Associations between Tobacco Smoking and Self-Reported SARS-CoV-2/COVID-19 Infections, Disease Severity, and Duration in the German Population

Background

The Coronavirus disease 2019 (COVID-19) is a contagious respiratory disease caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) virus (World Health Organization, 2021). Due to its high transmissibility, governments worldwide issued various behavioural restrictions such as periodic lockdowns to reduce the spread of the virus and to avoid overloading hospital systems from February 2020 (Taylor, 2021). About a year into the pandemic, on December 27, 2020, Germany’s vaccine program was rolled out, thus providing effective protection against severe disease and mortality, especially to those at high risk (e. g., older adults, men, individuals with comorbidities). While the COVID-19 vaccine campaign made its way slowly through the German population in 2021, surges of infections erupted due to the Delta variant of the SARS-CoV-2 virus. As of October 3rd, 2023, there have been an estimated total of 38 million cases and 168,935 deaths due to COVID-19 in Germany (John Hopkins University & Medicine, 2023).

There have been conflicting theories about the role tobacco smoking and/or nicotine plays in the susceptibility of COVID-19 infection, disease severity, and symptoms. One of the defining features of the SARS-CoV-2 virus is its spike protein, which is involved in receptor recognition, viral attachment, and entry into host cells via the host cell receptor angiotensin-converting enzyme 2 (ACE-2) (Huang et al., 2020). In some studies, active cigarette smoking has been found to upregulate ACE-2 expression, suggesting that smokers may be at an increased risk of a SARS-CoV-2 infection (Leung et al., 2020; Smith et al., 2020). However, reduced receptor levels in smokers have also been reported (Oakes et al., 2018). It has also been suggested that nicotine competes with SARS-CoV-2 for the nicotinic acetylcholine receptor, which acts as a co-receptor for viral cell entry (Farsalinos, Barbouni et al., 2020; Farsalinos, Niaura et al., 2020; Grundy et al., 2020). Furthermore, behavioural factors might play a role such as risk-reducing behaviour (e. g., meeting other people rather outdoors than indoors, or meeting less frequently) in smokers with pre-existing diseases (e. g., pulmonary or heart diseases) out of a fear of respiratory complications of COVID-19 (Richard et al., 2022; Wagner et al., 2021).

Current evidence about smoking and the risk of corona infection and disease outcomes includes a large living evidence review of over 500 studies from around the globe by Simons et al. (Simons et al., 2021). Findings from their unadjusted meta-analyses showed current smokers compared to never smokers were at a decreased risk of SARS-CoV-2 infection (Relative risk [RR] = 0.67, credible interval [CrI]= 0.60–0.75); and among hospitalised patients, current smokers compared to never smokers had an increased risk of greater COVID-19 severity (RR=1.3, CrI=1.01–1.71) (Simons et al., 2021). Mendelian randomization studies have further supported findings that smoking increases the risk of severe COVID-19 (Clift et al., 2022; Yeung et al., 2022).

Lacking in the literature are studies with random or representative population samples. The majority of studies have been conducted in hospital settings and with selected populations. The few high quality population studies did not primarily focus on the association between smoking and SARS-CoV-2 / COVID-19 (Barchuk et al., 2021; Carrat et al., 2021; Gornyk et al., 2021; Merkely et al., 2020; Radon et al., 2021; Richard et al., 2022; Wagner et al., 2021). Subsequent methodological limitations include incomplete data regarding smoking behaviour (in particular the distinction between recent vs. long-term ex-smoking), a lack of or incomplete adjustment for confounding factors, samples that did not capture the entire population or focused on only patient populations, and not capturing or reporting on asymptomatic infections.

The present study therefore aimed to add to the existing evidence by addressing the following research questions using self-reported data from a representative survey of the German population: (1) In the general German population aged 14+ years, compared with never smoking is (a) current and (b) former smoking associated with an increased risk of SARS-CoV-2 infection? (2) In people with a SARS-CoV-2 infection, compared with never smoking is (a) current and (b) former smoking associated with an increased risk of more severe COVID-19 symptoms? (3) In people who have COVID-19 with symptoms, compared with never smoking is (a) current and (b) former smoking associated with an increased risk of longer COVID-19 symptom duration? Evidence about the role of tobacco smoking and/or nicotine is potentially useful for future efforts of disease prevention and risk communication.

