Pleasure and Satisfaction as Predictors of Future Cigarette and E-cigarette Use: A Novel Two-Stage Modeling Approach

Abstract

Introduction

Subjective experience of e-cigarettes may be an important factor in helping people who use combustible cigarettes switch completely to e-cigarettes to reduce harm from smoking. This paper describes a novel two-stage analysis using pleasure and satisfaction responses from ecological momentary assessments (EMA) of both cigarette and e-cigarette use to predict future cigarette and e-cigarette tobacco use.

Aims and Methods

This observational study included adult users of cigarettes and e-cigarettes who provided 7 days of EMA, capturing cigarette and e-cigarette use, followed by biweekly reports of cigarette and e-cigarette use over 1 year. Participants were 279 adults who provided both cigarette and e-cigarette responses during the EMA. We employed a two-stage analytic approach in which EMA data were used to predict subsequent levels of cigarette and e-cigarette use. In the first stage, EMA responses to cigarette and e-cigarette events were modeled via a mixed-effects location scale model to yield summaries of participants’ means and variability on event-related ratings of pleasure and satisfaction. These EMA summaries served as predictors in the second stage analysis of the biweekly post-EMA longitudinal cigarette and e-cigarette use data.

Results

EMA pleasure and satisfaction ratings were similar for both products and predicted both longitudinal cigarette and e-cigarette use, even after controlling for baseline cigarette and e-cigarette dependence. Relatively higher levels of satisfaction with e-cigarettes were associated with greater decreases in cigarette use over time.

Conclusions

Pleasure and satisfaction are important predictors of subsequent cigarette and e-cigarette use.

Implications

Experienced subjective pleasure and satisfaction from e-cigarettes relative to cigarettes may be an important factor in helping individuals who smoke to switch completely to e-cigarettes as a harm reduction approach. In order to help sustain complete product switching and reduce dual use or relapse to smoking, e-cigarettes may need to deliver more satisfaction to the user compared to that experienced from cigarettes.

Introduction

Smoking cessation is the best way for individuals who smoke to reduce tobacco-related morbidity and mortality but remains an elusive achievement for many. E-cigarettes may have potential as harm reduction approaches for users of combustible cigarettes who switch completely from combustible cigarettes,^1–6 and a recent Cochrane review concluded that e-cigarettes are more effective than nicotine replacement therapy in helping people to quit smoking.⁷ Identifying factors that may enhance the probability of complete switching for individuals who smoke but who have lower motivation to quit or who have been unsuccessful with quitting may help to reduce combustible tobacco use.

To date, most of the research on factors associated with switching has focused on e-cigarette devices, product characteristics, or level of e-cigarette use. There are few good examples, though, of how best to facilitate complete switching or what factors most help to enhance switching.⁸ Among a sample of individuals who smoke with no intention of quitting, e-cigarettes with a nicotine delivery similar to cigarettes may be more effective in promoting switching.⁹ In a longitudinal study of smokers in Great Britain, smoking cessation or complete switching was found to depend on the type of e-cigarette device and frequency of use.¹⁰ Beyond nicotine content and delivery mode, flavors may also matter in switching away from cigarettes. Among individuals relatively early in their trials of e-cigarettes, sweeter (vs. tobacco) flavors were associated with greater likelihood of cigarette cessation at a 1-year follow-up.¹¹ This suggests that the pleasure of the use experience, perhaps enhanced by flavors, may be an important factor to consider in whether smokers can use e-cigarettes for cessation. One study that examined current experience with e-cigarettes in those who smoke found that early positive experiences with e-cigarettes predicted regular use, and about half of the regular users reported that e-cigarettes were substitutable for cigarettes. Poor satisfaction or dislike of the taste of e-cigarettes were the primary reasons why smokers did not become regular e-cigarette users.¹² In sum, pleasure and satisfaction with e-cigarettes are both key in determining whether smokers continue e-cigarette use and fully switch away from cigarettes. Abrams et al.’s (2018) conceptual model of how e-cigarettes may displace smoking highlighted the importance of appeal. They note that subjective satisfaction is an essential component of the appeal of a product, suggesting that products with minimal satisfaction will not be appealing and will be unlikely to be adopted or used enough to displace cigarettes.

