A Randomized Clinical Trial Examining the Effects of Instructions for Electronic Cigarette Use on Smoking-Related Behaviors and Biomarkers of Exposure

Abstract

Introduction

Electronic cigarettes (e-cigarettes) have the potential to significantly reduce exposure to harmful constituents associated with cigarette smoking when smokers completely substitute cigarettes with e-cigarettes. This study examined patterns of e-cigarette and cigarette use, and extent of toxicant exposure, if smokers were instructed and incentivized to completely switch to e-cigarettes compared to instructions to use the product ad libitum.

Aims and Methods

US adult daily smokers (n = 264; 49.2% female; M_age = 47.0), uninterested in quitting smoking immediately, were recruited from Minneapolis, MN, Columbus, OH, and Buffalo, NY. Participants were randomized to 8 weeks of instructions for (1) ad libitum use of e-cigarettes (AD-E), (2) complete substitution of cigarettes with e-cigarettes (CS-E), (3) complete substitution of cigarettes with nicotine gum or lozenge (CS-NRT), or (4) continue smoking of usual brand cigarettes (UB). Participants were incentivized for protocol compliance, including complete switching in the CS-E and CS-NRT groups. Outcome variables were cigarette smoking rate and tobacco-related biomarkers of exposure.

Results

Smokers in the CS-E and CS-NRT groups showed lower rates of smoking and lower exposure to carbon monoxide, tobacco carcinogens, and other toxicants than smokers in the AD-E group. In general, no significant differences were observed between CS-E versus CS-NRT or between AD-E versus UB for most biomarkers. Significantly higher 7-day point prevalence smoke-free rates were observed for CS-E versus CS-NRT.

Conclusions

Smokers instructed and incentivized to completely switch to e-cigarettes resulted in lower smoking rates and greater reductions in exposures to harmful chemicals than smokers instructed to use the product ad libitum.

Implications

Smokers instructed to completely substitute e-cigarettes for cigarettes displayed significantly lower levels of smoking and biomarkers of exposure to carcinogens and toxicants, compared to smokers instructed to use e-cigarettes ad libitum and similar levels as smokers instructed to completely substitute with nicotine replacement therapies. Furthermore, a higher rate of complete switching was achieved with e-cigarettes versus nicotine replacement therapies. Approaches to maximize complete substitution with e-cigarettes are an important area for future research.

Introduction

The National Academies of Sciences, Engineering, and Medicine (NASEM) report stated that electronic cigarettes (e-cigarettes) have substantially less harmful constituents when compared to cigarettes, and smokers are likely to experience reduced health risk in the short term if they completely switched over to these products.¹ Studies have shown that exclusive e-cigarette use results in substantially reduced levels of biomarkers of exposure to harmful constituents when compared to cigarette smoking,^1,2 but users of both cigarettes and e-cigarettes experienced less reduction^3–5 or similar if not higher^2,6 levels of toxicant and carcinogen exposures. Unfortunately, the majority of smokers who use e-cigarettes are dual users,^7,8 therefore they are unlikely to experience a substantial reduction of tobacco-related harm.

In this study, we determined the extent to which instructions and monetary incentives to completely switch to e-cigarettes among smokers can reduce smoking rate and biomarkers of exposure compared to instructions for ad libitum use of e-cigarettes and compared to complete substitution with nicotine replacement therapies (NRTs). We also determined how instructions for ad libitum e-cigarette use compare to continued smoking with usual brand cigarettes.

Methods

Participant Recruitment

Smokers were recruited via various media outlets across three institutions: (1) University of Minnesota, Twin Cities (lead site), (2) The Ohio State University, Columbus, OH, and (3) Roswell Park Cancer Center, Buffalo, NY. The advertisements stated that a study was recruiting smokers who were interested in trying a product that may reduce exposure to harmful tobacco smoke. Interested smokers called the respective study site and were provided a brief description of the study and prescreened for eligibility.

After initial telephone screenings, potential participants were asked to attend a screening visit, during which informed consent was obtained and further assessment of eligibility was made. Information on demographics, tobacco use history and dependence, medical history, vitals, and alveolar carbon monoxide (CO) levels was obtained and urine pregnancy tests were given to women of childbearing potential. Inclusion criteria included (1) at least18 years of age, (2) smoking at least 5 cigarettes/day (CPD) for the past year with a breath CO at least 10 ppm or NicAlert test = level 6 if CO less than 10 ppm, and (3) in stable physical and mental health. Exclusion criteria included (1) a serious quit attempt in the past 3 months, (2) recent (<3 months) alcohol or drug abuse problems, (3) regular use of other nicotine or tobacco products (eg, >9 days per month to minimize confounding effects of these products on biomarker outcomes), (4) planning to quit smoking in the next 3 months, (5) chronic conditions affecting results of biomarker analyses (eg, liver disease), (6) currently using NRT or other cessation medications, or (7) pregnant, planning to become pregnant, or breastfeeding.

Institutional Review Board approval of the study was obtained from each academic institution and a Data and Safety Monitoring Board met annually to monitor study progress and adverse events.