Methods

We conducted a cross-sectional analysis using data from the German Study on Tobacco Use (DEBRA: “Deutsche Befragung zum Rauchverhalten”): an ongoing representative household survey on tobacco use in the German population (Kastaun et al., 2017). The DEBRA study collects bimonthly data from computer-assisted face-to-face household interviews in a sample of approximately 2,000 persons aged 14+ per wave. Respondents were selected by using a dual frame design: a composition of random stratified sampling (50 % of the sample) and quota sampling (50 % of the sample). Details regarding this sampling design have been described in detail elsewhere (https://osf.io/e2nqr/). Data collection on COVID-19 infections and symptoms started in wave 28 (February/March 2021) of the DEBRA study and continued until wave 35 (March/April 2022). Additional data on corona vaccination were collected in waves 34 (January/February 2022) and 35. Respondents were not reimbursed for participation. The DEBRA study has been registered at the German Clinical Trials Register (registration numbers DRKS00011322, DRKS00017157, and DRKS00028054). We published a detailed study protocol a priori to analysing the data (https://osf.io/pzrv3).

Results

A total of 16,361 people were interviewed in the period between 18 February 2021 and 5 April 2022 (waves 28–35 of the DEBRA study), of which 16,028 were current smoker, long-term ex-smoker or never smoker who responded to the question regarding corona infection (79/16,107=0.5 % did not respond). The characteristics of the study population are shown in Table 1. Current smokers were somewhat younger and more frequently male and with a migration status. Furthermore, the rate of vaccination against SARS-CoV-2 (only measured in waves 34–35) was lower in current smokers (89.2 %) than in long-term ex-smokers (93.3 %) and never smokers (93.9 %).

Table 1 Characteristics of the study population by smoking status

Characteristic	Current smoker(n=5,242)	Long-term ex-smoker(n=2,731)	Never smoker(n=8,134)
Notes. Data presented as column percentage (number), unless stated otherwise. †In people ever infected with SARS-CoV-2 (314 current smokers, 134 long-term ex-smokers, 420 never smokers, 4 missings). ‡In people with at least mild symptoms (250 current smokers, 107 long-term ex-smokers, 315 never smokers, 52 missings). ¥Data on vaccination status only collected in 2 waves of the survey (waves 34–35: 1,437 current smokers, 628 long-term ex-smokers, 1,969 never smokers, 75 missings).
Age, years: mean (SD)	47.7 (16.1)	58.1 (16.3)	51.5 (20.3)
Female sex	46.4 (2,420)	43.2 (1,175)	58.7 (4,744)
Migration background	15.9 (788)	12.9 (327)	14.2 (1,092)
No. people in household >18 years: mean (SD)	1.8 (0.8)	1.8 (0.7)	1.8 (0.8)
No. people in household <18 years: mean (SD)	0.4 (0.8)	0.3 (0.7)	0.4 (0.8)
Monthly household income p. p. in €1000: mean (SD)	1.6 (0.9)	1.8 (0.8)	1.7 (0.9)
Educational attainment lowmiddlehigh	33.3 (1,712)43.6 (2,243)23.1 (1,185)	29.7 (807)39.0 (1,057)31.1 (849)	27.7 (2,130)34.8 (2,682)37.5 (2,886)
Rural region of living	39.8 (2,078)	34.2 (932)	35.1 (2,836)
Ever infected with SARS-CoV-2	6.1 (317)	4.9 (134)	5.2 (421)
COVID-19 without or with mild symptoms†	73.2 (230)	70.1 (94)	68.1 (286)
COVID-19 symptom duration >4.5 months‡	9.2 (23)	13.1 (14)	12.7 (40)
SARS-CoV-2 vaccination received¥	89.2 (1,282)	93.3 (586)	93.9 (1,848)

Table 1 Characteristics of the study population by smoking status

View as image HTML

A total of 872 people reported ever being infected with corona (5.4 %; Table 1 and ESM 3). A post-hoc ancillary analysis assessing the validity of this self-report showed that our estimated infection rates at the time points of the various surveys waves were comparable to the official infection rates from the Robert Koch Institute (see ESM 4). Among the 872 people with an infection, 610 (70.0 %) reported a low symptom severity (including n=148 without symptoms; Table 1 and ESM 5). Among the 724 people with an infection and with symptoms of any degree, 77 (10.6 %) reported a symptom duration of 4.5 months or longer (Table 1 and ESM 1).