The goal of the current study was to examine whether pleasure and satisfaction with both cigarettes and e-cigarettes, assessed in real-time with ecological momentary assessment (EMA) at the point of use, were associated with changes in subsequent levels of cigarette and e-cigarette use in a sample of smokers who had recently started using e-cigarettes, but who were not currently attempting to stop smoking. Most research to date has relied on cross-sectional or survey data to evaluate experiences with e-cigarettes, and no prior work has examined in-the-moment subjective experiences and how they may predict future levels of use. We hypothesized that those who experienced pleasure and satisfaction with e-cigarettes similar to or greater than with cigarettes would be more likely to decrease cigarette use and increase e-cigarette use. To address this question, we present and describe a novel analytical technique for the combined two-stage analysis of EMA data with subsequent longitudinal cigarette and e-cigarette use data.

Materials and Methods

Participants

Participants (N = 404) were recruited for a longitudinal observational study through social media, Craigslist, listservs, flyers, and word of mouth. Interested individuals completed an online screener and then a phone screening. The goal was to recruit participants who were primarily cigarette smokers and who were not established e-cigarette users in order to examine factors that might influence patterns of progression and dual use of the products. There was no criterion for length of time of use of either cigarettes or e-cigarettes. Eligibility criteria included: Residing in the Chicago area; aged 18 years or older; smoked cigarettes at least once a week in the past 30 days; used an e-cigarette within the last 14 days but not on a daily basis; and indicated a future intention to use e-cigarettes (ie, responding “moderately” or “very likely” to “How likely are you to use an e-cigarette in the next 2 weeks” and “How likely are you to purchase an e-cigarette in the next 2 weeks”). Exclusion criteria included: lack of English fluency and inconsistent responses to screening questions.

All procedures were approved by the University of Illinois Chicago Institutional Review Board. Eligible participants provided informed consent and completed baseline assessments of demographics, tobacco use history and patterns, motives, beliefs, and expectancies for e-cigarettes and cigarette use. Participants then completed 7 days of EMA using a study-provided phone. The current study focuses on participants (N = 279) who provided both a cigarette and e-cigarette EMA event response. The sample was 58% male, had a mean age of 35.6 years, and was racially diverse (39% White, 33% Black, 13% Asian/Pacific, 11% Hispanic, 1% American Indian, and 3% other).

Measures

Demographics

Participants self-reported their age, gender, race/ethnicity, and highest level of education at baseline (Table 1).

Table 1.

Participant Characteristics (N = 279)

Characteristic	n (%)/ M(SD)
Age	35.6(12.6)
Gender
Male	162(58)
Female	117(42)
Race
Non-Hispanic White	110(39)
Hispanic	30(11)
Black	92(33)
Asian/Pacific Islander	36(13)
American Indian/Alaskan Native	4(1)
Other	7(3)
Education
Grade 9–11	19(7)
Grade 12/GED	57(20)
1–3 years of college	131(47)
4 years of College or more	72(26)
Cigarette NDSS	2.9(0.68)
E-cigarette NDSS	2.5(0.76)
Baseline 30-day cigarette rate	8.8(7.4)
Baseline 30-day e-cigarette rate 12-month 30-day cigarette rate 12-month 30-day e-cigarette rate	5.5(7.9) 6.4(7.6) 4.8(8.0)

GED = General Educational Development Degree; NDSS = Nicotine Dependence Syndrome Scale; Cigarette Rate is the average number of cigarettes/day; E-Cigarette rate is the average number of E-Cigarette “sessions” (15-minute period) per day; N = 251 for 12-month rates.

Satisfaction and Pleasure

Participants completed EMA interviews in response to random prompts (5–6 times/day) and self-initiated tobacco use events (either cigarette or e-cigarette). The rate of missed prompts was low (median = 8.7) for the 7-day EMA. If participants indicated the use of a tobacco product during a randomly prompted interview, they received tobacco use questions. At each tobacco use event, participants were asked to rate pleasure and satisfaction (separate items) with using the cigarette or e-cigarette. Ratings were made on a scale from 1 to 10, with higher ratings indicating greater levels of pleasure and satisfaction. These two items were highly correlated: 0.705 at the event level and 0.988 at the subject level. Thus, for our analyses, we averaged the ratings of satisfaction and pleasure associated with each event. Hereafter, we refer to this average variable as “satisfaction.”

