Efficacy and Safety of Varenicline for Adolescent Smoking Cessation: A Randomized Clinical Trial

Key Points

Question  Is varenicline tartrate, when added to brief cessation counseling, efficacious and safe for smoking cessation in adolescents?

Findings  In this 12-week randomized clinical trial with 157 treatment-seeking adolescent cigarette smokers, abstinence rates at the end of treatment (primary outcome) did not differ between groups, but examination of secondary findings revealed that varenicline participants achieved abstinence earlier and had higher rates of overall abstinence during treatment and at posttreatment follow-up compared with placebo participants. Treatment-emergent adverse events did not differ between groups.

Meaning  Compared with placebo, varenicline was well tolerated but did not support improved end-of-treatment abstinence for adolescent smokers.

Importance  Cigarette smoking is the leading cause of preventable morbidity and mortality in the United States and worldwide, and most tobacco users begin smoking in adolescence. Although advances have yielded efficacious pharmacotherapies to complement smoking cessation counseling in adults, far less progress has been made in addressing tobacco use in adolescence.

Objective  To evaluate the efficacy and safety of varenicline tartrate for smoking cessation in adolescents and young adults.

Design, Setting, and Participants  This 2-group randomized, placebo-controlled, double-blind intention-to-treat clinical trial enrolled a volunteer sample of treatment-seeking adolescent and young adult cigarette smokers (n = 157) aged 14 to 21 years at an outpatient clinical site in Charleston, South Carolina, from August 15, 2012, to October 20, 2017. Follow-up was completed on January 25, 2018. Data were analyzed from March 19, 2018, to August 11, 2018, with further revisions completed April 10, 2019.

Interventions  Participants were randomized in a 1:1 ratio to a 12-week course of varenicline (n = 77) or placebo (n = 80). All participants received weekly smoking cessation counseling.

Main Outcomes and Measures  The preselected primary efficacy outcome was urine cotinine level–confirmed 7-day abstinence at the end of treatment. Secondary efficacy outcomes included weekly abstinence throughout active treatment, abstinence at posttreatment follow-up visits, and time to first 7-day abstinence. The primary safety outcome was the frequency of treatment-emergent adverse events.

Results  A total of 157 participants were enrolled (94 male [59.9%]; mean [SD] age, 19.1 [1.5] years). The varenicline and placebo groups did not differ in the primary outcome of cotinine-confirmed self-reported 7-day abstinence at the end of treatment (varenicline group, 4 of 45 [8.9%]; placebo group, 4 of 45 [8.9%]; risk ratio [RR], 0.97; 95% CI, 0.29-2.99; P = .96). However, among secondary outcomes, the varenicline group achieved self-reported earlier abstinence of at least 7 days (hazard ratio, 1.91; 95% CI, 1.12-3.27) and demonstrated higher rates of self-reported weekly abstinence during the full course of treatment (RR, 1.81; 95% CI, 1.09-2.99; P = .02) and at posttreatment follow-up (RR, 1.82; 95% CI, 1.01-3.28; P = .02). Study medication was generally well tolerated, and treatment-emergent adverse events did not differ between groups (any adverse events, 55 [71.4%] in the varenicline group vs 60 [75.0%] in the placebo group; RR, 0.95; 95% CI, 0.79-1.15; P = .61).

Conclusions and Relevance  When added to weekly cessation counseling for adolescent cigarette smokers, varenicline, compared with placebo, was well tolerated but did not improve end-of-treatment abstinence. However, varenicline may hasten abstinence and yield improvements in posttreatment abstinence outcomes.

Trial Registration  ClinicalTrials.gov identifier: NCT01509547

Tobacco use is the leading cause of preventable death in the United States and the world.¹ Although rates of smoking initiation in adolescence have declined in the last decade, most adult cigarette smokers began smoking before 21 years of age, and almost all adolescents who smoke regularly will continue smoking well into adulthood, leading to a life expectancy at least 10 years shorter than that of nonsmokers.^1-4 Almost two-thirds of adolescent smokers are interested in quitting, but only 4% to 6% of unassisted quit attempts are successful.^3-6

Few controlled studies have evaluated adolescent smoking cessation treatments, and almost all have focused on psychosocial interventions with generally discouraging results. A meta-analysis of 48 studies⁷ showed a mean quit rate of 9.1% compared with 6.2% among control groups. In the interest of enhancing these very modest quit rates, a handful of randomized trials have evaluated pharmacotherapies, including nicotine replacement therapy and bupropion hydrochloride, to complement psychosocial treatment.^8-17 Amid mixed findings, some evidence suggests modest efficacy during treatment but low rates of abstinence at posttreatment follow-up.

The introduction of varenicline tartrate, an α₄β₂ nicotinic acetylcholine receptor partial agonist, represented an advance in adult smoking cessation pharmacotherapy, demonstrating superiority to placebo (end-of-treatment odds ratio [OR] range, 3.6-5.9), bupropion (OR range, 1.7-2.2), and nicotine patch (OR, 1.7).^18-23 Postmarketing concerns regarding varenicline’s neuropsychiatric safety were addressed in a large trial of adults with and without stably treated psychiatric disorders, revealing no difference in neuropsychiatric adverse events compared with bupropion, nicotine patch, or placebo.²³

In contrast with the wealth of randomized clinical trials evaluating varenicline in adult smokers, little work has focused on varenicline in adolescent smokers. An initial 2-week adolescent laboratory study revealed a similar pharmacokinetic profile as in adults, with no serious adverse events or discontinuations due to adverse events,²⁴ and a subsequent proof-of-concept pilot adolescent smoking cessation trial²⁵ further supported the feasibility of this line of research. Encouraged by these findings, we sought to evaluate the efficacy and safety of varenicline for adolescent smoking cessation via a randomized clinical trial, hypothesizing that, when added to weekly cessation counseling, varenicline therapy would be well-tolerated and would yield higher rates of abstinence than placebo.

