Longitudinal associations between substance use problem severity and relative harm perceptions of e‐cigarettes compared with cigarettes: Results from the United States Population Assessment of Tobacco and Health study (2013–2023)

Abstract

Aims

This study examines, among adults who smoke: (1) the association between the harm perception of e‐cigarettes relative to cigarettes and substance use problem (SUP) severity, (2) whether changes in SUP severity over time are associated with changes in harm perceptions of e‐cigarettes relative to cigarettes and (3) whether associations between harm perceptions and vaping initiation are moderated by SUP.

Design

Longitudinal study.

Setting

The study setting was the United States (US) with data from the Population Assessment of Tobacco and Health (PATH) study waves 1–7 collected between 2013 and 2023.

Participants

The study population comprised non‐institutionalized US adults (18+) who smoked cigarettes in the past month.

Measurements

The primary predictor for aims 1 and 2 was SUP severity. The outcomes at follow‐up were: (1) relative harm perception of vaping compared with cigarette smoking [less harmful (accurate) versus more/same harm (inaccurate)], (2) change in SUP status (from no/low at baseline wave to moderate or high severity at follow‐up) and change in relative harm perceptions (from inaccurate at baseline to accurate at follow‐up). For aim 3, the primary predictor was relative harm perception, the outcome was nicotine vaping initiation and SUP was examined as a moderator.

Findings

A higher proportion and odds of respondents with high SUP (versus no/low SUP) had accurate harm perceptions [32.1% versus 28.5%; adjusted odds ratio (aOR) = 1.20; 95% confidence interval (CI) = 1.07–1.33]. Among individuals with no/low SUP at baseline with inaccurate perceptions, transitioning to high SUP at follow‐up was associated with higher odds of developing accurate harm perceptions (aOR = 1.63; 95% CI = 1.19–2.25). Among those who smoked but had no prior history of vaping, at baseline transitioning from inaccurate to accurate harm perception at follow‐up was associated with higher odds of vaping initiation (aOR = 2.08; 95% CI = 1.33–3.25).

Conclusion

People who smoke and have high substance use problem severity appear to perceive vaping as less harmful than cigarette smoking. Notably, among those who smoke but have never vaped, transitioning from inaccurate to accurate perceptions was associated with vaping initiation.

INTRODUCTION

Public perceptions of the relative harm between e‐cigarettes and cigarettes have undergone a significant shift in the past decade [1, 2, 3, 4], with increasing misperceptions that exclusive nicotine e‐cigarette use is more harmful than cigarette smoking [5, 6, 7, 8]. This shift aligns with major health events like the e‐cigarette or vaping product, use associated lung injury (EVALI) outbreak [misattributed to nicotine vaping and later attributed to Δ‐9‐tetrahydrocannabinol (THC)‐related contaminants] [9, 10] media portrayals [11, 12, 13] and regulatory actions relating to e‐cigarettes and cigarettes [14, 15]. Harm perceptions can influence the initiation and cessation of both e‐cigarette and cigarette use [16, 17], impacting overall consumption patterns. Therefore, furthering our understanding of people's perceptions of the relative harms of cigarettes and e‐cigarettes will promote efforts toward clear, evidence‐based communication to mitigate health harms associated with tobacco and nicotine product use.

Multiple independent reviews conclude that nicotine e‐cigarettes expose users to fewer carcinogens and toxicants than combustible cigarettes. For example, evidence reviews from England have concluded that vaping poses only a fraction of the health risks of smoking in the short‐ to medium‐term [18]. Similarly, the National Academies of Sciences, Engineering and Medicine [19] found evidence that switching completely from cigarettes to e‐cigarettes reduces exposure to several harmful and potentially harmful constituents. Although the long‐term effects of e‐cigarette use remain uncertain, weighing these unknowns against the well‐established and severe harms of continued smoking supports the conclusion that e‐cigarettes represent a lower‐risk alternative for adults who smoke. However, misperceptions about the harms of vaping relative to smoking still persist.

People with substance use problems (PWSUPs) represent a unique subpopulation whose perceptions of e‐cigarette harm might differ from the general population. This population includes individuals experiencing social, behavioral or functional challenges because of their use of psychoactive drugs such as alcohol, cannabis, illicit opioids or stimulants [20]. Differences in e‐cigarette harm perceptions among PWSUPs may arise from various factors shaped by their experiences with addiction and stigma [21, 22, 23, 24]. For example, PWSUPs might rely more on peer networks than on public health institutions for information about nicotine product harms, in part because of concerns about stigma or discrimination in medical settings [22]. Peer‐shared information is not necessarily more accurate, but it may become more influential when public health communications emphasize the absolute harms of e‐cigarettes (e.g. youth risk, nicotine exposure) while giving comparatively less attention to relative risk distinctions between e‐cigarettes and combustible cigarettes [19, 25], an imbalance that can potentially contribute to public misunderstanding of comparative harms. A separate mechanism might involve immediate and competing demands related to acquiring or managing their primary substance that reduce the perceived relevance of public health messages about nicotine and tobacco products. In addition, having potentially navigated the stigmatizing experiences within the healthcare system [22, 23, 24], PWSUPs might also be more doubtful of public health messages, including those about nicotine and tobacco products. This skepticism could lead to a different perception of e‐cigarette harms than the general population [26]. Furthermore, harm reduction models of substance use disorder (SUD) treatment are proliferating [27, 28], and PWSUPs may be familiar with harm reduction models of treatment [26, 27, 29], including medication treatment for opioid use disorder (OUD) [27, 29], needle exchange programs [27, 29] and, therefore, may view e‐cigarettes more favorably as tobacco cessation tools [30]. Taken together, these influences may contribute to patterns of e‐cigarette harm perception among PWSUPs that differ from those observed in the broader adult population.