Nicotine Dependence

At baseline, participants completed a 14-item version Nicotine Dependence Syndrome Scale (NDSS)¹³ to assess nicotine dependence for cigarettes and a modified version of NDSS to assess nicotine dependence for e-cigarettes.¹⁴ Internal consistency was good for both scales (α = 0.91 for e-cigarettes and α = 0.89 for cigarettes).

Cigarette and E-cigarette Use

At baseline, participants reported the number of cigarettes smoked/day and the number of daily e-cigarette “sessions” (ie, 15-minute period of e-cigarette use after a period of nonuse) for both the past 7 and 30 days. Following the week of EMA, subjects provided biweekly reports of their daily cigarette and e-cigarette use for 1 year (ie, a maximum of 26 biweekly assessments).

Overview of Analytic Approach

The first stage analysis examined participants’ cigarette and e-cigarette satisfaction during the EMA week. Overall, there were 8823 responses (6217 cigarettes and 2606 e-cigarettes), averaging 31.6 responses/participant (22.28 cigarettes [median = 19, range = 1–83], 9.34 e-cigarette [median = 6, range = 1–53]). Accounting for the clustering of responses within subjects, the average satisfaction rating was 7.46 (standard deviation [SD] = 2.07) across cigarette and e-cigarette reports.

Future Cigarette and E-cigarette Use

For the second stage analysis, we used data from the 1 year of post-EMA longitudinal biweekly assessments of daily cigarette and e-cigarette use. Overall, there were 5884 responses across 279 subjects (subject average = 21.1, median = 24). Accounting for the clustering of responses within subjects over time, participants reported 6.80 (SD = 7.27, range = 0–60) cigarettes/day and 3.99 (SD = 6.77, range = 0–60) e-cigarette sessions/day.

Baseline Covariates

To illustrate our approach, we selected a limited number of baseline covariates: Gender (0=male and 1=female) and age (mean centered at 35 and scaled so one unit is 10 years of age). We grand mean-centered participants’ baseline measurement of cigarette and e-cigarette dependence around their respective sample means.

Data Analysis

First Stage Analysis

For the EMA data, we used a mixed-effects location scale (MELS) model^15,16 to provide subject (random-effect) estimates of mean satisfaction from cigarette reports, mean difference in satisfaction from e-cigarette relative to cigarette reports, and variability (consistency/lability) in satisfaction reports for cigarettes and e-cigarettes. As this is a relatively new modeling approach, we describe it here in some detail. Specifically, consider the following model for the satisfaction measurement $y_{ij}$ of subject i (i = 1, 2, . . . N subjects) at occasion j (j = 1, 2,… $ni$ events):

$${\displaystyle \begin{array}{l}{\displaystyle y_{ij}=({\beta }_0+{\upsilon }_{0i})+({\beta }_1+{\upsilon }_{1i})ECigW{S}_{ij}+{\beta }_{2}ECigB{S}_i+{\epsilon }_{ij}}\end{array}}$$ (1)

The variable $ECi{g}_{ij}$ was an indicator of the response type (ie, cigarette = 0 or e-cigarette = 1 event). $ECi{g}_{ij}$ was decomposed into the between-subject (BS) variable $ECigB{S}_i={\overline{ECig}}_i$ (ie, a subject’s proportion of total events that were ECig) and the within-subject (WS) variable $ECigW{S}_{ij}=(ECi{g}_{ij}-{\overline{ECig}}_i)$ (ie, indicator of an ECig event relative to the subject’s overall proportion). This allowed separation of BS and WS effects of e-cigarette satisfaction, relative to cigarette satisfaction¹⁷; indicating whether participants who had more/less e-cigarette events reported higher/lower overall satisfaction, and the degree to which satisfaction was increased/decreased when a participant used an e-cigarette relative to a cigarette (controlling for overall e-cigarette and cigarette use), respectively. As $ECigB{S}_i$ was also centered around the sample mean, ${\beta }_0$ represents the average satisfaction from cigarette responses (for participants with average levels of e-cigarette events), ${\beta }_1$ is the average satisfaction difference associated with e-cigarette (vs. cigarette) events, and ${\beta }_2$ represents the average association of a participant’s proportion of e-cigarette events and their average satisfaction. The model included both a random subject intercept ${\upsilon }_{0i}$ (participant’s level of satisfaction associated with cigarette events) and slope ${\upsilon }_{1i}$ (difference in satisfaction associated with e-cigarette (vs. cigarette) events). These random effects are assumed to be (bivariate) normally distributed in the population of participants. Finally, ${\epsilon }_{ij}$ is an independent error term distributed normally with mean 0 and variance ${\sigma }_{\epsilon }^2$ to represent WS variance.

