https://orcid.org/0000-0001-6277-3912
Abstract
Current evidence indicates that smoking worsens COVID-19 outcomes. However, when studies restricted their analyses to current smokers, the risks for COVID-19 severity and death are inconsistent.
This meta-analysis explored the association between current smoking and the risk for mortality based on the studies that reported all three categories of smoking (current, former, and never smokers) to overcome the limitation of the previous meta-analyses which former smokers might have been classified as nonsmokers. We searched PubMed and Embase up to January 1, 2021. We included studies reporting all three categories of smoking behaviors of COVID-19 patients and mortality outcomes. A random-effects meta-analysis and meta-regression were used to examine relationships in the data.
A total of 34 articles with 35 193 COVID-19 patients was included. The meta-analysis confirmed the association between current smoking (odds ratio [OR] 1.26, 95% confidence interval [CI]: 1.01–1.58) and former smoking (OR 1.76, 95% CI: 1.53–2.03) with COVID-19 mortality. We also found that the risk for COVID-19 death in current smokers does not vary by age, but significantly drops by age in former smokers. Moreover, current smokers in non-high-income countries have higher risks of COVID-19 death compared with high-income countries (OR 3.11, 95% CI: 2.04–4.72 vs. OR 1.14, 95% CI: 0.91–1.43; p = .015).
Current and former smokers are at higher risk of dying from COVID-19. Tobacco control should be strengthened to encourage current smokers to quit and prevent the initiation of smoking. Public health professionals should take the COVID-19 pandemic as an opportunity to promote smoking prevention and cession.
This study makes an important contribution to the existing literature by distinguishing between current and former smoking and their separate effects on COVID-19 mortality. We also explore the effects by age of patients and country income level. Findings from this study provide empirical evidence against misinformation about the relationship between smoking and COVID-19 mortality.
Introduction
The world has been in a global health crisis with the spread of COVID-19 pandemic since March 2020.1 Risk factors associated with disease severity and mortality have been studied, including smoking.2 Multiple systematic reviews and meta-analyses indicate smoking worsens COVID-19 outcomes.3–6 However, these studies share similar limitations as their results are analyses of those with history of smoking (current and former smokers combined).3–5 Previous studies attempt to analyze the association of active smoking with COVID-19 outcomes, but the results remain inconclusive because of the limited number of studies included in the meta-analyses6–8 and lack of clarity on whether or not the unexposed group included former smokers.9,10 If former smokers were misclassified as never smokers, the prevalence of smoking and the effects of smoking on COVID-19 severity would be underestimated.6
Previous meta-analyses generally showed a significant association between former smoking and COVID-19 severity, but not for current smoking.7,8,11 This might be because the early studies after COVID-19 started came from more developed countries where prevalence of current smokers was often lower than low-income countries.12 In addition, access to care and capability of COVID-19 treatment could differ by countries’ income. These factors could result in underestimates of the effects of current smoking and COVID-19 mortality in higher income countries. Thus, the effects of country income level on an association between smoking and COVID-19 mortality should be further explored.
This study provides a comprehensive systematic review and meta-analysis to explore the association between current and former smoking as separate categories compared with never smoking with COVID-19 mortality. We also examined the effect of current and former smoking on the COVID-19 mortality by age of patients and country income level.
Methods
Data Source and Search Strategy
We conducted a systematic search using PubMed and Embase, with the search term: “(‘Coronavirus Infections’ OR ‘COVID-19’ OR ‘SARS-CoV-2’) AND (‘Smok*’ OR ‘Comorbidity’ OR ‘Risk Factors’ OR ‘Retrospective Studies’ OR ‘Cohort studies’ OR ‘Prospective Studies’ OR ‘Case-Control Studies’ OR ‘Treatment Outcome’ OR ‘Characteristics’) AND (‘Deceased’ OR ‘Death*’ OR ‘Survivor*’ OR ‘Dead’ OR ‘Died’ OR ‘Mortality’ OR ‘Expired’)” for studies published between January 1, 2020 and January 1, 2021. Medical Subject Headings (MeSH) were used whenever applicable.
Study Selection
Eligible studies included published peer-reviewed observational studies, retrospective cohort studies, prospective cohort studies, cross-sectional studies, case series, and case reports that reported demographic characteristics, comorbidities specifically smoking status, and outcomes of COVID-19 positive patients.
The outcome of interest in this study was mortality in COVID-19 positive patients. The exposed groups were current or former smokers as separate categories, and the unexposed group was never smokers. Therefore, studies which classified patient’s smoking status as current, former, or never smokers were eligible for this study.