Although we know that the misperception that e‐cigarettes are more harmful than cigarettes has become more prevalent in the general population of adults in the United States (US) and England in the past decade [3, 5, 6, 7, 8], these trends might not be transferable to PWSUPs. For example, some studies have shown that more than half of people on medication treatment for OUD considered e‐cigarettes less harmful than cigarettes [31, 32], and over two‐thirds indicated e‐cigarettes are helpful for quitting cigarette smoking [31, 32, 33]. Although these results suggest that PWSUPs have more accurate vaping perceptions (i.e. vaping is less harmful than smoking) and potentially more favorable views of e‐cigarettes as tobacco harm reduction tools, the findings of these studies may not be generalizable, because both were conducted at select addiction treatment centers (e.g. in a single state at a single institution) and only included adults in treatment for their OUD.

Some studies have demonstrated that the perception that vaping is less harmful than smoking is associated with the initiation of vaping among both people who do and do not smoke [16, 17]. However, this evidence is not consistent [34] and we do not know if this relationship holds in recent years, given increasing vaping misperceptions and other changes in the nicotine and tobacco landscape [2, 3, 5, 6] or how this relationship might be moderated by problematic substance use. This line of inquiry is important because it can inform targeted harm messaging for adults with substance use problems (SUPs) who also smoke at higher rates than adults without SUPs [35], and switching to exclusive use of less harmful non‐combustible tobacco products likely presents a second‐line alternative to cessation pharmacotherapy for adults who smoke combustible tobacco [36, 37, 38].

Importantly, the degree to which these factors influence harm perceptions may shift as SUP severity changes. Variations in substance use severity can alter individuals' competing priorities, reliance on peer networks and engagement with health information, suggesting that perceptions of the relative harms of e‐cigarettes compared to cigarettes may evolve over time. This changing context underscores the importance of examining whether changes in SUP severity correspond to changes in harm perceptions—a concept that has not been previously studied.

To address the gaps in our understanding of the harm perceptions of nicotine e‐cigarettes relative to cigarettes among people who smoke cigarettes with and without SUPs in a representative US sample, the aims of this study are: (1) to examine the association between e‐cigarette harm perceptions relative to cigarettes and substance use problems severity; and (2) to investigate whether changes in substance use problem severity are associated with changes in relative harmful perceptions of e‐cigarettes compared to cigarettes. For aims 1 and 2, we hypothesize that (1) higher SUP severity; and (2) a change from no/low SUP to high SUP severity, will be associated with higher odds of perceiving e‐cigarettes as less harmful than cigarettes. A third aim is to examine how changes in relative harm perceptions are related to vaping initiation among adults who smoke and the modifying effect of SUP. We hypothesize that a change from inaccurate (i.e. e‐cigarettes are equally or more harmful than cigarettes) to accurate (i.e. e‐cigarettes are less harmful than cigarettes) perceptions will be associated with higher odds of vaping initiation among adults who smoke, and SUP status will modify this association.

MATERIALS AND METHODS

Pre‐registration

This study was pre‐registered on the Open Science Framework (https://osf.io/g7va5).

Study design and population

The current observational study uses a longitudinal cohort design. Data from waves 1 to 7 of the Population Assessment for Tobacco and Health (PATH) study (collected between 2013 and 2023) public use files [39]. The PATH study is a longitudinal cohort study that collects information on tobacco use behaviors among non‐institutionalized youth and adults in the United States. Interviews used audio computer‐assisted interviews. More detailed information on the collection procedures are reported in the PATH study files [40, 41]. The authors followed Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines for observational studies.

Eligibility criteria

The study sample comprised adults (18+) reporting current (past month) cigarette smoking. We restricted the sample to adults reporting past‐30‐day cigarette smoking, which captures a broader fraction of people who smoke, including younger adults who may not yet meet established smoker criterion of 100+ lifetime cigarette sticks [42, 43], which is widely used but recognized as a pragmatic and somewhat arbitrary screener [44]. Past‐30‐day smoking is also the standard definition of current use in prior PATH methods description [40] and analyses examining smoking behavior and cigarette‐to‐e‐cigarette transitions, supporting comparability with existing research [45, 46, 47]. For aim 3, this restriction ensures that relative harm perceptions are examined among individuals for whom these perceptions are most behaviorally relevant—those currently making decisions about their own tobacco use and for whom switching to e‐cigarettes represent a potential harm reduction strategy.

Ethics

Written informed consent was obtained from all study respondents according to the Helsinki declaration of ethical principles for research involving human subjects. The study data were collected by Westat, and the Westat institutional review board approved the study protocol and the data collection procedures.

E‐cigarette relative harm perceptions (primary outcome)

All participants were asked: ‘Is using e‐cigarettes or other electronic nicotine products less harmful, about the same or more harmful than smoking cigarettes?’ However, in wave 1, the question was asked from only individuals who indicated they had heard or seen e‐cigarettes. For all waves, we dichotomized responses into ‘less harmful’ (i.e. accurate) (1) and ‘about the same’/ ‘more harmful’ (i.e. inaccurate) (0), to align with the existing evidence on relative risks demonstrating that although not harmless, e‐cigarettes are less harmful than combustible cigarettes [48]. We ran additional analysis where we treated participants who indicated ‘do not know,’ as holding ‘inaccurate’ relative harm perceptions.