Design

Eligible participants engaged in a 2-week baseline period (weeks −1 and 0) and 8 weeks of product assignment. During baseline, participants were asked to report the number of cigarettes smoked the previous day using an Interactive Voice Recording (IVR) system that called participants on a daily basis. At the baseline clinic visits, an inquiry was made about other tobacco or nicotine use and participants were asked to complete various subjective measures related to tobacco use. Vital signs and CO levels were also obtained. On the second baseline visit, first void urines were collected for biomarker analyses.

Participants who comprised the analysis for this study were randomized to one of four groups: (1) ad libitum use of e-cigarettes (ie, “use e-cigarettes whenever you like instead of a cigarette; can smoke as many or as few cigarettes as you want”; AD-E); (2) complete substitution with e-cigarettes (ie, “you will stop smoking cigarettes and use only e-cigarettes”; CS-E); (3) complete substitution with 4 mg nicotine gum or lozenge, with the participant choosing what product they would like to use (ie, “you will stop smoking cigarettes and use only nicotine gum or lozenge”; CS-NRT); and (4) continued smoking with usual brand cigarettes (UB). These four experimental groups in this study were culled from two sets of studies, one of which also examined groups comprised of complete substitution and ad libitum use of snus (a spitless smokeless tobacco product), thereby resulting in randomization to six experimental groups. Due to the challenges associated with recruiting smokers interested in being potentially randomized to using snus, the two snus groups were dropped about midway through the study, resulting in the four experimental groups described in this study. Both studies used almost identical study procedures. The study that included snus had a 1-week sampling period of their assigned product, which was eliminated in the subsequent study. Both studies involved initial weekly intervention visits followed by biweekly visits (eg, weeks 1, 2, 3, 4, 6, and 8); however, the week 3 visit was eliminated from the subsequent study.

After product assignment, participants continued to complete daily diaries via Interactive Voice Recording for cigarettes and assigned product use for the duration of the clinical trial and assessed for the measures obtained during baseline. All unused products and used e-cigarette cartridges were collected at each visit, with the exception of UB cigarettes. During the study, those in complete substitution groups received brief counseling on how to avoid smoking cigarettes. At the end of the clinical trial phase (week 8), smokers in the UB cigarette condition were offered e-cigarettes or NRT for up to 8 weeks, with a choice of product and no specific instructions for use. All subjects were strongly encouraged to quit using all tobacco products at the end of their study participation; treatment manual and resources were provided. A safety check for adverse events was conducted at a week 20 follow-up.

Compensation

Participants received compensation for attendance at each clinic visit, submission of biological samples, and completion of visit forms. Participants received a bonus if they were protocol compliant. Protocol compliance for the UB and AD-E group included completion of the study requirements such as keeping accurate record of product use, completion of daily diaries, and returning unused products. In order to maximize complete substitution, smokers in the CS-E and CS-NRT group also had to have a CO not more than 4 ppm at the visit for the bonus payment. If their CO was more than 4 ppm, they were provided half the bonus payment for that visit. The bonus payment was based on escalating payments ($30 at week 1 and $10 increments each subsequent visit, with increments continually increasing even if the participant did not meet study requirements in the prior visit). Total compensation was equivalent across all groups if all study requirements (including the additional CO levels for CS-E and CS-NRT arms) were met by the participant.

Products

The primary e-cigarette product was Vuse Solo (4.8% nicotine, manufactured by RJ Reynolds, Inc). Initially a choice of Blu cigarettes (cartridge-based system, marketed previously by Lorillard) and Fin (prefilled tanks system, manufactured by Fin Branding Group) was offered; but because Vuse attained the highest market share during the early phase of the study, we switched exclusively to Vuse. In total, only three participants chose Blu and six participants used both Vuse and Fin. Participants could choose one of four flavors: tobacco, mint, menthol, and berry. Participants were provided seven cartridges per week with the option of returning to the clinic before their next visit to obtain additional cartridges if needed. Nicotine gum or lozenge (participant’s choice) was provided in the 4 mg dose but down-titrated to 2 mg if adverse side effects were experienced. Nicotine gum came in mint, cinnamon, and fruit flavors, while the nicotine lozenge was mint or cherry flavors. All these products were provided free to the participants. Smokers assigned to the UB or AD-E cigarette condition purchased their own cigarettes.

Measures

Biomarker Analyses

The biomarker outcomes involved different categories of tobacco exposures including nicotine, expired CO, tobacco-specific nitrosamines, polycyclic aromatic hydrocarbons, and volatile organic compounds (VOCs). These biomarkers included (1) urinary total nicotine equivalents (total nicotine + total cotinine + total 3′-hydroxycotinine; TNE), (2) exhaled CO, (3) urinary 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol and its glucuronides (total NNAL, biomarker for NNK), (4) urinary phenanthrene tetraol (PheT, an indicator of carcinogenic polycyclic aromatic hydrocarbons), and (5) urinary metabolites of VOCs (mercapturic acids)—2-cyanoethylmercapturic acid (CEMA, biomarker for acrylonitrile), 3-hydroxypropylmercapturic acid (3-HPMA, biomarker for acrolein), 3-hydroxy-1-methylpropylmercapturic acid (HMPMA, biomarker for crotonaldehyde/methylvinyl ketone), 2-hydroxypropylmercapturic acid (2-HPMA, biomarker for propylene oxide), and N-acetyl-S-(carbamoylethyl)-L-cysteine (AAMA, biomarker for acrylamide).