A unique feature of the MELS model is the modeling of the WS variance, which allows one to examine whether covariates were related to consistency/lability of the satisfaction responses. Here, we model the WS variance as:

$${\displaystyle \begin{array}{l}{\displaystyle {\sigma }_{{\epsilon }_{ij}}^2=\exp ({\tau }_0+{\tau }_{1}ECigW{S}_{ij}+{\tau }_{2}ECigB{S}_i+{\omega }_i)}\end{array}}$$ (2)

The exponential function is used to ensure that the WS variance is always positive. ${\tau }_0$ is the satisfaction variability from cigarette responses (for participants with average levels of e-cigarette events); ${\tau }_1$ is the difference in satisfaction variability associated with e-cigarette events, relative to cigarette events; and ${\tau }_2$ represents the association of a participant’s proportion of e-cigarette events and their satisfaction variability. These WS regression coefficients can be exponentiated to yield variance ratio interpretations, which can be interpreted as the ratio of the WS variance for a unit change in the regressor. The random subject effect ${\omega }_i$ is assumed to follow a normal distribution and represents subject variability in satisfaction that is not explained by the covariates as a subject-level marker of consistency/lability in satisfaction reports. ${\omega }_i$ in equation 2 is referred to as a random scale effect, and ${\upsilon }_{0i}$ and ${\upsilon }_{1i}$ in equation 1 are random location effects. All three are allowed to be correlated with each other. Appendix A1 includes a figure of the stage 1 MELS model, and Appendix A2 lists all model terms and definitions.

Second stage Analysis

The random-effect estimates from the stage-1 analysis (${\upsilon }_{0i}$, ${\upsilon }_{1i}$, and ${\omega }_i$) were then used in a second-stage analysis predicting participants’ cigarette and e-cigarette daily use from the post-EMA longitudinal biweekly reports. Since these variables were right-skewed, a square-root transformation was used to normalize their distributions. Also, the stage-1 random effects were standardized (ie, mean = 0 and SD = 1) so that regression coefficients represented standard effects (ie, change attributable to SD = 1 change in the random effect). For both rate outcomes, linear mixed models were used incorporating a linear effect of time, baseline covariates, stage-1 random effects, and interactions of stage-1 random effects with time. Time was centered and scaled to represent an effect for 1 year, where main effects for the stage-1 random effects represent average effects over time (ie, grand means of cigarette and e-cigarette use) and the interactions with time represent changes per year (in cigarette and e-cigarette use) attributable to the stage-1 random effects. Finally, to account for the repeated longitudinal WS data, random subject intercepts and time effects were included in the linear mixed models.

It should be noted that we are using a simple linear trend for the effect of time. One might argue that this is overly simplistic and opt for more extended time effects like polynomials (eg, linear, quadratic, etc.), splines, or non-linear time effects. However, as we are interested in expressing whether subjects changed in terms of their cigarette and e-cigarette use across the year, the objective was not to precisely model cigarette and e-cigarette use across time, but to model coefficients reflecting use patterns in a basic and understandable way.

An important consideration in using stage-1 random effects as predictor variables in the stage-2 models is that these random effects are estimates and not known quantities. Thus, as has been recommended in the psychometric literature, we utilized the plausible value methodology¹⁸ to repeatedly impute stage-1 random effects in our stage-2 models. This multiple imputation properly accounts for the uncertainty in the random effect estimates, and therefore avoids type-I errors which could result by treating these as known predictors without error. Specifically, each stage-2 model (one for cigarette use, one for e-cigarette use) was repeated for each of 500 sets of imputed stage-1 random effect estimates, and then averaged to yield overall regression estimates.¹⁹ The analyses were carried out using the MixWILD software program,²⁰ which automates the two-stage modeling approach, including the use of the plausible value imputation. This freeware program and Supplementary Materials are available at https://reach-lab.github.io/MixWildGUI/.

