We consider the certainty of the evidence as it relates to primary outcomes for our three main comparisons: nicotine EC versus NRT; nicotine EC versus non‐nicotine EC; nicotine EC versus behavioural support only/no support (Table 1; Table 2; Table 3).

Our summary of findings tables and assessments of certainty are based on the evidence from randomised controlled trials (RCTs). The cohort studies that we include were all automatically classified as having high risk of bias on the basis of non‐random treatment allocation, irrespective of other methodological considerations. Data presented from these studies need to be interpreted with caution. However, data from cohort studies were consistent with data from RCTs.

Risk of bias did not impact on the certainty of evidence for comparisons between nicotine and non‐nicotine EC, or between nicotine EC and NRT. For the latter, we judged six out of seven studies to be at low or unclear risk of bias overall. For the former, removing one study at high risk of bias increased the effect estimate for our efficacy outcome. Risk of bias decreased our certainty in the effect estimates for our nicotine EC versus behavioural support only/no support comparison as, due to the nature of the comparison, blinding was not possible and different levels of support could lead to bias.

All but two of our primary outcomes for our main comparisons were downgraded for imprecision, due to wide confidence intervals and few events. Other than the risk of bias and imprecision, we identified no other issues that decreased the certainty of the primary outcomes for our main comparisons.

Due to the small number of studies contributing to individual analyses, we were unable to assess publication bias in most cases and cannot rule this out. For the comparison nicotine EC versus behavioural/no support, two of our outcomes had more than 10 studies contributing to meta‐analysis and, therefore, we were able to generate funnel plots. The funnel plot for the exhaled CO outcome did not show any evidence of asymmetry (Figure 13). However, the funnel plot for smoking cessation showed evidence of asymmetry (Figure 14), influenced by large estimated intervention effects in some studies that had either a small sample size or a very low cessation event rate in both study arms. This apparent asymmetry may be explained either by some level of publication bias (e.g. smaller studies that show no benefit of nicotine EC remaining unpublished), or by heterogeneity between studies in either the intervention effect or the underlying outcome event rate. We carried out a sensitivity analysis removing the two studies that showed the most marked effects of EC (Dawkins 2020; Halpern 2018 [335, 336, 337]), and this did not change the interpretation of the pooled result. Therefore, we did not downgrade the certainty of the evidence for publication bias; however, we will continue to monitor this as the evidence is updated.

This field of research and EC devices themselves continue to evolve rapidly. This is the fifth update conducted as part of our 'living systematic review' approach, which will proceed until at least 2027, meaning we can continue to rapidly incorporate new evidence (see Supplementary material 8).

This update incorporates data from 1 March 2024 to 1 March 2025. Subsequent monthly searches will keep the evidence in this review current. Although studies predominantly came from the USA and UK, this review covers data from 16 countries. Geographical range in studies may be particularly important, due to the marked differences in EC regulation between countries; for example, studies conducted in countries that limit nicotine dose in EC, or allow only certain EC devices to be tested, may observe less pronounced effects on quitting. This review includes studies on some under‐researched populations, including people not motivated to quit smoking, people with substance misuse disorders, people with serious mental health conditions, people living in socially deprived areas and people experiencing homelessness. Quit rates in these groups are traditionally lower, but these groups may particularly stand to benefit from EC if they are effective because, in absolute terms, conventional cessation methods are often not as effective for them.

As well as the rapid pace of research in this field, evolutions in EC technology pose a challenge when considering the applicability of our evidence to the present. We had downgraded the certainty of our data in the 2016 update, as the devices tested in the trials were first‐generation 'cig‐a‐like' devices which did not deliver nicotine well, meaning the studies may have yielded more conservative estimates than would be seen with newer models, as newer devices and models have tended towards improved nicotine delivery. Nicotine delivery is also relevant to the comparator NRT arms tested; use of both a shorter‐ and a longer‐acting form of NRT show the highest success, and it is important that, where possible, this be the comparator chosen for such trials [334]. We no longer downgrade the evidence on this basis as studies with newer device types are now included, although there will always be a time lag between current devices and the research evidence available. Within our primary comparisons, none of the analyses of our primary outcomes signified substantial levels of statistical heterogeneity, despite the fact that different devices were used in the included studies. However, this could be because CIs were wide for individual studies, and does not rule out clinically significant differences in effects between EC types. As further data emerge, we hope to be able to formally test for differences in subgroup analyses, and in head‐to‐head comparisons of different device types. A companion paper explores available data on flavours in more detail, but is up‐to‐date until February 2024 and so not as up‐to‐date as this review [318; 319].

The AEs described in both the RCT and cohort studies continue to look similar, regardless of the brand of EC used or nicotine content, with placebo and nicotine‐containing EC showing similar numbers and types of AEs in direct comparisons. They also reflect what is reported in survey data [48, 338].

The structure of our analyses follows the standard practice of the Cochrane Tobacco Addiction Group, i.e. evaluating outcomes on an intention‐to‐treat basis, meaning our pooled results represent the effect of offering an EC intervention. This is different from evaluating the per‐protocol effect, or the effect only on those who use the EC to quit smoking entirely, or continue to smoke whilst also using EC. Although pragmatic and hopefully of use to those designing and delivering interventions, we acknowledge that our intention‐to‐treat approach limits the ability to use the data presented here to draw conclusions about biomarkers in subgroups of participants based on subsequent EC use/smoking profiles. A companion publication, up to date to January 2022, attempts to address this deficit [333].