Participants provided exhaled CO using a Bedfont Smokerlyzer. Urinary TNE, total NNAL, and the mercapturic acids were analyzed using Liquid Chromatography–Mass Spectrometry.^9–11 PheT was analyzed by gas chromatography–tandem mass spectrometry as previously described.¹¹ Urinary concentrations of creatinine were analyzed using a colorimetric microplate assay (CRE34-K01) purchased from Eagle Bioscience (http://stores.eaglebio.com/creatinine-microplate-assay-kit). All biomarker analyses were adjusted for creatinine to account for urine dilution variability between participants.

CO was collected at each visit. Urinary TNEs were analyzed twice at baseline (weeks −1, 0) and weeks 4 and 8 during the experimental period. TNEs at weeks −1 and 0 were averaged to create a baseline TNE measurement. All other biomarkers were analyzed at weeks 0, 4, and 8.

Subjective Measures

Subjective measures that were administered and analyzed for this study included the Fagerström Test for Nicotine Dependence (FTND)¹² and the Center for Epidemiological Studies—Depression (CES-D) scale.¹³ Any changes in physical or mental health were assessed and documented as adverse events related to the study product when appropriate. Additional measures are described in Supplementary Table 1, and findings from some of the measures are provided in Supplementary Tables 4–6.

Data Analysis

Baseline demographics and tobacco use history variables were summarized for the four groups and compared using chi-square and Kruskal–Wallis tests. Biospecimen samples below the limit of quantification were imputed as 50% the limit of quantification. All biomarkers except TNE were transformed to the logarithmic scale and reported as geometric means with 95% confidence intervals (CIs). TNE was summarized by the mean, SD and analyzed in its original scale.

Poisson regression with repeated measures using generalized estimating equations evaluated the number of cigarettes and product used by group and week from baseline until week 8.¹⁴ A linear mixed model compared the groups and timepoints when analyzing the biomarkers (and subjective measures presented in the Supplementary Tables) assessed multiple times over the study period.¹⁵ These methods provide flexibility in their handling of missing data, such that participants who are missing at a given timepoint are not excluded from the analysis. The assumptions on the missing data for the mixed effects models and generalized estimating equation models are missing at random and missing completely at random, respectively. The following analytic steps were used for all the repeated measures analyses. The first step was the unadjusted model including only the group indicator, the week number, and their interaction. If the p-value for the interaction was more than .1, the interaction term was dropped, and the second step was the adjusted model that included a preselected set of baseline covariates. If the interaction term had a p-value less than .1, the four groups were analyzed separately with an adjusted model including the week and the preselected covariates. These baseline covariates included gender, race (White/non-White), age, employment (part/full time vs. other), FTND, CES-D, TNE, the study from which participants were recruited, and use of other combusted tobacco. These additional variables were included in the models in a step-down selection procedure to obtain a model containing only significant covariates (p-value < .05). The group and week indicators always remained in the model with week treated as a categorical variable. The exponentiated coefficients from the adjusted regression models for the cigarettes, product used, and biomarkers represent the estimated ratio (95% CI) of the outcome between levels of the predictor variables. The coefficients for TNE (and the subjective measures) estimate the mean difference for every one unit/level increase in the covariate.

Changes from baseline to 4 and 8 weeks were reported as the median difference (weeks 4 and 8 minus baseline) for CPD and TNE (and the subjective measures) and the geometric mean ratio (GMR), representing the ratio of weeks 4 and 8 over baseline, for all other biomarkers. Comparisons between the four groups were conducted with the Kruskal–Wallis test. Three selected pairs of groups of primary interest were further evaluated using the two-sample Wilcoxon rank-sum test, namely, AD-E versus CS-E, CS-E versus CS-NRT, and AD-E versus UB. No comparisons were made between UB with CS-E and CS-NRT because the latter groups are likely to have significantly reduced exposures than the UB group, and between AD-E and CS-NRT because of different instructions and products. Furthermore, we wanted to minimize multiple comparisons. Statistical significance was determined by a p-value cutoff of .05/3 = .017 to control for multiple comparisons for all outcomes, with the exception of the subscales for the Product Evaluation Scale. Since the UB group does not use any product, the p-value cutoff for two planned pairwise comparisons for these scales was .05/2 = .025.

Days when no cigarette smoking is reported were summarized by group for the percent of smoke-free days over the entire study period and compared between all groups with the Kruskal–Wallis test and pairwise with the Wilcoxon rank-sum test. Percent of days takes into account the duration the participant was in treatment. Missing values were considered to be smoking days. The 7-day CO-verified quit rate at 8 weeks was compared between the two complete substitution groups with Fisher’s exact test. Dropout rates between groups by week 8 were compared using a chi-squared test. All statistical analysis was performed using SAS 9.4 (SAS Institute, Inc, Cary, NC).