Results

At baseline, participants used cigarettes more frequently than e-cigarettes, and by the 12-month follow-up, the average rate of cigarette and e-cigarette use diminished somewhat (Table 1). Participants were mostly established cigarette users, with 97% reporting having smoked more than 100 cigarettes in their lifetimes (86% above 500), compared with 64.8% reporting have more than 100 e-cigarette lifetime sessions (36% above 500). At baseline, 51.7% used rechargeable e-cigarettes with refillable cartridges; 29.6% used rechargeables with pre-filled cartridges; and 18.5% used disposable e-cigarettes. The preferred e-cigarette flavors were tobacco flavor (12.6%), menthol or mint (34.7%), sweet flavors (eg, dessert, candy, fruit, and 44.8%), and other (7.9%).

Table 2 presents the results of the stage-1 MELS model. Here, the average satisfaction from the cigarette events was estimated as 7.46, and the WS difference for the e-cigarette ratings was 0.048 (p = .65). The BS effect of e-cigarettes was also not statistically significant (p = .67). Thus, on average, subjects reported similarly high levels of satisfaction to cigarette and e-cigarette events. In terms of the WS variance, the estimate for the cigarette reports was exp(0.308) = 1.360, and the estimate for the WS effect of e-cigarettes was estimated as exp(0.308−0.103) = 1.228. This difference in the WS variance was statistically significant (z = −2.277, p = .023) indicating that the satisfaction reports were more consistent (less labile) from e-cigarette events than from cigarette events. In terms of a variance ratio, exp(−0.103) = 0.902 indicates that satisfaction ratings were approximately 10% less varied (ie, more consistent) for e-cigarette, relative to cigarette events. The BS effect of e-cigarette events approached significance (p = .064) suggesting that participants with a higher proportion of e-cigarette events provided, on average, more consistent satisfaction ratings.

Table 2.

Mixed-Effects Location Scale Model of Ecological Momentary Assessments Satisfaction Ratings (279 Subjects; 8823 Observations)

	Estimate	SE	z value	p value
BETA (regression coefficients)
Intercept	7.456	0.090	82.392	.001
WS e-cigarette	0.048	0.106	0.450	.652
BS e-cigarette	−0.153	0.363	−0.421	.674
Random (location) effect variances and covariances
Intercept variance	2.183	0.192	11.371	.001
Covariance of intercept, WS e-cigarette	−0.455	0.169	−2.693	.007
WS e-cigarette variance	2.510	0.254	9.872	.001
TAU (WS variance regression coefficients; log-linear model)
Intercept	0.308	0.064	4.792	.001
WS e-cigarette	−0.103	0.045	−2.277	.023
BS e-cigarette	−0.471	0.254	−1.851	.064
Random scale variance and covariances
Covariance of Scale, intercept	−0.631	0.104	−6.077	.001
Covariance of Scale, WS e-cigarette	0.460	0.121	3.804	.001
Scale variance	1.043	0.097	10.798	.001

WS = Within-subject; BS = Between-subject; SE = standard error.

For random effects, there was considerable heterogeneity across participants in terms of their satisfaction ratings to the cigarette events (variance estimate = 2.183, z = 11.370, p < .001), difference in satisfaction to the e-cigarette, relative to cigarette, events (variance estimate = 2.510, z = 9.872, p < .001), and variability in their reports (scale variance estimate = 1.043, z = 10.798, p < .0001). Participants varied in terms of their satisfaction levels and in the variability of satisfaction. The covariances of these random effects were also highly significant, such that higher satisfaction ratings to cigarette events were associated with lower satisfaction ratings to e-cigarette events (covariance estimate = −0.455, z = −2.693, p = .007); higher satisfaction ratings to cigarette events were associated with reduced variability in satisfaction ratings (covariance estimate = −0.631, z = −6.077, p < .001); and higher satisfaction ratings to e-cigarette events, relative to cigarette events, were associated with increased variability in satisfaction ratings (covariance estimate = 0.460, z = 3.804, p = .001). As correlations, these three covariances were converted to −0.195, −0.418, and 0.284, respectively, reflecting moderate associations between these random effects.

