Background It is not currently clear what impact alternative nicotine-delivery products (electronic cigarettes, heated tobacco products and snus) have on smoking rates and cigarette sales. Objective To assess whether access to these products promotes smoking in the population. Design and data sources We examined associations of alternative nicotine product use and sales with smoking rates and cigarette sales overall, and in different age and socioeconomic groups, and compared smoking prevalence over time in countries with contrasting regulations of these products. For electronic cigarettes, we examined data from countries with historically similar smoking trajectories but differing current electronic cigarette regulations (United Kingdom and United States of America vs. Australia, where sales of nicotine-containing electronic cigarettes are banned); for heated tobacco, we used data from countries with state tobacco monopolies, where cigarette and heated tobacco sales data are available (Japan, South Korea), and for snus we used data from Sweden. Analysis methods We pre-specified dynamic time series analyses to explore associations between use and sales of alternative nicotine-delivery products and smoking prevalence and cigarette sales, and time series analyses to compare trends of smoking prevalence in countries with different nicotine product policies. Results Because of data and analysis limitations (see below), results are only tentative and need to be interpreted with caution. Only a few findings reached statistical significance and for most results the Bayes factor indicated inconclusive evidence. We did not find an association between rates of smoking and rates of the use of alternative nicotine products. The increase in heated tobacco product sales in Japan was accompanied by a decrease in cigarette sales. The decline in smoking prevalence seems to have been slower in Australia than in the United Kingdom overall, and slower than in both the United Kingdom and the United States of America among young people and also in lower socioeconomic groups. The decline in cigarette sales has also accelerated faster in the United Kingdom than in Australia. Limitations Most of the available data had insufficient data points for robust time series analyses. The assumption of our statistical approach that causal interactions are more likely to be detected when longer-term changes are screened out may not apply for short time series and in product interaction scenarios, where short-term fluctuations can be caused by, for example, fluctuations in prosperity or product supplies. In addition, due to dual use, prevalence figures for smoking and alternative product use overlap. The ecological study design limits the causal inferences that can be made. Longer time periods are needed for any effects of exclusive use of the new products on smoking prevalence to emerge. Conclusions We detected some indications that alternative nicotine products are competing with cigarettes rather than promoting smoking and that regulations that allow their sales are associated with a reduction rather than an increase of smoking, but the findings are inconclusive because of insufficient data points and issues with the assumptions of the pre-specified statistical analyses. Future work As further prevalence and sales data emerge the analyses will become more informative. Accessing sales figures in particular is the current research priority. Study registration The project is registered on Open Science Framework https://osf.io/bd3ah. Funding This project was funded by the National Institute for Health and Care Research (NIHR) Public Health Research programme (NIHR129968) and will be published in full in Public Health Research; Vol. 11, No. 7. See the NIHR Journals Library website for further project information. Plain language summary Alternative nicotine-delivery products are now available which are much less hazardous than cigarettes. These include electronic cigarettes (which contain no tobacco), Swedish snus (oral tobacco with low levels of cancer-causing chemicals) and heated tobacco products. There is concern that these products attract young people to smoking and discourage smokers from quitting (i.e. increase smoking), but it is also possible that they help smokers quit and steer young people who find nicotine attractive away from smoking, or that they have no effect on smoking. To clarify which of these end results is likely, we looked at data on smoking and on the use of these alternative products over time, and also compared data on smoking from countries that have similar tobacco control history, but that either allow (i.e. United Kingdom and United States of America) or ban the sale of electronic cigarettes (i.e. Australia). As the sale of heated tobacco products increased in Japan, sales of cigarettes seem to have gone down, suggesting that this product is competing with cigarettes rather than encouraging their use. We also found that the drop in smoking may have been slower in Australia