Our first regression model included 14,730 people after 1,298 (8.1 % of 16,028) with missing data on one or more of the potentially confounding factors included in the model had been excluded. The odds of an infection showed no relevant or statistically significant difference between current (adjusted odds ratio (aOR) = 1.02, 95 % confidence interval (95 %CI) = 0.86–1.20) and long-term ex-smokers (aOR=1.03, 95 %CI=0.83–1.28) compared with never smokers (Table 2).

Table 2 Associations between smoking status and corona infection, corona symptom severity, and corona symptom duration, adjusted for potential confounders

Smoking status	Infectionyes vs. no infectionaOR (95 %CI)n=14,730	Symptom severityhigh vs. lowaOR (95 %CI)n=800	Symptom durationin monthsaβ (95 %CI)n=626
Notes. aOR = adjusted odds ratio. aβ = adjusted β-coefficient of linear regression model. 95 %CI = 95 % confidence interval around OR or β. OR and β adjusted for: age, sex, migration background, people in household 18+ years, people in household aged <18 years, income, education, region of living, and wave of the survey.
CurrentLong-term ex-smokingNever smoking (reference)	1.02 (0.86–1.20)1.03 (0.83–1.28)1	0.84 (0.59–1.20)0.88 (0.55–1.38)1	-0.09 (-0.45–0.28)0.002 (-0.48–0.48)1

Table 2 Associations between smoking status and corona infection, corona symptom severity, and corona symptom duration, adjusted for potential confounders

View as image HTML

Our second regression model, in the sub-sample of people who ever had a corona infection, included 800 people after 72 (8.2 % of 872) with missing data had been excluded. Both current (aOR=0.84, 95 %CI=0.59–1.20) and long-term ex-smokers (aOR=0.88, 95 %CI=0.55–1.38) had a lower but statistically non-significant odds of high symptom severity compared with never smokers (Table 2).

Our third regression model, in the sub-sample of people who ever had a corona infection with symptoms included 626 cases. A total of 98 people (13.5 % of 724) with missing data had been excluded. The symptom duration between current smokers (adjusted β-coefficient (aβ)=-0.09, 95 %CI=-0.45–0.28) and long-term ex-smokers (aβ=0.002, 95 %CI=-0.48–0.48) showed no relevant difference from never smokers (Table 2).

Our a priori planned sensitivity analyses yielded partly different effect estimates, but none of the associations were statistically significant (ESM 6, 7, 8). Regarding our first outcome, our sensitivity analysis with restriction to waves 28–30 in which only a minority of the population had been vaccinated showed a lower but statistically non-significant odds of an infection both in current (aOR=0.86, 95 %CI=0.57–1.28) and in long-term ex-smokers (aOR=0.83, 95 %CI=0.51–1.37) compared with never smokers (ESM 7). Our sensitivity analysis with restriction to waves 34–35 which included additional adjustment for the vaccination status of the respondents showed a higher but statistically non-significant odds of an infection both in current (aOR=1.06, 95 %CI=0.84–1.33) and in long-term ex-smokers (aOR=1.10, 95 %CI=0.81–1.50) compared with never smokers (ESM 8). This led us to perform a post-hoc ancillary analysis using all data (waves 28–35) which showed a statistically significant interaction between wave of the survey (on a metric scale) and smoking status: the risk of an infection increased over time, but this increase was higher in current smokers compared with never smokers (aOR=1.10, 95 %CI=1.01–1.19: ESM 9). Subsequent analyses of the effect of time, stratified by smoking status, showed the following increases in the risk of an infection with increasing wave of the survey: aOR=1.48 (95 %CI=1.39–1.59) in current smokers, aOR=1.37 (95 %CI=1.25–1.50) in long-term ex-smokers, and aOR=1.36 (95 %CI=1.29–1.43) in never smokers.