Tables 3 and 4 list linear mixed model results of the longitudinal biweekly cigarette and e-cigarette reports, respectively, that used the random subject effect estimates from the stage-1 EMA analysis. The results are averaged over 500 solutions to the linear mixed models using the plausible value replications of the stage-1 random effect estimates, though it is important to note that the known covariates (year, gender, age, and NDSS) are not replicated (since they are known).

Table 3.

Linear Mixed Model for Longitudinal Analysis of Biweekly Daily Cigarette Use (Square Root Transformed); Averaged Estimates From 500 Plausible Value Replications (279 Subjects, 5884 Observations)

	Estimate	SE	z value	p value
Regression coefficients
Intercept	2.126	0.079	27.006	.001
yeara	−0.404	0.074	−5.500	.001
genderb	0.276	0.120	2.306	.021
agec	0.384	0.048	8.026	.001
CigNDSSd	0.656	0.088	7.458	.001
EMA_Cig	−0.076	0.060	−1.251	.211
EMA_Cig by year	0.119	0.073	1.625	.104
EMA_Ecig	−0.210	0.061	−3.458	.001
EMA_Ecig by year	−0.064	0.073	−0.872	.383
EMA_Scale	0.048	0.061	0.790	.430
EMA_Scale by year_c	0.006	0.073	0.086	.931
Variances and covariance
Intercept variance	0.938	0.082	11.436	.001
Covariance of intercept, year	0.269	0.075	3.567	.001
Year variance	1.221	0.129	9.453	.001
Residual variance	0.340	0.007	51.567	.001

SE = standard error; EMA = Ecological Momentary Assessments; Stage-1 EMA random effects (standardized with mean = 0 and SD = 1): EMA_Cig = satisfaction of EMA cigarette events, EMA_Ecig = satisfaction of EMA e-cigarette events relative to cigarette events, EMA_Scale = variability of EMA satisfaction reports.

^aMean-centered effect of time, scaled so that one unit = 1 year.

^b0=Male, 1=Female.

^cCentered around age 35, scaled so that one unit = 10 years of age.

^dParticipants’ centered baseline measurement of cigarette dependence.

Table 4.

Linear Mixed Model for Longitudinal Analysis of Biweekly Daily E-cigarette Use (Square Root Transformed); Averaged Results From 500 Plausible Value Replications (279 Subjects, 5884 Observations)

	Estimate	SE	z value	p value
Regression coefficients
Intercept	1.720	0.092	18.766	.001
yeara	−0.070	0.089	−0.785	.433
genderb	−0.025	0.125	−0.203	.839
agec	−0.113	0.050	−2.254	.024
ECigNDSSd	0.428	0.080	5.324	.001
EMA_Cig	−0.067	0.063	−1.065	.287
EMA_Cig by year	−0.175	0.088	−1.976	.048
EMA_Ecig	0.217	0.064	3.397	.001
EMA_Ecig by year	−0.048	0.089	−0.541	.588
EMA_Scale	−0.147	0.064	−2.315	.021
EMA_Scale by year	−0.028	0.089	−0.315	.753
Variances and covariance
Intercept variance	1.040	0.091	11.464	.001
Covariance of intercept, year	0.372	0.095	3.921	.001
Year variance	1.883	0.185	10.168	.001
Residual variance	0.369	0.007	51.566	.001

SE = standard error; EMA = Ecological Momentary Assessments; Stage-1 EMA random effects (standardized with mean = 0 and SD = 1): EMA_Cig = satisfaction of EMA cigarette events, EMA_Ecig = satisfaction of EMA e-cigarette events relative to cigarette events, EMA_Scale = variability of EMA satisfaction reports.

^aMean-centered effect of time, scaled so that one unit = 1 year.

^b0=Male, 1=Female.

^cCentered around age 35, scaled so that one unit = 10 years of age.

^dParticipants’ centered baseline measurement of e-cigarette dependence.

