Cigarette consumption in adult dual users of cigarettes and e-cigarettes: a review of the evidence, including new results from the PATH study

Analysis of adult data from waves 1, 2 and 3 of the PATH study

To be included in the analyses, individuals had to satisfy four criteria:

1. Be aged 18 to 54 years at Wave 1 (as e-cigarette use proved to be very uncommon in those aged 55+ years).

2. Be current every day cigarette smokers at each Wave with data available on CC.

3. Have data available on current and former established e-cigarette use at each Wave.

4. Have data available at Wave 1 on nine variables considered as potential predictors of CC (referred to subsequently simply as “predictors” – sex (male, female), age range (18–24, 25–34, 35–44, 45–54 years), highest grade of schooling (less than High School, GED, High School graduate, some college (no degree) or associates degree, bachelor’s degree, advanced degree), Hispanic origin (yes, no), race (white alone, black alone, other), age range when first started smoking cigarettes every day (<18, 18–24, 25–34, 35–44, 45–54, 55+ years), ever used alcohol (yes, no) ever used cocaine or crack (yes, no), ever use of other tobacco products (yes, no) – based on separate variables for traditional cigars, cigarillos, filtered cigars, pipes, hookah, smokeless tobacco, snus and dissolvable tobacco.

In the PATH dataset, the data on CC were recorded by two variables, a number and a unit (cigarettes per day or packs per day), the intention being to derive daily consumption directly from the number variable if the unit was cigarettes per day and by multiplying the number by 20 if the unit was packs per day. However, for some individuals the consumption data recorded in this way was highly implausible. For example, one individual had a recorded CC of 15 packs per day at Wave 1, 10 cigarettes per day at Wave 2 and 15 cigarettes per day at Wave 3. Here the implied consumption of 300 cigarettes per day is highly implausible and it seems much more likely that the Wave 1 consumption was actually 15 cigarettes per day. Accordingly, as described more fully in Extended data, Additional File 1 (Lee, 2020), a detailed investigation of the CC data at each Wave was carried out, and corrections were made for 84 individuals. There were 75 corrections relating to individuals where the consumption at one Wave was seven or more times the average consumption at the other two Waves. However, there were nine individuals where corrections were applied for other reasons, as described in the Additional File (see Extended data; Lee, 2020). All the analyses reported are based on the corrected CC data.

Analyses were carried out relating to current and ever established e-cigarette use, based on data for a current established e-cigarette user (one who has ever used, has used fairly regularly, and uses every day or some days) and a former established e-cigarette user (one who has ever used, has used fairly regularly, and currently does not use at all), with ever users being those who were either current or former users. Analyses were also carried out relating to current every day e-cigarette use. Note that for Wave 3, the data used were for e-product use (which also includes e-cigars, e-pipes and e-hookahs) rather than e-cigarette use, data for e-cigarette use not being available for this Wave.

All the analyses were weighted, based on the weighting factors for Wave 1, and used weighted linear regression analyses. Analyses were carried out with no adjustment for other variables and with adjustment for the other nine predictor variables selected, using a forward stepwise approach, including variables significant at p<0.05.

To answer Q1, the corrected CC was compared in exclusive cigarette smokers and in dual users, separately at each Wave.

To answer Q2 (and to provide some additional information), the change in corrected CC between two Waves was compared in four groups – exclusive cigarette smokers at both Waves, dual users at both Waves, dual users only at the second Wave, and dual users only at the first Wave. Separate analyses were carried out for each pair of Waves.

To answer Q3, the corrected CC at the first of two Waves in those who were then exclusive cigarette smokers was compared in those who did or did not become dual users. Separate analyses were carried out for each pair of Waves.

Software used to analyse data from the PATH study

Relevant data were transferred for analysis to a ROELEE database, and analysed using the ROELEE program (Release 59, Build 49). All these analyses could be run using the R-program (https:/www.r-project.org), using the "lm" function including the "weights=" option for weighted linear regression, and for stepwise regression using the "step" function specifying "method=forward" and test="F".

Literature searches

The search was limited to publications from 2008 onwards, e-cigarette usage not being widely established before then (Barrington-Trimis et al., 2015; Bauld et al., 2014; Bell & Keane, 2014). Potential publications were initially obtained from a search on PubMed on 22 October 2018:

Subsequently, on 26th July 2019, the search was repeated, using the same keywords as before, but restricted to papers published since 1st January 2018.