We excluded studies of other coronavirus that were not specifically to COVID-19, studies that the number of smokers was zero or omitted, studies in which all patients had the same outcome, studies that all patients were children or pregnant, studies explicitly supported by the tobacco industry, and studies that the full text could not be retrieved. There were no language restrictions.
Data Extraction and Quality Assessment
RA, TS, and SA independently screened the abstract or the full text, with disagreements resolved through discussion among all authors, and extracted information for each study. The information extracted from the included studies were: first author, location, setting, type of study design, date of data collection, characteristics of included patients, disease outcome, smoking status, sample size, number of patients in the severe and less-severe group, mean or median age of study population, effect estimates, and adjustment for confounders.
We evaluated the quality of studies using the Newcastle-Ottawa Scale (NOS).13 Based on the NOS’s three elements (patient selection, comparability, and assessment of outcome), the maximum score was 9 points, and the studies with a score of 6 or above were considered as high quality.6
This study followed the Preferred Reporting in Systematic Reviews and Meta-Analyses (PRISMA) guidelines and is registered with PROSPERO (CRD 42021272760).
Statistical Analyses
Our meta-analyses were based on unadjusted odds ratios (ORs) and 95% confidence intervals (CIs) that were either reported in the studies14 or computed using the number of current, former and never smokers with and without COVID-19 mortality.15–47 We did a sensitivity analysis to determine the change in results when studies with NOS score below 6 were excluded.14,15,23,45 We stratified the analyses by countries’ income level based on the World Bank’s classification.48 We also use a meta-regression analysis to evaluate whether the relationship between current and former smoking and COVID-19 death varied by age using (mean 32 studies) or median (2 studies) age reported in each study as a continuous variable. To make the ORs associated with age more interpretable we divided the age by 10, so the ORs are for every 10-year increase in age rather than every year increase in age.
We also computed the pooled adjusted effect estimates using the studies that reported adjusted effect estimates (OR, HR, or RR) and 95% CI and compared it with the pooled unadjusted estimates.
The results of the included studies were performed with random-effect models using the Stata version 14.0 metan command and metabias command with Egger’s test for the presence of publication bias. A p < .05 in publication bias tests was suggestive of publication bias. To determine the heterogeneity across the included studies, we relied on the I2 statistic with 25%, 50% and 75% representing low, moderate, and high heterogeneity, respectively.49 We also used metareg command to determine whether the adjustment of OR and countries’ income level affected the results.
Results
7141 unique studies were retrieved. After screening of titles and abstracts, 173 studies remained for full-text assessment from which 139 were subsequently excluded. Eventually, 34 studies with 35 193 COVID-19 patients met the eligibility criteria and were included in this meta-analysis (Figure 1). Overall, the prevalence of active smokers was 0.08 (95% CI: 0.07–0.09) and the prevalence of former smokers was 0.24 (95% CI: 0.20–0.28).
PRISMA diagram.
Of the 34 studies, 3014–18,22–32,34–47 were from high-income countries and 419–21,33 were from upper-middle-income countries. The mean age of the patients in the included studies was 63.5 years (assuming the reported median ages as the mean) (Supplementary Table S1).
Active Smoking and COVID-19 Mortality
Overall, current smokers were at higher risk of COVID-19 death compared with never smokers (OR 1.26, 95% CI: 1.01–1.58, p = .043; I2 = 62.2%, p < .001). There was no evidence of significant publication bias (p = .569) (Figure 2). A meta-regression of the odds of COVID-19 death and the patients’ mean age did not show a statistically significant relationship for current smokers (p = .901).
Active smoking and COVID-19 mortality.
The association between current smoking and COVID-19 death was statistically significant in non-high-income countries (OR 3.11, 95% CI: 2.04–4.72, p < .001; I2 = 0.0%, p = .873), but not in high-income countries (OR 1.14, 95% CI: 0.91–1.43, p = .254; I2 = 57.9%, p < .001). The difference was statistically significant (p = .015). No significant publication bias was found for either subgroup (p = .241 and p = .373, respectively) (Supplementary Figure S1).
Former Smoking and COVID-19 Mortality
Former smokers were also at higher risk of COVID-19 mortality (OR 1.76, 95% CI: 1.53–2.03, p < .001; I2 = 68.1%, p < .001) than never smokers. No evidence of significant publication bias was found (p = .505) (Figure 3). A meta-regression of the odds of COVID-19 death for former smokers against the patients’ mean age showed that the odds of COVID-19 mortality dropped statistically significantly by a factor of 0.75 (95% CI: 0.63–0.91, p = .004) per 10 years.
