In each case, these data were compared to the prevalence of smokers in each country by considering the proportion of males and females whenever possible. In every case except one, which had the fewest patients, very statistically significant differences were observed (p < 0.001) and would indicate that something is at play with regard to COVID-19 incidence in smokers.
To assure that only peer-reviewed data were analysed and therefore any bias due to poor-quality data was avoided in our meta-analyses, no pre-prints databases (vg. arXiv; bioarViv, etc.) were searched in the present work. However, we believe that, for future revisions, the inclusion of new studies and the extension of the search to other databases (vg. Pubmed), in which some current high-quality preprints would have developed into peer-reviewed works would be very interesting and beneficial.
Nevertheless, when a disease begins to spread in the population, the corresponding information is also transmitted between individuals, which in turn influences the pattern of the disease spread [63
]. In this context, the responsible use and dissemination of some preprints may be of interest. Additionally, although there is a wide class of research that studies the dynamics of the dissemination of information, most of them are based on the classic spread of epidemics. Currently, the transmission of information requires a much lower cost and varies much faster than physical contagion, therefore the modelling of the dissemination of information should also help to understand the spread of epidemics [65
] and the interpretation of the meta-analytic results of diseases.
4.1. Limitations and Biases
Both the systematic review and the presented meta-analyses have some limitations. Heterogeneity in the meta-analysis (i.e., variation in the study outcomes between the studies) was determined as I2 = 64% in the Chinese studies and as I2 = 69% when summing the US and Italian works. However, when heterogeneity was further explored through L’Abbé plotting, and the only outlier study was removed from both analyses, the I2 values decreased notably (I2 = 36.5% for the meta-analysis of Chinese works and I2 = 55.7% in the global meta-analysis) confirming the validity of our main results and conclusions.
It was not possible to perform a detailed study using the age groups of current smokers, although all patients were adults. As smoking habit prevalence changes with age, mean values were used. With males, this value could vary with age from 41.5% (males aged 70 years) and 60.3% (males between 40–49 years old) in China [20
]. Conversely, these values for females were much lower, and varied between 1.2% (aged 18–29 years) and 5.8% (older than 70). The number of males and females was similar in practically all the studies. More male patients were included in all the studies, they smoked more heavily and were at higher risk of suffering the disease [66
]. If tobacco, or some of its components, or smoking habit had some protective effect, more females would be expected to be hospitalised, but this was not the case. What we doubtlessly observed was that the difference between smokers hospitalised for COVID-19 and the expected values was very significant.
Other factors or artefacts could bias this study. For instance, as smokers know they are an at-risk population (as they are more likely to catch the disease from their habits: touching cigarettes and cigarette packets, exchanging tobacco, touching their face or placing cigarettes in their mouths, etc.; apart from the respiratory effects of tobacco itself), they could have been more aware of taking social distancing and hygiene measures. Nonetheless, as the temporal frame within which the studies were conducted was an early stage of today’s pandemic, and no differences were observed among them, this would not appear to be a plausible hypothesis.
Another possible bias may be related with data quality. We believe that smokers could have attempted to hide this characteristic given the alarm of these characteristics, and the threat of hospitals and ICUs being overcrowded. Nonetheless, most data were taken from electronic medical records, which meant that we had access to patients’ smoking background in many cases. Given the serious nature of the pandemic, in other cases we could presume that many smoking patients had stopped smoking before being hospitalised and were thus included in the groups of non-smokers or former smokers. So, it would be very interesting to specify the exact time when these data were collected, for example during a medical interview when admitted to hospital or from patients’ previous medical records. Moreover, the definition of “smoker” in such studies is not clear because heavy smokers are not distinguished from occasional smokers.
Epidemiologic studies could in some cases be inaccurate due to unrecognised bias. For example, while several case-control studies documented a “protective” effect of smoking on Alzheimer’s disease, subsequent cohort studies showed this was not the case and smoking may not be related to the onset of Alzheimer’s disease or possibly lead to a moderate increase in risk. Biased due to higher mortality in smoking AD patients, resulting in a lesser probability to catch them as cases in case-control studies was unveiled and could explain the inaccuracy [67
]. In our work, data were included based on hospitalization records, therefore it was very unlikely that higher mortality in smoking COVID-19 patients would prevent them being selected in case-control studies biasing our meta-analysis. Current scientific evidence suggests that active smokers hospitalised for COVID-19 have a worse prognosis [14
], and current smoking does not seem to be associated with an adverse outcome [24
]. It must also be considered that current smokers cease to be so when entering the hospital, as far as nicotine is concerned. In our work, we only refer to the fact that there are less hospitalised current smokers than expected, which is why nicotine has been suggested to very likely have a protective effect against serious symptoms, calming the cytokine storm (see Section 4.2
]). This might be a cause of underrepresentation among hospitalised patients.