Past‐year substance use problem severity (primary predictor)

The primary predictor variable was severity of SUPs using the Global Appraisal for Individual Needs‐short screener (GAINS‐SS) [20, 35]. The GAIN‐SS SUP subscale comprises seven questions (possible score 0–7) that assess past‐year substance use‐related social and functional problems. The score groups respondents into no/low (0–1), moderate (2–3) and high (4 or more) past‐year SUPs [35]. Respondents having low severity are unlikely to receive a diagnosis or require treatment services, while moderate severity suggests a possible diagnosis and the need for services, and high severity indicates a high chance of a diagnosis and a need for treatment services [20, 35]. In addition to using the 3‐level variable, we also treated the SUP scores as a continuous response using each score (0–7) with ‘0’ being participants with the lowest possible SUP score, and ‘7’ for respondents with the highest SUP severity score.

Lifetime e‐cigarette use

Respondents indicated ‘yes’ (1) or ‘no’ (0) to lifetime e‐cigarette use (‘ever used any electronic nicotine product’). The measure was treated as a potential moderator of the relationship between SUPs and relative harm perceptions [49, 50], as well as an outcome for the third research aim where we examined how a change in relative harm perceptions was associated with vaping initiation at time point 2, among never vapers (e‐cigarette naïve adults) at time point 1 and the modifying effect of SUP. Additional analysis was done using current established e‐cigarette use as a moderator. This was done to distinguish the effect in established e‐cigarette users from the lifetime use sample that would include experimental users.

Socio‐demographic covariates

We assessed the socio‐demographic characteristics of respondents using age categories (18–24 years, 25–34 years, 35–44 years, 45–54 years, 55–64 years and ≥65 years); sex assigned at birth (male, female); race/ethnicity (non‐Hispanic White, non‐Hispanic Black, Hispanic and other groups)/multi‐racial groups; and educational attainment (less than high school, high school graduate/General Educational Development and beyond high school).

Statistical analysis

We first described the weighted proportions of study participants' socio‐demographic, primary independent [SUP severity (no/low, moderate, high)], and outcome variable [relative harm perception of e‐cigarettes to cigarettes—more/same harm (ref) versus less harm]. Next, we described the proportional differences in less harmful perception across time (each wave) comparing individuals with no/low SUP to those with moderate and those with high severity SUP.

For the first aim, which examines the difference in relative harm perceptions among people who smoke cigarettes with varying levels of SUP severity, we used generalized estimating equations (GEE) to examine the association between SUP severity and relative harm perceptions within each wave of the PATH study. We included an interaction term between SUP status and lifetime e‐cigarette use to determine how lifetime e‐cigarette use moderates the relationship between SUP and relative harm perceptions. Statistical significance of interaction terms (SUP × lifetime e‐cigarette use) was assessed using Wald tests within the GEE models.

The second aim was to examine how a change in SUP severity is associated with a change in relative harmful perceptions. To examine changes over time, we structured the dataset into biennial wave pairs to approximate 2‐year intervals. Each pair consisted of a baseline (time point 1) and a follow‐up (time point 2) observation, allowing us to examine changes in SUP severity and harm perceptions. The wave pairs are as follows: wave pair 1 [2013/2014 (time point 1) to 2015/2016) time point 2)]; wave pair 2 [2014/2015 (time point 1) to 2016/2018 (time‐point 2)]; wave pair 3 [2016/2018 (time point 1) to 2018/2019 (time point 2)]; wave pair 4 [2018/2019 (time point 1) to 2021 (time point 2)]; wave pair 5 [2021 (time point 1) to 2022/2203 (time point 2)]. The unit of analysis was person‐years, and each observation was structured by wave pair (year). We used adjusted GEE models to estimate the odds of developing a less harmful perception at follow‐up as a function of developing a moderate or severe SUP, given no SUP and a more harmful/same harm perception at baseline. This way we could estimate how relative harm perception changes as SUP changes, using multi‐wave metrics that are representative of the US population [51]. For the third aim, in a subsample of never vapers with inaccurate relative harm perceptions at time point 1, we examined how the interaction between SUP and relative harm perception at time point 2 was associated with vaping initiation at time point 2. This allowed us to examine how relative harm perceptions might influence subsequent vaping initiation and the modifying role of SUP. Based on the 3‐level SUP variable and 2‐level harm perception variable, the interaction terms were jointly tested using a Wald test (testparm) to assess overall interaction, and predicted marginal probabilities were generated using margins command. Although individual interaction coefficients are reported for completeness, interpretation was based on the overall Wald test for interaction and the predicted marginal probabilities because this better characterizes whether the association between harm perception and vaping initiation differs across SUP categories.

The models for aim 2 and aim 3 required a different data management approach to account for baseline and follow‐up time points. As a result, a higher rate of missingness was observed because some participants did not provide responses at both time points, limiting the availability of complete cases. Missingness in the analytic sample following this data management approach is reported in Table S1. To account for the missing observations, we conducted sensitivity analysis using multiple imputations under the missing at random (MAR) assumption (Table S2). Results from the sensitivity analyses were consistent with the main results (Tables S3–S5. At wave 1, 793 respondents were ineligible to answer the relative harm perception question because of lack of awareness of e‐cigarettes and non‐past month cigarette smoking. Among those eligible to respond (past month cigarette users who had seen or heard of e‐cigarettes; n = 13 426), 431 respondents (3.21%) selected ‘do not know’. The number of respondents selecting ‘do not know’ to the relative harm item at each wave is reported in Table S6. We ran an additional analysis where we grouped these responses as ‘inaccurate’ (i.e. more harm/same harm/do not know). The results were also consistent with our main findings and reported in Supporting information (Tables S7–S9).

All models were weighted and adjusted for age, sex, education and race/ethnicity. We used GEE to obtain population‐averaged estimates while accounting for repeated observations within individuals across waves, and all models incorporated the PATH longitudinal survey weights to appropriately account for the study's complex sampling design [39, 40]. The GEE models used an unstructured covariance and within‐person correlation matrices, with a binomial distribution of the outcome variable using the logit link function and robust standard errors. We reported odds ratios using exponentiated coefficients (eform). The study used data from the PATH longitudinal cohort, applying complete case analyses. Sensitivity analysis was conducted using multiple imputations for missing observations assuming MAR. Analyses were weighted using the wave 7 adult, wave 1 cohort all‐waves weights and Fay's adjustment was set at 0.3 to increase the stability of estimates. P‐values were considered significant at <0.05. Data analysis was conducted using Stata 18 software (Stata Corp. 2023).