Discussion

Our study using representative data from the German population collected in the period between February 2021 and April 2022 showed no relevant and statistically significant differences in self-reported corona infections, corona symptom severity, and corona symptom duration between current smokers, long-term ex-smokers, and never smokers.

Only few population-based studies have investigated the association between smoking status and SARS-CoV-2 infection so far. These studies were conducted in Germany (Gornyk et al., 2021; Radon et al., 2021; Wagner et al., 2021), France (Carrat et al., 2021), Russia (Barchuk et al., 2021), and Switzerland (Richard et al., 2022) in a period between April 2020 and February 2021. All studies used SARS-CoV-2 antibodies from blood samples as outcome measure and consistently reported lower seropositivity in current smokers compared with never smokers. We used self-reported SARS-CoV-2 infection detected with a positive test by healthcare personnel as an outcome measure and found that smokers were at the same odds of an infection as never smokers. Such tests are usually rapid antigen tests aimed at detecting SARS-CoV-2 virus load. Hence, it may be that smokers are just as likely to acquire a SARS-CoV-2 infection (measurable with an antigen test) but are less likely to produce sufficient antibodies after an infection which then results in a lower seropositivity. This may be one explanation why studies using antibodies as the outcome measure reported lower infection rates in smokers. This is supported by the consistent finding from various vaccination studies that smokers show lower SARS-CoV-2 antibody titres compared with non-smokers (Ferrara et al., 2022; Herzberg et al., 2022; Swartz et al., 2023; Toda et al., 2022; Tsatsakis et al., 2021; Uysal et al., 2022; Watanabe et al., 2022; Yamamoto et al., 2022). However, in a series of planned and unplanned sensitivity analyses, there was some indication (albeit non-significant) that current compared with never smokers had reduced odds of infection when restricting the analyses to the survey waves prior to widespread vaccination – which is consistent with findings from a recent living review of >500 observational studies (Simons et al., 2021). In addition, a significant interaction between survey wave and smoking status was observed, with the risk of infection increasing over time across all levels of smoking status but with the increase being more pronounced in current compared with never smokers. This could be interpreted to suggest that the initially observed negative association between current (compared with never) smoking and infection has attenuated over the course of the pandemic due to mass infection and/or current smokers being less likely to produce a sufficient immune response following vaccination. Furthermore, the pronounced increase in infection rate in current smokers may be due to a lower vaccination rate in this very group.

We also did not find a statistical significant association between smoking status and corona symptom severity. Two large-scale observational and Mendelian randomisation studies reported an increased risk of hospitalisation and COVID-19-related mortality in current smokers compared with never smokers (Clift et al., 2022; Yeung et al., 2022). Previous studies conducted in hospitalised patients also found an increased risk of greater COVID-19 severity among current smokers compared with never smokers (Simons et al., 2021). Our study, however, was based on a general population sample in which the majority of people with a self-reported infection (70.1 % of 893) reported no or mild symptoms. Only 25 people (2.8 %) reported treatment in a hospital. Hence, our sample size was probably too low to detect any meaningful differences. Furthermore, the highest degree of corona symptom severity we were able to measure with our survey was intensive care treatment in hospital, and this only in those who recovered in such a way that allows living at home and responding to an interview survey. If smokers are more likely to being hospitalised and to die from COVID-19 than never smokers, as the above-mentioned studies suggest, our analysis of the association between smoking status and corona symptom severity might have been biased due to selection.

There is very little evidence about the association between smoking status and corona symptom duration in the general population. In our sample, 10.6 % reported a symptom duration of 4.5 months or longer, which can be regarded as indicative of long COVID (Shah et al., 2021). Longitudinal population studies from the UK and the US reported prevalence rates for long COVID between 10–38 % (Whitaker et al., 2022; Wu et al., 2022). The UK study found an increased risk of persistent symptoms in current smokers compared with non-current smokers (Whitaker et al., 2022). On the contrary, the US study, which also took pre-infection symptoms and existing health conditions at baseline into account, did not find an increased risk in current smokers (Wu et al., 2022). Hence, current smoking may not be a risk factor for corona symptom duration, or at least less important than other risk factors such as obesity (Whitaker et al., 2022; Wu et al., 2022).