Results

Participant Characteristics

This study was initiated on November 25, 2014 and the last follow-up occurred on December 2, 2018. Of the 292 participants randomized, 264 entered into the trial. Table 1 provides baseline demographic information and tobacco use history of trial-entered participants across groups. Over half of the participants were White (53.4%), with 49.2% female and a median age of 47.0 (range 19–74) years. The median number of CPD was 15 (range 5–40), the median FTND score was 3.0 (range 0–8), and the median TNE concentration was 62.7 nmol/mg creatinine (range 5.4–236.6). There was a significant difference in dropout rates across groups following study entry (p = .041), with the highest dropout rates observed in the complete substitution groups, particularly in the NRT group (see Supplementary Figure 1). Supplementary Figure 1 shows reasons for dropping out after entry into the clinical trial.

Table 1.

Demographics and Smoking History Overall and by Study Groups

Variable	All arms (N = 264)	AD-E (N = 76)	CS-E (N = 76)	CS-NRT (N = 76)	Usual brand (N = 36)	p +
Study site, (n, %)
UMN	126 (47.7)	36 (47.4)	37 (48.7)	36 (47.4)	17 (47.2)	.947
OSU/APP	102 (38.6)	31 (40.8)	30 (39.5)	29 (38.2)	12 (33.3)
Roswell	36 (13.6)	9 (11.8)	9 (11.8)	11 (14.5)	7 (19.4)
Age, median [min/max]	47.0 [19/74]	48.0 [19/68]	45.0 [20/71]	50.5 [21/74]	46.5 [21/64]	.626
Sex, female (n, %)	130 (49.2)	38 (50.0)	36 (47.4)	39 (51.3)	17 (47.2)	.957
Race, three-category (n, %)
White	141 (53.4)	38 (60.0)	48 (63.2)	37 (48.7)	18 (50.0)	.363++
Black	100 (37.9)	28 (36.8)	24 (31.6)	31 (40.8)	17 (47.2)
Other	23 (8.7)	10 (13.2)	4 (5.3)	8 (10.5)	1 (2.8)
Education (n, %)
Eighth grade or less	1 (0.4)	1 (1.3)	0 (0.0)	0 (0.0)	0 (0.0)	—
Some high school	24 (9.1)	7 (9.2)	7 (9.2)	7 (9.2)	3 (8.3)
High school	85 (32.2)	27 (35.5)	22 (29.0)	27 (35.5)	9 (25.0)
Some college	126 (47.7)	32 (42.1)	40 (52.6)	35 (46.1)	19 (52.8)
College grad	21 (8.0)	6 (7.9)	5 (6.6)	6 (7.9)	4 (11.1)
Grad./Prof. degree	7 (2.7)	3 (4.0)	2 (2.6)	1 (1.3)	1 (2.8)
Education, two-category (N, %)
High school grade or less	110 (41.7)	35 (46.1)	29 (38.2)	34 (44.7)	12 (33.3)	.511
Some college or more	154 (58.3)	41 (54.0)	47 (61.8)	42 (55.3)	24 (66.7)
Income (n, %)
Less than $30 000	180 (68.2)	53 (69.7)	48 (63.2)	53 (69.7)	26 (72.2)	.723
More than $30 000	84 (31.8)	23 (30.3)	28 (36.8)	23 (30.3)	10 (27.8)
Current employment
Full/part-time (n, %)	90 (34.1)	28 (36.8)	28 (36.8)	23 (30.3)	11 (30.6)	.751
Cigarettes per day (CPD)* Median [min/max]	15.0 [5/40]	13.0 [5/40]	16.5 6/35]	15.0 [5/40]	14.5 [7/25]	.266
Baseline TNE nmol/mg creatinine Median [min/max]	62.7 [5.4/236.6]	59.6 [8.9/126.5]	62.5 [6.6/182.5]	78.0 [9.9/236.6]	60.1 [5.4/148.1]	.071
Usual brand cigarette type**
Menthol	148 (56.3)	43 (56.6)	38 (50.0)	44 (58.7)	23 (63.9)	.524
Tobacco	115 (43.7)	33 (43.4)	38 (50.0)	31 (41.3)	13 (36.1)
FTND total score Median [min/max]	3.0 [0/8]	3.0 [0/7]	4.0 [1/8]	3.0 [0/7]	3.0 [1/7]	.118
CES-D (depression) Median [min/max]	6.0 [0/31]	7.0 [0/29]	6.5 [0/31]	6.0 [0/25]	7.0 [0/20]	.729
Study dropouts	56/264 (21.2)	11/76 (14.5)	18/76 (23.7)	23/76 (30.3)	4/36 (11.1)	.041

AD-E = ad libitum e-cigarette; CS-E = complete substitution e-cigarette; CS-NRT = complete substitution nicotine replacement therapies; UMN = University of Minnesota; OSU/APP = The Ohio State Universality/Appalachia; TNE = total nicotine equivalents; FTND = Fagerström Test for Nicotine Dependence; CES-D = Center for Epidemiological Studies—Depression.

⁺The p-values were derived from the chi-square test or the Kruskal–Wallis test.

⁺⁺This p-value compares Whites and Blacks only.

*CPD from the Tobacco History Questionnaire.

**One person in CS-NRT said “no usual type.”

Product Use

Cigarettes per Day

Average self-reported CPD by groups and by weeks is shown in Figure 1. Significant group (p < .001), week (changes over time; p < .001), and group by week interaction (p < .001) effects were observed using an unadjusted model. Because of the significant interaction effect, the four groups were analyzed separately for trends over study visits using the adjusted model and baseline as the reference point (see Table 2). For the AD-E, CS-E, and CS-NRT groups, significant reductions in CPD was observed for each intervention week (ps < .001). For the UB group, slight but significant reductions were observed for CPD only at week 1 (p = .013) and week 2 (p = .03).