than in the United Kingdom. For young people and those on low income specifically, the reduction in smoking was slower in Australia than in both the United Kingdom and United States of America. Allowing alternative nicotine products to be sold seems to have been linked with lowered rather than increased rates of smoking. Our findings, however, are uncertain because only limited data were available. Clearer conclusions will become possible as more data on the use and especially on the sales of alternative nicotine products are collected. Scientific summary Background The key controversy surrounding reduced-risk alternative nicotine-delivery products (ANDs), comprising electronic cigarettes (ECs), snus and heated tobacco products (HTPs), concerns their effects on smoking prevalence. They may promote smoking by enticing adolescents to cigarette use and by reducing quitting in smokers (e.g. by allowing them to circumvent smoke-free regulation), but it is also possible that ANDs redirect young nicotine seekers away from smoking and help smokers quit, or they may have no clear impact on smoking. Alternative nicotine-delivery systems have been used for a number of years, and some information may already exist in data on smoking prevalence and on AND and cigarette sales that can contribute to answering this question. If, in countries with substantial AND use, the trends in the historical decline in smoking rates and cigarette sales have slowed down or reversed, and this is not happening in countries that restrict or ban ANDs, this would be consistent with the ‘gateway into smoking’ hypothesis, while the opposite finding would suggest that these products are replacing rather than promoting smoking. A finding that ANDs have no net effect on smoking prevalence and cigarette sales, or that currently available data do not provide a clear answer, would also have implications both for policies (which will need to take account of the uncertainty) and for future research (which can aim to fill the gaps). Objectives We aimed to answer the following research questions (RQs): Is there an association between the prevalence of smoking and prevalence of AND use (in the population in general, in young adults and in different socioeconomic groups), in different countries and for different AND products? Is there an association between sales of different AND products and cigarette sales? Are there any differences in time trends of smoking prevalence (overall, in young adults and in different socioeconomic groups) and cigarette sales following the introduction of ANDs between countries that had similar tobacco control measures prior to ANDs emerging locally, but that have allowed EC sales [United Kingdom (UK) and United States of America (USA)] or banned them (Australia)? Methods We examined data from the countries and data sets described below that allow examinations of links between AND use and smoking rates, and between cigarette and AND sales. We also compared trajectories of smoking prevalence over time in countries with contrasting legislative and policy frameworks concerning ANDs. Countries selected Regarding prevalence data (RQ1), we explored links between the prevalence of EC use and cigarette smoking in the UK and USA (where data on EC use and cigarette use are available), Sweden (the only EU country where snus use is allowed and where data on snus and cigarette use are available) and Japan (where in addition to smoking prevalence data, at least some data on HTP use are available). We also looked at links between prevalence of smoking and sales of ANDs in Japan and South Korea, where data on HTP sales were available. Regarding sales data (RQ2), we explored links between sales of cigarettes and snus in Sweden, and between sales of cigarettes and sales of HTPs in Japan and South Korea. No comprehensive data on EC sales are available for the UK and USA, because ECs are produced by a large number of manufacturers, but we looked at ‘hybrid’ links between the prevalence of EC use and sales of cigarettes in the UK and USA. We also explored whether time trends in cigarette use and sales in the UK and USA, where ECs are allowed, differ from those in Australia, where sales of nicotine-containing ECs are banned (RQ3). The three countries have a broadly similar history and formats of tobacco control otherwise. Data sources Where more than one survey was available per country, we selected the survey to use based on: quality of smoking and AND use measures years covered representativeness of samples and weighting number of time points (monthly/quarterly breakdown) sample size quality of socioeconomic status (SES) information. For the UK, we used the Health Survey for England (HSE) for prevalence and HM Revenue and Customs National Statistics Tobacco Bulletin for cigarette sales data. For the USA, we used the National Health Interview Survey (NHIS) and the National Youth Tobacco Survey (NYTS) for prevalence of smoking. NielsenIQ data on cigarette sales were purchased from Chicago Booth. For