Former smoking and COVID-19 mortality.
Unlike current smoking, the association between former smoking and COVID-19 death was statistically significant in high-income countries (OR 1.72, 95% CI: 1.52–1.95, p < .001; I2 = 59.4%, p < .001), but not in non-high-income countries (OR 1.92, 95% CI: 0.52–7.10, p = .330; I2 = 86.6%, p < .001). The difference was not statistically significant (p = .122). No significant publication bias was found for either subgroup (p = .330 and p = .320, respectively) (Supplementary Figure S2).
Unadjusted Versus Adjusted Effect Estimates
Eight studies (Supplementary Table S219,26–29,38,41,43) that reported adjusted effect estimates (common adjustments were age, sex, race, and disease comorbidities) yielded an elevated point estimate for the effect of current smokers on COVID-19 death for current smoking (OR 1.98, 95% CI: 0.99–3.93, p = .052; Supplementary Figure S3, top), but it did not reach conventional statistical significance. When comparing the adjusted ORs with the unadjusted (OR 1.81, 95% CI: 0.99–3.31, p = .054; Supplementary Figure S3, bottom) from the same studies, the result was not significantly different (p = .800). For the adjusted ORs, the heterogeneity among the studies was high and statistically significant (I2 = 85.1%, p < .001) with no evidence of publication bias (p = .197). For the unadjusted ORs, the heterogeneity among the studies was high and statistically significant (I2 = 82.6%, p < .001) with no evidence of publication bias (p = .569).
For former smoking, the point estimate for the adjusted ORs (OR 1.44, 95% CI: 1.03–2.00, p = .031; Supplementary Figure S4, top) was significantly lower (p < .001) than the point estimate for unadjusted ORs (OR 2.69, 95% CI: 1.80–4.03, p < .001; Supplementary Figure S4, bottom). For the adjusted ORs, the heterogeneity among the studies was high and statistically significant (I2 = 81.3%, p < .001) with evidence of publication bias (p = .021). For the unadjusted ORs, the heterogeneity among the studies was high and statistically significant (I2 = 88.8%, p < .001) with no evidence of publication bias (p = .505).
Sensitivity Analysis
Dropping the 4 studies14,15,23,45 with the NOS score below 6 had little effect on the odds of COVID-19 mortality among current smokers (OR 1.32, 95% CI: 1.04–1.67, p = .025; I2 = 64.8%, p < .001), and former smokers (OR 1.82, 95% CI: 1.57–2.12, p < .001; I2 = 70.1%, p < .001). No significant publication bias was found for either group (p = .909 and p = .817, respectively) (Supplementary Figure S5).
Discussion
Our analysis included many more studies than the previous meta-analyses4,7,10 that examined the association between current smokers and COVID-19 mortality. Including only studies that reported all three categories of smoking (current, former, and never smokers) allowed us to conduct the meta-analysis for current and former smokers separately to overcome the limitation of the previous meta-analyses which former smokers might have been included in the unexposed group.4,6,10 Our findings that both current and former smoking significantly increase the risk of COVID-19 death are consistent with well-established knowledge that smoking impairs lung function and the body’s defense mechanisms against infections.50 Nicotine and other constituents in tobacco smoke could also potentially worsen COVID-19.51,52 Specifically, nicotine involves in downregulating interferon regulatory factor 7 (IRF7) and suppresses antiviral immune responses.51 Non-nicotine components of tobacco smoke contain numerous respiratory toxicants and irritants that could accelerate COVID-19 mortality.52
We also found that the risk of COVID-19 death does not vary by age among current smokers, but the risk significantly drops as age increases among former smokers. This finding on former smokers is similar to a previous meta-analysis of ever-smoking on disease progression when former smokers were included in the exposed group.8 These findings are also consistent with the previous studies that the lungs of current smokers contain more angiotensin-converting enzyme 2 (ACE2) receptors, the receptor for SARS-CoV-2.53–55 ACE2 expression has a dose-dependent relationship with smoking, and the ACE2 decreases by 40% among former smokers who had quit for at least 12 months.53 We did not have information on duration of smoking abstinence, but it is possible that the older patients had a longer time since quitting.