In terms of daily cigarette use (Table 3), there was a significant decrease in use across the year (p = .001), but increased use for females, with age, and for higher levels of cigarette dependency (p = .021, p = .001, p = .001, respectively). Higher levels of satisfaction during e-cigarette (vs. cigarette) events predicted significantly lower levels of daily cigarette use (p = .001). As the interaction with the year was not significant, this effect was indicative of a decrease in daily cigarette use across the entire year. Specifically, 70% of individuals who had higher cigarette (vs. e-cigarette) satisfaction smoked on average more than 5 cigarettes/day over time, while 64% of individuals who had higher e-cigarette (vs. cigarette) satisfaction smoked less than 5 cigarettes/day over time (p = .001). Regarding variances, there was considerable heterogeneity across subjects both in daily cigarette use and cigarette use time trends across the year (ps = .001). Expressing the covariance as a correlation (r = 0.251) indicated a moderate positive association between the yearly average level of cigarette use and the change across the year in cigarette use—where those that had overall higher/lower average levels of cigarette use during the year, also had increased/decreased levels of cigarette use across the year.

Regarding daily e-cigarette use (Table 4), there were no significant effects of year or gender (ie, no significant change across time or differences between males and females in e-cigarette vs. cigarette use). However, e-cigarette use significantly decreased with age (p = .024) and increased with e-cigarette dependency (p = .001). Higher satisfaction levels with EMA cigarette events predicted a reduction in daily e-cigarette use across time (p = .048). Alternatively, higher levels of satisfaction with the EMA e-cigarette (vs. cigarette) events predicted significantly higher levels of daily e-cigarette use (p = .001). Higher levels of consistency in the EMA satisfaction reports (ie, more negative EMA scale random effects) predicated higher daily e-cigarette use (p = .021). Regarding variances, there was considerable heterogeneity across subjects in both daily e-cigarette use and in e-cigarette use time trends across the year (ps = .001). Expressing the covariance as a correlation (r = 0.266) indicated that increased/decreased e-cigarette use across the year was moderately associated with overall higher/lower average levels of e-cigarette use during the year.

Discussion

Identifying factors that predict both cigarette and e-cigarette use for individuals who smoke or use both products may aid in developing interventions to promote complete switching and harm reduction. We hypothesized that an individual’s subjective satisfaction with e-cigarettes, relative to their satisfaction with cigarettes, would predict future reductions in cigarette use. We employed a novel two-stage approach to using satisfaction levels from each product, captured in real-time, to predict future cigarette and e-cigarette use. EMA satisfaction reports were significantly related to future use of both products. In particular, satisfaction with e-cigarettes, relative to cigarette, use was related to a reduction in cigarette use across the year and was strongly related to higher levels of future e-cigarette use. Over the course of the year, in this observational study, participants’ average level of daily cigarette use declined from 7.5 to 6.33 cigarettes/day, with ~18% of participants reporting no cigarette use by the end of the year follow-up. E-cigarette use increased only minimally over the year from an average of 4 to 4.25 e-cigarette sessions/day. Notably, the importance of satisfaction ratings was held even after controlling for baseline levels of cigarette and e-cigarette dependence.

Results from the variance modeling found that increased consistency of the subjective satisfaction EMA reports was also predictive of higher e-cigarette use. This consistency was for both cigarette and e-cigarette EMA events, as we could not separate the consistency of ratings for both events (ie, they were collinear). We also did not track specific e-cigarette devices or flavors used at the event level, so it is not known whether the consistency in satisfaction rating for the e-cigarettes varied as a function of the product or device characteristics (eg, specific device or flavor changes) or whether it was more a function of capturing e-cigarette use that was still in its early stages.

To the best of our knowledge, this is the first report using real-time data capturing the subjective experience of cigarettes and e-cigarette use to predict future levels of product use. The novel two-stage modeling approach allowed us to connect data from an EMA measurement period to subsequent product use via the first stage MELS model, which estimated random subject effects as summaries of each participant’s EMA data. These random effects were then used in the second stage models of future product use, utilizing plausible value imputations to properly account for the uncertainty in these estimates. Plausible value imputations allow the uncertainty in these estimates to vary across participants, as participants vary in the amount of EMA data. Another advantage of using the random effects as summaries of the EMA data, rather than simple calculated quantities (eg, subject means and SDs), is the assumption made about the missing EMA data. The random effects are estimated under the missing at random assumption, whereas the calculated quantities would be based on the more restrictive missing completely at random assumption. For longitudinal data, it is widely known that missing at random is much more plausible than missing completely at random assumption.²¹