Figure 1.

Median cigarettes per day (CPD), median e-cigarette puffs, and geometric mean carbon monoxide (CO) by study groups over time.

Table 2.

Extent of Reduction in Cigarettes and Biomarker Levels by Week with Baseline as a Reference Point

Visit		AD-E		CS-E		CS-NRT		UB
Cigarettes/day (n, estimated ratio, 95% CI)
Baseline	n = 76	1.00	n = 76	1.00	n = 76	1.00	n = 36	1.00
1	n = 70	0.81 (0.75, 0.87)**	n = 65	0.30 (0.25, 0.38)**	n = 69	0.45 (0.37, 0.55)**	n = 35	0.94 (0.89, 0.99)+
2	n = 71	0.80 (0.73, 0.88)**	n = 64	0.23 (0.16, 0.32)**	n = 62	0.35 (0.26, 0.46)**	n = 33	0.94 (0.89, 0.99)+
4	n = 69	0.76 (0.69, 0.84)**	n = 61	0.25 (0.17, 0.36)**	n = 60	0.35 (0.25, 0.48)**	n = 33	0.96 (0.90, 1.04)
6	n = 65	0.78 (0.70, 0.85)**	n = 60	0.21 (0.14, 0.33)**	n = 54	0.28 (0.20, 0.38)**	n = 33	0.92 (0.84, 1.00)
8	n = 65	0.73 (0.65, 0.82)**	n = 58	0.25 (0.17, 0.37)**	n = 53	0.29 (0.21, 0.39)**	n = 32	0.93 (0.84, 1.02)
Carbon monoxide (estimated ratio, 95% CI)
Baseline	n = 75	1.00	n = 76	1.00	n = 76	1.00	n = 36	1.00
1	n = 70	0.74 (0.64, 0.86)**	n = 65	0.40 (0.33, 0.49)**	n = 69	0.53 (0.44, 0.64)**	n = 35	1.01 (0.92, 1.11)
2	n = 71	0.79 (0.70, 0.90)**	n = 64	0.40 (0.32, 0.48)**	n = 62	0.53 (0.43, 0.65)**	n = 33	0.95 (0.78, 1.15)
4	n = 69	0.79 (0.68, 0.91)*	n = 61	0.39 (0.31, 0.47)**	n = 60	0.52 (0.43, 0.63)**	n = 33	0.86 (0.71, 1.03)
6	n = 65	0.81 (0.72, 0.93)*	n = 60	0.41 (0.33, 0.52)**	n = 54	0.51 (0.42, 0.62)**	n = 33	0.79 (0.64, 0.96)+
8	n = 65	0.76 (0.65, 0.90)*	n = 58	0.43 (0.34, 0.54)**	n = 62	0.55 (0.44, 0.69)**	n = 32	0.87 (0.71, 1.07)
Total nicotine equivalents (estimated mean difference, 95% CI)
Baseline	n = 76	0.00	n = 76	0.00	n = 76	0.00	n = 36	0.00
4	n = 68	−5.19 (−11.30, 0.93)	n = 59	−11.23 (−18.93, −3.54)*	n = 59	4.91 (−6.11, 15.92)	n = 33	1.33 (−11.06, 13.72)
8	n = 65	−6.00 (−11.34, 0.67)+	n = 58	−9.13 (−16.89, −1.36)+	n = 53	−18.61 (−29.93, −7.29)*	n = 32	−5.69 (−16.76, 5.39)
Total NNAL (estimated ratio, 95% CI)
Baseline	n = 71	1.00	n = 76	1.00	n = 72	1.00	n = 35	1.00
4	n = 67	0.87 (0.76, 0.99)+	n = 59	0.44 (0.36, 0.53)**	n = 59	0.59 (0.46, 0.75)**	n = 32	0.89 (0.71, 1.13)
8	n = 65	0.75 (0.63, 0.89)*	n = 57	0.47 (0.35, 0.64)**	n = 53	0.48 (0.37, 0.63)**	n = 31	0.81 (0.66, 1.00)+
PheT (estimated ratio, 95% CI)
Baseline	n = 66	1.00	n = 62	1.00	n = 57	1.00	n = 31	1.00
4	n = 67	1.06 (0.92, 1.21)	n = 59	0.80 (0.69, 0.93)*	n = 57	0.90 (0.73, 1.10)	n = 33	1.21 (0.95, 1.56)
8	n = 65	1.08 (0.93, 1.25)	n = 56	0.79 (0.69, 0.91)*	n = 53	0.95 (0.77, 1.18)	n = 32	1.34 (1.03, 1.75)+
CEMA (estimated ratio, 95% CI)
Baseline	n = 66	1.00	n = 63	1.00	n = 57	1.00	n = 31	1.00
4	n = 68	0.88 (0.69, 1.12)	n = 59	0.30 (0.21, 0.44)**	n = 59	0.48 (0.34, 0.69)**	n = 33	0.96 (0.66, 1.40)
8	n = 65	0.80 (0.62, 1.03)	n = 58	0.34 (0.23, 0.51)**	n = 53	0.46 (0.32, 0.67)**	n = 32	0.95 (0.67, 1.34)
3-HPMA (estimated ratio, 95% CI)
Baseline	n = 66	1.00	n = 63	1.00	n = 57	1.00	n = 31	1.00
4	n = 68	0.84 (0.71, 1.01)	n = 59	0.44 (0.34, 0.58)**	n = 59	0.66 (0.51, 0.86)*	n = 33	0.94 (0.67, 1.32)
8	n = 65	0.78 (0.62, 0.97)+	n = 59	0.53 (0.41, 0.69)**	n = 53	0.62 (0.48, 0.80)**	n = 32	1.12 (0.82, 1.53)
HMPMA (estimated ratio, 95% CI)
Baseline	n = 69	1.00	n = 63	1.00	n = 59	1.00	n = 31	1.00
4	n = 68	0.91 (0.75, 1.10)	n = 59	0.40 (0.31, 0.53)**	n = 59	0.65 (0.52, 0.82)**	n = 33	0.93 (0.69, 1.27)
8	n = 65	0.86 (0.70, 1.05)	n = 58	0.53 (0.40, 0.69)**	n = 53	0.64 (0.50, 0.83)**	n = 32	1.15 (0.85, 1.56)
AAMA (estimated ratio, 95% CI)
Baseline	n = 68	1.00	n = 61	1.00	n = 58	1.00	n = 30	1.00
4	n = 65	0.78 (0.66, 0.92)*	n = 59	0.53 (0.42, 0.66)**	n = 58	0.68 (0.53, 0.86)*	n = 32	0.85 (0.64, 1.13)
8	n = 64	0.73 (0.61, 0.86)**	n = 58	0.68 (0.54, 0.86)*	n = 51	0.61 (0.47, 0.78)**	n = 32	1.01 (0.76, 1.34)