RESULTS

The study analytic sample comprised 22 612 adults and 72 668 observations who indicated current past 30 day cigarette smoking in the aggregated sample between 2013 and 2014 and 2022 to 2023 (waves 1–7) (Figure S1).

In the aggregated sample (unique participants pooled across waves), 71.24%, 18.81% and 9.95% indicated no/low, moderate and high SUP severity, respectively. Overall, the majority (71.55%) indicated inaccurate perceptions (i.e. e‐cigarettes are more harmful than cigarettes) while 28.45% indicated accurate perceptions (i.e. e‐cigarettes are less harmful than cigarettes). Most participants had a lifetime history of ever e‐cigarette use (72.33%), but less than a quarter had used e‐cigarettes in the past month (21.98%) (Table 1).

TABLE 1.

Description of adults who currently smoke cigarettes in the PATH study waves 1–7 (2013–2014 to 2022–2023).

N = 22 612 participants
Variable	Unweighted n	Weighted % (95% CI)
Age, years
18–24	8353	13.05 (12.50–13.62)
25–34	4462	23.79 (22.78–24.84)
35–44	3167	20.17 (19.04–21.36)
45–54	3045	18.22 (17.24–19.25)
55–64	2377	16.17 (15.14–17.25)
65+	1206	8.6 (7.81–9.46)
Missing	2	–
Sex
Male	12 122	53.94 (52.76–55.12)
Female	10 473	46.06 (44.88–47.24)
Missing	17	–
Race/ethnicity
Non‐Hispanic White	13 121	65.56 (64.22–66.88)
Non‐Hispanic Black	3146	14.79 (13.84–15.79)
Hispanic	4132	13.81 (12.98–14.69)
Other groups	1801	5.83 (5.24–6.49)
Missing	412	–
Education
Less than high school	3948	15.54 (14.69–16.43)
High school/GED	8343	38.78 (37.61–39.97)
Beyond high school	9692	45.68 (44.38–46.98)
Missing	629	–
SUP status
No/low	15 281	71.24 (70.25–72.2)
Moderate	4557	18.81 (18.19–19.46)
High	2292	9.95 (9.37–10.56)
Missing	482	–
Relative harm perceptions
Inaccurate (more/same harm)	11 916	71.55 (70.73–72.37)
Accurate (less harmful)	9332	28.45 (27.63–29.27)
Missing	1364	–
Lifetime e‐cigarette use
Never use	7847	27.67 (26.52–28.85)
Ever use	14 660	72.33 (71.15–73.48)
Missing	105	–
Past month e‐cigarette use
Never use	15 860	78.02 (77.24–78.78)
Ever use	6683	21.98 (21.22–22.76)
Missing	69	–
New participants in each wave of collection
2013–2014	14 219
2014–2015	1577
2015–2016	982
2016–2018	2997
2018–2019	1304
2021	602
2022–2023	931

Note: Table 1 presents characteristics of unique individuals (n = 22 612). These individuals contributed repeated observations across waves 1–7 (72 668 total observations) as shown in Figure S1.

Abbreviations: ENDS, electronic nicotine delivery system; GED, General Educational Development; PATH, Population Assessment of Tobacco and Health; SUP, substance use problem.

Differences in accurate relative harm perceptions across time (each wave) by SUP status (aim 1)

Among adults who currently smoked, the prevalence of accurate relative harm perceptions declined over time (Figure S2). Data from waves 1 to 7 show a decreasing trend in the prevalence of accurate perceptions, with a steeper decline observed in individuals with no/low SUP and moderate SUP (Figure 1). Specifically, in wave 1 (2013–2014), there was no significant difference in the prevalence of accurate relative harm perception between e‐cigarettes and cigarettes by SUP status (P‐value: 0.635) (Figure 1 and Table S10). However, from wave 2 to 7, individuals with high SUP had significantly higher prevalence of accurate relative harm perceptions compared to those with no/low or moderate SUP (Figure 1 and Table S10).

FIGURE 1.

Weighted prevalence of accurate relative harm perceptions by substance use problem (SUP) status among United States (US) adults who currently smoke cigarettes in The Population Assessment of Tobacco and Health (PATH) study waves 1–7 (2013–2014 to 2022–2023). Note: Accurate harm perception was defined as ‘less harmful’ perception of e‐cigarettes compared to combustible cigarettes.

The GEE model in Table 2 showed that individuals with high SUP had higher odds of accurate relative harm perceptions than those with no/low [adjusted OR (aOR) = 1.10; 95% CI = 1.02–1.20]. Likewise, those with moderate SUP (versus no/low) had higher odds of accurate relative perceptions (aOR = 1.20; 95% CI = 1.07–1.33). Individuals with no lifetime history of electronic nicotine delivery system (ENDS) use had lower odds of accurate relative harm perceptions (aOR = 0.60; 95% CI = 0.53–0.67). Although individual coefficients were not statistically significant, the joint Wald test of interaction was significant (P‐value < 0.001) (Table 2). Therefore, interpretation focused on the overall interaction text and model‐based predicted probabilities. Predicted probability estimates showed that in all SUP categories, those with high SUP and lifetime ever ENDS use (34.15%) had the highest prevalence of accurate relative harm perceptions, whereas those with no/low SUP and lifetime never ENDS use (22.18%) had the least (Table 2).

TABLE 2.