The majority of smoking cessation approaches emphasize withdrawal mitigation and increasing motivation for cessation. Much less attention has been paid to the role that satisfaction from smoking plays in cessation, especially in the context of switching to e-cigarettes. Our results suggest that satisfaction or pleasure is important in the likelihood that e-cigarettes can potentially replace smoking. Other studies have found that more frequent use of e-cigarettes is associated with perceived satisfaction^22–25 and highlighted that satisfaction is important in a measure of sensory experiences associated with vaping e-cigarettes.²⁴ Therefore, individuals who want to reduce or stop smoking may be more willing to try a substitute product that provides as strong a level of satisfaction as cigarettes, and adherence to a regimen of switching may be enhanced with higher levels of e-cigarette satisfaction. Having a viable alternative approach to reducing smoking that provides at least as much satisfaction as smoking may help to advance cessation efforts.

As newer noncombustible tobacco products enter the market, a key regulatory issue is considering the net population health advantage, balancing the potential for adults who smoke combustible cigarettes to switch to potentially less harmful products while not enticing youth to start using these products. Thus, balancing satisfaction and level of appeal is critical to achieving this goal, and caution is needed not to reduce the levels of satisfaction so low that these products are not viable substitutes for cigarettes. Achieving an optimal level of satisfaction, which may be a level slightly higher than that for cigarettes, maybe the goal.

Our study employed a novel analytic approach using EMA data and highlighted the importance of relatively high levels of satisfaction with e-cigarettes, compared to cigarettes, in predicting reductions in cigarette use and continued use of e-cigarettes. Limitations include the reliance on self-reports of cigarette and e-cigarette use; however, given this was an observational study, there was little incentive for participants to under- or over-report a particular product use. Defining e-cigarette use episodes was necessary to capture subjective experiences before and after use; however, our 15-minute discrete episode definition may not capture patterns of “grazing” sometimes found with use. In our adult sample, though, “grazing” patterns were not common by participant reports, and defining episodes of use remains a challenge for the field. Additionally, we focused on only a few variables that might influence the relationship between satisfaction and use (eg, age, gender, and dependence), when there may be other potential factors that could influence the association between satisfaction and use. For example, the history of use of e-cigarettes or contextual patterns of use (eg, using only in prior smoking contexts as a “bridge” product or using more spaced throughout the day) might be related to longitudinal changes in patterns of both cigarettes and e-cigarettes use. These are questions for future investigations. In sum, regulatory policies need to consider carefully the balance of the experience of satisfaction/pleasure from a product in evaluating its potential net public health benefit.

Supplementary material

Supplementary material is available at Nicotine and Tobacco Research online.

Funding

Dr. Hedeker was supported by NCI grant R01 CA240713 and NIDDK grant R01DK125414. Dr. Mermelstein was supported by P01-CA180945, U01-DA045524 and 1UL1TR002003, from NCI, NIDA, and NCATS at the NIH. Julia Brooks was supported by F31DA057064 from NIDA. Dr. Jao was supported by NHLBI grant K01HL164670. The other authors report no other grant funding. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.The NIH had no role in the study design, collection, analysis, and interpretation of data, writing the report, or the decision to submit the report for publication.

Declaration of interest

The authors have no conflicts of interest to report. The authors have no financial disclosures to report.

Author Contributions

Donald Hedeker (Conceptualization [lead], Formal analysis [lead], Methodology [lead], Software [lead], Writing—original draft [lead]), Julia Brooks (Writing—review & editing [supporting]), Kathleen Diviak (Project administration [equal], Writing—review & editing [equal]), Nancy Jao (Writing—review & editing [equal]), and Robin Mermelstein (Conceptualization [equal], Funding acquisition [lead], Project administration [lead], Writing—review & editing [equal]).

Data availability

The data underlying this article cannot be shared publicly for the privacy of individuals who participated in the study. The data will be shared on reasonable request to Dr. Robin Mermelstein.

Previous Presentations

Work-related to this paper was presented by the first author at the 2023 meeting of the Society of Multivariate Experimental Psychology, Iowa City, Iowa.