Study group abbreviations. AD-E = ad libitum e-cigarette; CS-E = complete substitution e-cigarette; CS-NRT = complete substitution nicotine replacement therapies; UB = usual brand. Biomarker abbreviations. Total NNAL = 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol and its glucuronides, biomarker for NNK exposure; PheT = phenanthrene tetraol, indicator of exposure to polycyclic aromatic hydrocarbons; CEMA = 2-cyanoethylmercapturic acid, biomarker for acrylonitrile; 3-HPMA = 3-hydroxypropylmercapturic acid, biomarker for acrolein; HMPMA = 3-hydroxy-1-methylpropylmercapturic acid, biomarker for crotonaldehyde/methylvinyl ketone; AAMA = N-acetyl-S-(carbamoylethyl)-l-cysteine, biomarker for acrylamide; 2-HPMA = 2-hydroxypropylmercapturic acid, biomarker for propylene oxide. Biomarkers are expressed per milligram creatinine.

Each study group was analyzed separately due to a significant interaction between group and week.

Cigarettes per day were analyzed by the GEE model and the biomarkers were analyzed using a linear mixed model. All results in the table represent the adjusted models; however, the significant covariates in the final models are not reported in the table.

Statistical analysis shows significant differences by week with baseline as a reference: **p < .001, *p < .01, +p < .05.

Significant differences were observed across the four groups when examining median changes in CPD at weeks 4 and 8 minus baseline (ps < .001; see Supplementary Table 3). Pairwise comparisons for the three planned comparisons showed significantly greater CPD reductions at both weeks for CS-E versus AD-E (ps < .001, respectively) and AD-E versus UB (ps < .001, respectively) but no differences between CS-E versus CS-NRT (p = .122 at week 4, p = .185 at week 8).

E-Cigarettes and NRT

The flavor choices for e-cigarettes (n, %) were the following: AD-E—tobacco (13, 17.3%), menthol (22, 29.3%), mint (18, 24.0%), berry (21, 28.0%), other (1, 1.3%, own brand) and for CS-E—tobacco (16, 21.1%), menthol (18, 23.7%), mint (16, 21.1%), and berry (26, 34.2%). Average estimated puffs per day between AD-E and CS-E are shown in Figure 1. Significantly lower levels of e-cigarette use were reported in the AD-E group for each intervention week compared to the CS-E group (ps < .001). For the entire study period, AD-E demonstrated fewer number of puffs per day compared to the CS-E group (ratio = 0.42, 95% CI 0.27 to 0.66, p < .001). The median (range) number of nicotine gum or lozenge used varied from a minimum of 3.9 (0–44) to a maximum of 4.4 (0–21) for all visits.

Smoke-Free Days

Significant overall difference (p < .001) was found for the median (range) percent of smoke-free days: AD-E (0% [0–96]), CS-E (59.6% [0–100]), CS-NRT (24.3% [0–100]), UB (0% [0–68]), but only AD-E versus CS-E reached significance (p < .001). A total of 193 visits had self-reported smoke-free days since the previous visit with 7 in AD-E, 104 in CS-E, 80 in NRT, and 2 in UB; in 163 out of the 193 visits (84.5%), abstinence was verified based on CO of not more than 6 ppm. Significantly more participants experienced CO-verified end-of-treatment 7-day point prevalence abstinence in the CS-E (25/76, 32.9%) versus CS-NRT (13/76, 17.1%, p = .039).