GEE model examining the association between SUP status and relative harm perceptions, and the modifying role of lifetime vaping history among US adults who currently smoke cigarettes in the PATH study waves 1–7 (2013–2014 to 2022–2023).

Variable, n = 7405 participants	Weighted % a	aOR	LL 95% CI	UL 95% CI	P‐value
SUP status
No/low	28.5	ref
Moderate	30.1	1.10	1.02	1.20	0.018 *
High	32.1	1.20	1.07	1.33	0.001 *
Lifetime ENDS use status
Ever ENDS use	22.8	ref
Never ENDS use	31.5	0.60	0.53	0.67	<0.001 *
SUP status lifetime ENDS use status
No/low SUP ever ENDS use		ref
Moderate SUP never ENDS use		0.96	0.80	1.17	0.706
High SUP never ENDS use		1.08	0.81	1.45	0.596
Wald b joint test of interaction $X2$ = 108.18, P‐value = <0.001
Age, years
18–24	33.0	ref
25–34	30.8	0.89	0.80	0.99	0.025 *
35–44	27.6	0.75	0.66	0.84	<0.001 *
45–54	29.2	0.81	0.71	0.92	0.002 *
55–64	26.6	0.70	0.61	0.81	<0.001 *
65+	25.3	0.65	0.53	0.79	<0.0018
Sex
Male	32.6	ref
Female	25.3	0.66	0.60	0.73	<0.001 *
Race/ethnicity
Non‐Hispanic White	31.8	ref
Non‐Hispanic Black	23.9	0.64	0.56	0.73	<0.001 *
Hispanic	21.0	0.53	0.46	0.60	<0.001 *
Other	31.6	0.99	0.82	1.20	0.907
Education
Less than high school	24.2	ref
High school/GED	26.1	1.12	0.997	1.26	0.055
Beyond high school	33.2	1.67	1.48	1.88	<0.001 *
Year
2013–2014	57.2	ref
2014–2015	41.5	0.51	0.47	0.55	<0.001 *
2015–2016	31.8	0.33	0.30	0.36	<0.001 *
2016–2018	24.6	0.22	0.20	0.25	<0.001 *
2018–2019	20.2	0.17	0.16	0.19	<0.001 *
2021	14.7	0.12	0.10	0.13	<0.001 *
2022–2023	11.9	0.09	0.08	0.10	<0.001 *
Predicted prevalence of accurate harm perception
No/low SUP and never lifetime ENDS use	22.2
No/low SUP and ever lifetime ENDS use	30.8
Moderate SUP and never lifetime ENDS use	23.2
Moderate SUP and ever lifetime ENDS use	32.6
High SUP and never lifetime ENDS use	26.3
High SUP and ever lifetime ENDS use	34.2

Abbreviations: aOR, adjusted ORs; ENDS, electronic nicotine delivery system; GEE, generalized estimating equation; GED, General Educational Development; LL 95% CI, lower limit 95% CI; PATH, Population Assessment of Tobacco and Health; SUP, substance use problem; UL 95% CI, upper limit 95% CI; US, United States.

^a Adjusted proportions represent predicted proportional probabilities of the outcome for the GEE model. Estimates were obtained using margins command after the GEE model and accounted for survey weights and adjusted all covariates in the model.

^b The joint test of interaction tested the joint significance of all terms (main effects and the interaction term) in the interaction specification.

^* P‐value: <0.05. Italicized values indicate statistical significance at P < 0.05. The model has 31 123 observations.

Association between change in SUP severity and change in relative harmful perceptions (aim 2)

In the subsample of adults who currently smoked and had complete data at both time points. Table 3 shows that as individuals with no/low SUP and inaccurate harm perceptions transition to high SUP between survey waves, and they have 63% higher odds of developing accurate relative harm perceptions compared to those whose SUP status remained no/low (95% CI = 1.19, 2.25). For those who transitioned to moderate SUP status, there was no difference in their relative harm perception compared to those who remained no/low SUP (aOR = 1.12; 95% CI = 0.89–1.41).

TABLE 3.

GEE model examining the association between change in SUP status and change in relative harmful perception among US adults who currently smoke cigarettes in the PATH study waves 1–7 (2013–2014 to 2022–2023).

Odds of accurate relative harm perceptions at follow‐up
Variable, n = 4782 participants	Weighted % a	aOR	LL 95% CI	UL 95% CI	P‐value
SUP status at follow‐up
No/low SUP	9.1	ref
Moderate	10.1	1.12	0.89	1.41	0.322
High	13.9	1.63	1.19	2.25	0.003 *
Age at baseline, years
18–24	14.3	ref
25–34	10.1	0.67	0.53	0.84	0.001 *
35–44	8.7	0.57	0.43	0.74	<0.001 *
45–54	8.6	0.56	0.43	0.72	<0.001 *
55–64	7.8	0.50	0.38	0.66	<0.001 *
65+	7.9	0.51	0.35	0.74	<0.001 *
Sex at baseline
Male	10.1	ref
Female	8.7	0.85	0.72	1.00	0.055
Race at baseline
Non‐Hispanic White	10.3	ref
Non‐Hispanic Black	7.3	0.68	0.54	0.86	0.001 *
Hispanic	6.7	0.63	0.48	0.81	<0.001 *
Other	10.9	1.08	0.73	1.58	0.709
Education at baseline
Less than high school	7.6	ref
High school/GED	8.0	1.06	0.82	1.38	0.652
Beyond high school	11.3	1.56	1.21	2.02	0.001 *
Year—time point 1 to time point 2
2013/2014 to 2015/2016	14.5	ref
2014/2016 to 2016/2018	11.2	0.74	0.58	0.94	0.014 *
2016/2018 to 2018/2019	10.6	0.69	0.55	0.88	0.002 *
2018/2019 to 2021	8.3	0.53	0.41	0.68	<0.001 *
2021 to 2022/2023	5.1	0.31	0.24	0.41	<0.001 *

Abbreviations: aOR, adjusted ORs; ENDS, electronic nicotine delivery system; GEE, generalized estimating equation; GED, General Educational Development; LL 95% CI, lower limit 95% CI; PATH, Population Assessment of Tobacco and Health; SUP, substance use problem; UL 95% CI, upper limit 95% CI; US, United States.