In any case, it is necessary to remember that tobacco causes 20,000 deaths a day all over the world [15
] and, with COVID-19 patients, it generally comes with comorbidities, which means a worse prognosis [14
Finally, in must be also noted that a potential threat to the validity of the meta-analytic results is the so-called publication bias
, meaning that the publication of studies depends on the direction and statistical significance of the results. Studies with statistical significance are more likely to be published than those with non-significant results (which would be published less often) [68
]. However, in the studies of the present meta-analyses, at the time of their publication, the fact that patients are smokers, ex-smokers or non-smokers is secondary information and therefore does not influence our results.
4.2. Physiological Substrate for Anti-Inflammatory Pulmonary Effect
SARS-CoV-2 causes varying degrees of illness. Fever and cough are the dominant symptoms, but severe disease also occurs. When COVID-19 patients’ aggravation takes place, lung hyperinflammation may appear due to a virus-activated “cytokine storm” or CRS [69
]. Of the different cytokines that increase and reach such an exacerbated response [70
], Interleukin-6 (IL-6) in serum is mainly expected to predict SARS-CoV-2 pneumonia severity as the suppression of pro-inflammatory IL-6 has been demonstrated to have a therapeutic effect on many inflammatory diseases, including viral infections [71
]. In severe cases, SARS-CoV-2 has been shown to activate both innate and adaptive immune systems in alveolar tissue by inducing the release of many cytokines and subsequent cytokine release syndrome [72
]. During this response, levels of pro-inflammatory cytokines (include TNFα, interleukin (IL)-1b, IL-6, and IL-8) rise [70
], which is an important cause of death [73
]. Therefore, it is believed that controlling such crucial inflammatory factors could be a successful approach to reducing mortality in severe patients.
The existence of a cholinergic anti-inflammatory pathway has been demonstrated, which modulates inflammatory responses during systemic inflammation [74
]. The α7-nicotinic acetylcholine receptors (α7nAChR) are known to be expressed in macrophages and are essential for attenuating the inflammatory response by their activation during systemic inflammation [75
]. The underlying mechanism conveys that α7nAChR activation in infiltrated inflammatory cells, including macrophages and neutrophils, induces not only the suppression of NF-kB activation [76
], but also the secretion of pro-inflammatory cytokines and chemokines from inflammatory cells, including alveolar macrophages [77
]. In lungs, this process involves a physiological feedback mechanism as it has been demonstrated that pulmonary injury signals produced by inflammation are transmitted by vagal sensory neurons to the central nervous system [78
], where they are integrated and transformed into a vagal reflex [79
]. This response activates the parasympathetic neurons innervated by the efferent vagus nerve, which results in a higher ACh concentration in the lungs [80
]. Interestingly, it has been reported that nicotine, an α7nAChR agonist, exerts an anti-inflammatory effect of acute lung injury in a murine model [75
]. In other inflammatory diseases, such as ulcerative colitis (UC), smoking or treatment with nicotine has been demonstrated to significantly reduce the risk of developing the disease [76
]. Indeed, nicotine has been shown to reduce acute colonic inflammation severity with the concomitant inhibition of IL-6 mRNA expression [81
]. So, nicotine, an exogenous α7nAchR agonist, has already been demonstrated to selectively downregulate the inflammatory response in a number of infection and inflammatory diseases and it has also been suggested that smoking could interact with susceptibility to SARS-CoV-2 infection through the renin–angiotensin system [84
SARS-CoV-2 has been proposed to use the ACE2 receptor located at the surface of host cells to facilitate virus entry [85
]. On the one hand, it has been suggested that smoking may upregulate ACE2 expression [13
] and also that SARS-CoV-2 infections could be positive feed-back loops to increase ACE2 levels and facilitate virus dissemination [86
]. On the other hand, evidence suggests that nicotine downregulates ACE2 expression [13
]. In any case, the exact role of ACE2 as a mediator of disease severity remains to be determined. As ACE2 expression is necessary and sufficient for SARS-CoV-2 infections, it seems highly likely that an expansion of ACE2-expressing cells in the lungs facilitates viral shedding. However, it is possible that ACE2 expression also has some beneficial consequences. ACE2 has strong vasodilatory, anti-inflammatory, and antioxidant properties. Based on these properties, increased ACE2 levels have been proposed to be more beneficial than harmful, particularly in patients with lung injury. In this sense, children, and younger adults, who have milder COVID-19 symptoms, have higher ACE2 levels compared to older people [38
]. Therefore, even if smoking upregulates ACE2, this does not necessarily imply an adverse prognosis [39
]. For the above reasons, further research will be required to determine the precise impact of ACE2 levels on the clinical course of COVID-19 and its relationship to smoking and nicotine.