Biomarkers of Exposure

Group, Week, and Group by Week Effects

See Supplementary Table 2 for mean and median values for biomarkers. For CO and TNE using the unadjusted model, significant group (p < .001 and p = .028, respectively), week (ps < .001), and group by week interaction effects (p < .001 and p = .007, respectively) were observed. For total NNAL, CEMA, 3-HPMA, HMPMA, and AAMA, there were no significant group effects (ps > .05), but significant week (ps ≤ .001) and group by week interaction effects (ps < .001, except AAMA with p = .014). For 2-HPMA, no significant effects were observed.

Week Effects Within Groups

Table 2 shows significant week effects relative to baseline within each group for biomarkers where significant group by week interactions were observed. For AD-E, each intervention visit showed significantly lower levels of CO compared to baseline (p = .002 to p < .001; see Figure 1). Significant reductions compared to baseline were also observed for other biomarkers, specifically weeks 4 and 8 for total NNAL (p = .041 and p = .002, respectively) and AAMA (p = .005 and p < .001, respectively), and week 8 for TNE (p = .028) and 3-HPMA (p = .029). No significant reductions were observed for PheT, CEMA, and HMPMA. For CS-E and CS-NRT, significant reductions were observed for all biomarkers at each intervention week (p = .022 to p < .001 and p = .002 to p < .001, respectively), with the exception that no significant reduction for CS-NRT was observed for TNE at week 4 and PheT at weeks 4 and 8 (ps > .05, respectively). For UB, significant reductions were only observed at week 6 for CO (p = .018) and week 8 for total NNAL (p = .046) and an increase for PheT (p = .03).

Group Effects for Changes From Baseline

In Supplementary Table 3 for TNE, the group effects for medians for weeks 4 and 8 minus baseline are displayed. For other biomarkers, the group effects for the GMR for the weeks 4 and 8 over baseline are presented. Significant group differences were observed for weeks 4 and 8 for CO (ps < 0.001), total NNAL (p < .001 and p = .006, respectively), PheT (p = .005 and p = .003, respectively), CEMA (ps < .001), 3-HPMA (p < .001 and p = .003, respectively), and HMPMA (ps < .001). Significant group differences were only observed for week 4 for TNE (p = .013) and 2-HPMA (p = .022) and no differences at week 4 or 8 for AAMA (p > .05).

Pairwise Comparisons Between Groups for Changes From Baseline

Significant differences in the magnitude of the change in biomarker levels from baseline were observed for some of the biomarkers at weeks 4 and 8, predominantly between AD-E versus CS-E (Supplementary Table 3). At week 4, GMRs (week 4/baseline) of CO (p < .001), total NNAL (p < .001), PheT (p = .007), CEMA (p < .001), 3-HPMA (p < .001), and HMPMA (p < .001) were significantly lower for CS-E versus AD-E. Significant median differences were observed at week 4 for CS-NRT versus CS-E for TNE, with an increase in TNE in the CS-NRT group (p = .005; perhaps due to higher smoking rates in CS-NRT); no differences in GMRs were observed between CS-E versus CS-NRT for all other biomarkers (p > .017) except HMPMA for which CS-E had lower GMR than CS-NRT (p < .003). No significant week 4 differences were observed for AD-E versus UB.

At week 8, GMRs (week 8/baseline) for CO (p < .001), PheT (p = .008), CEMA (p = .001), and HMPMA (p = .006) were significantly lower for CS-E versus AD-E. No significant differences in GMRs were observed between CS-E versus CS-NRT and AD-E versus UB for any of the biomarkers at week 8.

Adverse Events

The most frequent adverse events (occurred in three or more people) that were considered definitely related, possibly related, or unknown relationship to e-cigarettes were cough (n = 15), headache, non-migraine (n = 8), sore/itchy burning throat (n = 8), nausea (n = 5), and dizziness/lightheadedness (n = 3). For NRT, the adverse events included nausea (n = 11), vomiting (n = 3), diarrhea (n = 3), sore/itchy/burning throat (n = 3), indigestion (n = 3), and dizziness/lightheadedness (n = 3).

Selection of Products by UB Group

Following completion of the 8-week trial, 11/32 completers in the UB group chose to use e-cigarettes and 2 to use NRT for the following 8 weeks. None achieved 7-day point prevalence abstinence at the end of the 8 weeks of follow-up.

Discussion

Smokers who were instructed and reinforced for complete substitution of cigarettes with e-cigarettes experienced significantly greater reductions in cigarette smoking and biomarkers of exposure than those who were instructed to use e-cigarettes ad libitum. Furthermore, biomarkers of exposure for complete substitution of cigarettes with NRT were not significantly different than complete substitution with e-cigarettes. Another relevant observation was that 33% of participants achieved CO-verified 7-day smoking abstinence in the CS-E group at the end of treatment versus 17% in the CS-NRT group. These results suggest that complete substitution is difficult to achieve with both the e-cigarette tested in this study and NRT, but the e-cigarette may be more effective for abstinence and higher compliance than the use of NRT.

