine might offer a protective effect against severe disease. However, these early studies were often limited by small sample sizes and "collider bias"—a statistical distortion occurring when researchers only analyze patients already admitted to the hospital.

This protocol outlines a massive, prospective cohort study designed to rigorously investigate the relationships between tobacco smoking, nicotine use, inhaled corticosteroids (ICS), and COVID-19 severity. By leveraging a population-scale dataset of 8.3 million people in England, the study aims to provide definitive evidence to either validate or debunk these early observations.

Study Objectives

The primary goal of this research is to characterize the association between smoking status and COVID-19 outcomes (hospitalization, ICU admission, and death). Specifically, the study seeks to:

  1. Assess Confounding: Determine how smoking-related comorbidities influence the risk of severe COVID-19.
  2. Dose-Response Analysis: Determine if a relationship exists between smoking intensity and disease severity, which would strengthen the case for causality.
  3. Isolate Nicotine Effects: Examine former smokers who use Nicotine Replacement Therapy (NRT) or electronic cigarettes to separate the effects of nicotine from tobacco smoke.
  4. Evaluate Respiratory Treatments: Assess whether the use of inhaled corticosteroids (ICS) for conditions like asthma or COPD modifies the risk of severe infection.

Methodology and Data Linkage

The strength of this study lies in its comprehensive data-linkage framework, which connects five major UK-based datasets to create a high-resolution picture of patient health:

  • QResearch: Anonymized primary care records from 1,205 general practices.
  • Public Health England (PHE): National records of SARS-CoV-2 positive tests.
  • Hospital Episode Statistics (HES): Comprehensive data on all NHS hospital admissions.
  • ICNARC Case Mix Programme: High-resolution data specifically from intensive care units.
  • Office for National Statistics (ONS): Official death certificate registrations.

By using a population-wide "denominator" of 8.3 million patients rather than just those already in the hospital, the researchers can avoid the selection biases that plagued earlier, smaller studies.

Statistical Rigor

To ensure causal clarity, the study employs Directed Acyclic Graphs (DAGs). These visual models map out potential confounding paths—such as how age or BMI might influence both the likelihood of smoking and the severity of COVID-19—allowing researchers to identify the "minimally sufficient" set of variables for adjustment. The primary analysis utilizes Cox proportional hazard models to estimate risk.

Furthermore, the study addresses common data challenges through:

  • Multiple Imputation with Chained Equations (MICE): This technique handles missing data in GP records (e.g., missing BMI or smoking status) without introducing the bias that comes from simply excluding patients with incomplete records.
  • E-values: These quantify how strong an unmeasured confounder would need to be to explain away the observed associations, providing a "stress test" for the findings.

The Nicotine Hypothesis

A central focus of the study is the biological plausibility of nicotine as a potential treatment. Nicotine interacts with the "cholinergic anti-inflammatory pathway," which may suppress the pro-inflammatory cytokines responsible for the "cytokine storm" seen in severe COVID-19 cases. By distinguishing between tobacco smoke (which contains thousands of chemicals) and pure nicotine (via NRT or vaping), the researchers hope to identify whether any observed protective effect is attributable to nicotine itself.

Limitations and Considerations

As a study protocol, the authors acknowledge several inherent challenges in the data:

  • NRT Under-ascertainment: Because many users purchase NRT over-the-counter rather than via prescription, GP records likely underestimate the true prevalence of nicotine use.
  • Testing Constraints: During the study period (January–May 2020), COVID-19 testing in the UK was largely restricted to hospital settings, potentially missing many mild or asymptomatic cases in the community.
  • Clinical Decision Bias: The decision to admit a patient to the ICU is a subjective clinical choice. If clinicians were less likely to admit smokers or those with severe comorbidities to intensive care due to a perceived lower chance of survival, the data might show a spurious "protective" effect for smoking.

Conclusion

This protocol represents one of the most comprehensive efforts to date to untangle the complex interplay between respiratory health, smoking, and COVID-19. While preliminary findings from related work suggested hazard ratios for ICU admission among smokers as low as 0.11 to 0.27, this study is designed to formally investigate whether such associations are truly protective or the result of systemic biases. The results will provide critical insights into the potential role of nicotine and inhaled steroids in managing pandemic-scale respiratory infections.

Summary updated

Abstract

Abstract Introduction Epidemiological and laboratory research seems to suggest that smoking and perhaps nicotine alone could reduce the severity of COVID-19. Likewise, there is some evidence that inhaled corticosteroids could also reduce its severity, opening the possibility that nicotine and inhaled steroids could be used as treatments. Methods In this prospective cohort study, we will link English general practice records from the QResearch database to Public Health England’s database of SARS-CoV-2 positive tests, Hospital Episode Statistics, admission to intensive care units, and death from COVID-19 to identify our outcomes: hospitalisation, ICU admission, and death due to COVID. Using Cox regression, we will perform sequential adjustment for potential confounders identified by separate directed acyclic graphs to: Assess the association between smoking and COVID-19 disease severity, and how that changes on adjustment for smoking-related comorbidity. More closely characterise the association between smoking and severe COVID-19 disease by assessing whether the association is modified by age (as a proxy of length of smoking), gender, ethnic group, and whether people have asthma or COPD. Assess for evidence of a dose-response relation between smoking intensity and disease severity, which would help create a case for causality. Examine the association between former smokers who are using NRT or are vaping and disease severity. Examine whether pre-existing respiratory disease is associated with severe COVID-19 infection. Assess whether the association between chronic obstructive pulmonary disease (COPD) and asthma and COVID-19 disease severity is modified by age, gender, ethnicity, and smoking status. Assess whether the use of inhaled corticosteroids is associated with severity of COVID-19 disease. To assess whether the association between use of inhaled corticosteroids and severity of COVID-19 disease is modified by the number of other airways medications used (as a proxy for severity of condition) and whether people have asthma or COPD. Conclusions This representative population sample will, to our knowledge, present the first comprehensive examination of the association between smoking, nicotine use without smoking, respiratory disease, and severity of COVID-19. We will undertake several sensitivity analyses to examine the potential for bias in these associations.
BibTeX
@misc{lindson2020associations,
  author = {Nicola Lindson and Min Gao and Jamie Hartmann‐Boyce and Margaret Smith and Paul Aveyard and Duncan Young and Carol Coupland and Pui San Tan and Ashley Kieran Clift and David A Harrison and Doug W Gould and Ian Pavord and Peter Watkinson and Julia Hippisley–Cox},
  title = {{Associations between COVID-19 infection, tobacco smoking and nicotine use, common respiratory conditions and inhaled corticosteroids: a prospective QResearch-Case Mix Programme data linkage study January-May 2020}},
  year = {2020},
  doi = {10.1101/2020.06.05.20116624},
  howpublished = {medRxiv},
}
Figure