An initial trial was undertaken on the first 100 papers for which free copies were available via PubMed, with PNL screening using the abstract only, and BAF independently screening using the full paper. Based on the high level of agreement between the two methods, it was decided that initial screening on the abstract was satisfactory, but that where there was any doubt the papers would need to be obtained and examined in full. In particular, any paper apparently reporting an epidemiological survey in a general adult population which had enquired about both cigarettes and e-cigarettes should be obtained, even if there was no mention in the abstract of dual use or of the number of cigarettes smoked.

All papers were then screened by BAF or AJT, with the full paper obtained wherever necessary. Papers were accepted as having data, or as being relevant reviews. No restriction was made as to the type of device referred to by the original papers, whether e-cigarettes, ENDS, ANDS, cigalike or other terminology, or whether containing nicotine or not, except that results specifically referring to heated tobacco products were excluded.

For each paper, rejected papers were marked in two ways. First, whether they were rejected on the basis of the full paper, abstract or title, this final option only being used where no abstract was available through PubMed and the title and publication type clearly indicated it was not relevant. Papers which were at all debatable or borderline for acceptance were also examined by KJC and PNL before being finally rejected.

Second, the reason for rejection was coded as follows:

1. Not a study asking individuals about their smoking and e-cigarette usage

2. Study of adolescents, young adults (age 25 or under), or students

3. Study in persons with a pre-existing medical condition

4. Study in persons likely to have unusual smoking and/or e-cigarette habits (e.g. studies of illicit drug users, prisoners, homeless persons, e-cigarette retail employees, medical personnel)

5. Trial of smoking cessation methods or of switching to e-cigarettes

6. Study with <100 participants

7. Study restricted to e-cigarette users (or dual users), or inappropriately restricted on smoking habits or quitting status/intention

8. No results presented distinguishing exclusive cigarette smokers and dual users

9. No results on CC

10. Results on CC not in a useable format (e.g. studies which reported amount smoked overall, but not by exclusive/dual use).

11. Report on the PATH study.

Only one reason was recorded per paper, usually (but not necessarily) the first from the list above. A note on the reason for rejection was also usually recorded.

The papers identified as review papers from the PubMed search were examined, looking for references to papers apparently describing an epidemiological survey which had enquired about both cigarettes and e-cigarettes. Except where either the context of the review paper or the title of the secondary paper indicated that the survey was in an ineligible population (corresponding to rejection codes 2–7), new papers were added to the list of candidate papers and at least the abstract was examined. As previously, unless the abstract made clear that the study population was ineligible, the full paper was obtained, regardless of whether dual use or amount smoked were mentioned in the abstract. Further review papers cited in the original set of reviews were also considered, but only if the context in the original review indicated that they referred to amount smoked.

Extraction and analysis of information from the accepted papers

For each publication, information was collected on the study location, title and type (prospective or cross-sectional), the period when the study was conducted, the age of the population considered, the number of smokers, the definition of e-cigarette use (and whether it contained nicotine) and which of the three questions it could provide answers to. For Q1, the data extracted (separately for current and ever e-cigarette use) included the numbers of dual users of cigarettes and e-cigarettes, and of exclusive cigarette smokers, the CC for each group and the difference, with its statistical significance, where possible. Where distributions were given by grouped amount smoked, the estimates of CC were derived using assumed midpoints. For studies where differences could be estimated, comparisons were made, again separately for current and ever e-cigarette use, of the number showing positive and negative differences, with unweighted means and standard errors estimated overall, and separately for US and European studies. For Q2, data were extracted for studies comparing changes in CC in those becoming or not becoming current e-cigarette users, with the number showing significant differences in each direction summarized. Data were also extracted for studies reporting changes in CC only in those becoming current dual users, with the numbers showing reductions or increases summarized. For Q3, evidence was summarised comparing CC in exclusive cigarette smokers according to whether they later used e-cigarettes, there being too few studies to allow more than a description of the results.

Results from the PATH study

A total of 3,543 adults were eligible for the analysis. Table 1 summarizes the results relating to Q1. They show no significant tendency for CC to be associated with current or ever established e-cigarette use, with six of the differences being positive and six negative. In contrast, though based on much smaller numbers of dual users, current every day e-cigarette use is associated with a consistently lower CC, with five of the six estimates significant at p<0.05, and the estimated difference ranging from 1.50 to 2.92 cigarettes per day.

Table 2 summarizes the results relating to changes in e-cigarette status between Waves, with those who are exclusive cigarette smokers at each Wave (“No No”) used as the base for comparisons. The analyses comparing the “No Yes” with the “No No” group are directly relevant to Q2 as defined, while those comparing “Yes No” with “No No” provide information on the reverse association.