^a Adjusted proportions represent predicted proportional probabilities of the outcome for the GEE model. Estimates were obtained using margins command after the GEE model and accounted for survey weights and adjusted all covariates in the model.

^* P‐value: <0.05. Italicized values indicate statistical significance at P < 0.05. The model has 10 861 observations.

Modifying role of SUP on the association between relative harm perceptions and vaping initiation at follow‐up among vaping naïve individuals who held inaccurate harm perceptions at baseline (aim 3)

Among a subsample of individuals who currently smoked, had no lifetime history of vaping and held inaccurate relative harm perceptions at baseline and those who transitioned to accurate beliefs at follow‐up had higher odds of vaping initiation compared to those who continued to hold inaccurate relative harm perceptions (Table 4). Similarly, among this subsample, those who had moderate SUP (aOR = 1.58; 95% CI = 1.20–2.08) or high SUP (aOR = 2.15; 95% CI = 1.46–3.16) had higher odds of vaping initiation compared to those with no/low SUP. The joint test of interaction terms was significant (P‐value < 0.001). Predicted probability estimates of the interaction between SUP status and relative harm perceptions demonstrated that regardless of SUP status, those with accurate relative harm perceptions had a higher probability of initiating vaping than those with inaccurate relative harm beliefs [no/low SUP and inaccurate perceptions (10.37%) versus no/low SUP accurate perceptions (18.38%), high SUP and inaccurate perceptions (18.80%) versus high SUP and accurate perceptions (39.75%)] (Table 4).

TABLE 4.

GEE model examining the modifying role of SUP on the association between relative harm perceptions and vaping initiation at follow‐up among vaping naïve individuals who held inaccurate harm perceptions at baseline, among US adults who currently smoke cigarettes in the PATH study waves 1–7 (2013–2014 to 2022–2023).

ENDS initiation at follow‐up, n = 1718 participants
Variable	Weighted % a	aOR	LL 95% CI	UL 95% CI	P‐value
SUP status at follow‐up
No/low SUP	11.0	ref
Moderate	15.9	1.58	1.20	2.07	0.001 *
High	20.3	2.15	1.46	3.16	<0.001 *
Relative harm perceptions at follow‐up
Inaccurate (more/same harm)	11.6	ref
Accurate (less harmful)	21.4	2.08	1.33	3.25	0.001 *
SUP status and relative harm perceptions at follow‐up
No/low SUP and inaccurate harm perceptions		ref
Moderate SUP and accurate harm perceptions		1.24	0.53	2.92	0.623
High SUP and accurate harm perceptions		1.59	0.46	5.46	0.463
Wald b joint test of interaction $X2$ = 46.16, P‐value: <0.001
Age at baseline, year
18–24	30.4	ref
25–34	16.1	0.41	0.28	0.61	<0.001 *
35–44	14.2	0.35	0.24	0.52	<0.001 *
45–54	11.5	0.27	0.19	0.40	<0.001 *
55–64	7.7	0.17	0.11	0.27	<0.001 *
65+	4.4	0.09	0.05	0.19	<0.001 *
Sex at baseline
Male	12.5	ref
Female	12.2	0.97	0.76	1.24	0.836
Race at baseline
Non‐Hispanic White	13.2	ref
Non‐Hispanic Black	11.8	0.87	0.64	1.18	0.369
Hispanic	11.4	0.83	0.61	1.13	0.239
Other	9.6	0.67	0.38	1.19	0.173
Education at baseline
Less than high school	11.1	ref
High school/GED	11.8	1.08	0.80	1.47	0.605
Beyond high school	13.7	1.31	0.96	1.80	0.091
Year—time point 1 to time point 2
2013/2014 to 2015/202016	21.1	ref
2014/2016 to 2016/2018	14.4	0.60	0.50	0.74	<0.001 *
2016/2018 to 2018/2019	8.7	0.33	0.23	0.47	<0.001 *
2018/2019 to 2021	9.6	0.37	0.26	0.54	<0.001 *
2021 to 2022/2023	6.4	0.23	0.15	0.36	<0.001 *
No/low and more harmful/same harm perceptions	10.4
No/low and less harmful perceptions	18.4
Moderate and more harmful/same harm perceptions	14.9
Moderate and less harmful perceptions	28.9
High and more harmful/same harm perceptions	18.8
High and less harmful perceptions	39.8

Abbreviations: aOR, adjusted ORs; ENDS, electronic nicotine delivery system; GEE, generalized estimating equation; GED, General Educational Development; LL 95% CI, lower limit 95% CI; PATH, Population Assessment of Tobacco and Health; SUP, substance use problem; UL 95% CI, upper limit 95% confidence interval; US, United States.

^a Adjusted proportions represent predicted proportional probabilities of the outcome for the GEE model. Estimates were obtained using margins command after the GEE model and accounted for survey weights and adjusted all covariates in the model.

^b The joint test of interaction tested the joint significance of all terms (main effects and the interaction term) in the interaction specification.

^* P‐value: <0.05. Italicized values indicate statistical significance at P < 0.05. The model has 4131 observations.

