Chronic health effects associated with electronic cigarette use: A systematic review

Introduction Over the last decade, e-cigarette use has been on the rise but with growing health concerns. The objective of this systematic review was to update findings for chronic health outcomes associated with e-cigarette use from the 2018 National Academies of Sciences, Engineering, and Medicine (NASEM) report. Methods Three bibliographic databases were searched to identify studies comparing the chronic health effects of e-cigarette users (ECU) to non-smokers (NS), smokers, and/or dual users indexed between 31 August 2017 and 29 January 2021. Two independent reviewers screened abstracts and full texts. Data were extracted by one reviewer and verified by a second one. Outcomes were synthesized in a narrative manner using counts and based on statistical significance and direction of the association stratified by study design and exposure type. Risk of bias and certainty of evidence was assessed. The protocol was prospectively registered on Open Science Framework https://osf.io/u9btp. Results A total of 180 articles were eligible. This review focused on 93 studies for the 11 most frequently reported outcomes and from which 59 reported on daily e-cigarette use. The certainty of evidence for all outcomes was very low because of study design (84% cross-sectional) and exposure type (27% reported on exclusive ECU, i.e., never smoked traditional cigarettes). Overall, the summary of results for nearly all outcomes, including inflammation, immune response, periodontal and peri-implant clinical parameters, lung function, respiratory symptoms, and cardiovascular disease, suggested either non-significant or mixed results when daily ECU was compared to NS. This was also observed when comparing exclusive ECU to NS. The only notable exception was related to oral health where most (11/14) studies reported significantly higher inflammation among daily ECU vs. NS. Compared to the smokers, the exclusive-ECUs had no statistically significant differences in inflammation orperiodontal clinical parameters but had mixed findings for peri-implant clinical parameters. Conclusions This review provides an update to the 2018 NASEM report on chronic health effects of e-cigarette use. While the number of studies has grown, the certainty of evidence remains very low largely because of cross-sectional designs and lack of reporting on exclusive e-cigarette exposure. There remains a need for higher quality intervention and prospective studies to assess causality, with a focus on exclusive e-cigarette use.

High level of certainty means that we are confident that the true direction of association between ECU and the health outcome lies close to the association that we have estimated and further research is very unlikely to change our confidence in the effect.
 If the initial level of certainty was high, but was discounted by 0.5 or 1 point for any reason based on Table 1, or the initial level of certainty was low, but was eligible for an upgrade (higher the level of certainty) then the final rating will be moderate level of certainty.  If the initial level of certainty was high, but was discounted by 1.5 to 2 points, or the initial level of certainty was low and there was no reason to downgrade it, then the final rating will be low level of certainty.
 If the initial level of certainty was high, but was discounted by 2.5 points or more, or the initial level of certainty was low and was discounted by 0.5 points or more, then the final rating will be very low level of certainty.
For certainty of evidence from moderate to very low, include an interpretation for the reason you discounted the points for, discuss the limitation in the evidence that made you downgrade the certainty of evidence based on Table 1.

Inconsistency
No serious inconsistency 0 Look at the direction of effect.
 If the judgment of heterogeneity is considered expected or acceptable (≥75% are in the same direction (e.g., positive).
Use the vote counting tables (Table 3 in manuscript) to evaluate consistency of the direction.
The magnitude of effect is not used in this systematic review. We only consider direction, and statistical significant differences for each outcome comparing ECU to traditional smokers (TS, non-smokers (NS), or dual users (DU).
Judge inconsistency by evaluating the consistency of the direction and primarily the difference in the magnitude of effects across studies (since statistical measures of heterogeneity are not available). Widely differing estimates of the effects indicate inconsistency (Murad, 2017 (Guyatt GH, et al, 2011) 4  For observational studies, we will set an OIS at 10,000, which is equivalent to the minimum sample size for national surveys (Daniel, 2012).
Add the studies sample sizes for each comparison group and make the judgement relying only on the OIS.
We will not evaluate impression based on confident intervals. We will only reply on the OIS in this item.
Because of the wide range of outcomes, it is not feasible to set threshold per outcome.
Consider the optimal information size (OIS) (or the total number of events for binary outcomes and the number of participants in continuous outcomes) across all studies.
Results may also be imprecise when the CIs of all the studies or of the largest studies include no effect and clinically meaningful benefits or harms (Murad, 2017).
Serious imprecision -1 points If OIS is less than 1,250 then rate down for serious imprecision A threshold to evaluate 95% CI imprecision will not be used in this systematic review. We will only evaluate impression based on a fixed OIS across domains. If the total number of participants/ patients included in the rapid review < the number of patients generated by a conventional sample size calculation for a single adequately powered trial consider downgrading evidence. We will consider the same OIS threshold across all health domains, assuming α of 0.05, and β of 0.2 for relative risk ratio (RRR) of 20%, and the best estimate of control event rate was 0.2, the OIS threshold will be approximately 1, 250 participants (Guyatt GH, et al, 2011).  5 Typical sample size for different research designs include: (sample size = 15 to 30 participants per group for experimental research, 400 to 2,500 participants for community surveys, and 10,000 to 15,000 for national surveys (Daniel, 2012).  For studies only assessing prevalence (i.e. injuries), if a study have a potential conflict of interest related to funding, and were responsible for either the highest or lowest prevalence estimate in the range of estimates for the outcome comprehensive (Murad, 2017). CASP = Critical Appraisal Skills Programme, CI = confidence interval; GRADE = Grading of Recommendations Assessment, Development and Evaluation; OIS = optimal information size; RoB = risk of bias.