Australia, we used the National Drug Strategy Household Survey (NDSHS) for prevalence of smoking and Euromonitor International reports for cigarette sales data. For Japan, we used the National Health and Nutrition Survey for prevalence, Tobacco Institute of Japan data for cigarettes sales and Philip Morris International reports for HTP sales data. For South Korea, we used the Korea National Health and Nutrition Examination Survey (KHANES) for prevalence, and Korea Tobacco and Ginseng Corporation (KT&G) reports for sales data. For Sweden, we used the Swedish Council for Information on Alcohol and Other Drugs (CAN): Monitor Studies for prevalence and Swedish Match AB provided data for cigarette and snus sales directly by e-mail. See Chapter 2 – Methods, Survey data chosen for details of where these surveys were available. For analyses looking at the interplay between cigarettes and EC/HTP use (RQ1), we included data on smoking from 2005 until the most recent available year up to 2019, before the COVID-19 epidemic started. The same timeframe was used, where available, for analyses exploring the interplay of cigarettes and EC/HTP sales (RQ2). This allowed for modelling of smoking and cigarette sales for at least 5 years prior to the emergence of these newer products. For snus, we analysed the data from 2007 to 2019. For analyses looking at trends in cigarette smoking in the UK, USA and Australia we used data from 2004 onwards as the NDSHS is triennial and we wanted to include 2004 as a proxy for 2005. Similarly, for the NYTS, which used to be biannual, we included 2004 as a proxy for 2005. For analyses looking at trends in cigarette sales for the UK, USA and Australia we looked at data from 2010 as sales data for Australia were only available from that year. Statistical methods We pre-specified in our statistical analysis plan that we would use dynamic time series applying generalised least squares (GLS) models to explore the association between prevalence of smoking and product use, as well as cigarette and product sales. Regression models were used to compare trends in smoking prevalence between countries where EC use is allowed or banned. The analyses were adjusted for two policy variables which are believed to have the strongest effect on smoking prevalence and cigarette sales and that have been implemented at different time points in most countries: the introduction of bans on smoking in public spaces where they occurred during the study period; and increases in cigarette prices exceeding normal trends (25% increase in 2010 in Australia, 37% in October 2010 in Japan, 80% increase in 2015 in South Korea, 14% increase in 2009 in the USA and 11% increase in 2015 in the UK). Each policy was modelled as a step level change, coded as one after the introduction of the policy and zero before. Unfortunately, the available data had insufficient data points for robust time series analyses. We carried out the planned analyses to adhere to the original plan, but the results need to be considered as only tentative and interpreted cautiously. Due to the short time series, it is possible that underlying trends were not correctly identified and removed, resulting in confounding. For example, in the majority of analyses, trends were identified as linear and first-order differencing was used. However, studies suggest that the decline in smoking prevalence follows a more monotonic curvilinear decline over time, which was not identified in the short series and would have required higher differencing. Conversely, with short-term non-granular time series, there is a risk of removing trends which in fact reflect short-term changes and perhaps a casual association between series. In addition, because of the short series, more sophisticated time series models, such as autoregressive integrated moving average with explanatory variable (ARIMAX), could not be used for the majority of analyses because these require more time-points. ARIMAX models have several benefits over GLS, including the use of transfer functions to model lags between multiple time series and inclusion of seasonal autocorrelation. Finally, short time series are highly influenced by outliers, and these can dominate the output from models. Longer time series are not influenced in the same way. These factors can all result in spurious associations. We used Bayes factors (BF) to help us to determine whether non-significant results provided evidence of no effect or were due to data being insensitive, with BF between 1/3 and 1 suggesting no or weak evidence for the null hypothesis, BF between 1/3 and 1/10 suggesting moderate evidence, and 1/10 or lower as strong evidence for the null hypothesis. Results Only a few associations were detected. Regarding RQ1, we did not find an association between rates of smoking and rates of alternative nicotine product use, potentially due to the analyses being under-powered. Regarding RQ2, the increase in sales of HTPs in Japan was accompanied by a decrease in sales of cigarettes (b = −1.09, p