DISCUSSION

The majority of US adults who smoke cigarettes in the PATH study held inaccurate perceptions of the relative harms of vaping and smoking, and an increasing trend toward inaccurate perceptions were observed between 2013 and 2023, regardless of substance use problem status. Our hypothesis (aim 1) that people with SUPs might have different e‐cigarette harm perceptions than those without SUPs was partially supported. Specifically, as hypothesized, individuals with high SUP severity (compared to low/no SUP) had higher odds of accurate perceptions of relative harms of vaping and smoking, and this relationship was significantly moderated by e‐cigarette use history. However, the prevalence of accurate perceptions among people with high SUP severity also declined over time, suggesting that they are still affected by the broader shift toward increasingly inaccurate perceptions, although to a lesser extent than adults with no/low SUP. As hypothesized, we also observed that among adults who smoke, those who transitioned from no/low SUP to high SUP between survey waves had higher odds of developing accurate perceptions than those whose SUP status did not change. In addition, our third hypothesis was supported that among individuals who smoke with no prior history of vaping, transitioning from inaccurate perceptions to accurate perceptions was associated with higher odds of initiating vaping.

Our findings are similar to studies that were conducted among people receiving treatment for OUDs [31, 32], which also identified more accurate perceptions among those with higher substance use severity. Additionally, our finding that more people who smoked with a lifetime history of vaping hold accurate perceptions compared to those who had never used e‐cigarettes are consistent with existing studies [52, 53, 54]. However, our study extends this literature by examining these trends over a decade in a nationally representative US sample and by using a comprehensive assessment of problematic substance use symptoms associated with a broad range of licit and illicit psychoactive drugs [20, 35]. These results indicate that relative e‐cigarette harm perceptions are likely different among people with substance use problems compared to those without, potentially reflecting how lived experience might shape comparative beliefs about tobacco products and their risk.

The current study also shows that transitioning from inaccurate to accurate harm perceptions was associated with higher odds of initiating vaping among individuals who had never vaped, but had smoked at baseline. This relationship and other results may reflect a complex interplay between the lived experience of substance use, harm perceptions and their influence on vaping initiation. However, because we could not determine the exact timing of relative e‐cigarette harm perceptions and vaping initiation between survey waves, it is possible that some individuals had initiated vaping before developing accurate perceptions. Future longitudinal research can investigate precisely how changing harm perceptions influence vaping initiation among people who smoke and how this also relates to SUP. This line of inquiry is important given the decline of cigarette smoking among US adults, but concurrent increase in vaping over the past decade, and the goal of supporting at‐risk populations to switch away from smoking.

Harm perception mechanisms within cognitive and emotional dimensions may help explain the study findings [55]. Cognitively, PWSUPs may be less trusting of established public health institutions and consequently less exposed to mainstream media messages about the harms associated with e‐cigarettes, such as EVALI [9, 10, 11]. PWSUPs may also rely more on information from peer networks [33, 56]. This combination may impact their processing of information on harms [21]. Emotionally, because of their risk environments, PWSUPs may prioritize certain harms differently [26, 57], using availability heuristics [55]. For example, the high prevalence of cigarette use within their social networks often leads to more frequent encounters with tobacco‐related illnesses [58, 59], increasing the salience and familiarity with combustible cigarette harm over those associated with e‐cigarettes, which are less salient. In addition, several studies have shown that PWSUPs are interested in switching from cigarettes to e‐cigarettes to reduce the harm from combustible tobacco use [30, 33, 60]. Recent evidence also suggests that e‐cigarettes are an effective cigarette cessation aid for adults who smoke in the general adult population and some preliminary data suggest among adults who smoke with SUD [36, 56, 61]. Therefore, interest in e‐cigarettes for tobacco harm reduction may contribute to more prevalent lower harm beliefs about e‐cigarettes among PWSUPs [33, 60, 62].

Our findings have important implications for clinical practice. The higher prevalence of accurate perceptions of the relative harm of vaping and smoking among adults with high SUP who smoke could present opportunities for harm reduction intervention, particularly for those who have been unsuccessful with traditional cessation therapy. Treatment seeking settings for substance use may also serve as valuable channels for providing harm reduction education about switching, as PWSUP often use and trust these services. However, because most adults with high SUP still hold inaccurate perceptions overall, these settings remain critical venues for delivering accurate and tailored harm‐reduction messaging. Health care providers can also use this opportunity to encourage the use of the US Food and Drug Administration‐approved first‐line cessation treatments (behavioral counseling with pharmacotherapy) among individuals with SUP who smoke [63]. However, if unsuccessful with these first‐line treatments, individuals might consider switching to exclusive use of less harmful non‐combustible nicotine products like e‐cigarettes [36, 64].

The results from this study also suggest implications that are consistent with the Capability‐Opportunity‐Motivation Behavior change (COM‐B) framework [65]. For example, vaping may be more compatible with the ‘capability’ and habits of individuals who smoke, since vaping and smoking share similar physical, behavioral and social features. Our results also show that a significant proportion believes e‐cigarettes are less harmful than cigarettes, offering an ‘opportunity’ for behavior change. Because existing studies demonstrate that, among those who smoke, people with SUPs have interest in quitting [33, 60], and some use e‐cigarettes as a cessation tool [30], this reflects ‘motivation’ for behavior change in this population. Therefore, our observations indicate an opportunity to reduce harm in this population disproportionately affected by tobacco use by leveraging the opportunity to correct misperceptions.