The results relating to established e-cigarette use, with one marginal exception, show no significant tendency for changes in CC to be associated with changes in e-cigarette use. The results for every day e-cigarette use show more evidence of a relationship. While they show no consistent tendency for changes to differ between those who are exclusive cigarette smokers at each Wave (“No No”), and those who are dual users at each Wave (“Yes Yes”), there is a consistent decrease in CC in those who became dual users at the second Wave (“No Yes”), and a consistent increase in those who switch from dual use to exclusive smoking (“Yes No”). While none of the decreases are significant (at p<0.05), four of the increases are highly significant (p<0.001). However, the numbers becoming or ceasing to be dual users are small.

Table 3 summarizes the results relating to Q3. The results do not show any consistent difference in baseline CC in those who were not e-cigarette users at baseline according to their subsequent e-cigarette use, whether based on current or ever established or current every day use.

Literature search

Our original PubMed search identified 1,403 publications, a number which extended to 1,434 from examination of reference lists in review papers and secondary references. From these 55 papers were identified as relevant, which described results from 45 apparently relevant studies. The later search identified 697 publications, of which 342 had already been identified by the first search. Nine of the remaining 355 were selected as relevant, with eight describing new studies and one providing additional results from a study identified earlier.

Further examination of the papers led to seven of the 53 studies being rejected: the Smoking Toolkit Survey (Beard et al., 2018) a time-series analysis linking prevalence of e-cigarette use with average CC in England; a paper (Morean et al., 2018) describing results from four studies, with the data for dual use and for cigarette use coming from two different surveys (and the other two not concerning tobacco); analyses of multiple internet and mail surveys (Etter & Eissenberg, 2015) where the surveys of e-cigarette users and of smokers were conducted at different times and in different national populations; the Mexico ITC survey (Lozano et al., 2019) as its analysis relating to Q2 was not restricted to those not using e-cigarettes at baseline; the Korea KCHS study (Han, 2019) which only related smoking to use of e-cigarettes due to a price increase; a US GFK study (Weaver et al., 2018) which did not relate CC to e-cigarette use at the same time point; and the California CSC study (Al Delaimy et al., 2015) which only compared smokers who either reported they have used e-cigarettes or will never use e-cigarettes at both surveys.

Study characteristics

Table 4 summarizes the characteristics of the 46 studies considered, in order of country within continent. Of the 46 studies, 22 were conducted in the US, five in multiple countries, three in the United Kingdom, three in Italy, two in France and one in each of 11 other countries. Of these, 16 studies involved an element of follow-up, the remaining 30 being of cross-sectional design, with three of these involving multiple cross-sectional studies. The studies mainly covered the whole adult population though some were restricted to younger adults. The number of smokers included varied considerably between studies with 10 involving more than 4,000 and seven less than 250. As will become evident, the studies varied in how e-cigarette use was defined and the questions they could answer. All the papers described the product as e-cigarettes, and it has been assumed that all the products contained nicotine, no study stating otherwise, though many did not give details.

Table 4 also gives references for the publications considered in our analyses. For some studies there were additional publications which provided no useful extra results to those cited: Study 2 (Laverty et al., 2018), study 5 (Andler et al., 2015; INPES, 2014), study 9 (Office for National Statistics, 2016), study 10 (Nelson et al., 2015), and study 27 (Shi et al., 2015).

Results related to Q1

Table 5 summarizes the results from the 39 studies that presented evidence comparing CC in dual users with that in exclusive cigarette smokers. Of these studies, the evidence related to current e-cigarette use in 25, to ever e-cigarette use in eight and to both current and ever use in six. In some studies, results were presented in terms of the distribution of the subjects by broad categories of cigarette consumption and (as described in the footnotes to Table 5) the overall mean CC could only be estimated by making assumptions about the mean CC for each category.

Of the 30 studies providing estimates relating to current e-cigarette use, two studies (18 and 22) merely stated there was a lack of association between e-cigarette use and amount smoked. In the other studies, estimates of the difference in CC between dual users and exclusive e-cigarette use could be derived, with two being from study 38 at two different time points. The evidence was conflicting, with 11 estimates showing a lower CC in dual users and 18 a higher CC. Omitting the estimate from study 29 (Pearson et al., 2015) (as it only related to e-cigarette use as a cessation aid) and the earlier estimate from study 37 (Grana et al., 2014) to avoid double-counting, the unweighted mean difference was estimated as +0.03 cigarettes per day reduction (Standard error [SE] 0.36). While there was some evidence of a reduction in the 11 estimates from Europe (1.04) cigarettes per day reduction, standard error [SE] 0.72), the 14 estimates from the USA did not confirm this (0.56 increase, SE 0.29).