Limitations and Strengths

Our study has several limitations. First, our study had a cross-sectional design which limits the ability to assess temporal associations between exposures and outcomes. For example, it is possible that a corona infection with symptoms affects smoking behaviour, in particular that it triggers smoking cessation. We tried to limit this risk of bias by excluding recent ex-smokers (those who stopped smoking <12 months) from our analyses. Second, our outcome measures were prone to bias because they relied on self-report and recall of corona infections that occurred in the past and corona symptoms, most of which are unspecific, although we have no reason to assume that recall differs by smoking status. Third, for the measurement of our outcome corona infection, we asked for a positive test by healthcare personnel in order to differentiate such a test from self-tests at home which are more prone to errors in handling and interpretation. However, there is a chance of misinterpretation because tests are often performed at test stations by persons who may or may not be healthcare professionals. Also, we did not have any information on the type of tests used, which would impact sensitivity and specificity of viral detection. Fourth, our measurement of exposure was restricted to smoked tobacco and did not include the use of other nicotine products such as e-cigarettes or heated tobacco products which may also have adverse effects on respiratory health. Finally, some relevant potential confounding variables were not measured such as comorbidities, place of work (home working), and key worker status, and vaccination status was not measured during the entire observational period. Our sensitivity analyses with data restricted to waves with information on vaccination status yielded similar results as our main analyses, though. Nevertheless, residual confounding may have occurred. Strengths of our study include the use of a representative sample of the general population (as indicated by the self-reported infection rates which are comparable with official infection rates from the Robert Koch Institute during most of the study period; see ESM 2); however, given the dynamic nature of a communicable disease like SARS-CoV-2, which moves through the population at varying rates depending on the number of infections, susceptible individuals and recovered individuals at each time point, the degree of representativeness of our survey with respect to SARS-CoV-2 infection dynamics remains unknown. Another strength of the study includes having followed a well-planned and a priori published analysis plan including well-founded adjustment for various important confounders.

Conclusion and Recommendations

Based on our study findings and in light of previous research we conclude that – in the general German population – smokers appear to be just as likely to acquire a corona infection as long-term ex-smokers and never smokers. The majority of participants experienced mild symptoms and symptoms that last less than three months. Our findings regarding the association between smoking status and symptom severity and duration are inconclusive due to methodological limitations. More longitudinal studies in representative samples of the population and with extended measurement of prognostic factors of corona disease progression are needed to disentangle the complex relationships with smoking.

Implications

• •
  Current tobacco smokers appear to be just as likely to acquire a corona infection as long-term ex-smokers and never smokers.
• •
  The finding from previous studies reporting a reduced risk of corona infection in current smokers based on SARS-CoV-2 antibodies from blood samples as outcome measure may have been biased. One explanation could be that smokers are less likely to produce sufficient antibodies after an infection which then results in a lower seropositivity.
• •
  The majority of smokers with a corona infection experiences mild symptoms and symptoms that last less than three months.

Electronic Supplementary Material

The electronic supplementary material (ESM) is available with the online version of the article at https://doi.org/10.1024/0939-5911/a000858

• ESM 1. Responses to the outcome corona symptom duration in subsample of people with a corona infection and symptoms of any degree (Table)
• ESM 2. Causal diagrams indicating the hypothetical associations between exposure, the outcomes and potential confounding factors (Figures)
• ESM 3. Responses to the outcome corona infection in total sample (Table)
• ESM 4. Weighted lifetime prevalence of corona infection rates at different time points of the DEBRA waves estimated with DEBRA data compared with official cumulative corona infection rates from the Robert Koch Institute (Figure)
• ESM 5. Responses to the outcome corona symptom severity in subsample of people with a corona infection (Table)
• ESM 6. Sensitivity analyses of the associations between smoking status and corona infection and corona symptom severity, adjusted for potential confounders (Table)
• ESM 7. Sensitivity analyses (restriction to waves 28–30 in which only a minority of the population had been vaccinated) of the associations between smoking status and corona infection, corona symptom severity, and corona symptom duration, adjusted for potential confounders (Table)
• ESM 8. Sensitivity analyses of the associations between smoking status and corona infection, corona symptom severity, and corona symptom duration, adjusted for potential confounders (Table)
• ESM 9. Post-hoc ancillary sensitivity analysis of the associations between smoking status and corona infection, adjusted for potential confounders (Table)