Our study has several limitations. First, we use a single measure to assess relative harm perceptions. However, harms (or risk) can be interpreted across various dimensions (e.g. absolute, relative and perceptions of nicotine as harmful) [55]. Second, the PATH study wave 6 data were collected during the coronavirus disease 2019 (COVID‐19) pandemic, and harm perceptions might have been broadly influenced by pandemic‐related changes in the social environment and the public health messaging during this period [66, 67]. In addition, the use of wave pairs does not account for period effects over chronological time, for example, perceptions during wave pairs overlapping with the COVID‐19 pandemic or EVALI might differ from those observed in other wave pairs (or periods). Third, measures of cigarette and substance use behaviors were based on self‐reports and not biochemically verified assessments, therefore, they may be prone to recall bias. Fourth, we were unable to determine the exact time points when perceptions or behaviors changed and the order in which they changed, therefore, we can infer an association, but not causal direction of the relationships. Last, because the MAR assumption underlying our imputations cannot be empirically verified, some residual bias may remain [68]. Although we included comprehensive observed covariates in our imputation models to make MAR more plausible, we cannot rule out that missingness may be related to unobserved factors, which could introduce some residual bias. However, the consistency between our complete case analyses and imputed results provides some indication regarding the robustness of our findings. Despite these limitations, our results contribute to the discourse on harm perceptions of e‐cigarettes compared to cigarettes among people with and without SUP over a decade by using a weighted non‐institutionalized national sample [67], with a scale that comprehensively assesses individuals on a range of symptoms for substance use problem severity [35].

CONCLUSION

The findings from this study demonstrate that, among people who smoke cigarettes, high SUP severity was associated with perceiving e‐cigarettes to be less harmful than cigarettes. A higher proportion of individuals with inaccurate perceptions who transitioned from no/low SUP to high SUP developed accurate perceptions, and a higher proportion of those who transitioned from inaccurate to accurate perceptions initiated vaping. Although these associations cannot establish directionality [69], identifying subgroups where accurate harm perceptions are more prevalent may help clinicians tailor communication and address misperceptions within substance use treatment settings. These findings also highlight opportunities to promote accurate harm perceptions, specifically among adults who exclusively smoke combustible tobacco and might be interested in switching to less harmful nicotine products like e‐cigarettes as a second‐line approach after unsuccessful attempts with standard cessation pharmacotherapy.

AUTHOR CONTRIBUTIONS

Olufemi Erinoso: Conceptualization (lead); formal analysis (lead); methodology (equal); resources (lead); writing—original draft (lead); writing—review and editing (lead). Katherine East: Conceptualization (lead); methodology (equal); validation (equal); writing—review and editing (equal). Joanna Streck: Conceptualization (equal); methodology (equal); validation (equal); visualization (equal); writing—review and editing (equal). Karin Kasza: Conceptualization (equal); methodology (equal); supervision (equal); validation (equal); visualization (equal); writing—review and editing (equal). Andrew Hyland: Conceptualization (equal); supervision (equal); validation (equal); writing—review and editing (equal).

DECLARATION OF INTERESTS

None.

Supporting information

Table S1. Description of analytic sample for aims 2–3 examining longitudinal associations between SUP status and accurate perceptions, and change in relative perceptions and ENDS initiation among US adults who currently smoke cigarettes at baseline between waves 1 and 7 (2013–2014 to 2022–2023).

Table S2. Missing imputations assuming Missing at Random (MAR) among US adults who currently smoke cigarettes.

Table S3. Missing imputations assuming Missing at Random (MAR) to examine the modifying role of lifetime ENDS use on SUP status and relative harm perceptions among US adults who currently smoke cigarettes in the PATH Study waves 1–7 (2013–2014 to 2022–2023).

Table S4. Missing imputations assuming Missing at Random (MAR) to examine association between change in SUP status and change in relative harmful perception among US adults who currently smoke cigarettes in the PATH Study waves 1–7 (2013–2014 to 2022–2023).

Table S5. Missing imputations assuming Missing at Random (MAR) to examine the modifying role of relative harm perceptions on the relationship between SUP status and ENDS initiation among ENDS abstinent among US adults who currently smoke cigarettes in the PATH Study waves 1–7 (2013–2014 to 2022–2023).

Table S6. Wave‐Specific “Do not Know” Responses to the Relative Harm Perception Question (waves 1–7) among respondents who are aware of e‐cigarettes and consumed cigarettes in the past month.

Table S7. Modifying role of lifetime ENDS use on SUP status and relative harm perceptions among US adults who currently smoke cigarettes in the PATH Study waves 1–7 (2013–2014 to 2022–2023) in regrouped inaccurate perception (more harm/ same harm/ do not know) category.

Table S8. Association between change in SUP status and change in relative harmful perception among US adults who currently smoke cigarettes in the PATH Study waves 1–7 (2013–2014 to 2022–2023) in regrouped inaccurate perception (more harm/ same harm/ do not know) category.

Table S9. Modifying role of SUP on the association between relative harm perceptions and vaping initiation at follow‐up among vaping naïve individuals who held inaccurate harm perceptions at baseline, among US adults who currently smoke cigarettes in the PATH Study waves 1–7 (2013–2014 to 2022–2023) in regrouped inaccurate perception (more harm/ same harm/ do not know) category.

Table S10. Weighted prevalence of accurate relative harm perceptions by SUP status among US adults who currently smoke cigarettes in the PATH Study waves 1–7 (2013–2014 to 2022–2023).

Figure S1. Flow chart showing analytic sample selection for US adults who currently smoke cigarettes in the PATH study waves 1–7.

Figure S2. Study conceptual model showing connections between analyzed primary variables in adults who currently smoke cigarettes.

Erinoso O, East K, Streck J, Kasza K, Hyland A. Longitudinal associations between substance use problem severity and relative harm perceptions of e‐cigarettes compared with cigarettes: Results from the United States Population Assessment of Tobacco and Health study (2013–2023). Addiction. 2026;121(8):2210–2224. 10.1111/add.70426

DATA AVAILABILITY STATEMENT

The PATH study waves 1–7 public use files data may be obtained from a third‐party following application request here: https://www.icpsr.umich.edu/icpsrweb/NAHDAP/studies/36231.