Other studies have observed similar trends in biomarker reduction when comparing complete with partial substitution or dual use. In a national cross-sectional survey, substantially lower exposure biomarker levels were observed for several harmful constituents among exclusive e-cigarette users compared to exclusive smokers.² For example, exclusive e-cigarette users experienced lower total NNAL (98% lower), cadmium (30% lower), biomarkers of polycyclic aromatic hydrocarbons (34%–62% lower), and some biomarkers for VOCs (59%–97% lower). On the other hand, dual users experienced significantly higher concentrations on most biomarkers compared to exclusive cigarette users. In a study that recruited users of different products,⁶ long-term exclusive users of NRT or e-cigarettes had substantially lower exposures to tobacco-specific nitrosamines and VOCs than exclusive cigarette smokers, but dual use of these products with cigarettes showed minimal differences in exposure levels compared to exclusive cigarette smokers. On some measures, exclusive e-cigarette users showed lower levels of carcinogen or toxicant exposures than exclusive NRT users.

Several types of switching studies have shown comparable results. In a short-term switching study involving 5 days in a confined unit,³ smokers allowed to only use e-cigarettes experienced reduced exposure to polycyclic aromatic hydrocarbons (pyrene), tobacco-specific nitrosamines (NNAL and NNN), and VOCs, with levels similar to those who assigned to being abstinent from smoking. The dual user group, who were instructed to reduce their cigarette smoking by half, experienced a quarter to a third the reduction of the complete substitution group. In a stop smoking study conducted in the United Kingdom, biomarker levels (CO and 3-HMPMA) decreased substantially among smokers who switched completely to e-cigarette use and were reduced but higher among dual users.⁴ In a 12-week open blind trial with a 1-week confinement for a subgroup of the participants, smokers who were not interested in quitting were randomized to e-cigarettes or continued smoking.⁵ Results showed greatest reductions ranging from over a third to over a half observed at week 4 for total NNAL, SPMA (biomarker for benzene), 3-HPMA, and CO among those assigned to e-cigarettes. Greater reductions were observed among compliant e-cigarette users compared to the total samples of e-cigarette users and among the subgroup of participants assigned e-cigarettes during confinement. Both the observational studies and clinical trials confirm that switching completely to e-cigarettes or NRT is important to see substantial reductions in exposure to harmful constituents, which are not always observed with dual use.

With regard to e-cigarettes as a means to achieve complete switching and thereby smoking abstinence among nontreatment seeking smokers, the studies on the use of e-cigarettes as a cessation tool might have relevance. The NASEM¹ concluded that there is “limited evidence that e-cigarettes may be effective aids to promote smoking cessation.” The NASEM report also stated that there was “insufficient evidence from randomized controlled trials about the effectiveness of e-cigarettes as cessation aids compared to no treatment or US FDA-approved smoking cessation treatments.” Whereas an early clinical trial showed no differences in abstinence rates among smokers, motivated to quit, who were randomly assigned to first-generation e-cigarettes or nicotine patch,¹⁶ a large cross-sectional study showed smokers who used e-cigarettes during a quit attempt were 1.6 times more likely to be abstinent compared to those who used over-the-counter NRT or had not used smoking cessation aids.¹⁷ In a recent large clinical trial in which smokers were randomized to nicotine replacement products (including combination NRT) versus e-cigarettes to quit smoking and received behavioral support, those assigned to e-cigarettes were almost twice as likely to experience 1-year sustained abstinence.¹⁸ The current study similarly demonstrated that e-cigarettes resulted in significant higher end-of-treatment 7-day point prevalence cigarette-free days (complete switching) compared to NRT when nontreatment seeking participants were instructed to avoid cigarettes and provided brief counseling and monetarily incentivized to completely switch. Clearly larger clinical trials need to be conducted that examine how to facilitate complete switching (eg, messaging, psychosocial, and policy interventions) in smokers uninterested in total abstinence from nicotine products.

Limitations for this study included small sample sizes, lack of compliance to complete substitution (which would affect the biomarkers), and use of incentives in the CS-E and CS-NRT conditions (which might limit the generalizability of this study in a real-world environment). Additionally, dropout rates were higher in the complete substitution groups. Given that these are study participants, extrapolation to the general population, as with all clinical trials, is limited.

In summary, this study showed that instructions, brief counseling, and monetary incentives for complete switching to e-cigarettes would lead to significant reductions in the number of cigarettes smoked and biomarkers of exposure, at levels similar to NRT. On the other hand, instructions for ad libitum use might lead to reduced levels of exposure for some biomarkers, but at levels not significantly different than UB cigarettes. Labeling e-cigarettes with appropriate instructions for use (eg, exclusive use of the e-cigarette) and exploring approaches that increase motivation and facilitate complete substitution are an important area of future research. Furthermore, the results showed that e-cigarettes are just as, if not more, effective as NRT in leading to 7-day abstinence. Evaluating e-cigarettes as a cessation tool should be pursued.

Funding

The research reported in this publication was supported by grants U19CA157345 from the National Cancer Institute (DKH/PS), UL1 TR000062 and UL1 TR002494 from the National Center for Advancing Translational Science of the National Institutes of Health, and T32 DA007097 from the National Institute of Drug Abuse (EM). The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding agencies.

Declaration of Interests

RJC is a member of the FDA Tobacco Products Scientific Advisory Committee. PGS serves or has served as an expert witness in tobacco company litigation on behalf of plaintiffs.