The 14 studies providing estimates relating to ever e-cigarette use were equally inconsistent, with four of the ten estimated differences showing a reduction, and six an increase.

Results related to Q2

Our searches revealed 17 studies that presented evidence on whether cigarette-only smokers who become dual users change their CC. The top half of Table 6 summarizes results for five studies, all of prospective design, where dual use refers to current use and where changes are compared in those who remain exclusive cigarette smokers (“No No”) and those who become dual users (“No Yes”). All these studies present evidence that reductions in CC are significantly greater in dual users. The bottom half of Table 6 also presented evidence from a further eight studies (one prospective and seven cross-sectional), where changes were only recorded in those who became dual users. Of these, seven showed large reductions, which seem likely to be substantially greater than in those remaining exclusive cigarette smokers, although lack of the relevant data prevents this from being tested, while the other (study 41 (Seto et al., 2016)) showed a smaller reduction. The evidence from these 13 studies clearly indicates that becoming a current e-cigarette user leads to a reduction in CC.

The other four studies provided less useful results. Study 22 (Choi & Forster, 2014a) reported that changes in CC were almost identical between those who had never used e-cigarettes and those who had used e-cigarettes for at least one day in the past 30 days at baseline, after adjustment for demographics and baseline CC. These results neither relate to a proper definition of current e-cigarette use, nor to taking up e-cigarettes between baseline and follow-up. Study 27 (Shi et al., 2016) concluded that “among early adopters, ever-use of first generation e-cigarettes to aid quitting cigarette smoking was not associated with improved cessation or with reduced CC, even among heavier smokers” but the basis for this conclusion is most unclear, with no formal statistical analysis described and it being unclear when the e-cigarette use occurred, and whether those who quit during follow-up were excluded. Study 37 (Grana et al., 2014) reported that “e-cigarette use at baseline was not associated with a change in cigarette consumption, controlled for baseline cigarette consumption”. In study 43 (Wu et al., 2015), no difference was seen between exclusive cigarette smokers and dual users at six months follow-up in individuals who continued to smoke. These analyses related to ever rather than current use of e-cigarettes.

It can also be noted that, as studies restricted to dual users were excluded (see Literature searches, rejection reason 7), the results shown in the bottom half of Table 6 may not be comprehensive.

Results related to Q3

Our searches revealed only three published studies that presented evidence relevant to whether CC in exclusive cigarette smokers differed according to whether they later used e-cigarettes. Study 5 (Andler et al., 2016), a cross-sectional study which involved 4,752 current smokers of whom 759 also currently used e-cigarettes, noted that, before they started to vape, these individuals smoked 21 cigarettes per day as compared to 11.3 in those who continued to smoke. Though the results presented did not allow precise estimation of the difference in CC between smokers according to whether they later used e-cigarettes, the results clearly suggested that heavier smokers were more likely to start vaping. Study 35 (English et al., 2018), a cross-sectional study involving 245 current smokers, of whom 103 also currently used e-cigarettes, reported that, for each additional cigarette smoked per day prior to using e-cigarettes, individuals were 4.0% (p = 0.001) more likely to use e-cigarettes. Study 45 (Chan et al., 2019), a prospective study, presented results from a logistic regression predicting uptake of vaping in 3,797 smokers. After adjustment for a range of smoking-related and demographic variables, they reported an increasing likelihood of uptake with increasing cigarette consumption, with ORs of 1.13 (95% 0.95-1.35), 1.41 (1.12-1.78) and 1.69 (1.19-2.39) for smokers of 11-20, 21-30 and 31+ cigarettes per day, as compared to 0-10 cigarettes per day.

Source data

PATH study data: National Addiction & HIV Data Archive Program: Population Assessment of Tobacco and Health (PATH) Study [United States] Public-Use Files (ICPSR 36498). https://www.icpsr.umich.edu/icpsrweb/NAHDAP/studies/36498/versions/V9 (United States Department of Health and Human Services (USDHHS)).

The data are available under the Terms of Use as set out by ICPSR, which can be accessed when users start the process of downloading the data.

Extended data

Open Science Framework: Cigarette consumption in adult dual users of cigarettes and e-cigarettes. A review of the evidence, including new results from the PATH study. https://doi.org/10.17605/OSF.IO/QH8ZN (Lee, 2020).

This project contains the following extended data file:

Extended data are available under the terms of the Creative Commons Zero “No rights reserved” data waiver (CC0 1.0 Public domain dedication).