In non-high-income countries we found current smoking was significantly associated with COVID-19 mortality. In contrast, in high-income countries, we found former smoking was significantly associated with COVID-19 mortality. These findings may be explained by the difference of prevalence of current and former smoking between high-income and non-high-income countries as the tobacco epidemic is moving from high-income countries to poorer countries.56 Studies show that current smokers in high-income countries tend to quit at a younger age than non-high-income countries. For example, the mean age of quit smoking was 39.5 years old in United States,57 but 48.7 years old in China.58 While, the number of current smokers in high-income countries has been replaced by former smokers, and tobacco-caused deaths are decreasing, the situation in non-high-income countries has gone in the completely opposite direction.59 Tobacco control plays a vital role in turning current smokers to former smokers. Tobacco control movements have started from developed countries since the US Surgeon General’s report concluded that cigarette smoking caused lung cancer in 1964.60 The strengthening tobacco control coincided with the tobacco industry seeking new business in developing countries where their tobacco control programs were weaker.61
Another rationale why the association between current smoking and COVID-19 mortality is not significant in high-income countries could be current smokers in high-income countries tend to be younger than non-high-income countries. A recent study using data from 204 countries shows that the mean age at smoking initiation among high-income countries tends to be lower than non-high-income countries (17.0 years old in high-income countries vs. 19.2 years old globally) and youth smoking prevalence was also higher in high-income countries (20.8% in high-income countries vs. 12.7% globally).62 Younger age is associated with decreased mortality from COVID-19,9,17,18 thus, potentially explaining our findings. In addition, high-income countries are also more likely to have better smoking cessation programs63 as well as capability for COVID-19 treatment than lower-income countries.
Some have argued that current smoking is not associated with COVID-19 and even protective against COVID-19 due to low percentages of reported smoking among COVID patients.64,65 Such claims have been quickly picked up via media and found to be associated with self-reported increases in tobacco consumption.66,67 Recently, these unproven claims have been found to be related with the tobacco industry68 who benefited from manipulating public opinion through media hypes and disseminating misinformation.67 Our meta-analysis provides empirical evidence against these claims.
Limitations
Although we selected only studies that classified smoking status into three categories: current, former, and never smokers, the reported smoking status may be incomplete under the pandemic. From the 34 studies included in the meta-analysis, 15 studies14–18,23,24,30,36,38,41,42,44–46 had more than 10% of missing reported smoking status. If missing smoking data are not removed from the denominator, the estimates can be wrong that smokers are less likely to develop severe disease.69 We were aware of this error, so we excluded patients with missing smoking data from our calculation of the ORs.
Moreover, definitions of smoking, duration of smoking, types of tobacco product used, and abstinence and pack-year of cigarette smoking were not reported in the included studies. Specifically, only two19,20 of the 34 studies included in our meta-analysis provided a definition of former smokers. In Caliskan and Saylan,19 former smokers were those who smoked at least 100 cigarettes during lifetime and reported not smoking at least the past 3 months. In Chen et al.,20 former smokers were those who stopped smoking more than 5 years. This information should be collected for the future analysis to obtain more accurate results.
Due to the limited number of studies reported adjusted effect estimates and a variety of covariates, the main results in this meta-analysis are unadjusted ORs. We did a sensitivity analysis of the adjusted effect estimates and found no statistically significant difference between adjusted and unadjusted ORs of current smoking and COVID-19 death. However, the adjusted effect estimate of former smoking and COVID-19 death is statistically significantly lower than the unadjusted effect (former smoking is still statistically significant associated with COVID-19 death after adjustment). The inconsistent results between the adjusted and unadjusted effects could be a variety of covariates and statistical measures which differ among studies which means that the resulting ORs are not strictly comparable,8 and the pooling analysis could be bias.70 Therefore, when estimating adjusted effect estimates in nonrandomized control studies, further methodological research is required.8,70
Lastly, even though our analysis was based on studies published up to 1 year since the pandemic started, most studies were conducted during the first 6 months of the pandemic and no studies were from low-income countries. Therefore, the results might not reflect the latter wave of the pandemic.
Conclusions
Current and former smokers diagnosed with COVID-19 are at higher risk of dying. The effects seem to decline with age among adults who had quit smoking. Tobacco control should be strengthened to encourage current smokers to quit and prevent initiation of smoking. The government, physicians, and public health professionals should take the COVID-19 pandemic as an opportunity to promote smoking prevention and cession.
Supplementary Material
A Contributorship Form detailing each author’s specific involvement with this content, as well as any supplementary data, are available online at https://academic.oup.com/ntr.
Funding
None declared.
Declaration of Interests
None declared.
Authors’ Contributions
RP developed the idea for the study, collected, analyzed the data, and wrote the first draft of the manuscript. TS and SA assisted with data collection and extraction, review and editing the manuscript. SAG assisted with revising and refining the manuscript.
Data Availability
All data used to prepare this paper are available from the cited sources.
Comments