Treatment-seeking adolescent smokers were randomized in 1:1 parallel group allocation to receive a double-blind 12-week course of varenicline or placebo added to weekly cessation counseling. Self-reported smoking data were collected via daily diaries and timeline follow-back methods.²⁶ For biomarkers of smoking, breath carbon monoxide (CO) level was measured at all visits, and urine cotinine level was measured at the end-of-treatment visit. The study was conducted within an approved US Food and Drug Administration Investigational New Drug application. The institutional review board of the University of South Carolina, Charleston, approved the study protocol (Supplement 1), which was overseen by an independent National Institute on Drug Abuse–appointed data and safety monitoring board. If the participant was younger than 18 years, a parent or guardian provided written informed consent and the adolescent provided assent. Individuals 18 years or older provided informed consent. We followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline.

The trial was implemented in a university outpatient clinic in Charleston, South Carolina. The trial enrolled adolescent and young adult cigarette smokers aged 14 to 21 years who smoked daily for at least 6 months, had at least 1 failed prior quit attempt, and were seeking to quit. Female participants agreed to use reliable birth control methods. Individuals with a history of DSM-IV-TR mood or psychotic disorders, suicidality, homicidality, or significant hostility or aggression; with current DSM-IV-TR substance dependence other than nicotine, unstable medical disorder, pregnancy, breastfeeding, or use of medications with smoking cessation efficacy; or with known hypersensitivity to varenicline use were excluded. Recruitment was conducted from August 1, 2012, through October 31, 2017, primarily through media advertisements. Participant enrollment, occurring within this active recruitment period, spanned August 15, 2012, through October 20, 2017. Interested individuals were prescreened by telephone, and those meeting general entry criteria were scheduled for consent and screening procedures in the clinic.

At the baseline visit, comprehensive psychiatric and substance use diagnostic assessments, medical history, physical examination, and laboratory testing (urine pregnancy and drug tests) were performed to evaluate suitability for inclusion based on the aforementioned criteria.^27,28 We used the timeline follow-back method to assess past 30-day self-reported tobacco and other substance use.²⁶

Participants were seen weekly during the 12-week treatment, and follow-up posttreatment assessments were conducted at weeks 18 and 26. During weekly treatment visits, brief smoking cessation counseling, medication management, and adverse event assessment were provided by a study physician or physician assistant; participants completed self-assessments; and breath CO level was assessed. Follow-up was completed on January 25, 2018.

Participants were randomized to a double-blind 12-week course of orally administered varenicline tartrate or placebo. Randomization was on a 1:1 ratio, with stratification by the prognostic covariates of age (14-17 vs 18-21 years) and baseline smoking level (<12 vs ≥12 cigarettes per day [CPD], based on median baseline smoking in a previous randomized adolescent smoking cessation trial).¹⁶ Pfizer, Inc supplied varenicline tartrate 0.5-mg and 1.0-mg tablets and matched placebo tablets to the university investigational pharmacy. All tablets were packaged in blister packs with individual labels for time and date of each dose and were dispensed in weekly supplies at study visits; medication diaries and pill counts were used to assess adherence. Consistent with recommendations from an adolescent varenicline pharmacokinetic study,¹⁴ participants weighing more than 55 kg received varenicline tartrate or placebo in a dosage of 0.5 mg once daily for 3 days, titrated to 0.5 mg twice daily for 4 days, and titrated to 1.0 mg twice daily thereafter. Participants weighing 55 kg or less received varenicline tartrate or placebo in a dosage of 0.5 mg once daily for 7 days and titrated to 0.5 mg twice daily thereafter.

During the assessment visit, participants were given adolescent-targeted smoking cessation brochures and briefly counseled on cessation strategies. Participants were instructed to set a quit date 1 week after therapy initiation. At weekly visits, participants were provided with brief individual skills-based cessation counseling that paralleled psychosocial interventions in similarly structured adolescent and adult cessation pharmacotherapy studies.^14,18-20 If participants did not succeed with the initial quit date, they were advised to select another quit date.

The primary efficacy outcome was urine cotinine level–confirmed (≤50 ng/mL) 7-day self-reported abstinence at the week 12 end-of-treatment visit. Secondary efficacy measures included weekly self-reported and breath CO–confirmed (≤8 ppm) abstinence (since the prior weekly visit) during study treatment (weeks 1-12) and 7-day abstinence at follow-up (weeks 18 and 26). In addition, time from therapy initiation to the first period of at least 7 days of sustained abstinence was measured. Timeline follow-back methods allowed for reliable collection of smoking data even during periods when study visits were missed.²⁶

The primary safety outcome was frequency of treatment-emergent adverse events. General and neuropsychiatric adverse events were systematically assessed by a study physician or physician assistant at all visits.

Data were analyzed from March 19, 2018, to August 11, 2018, with further revisions completed April 10, 2019. All statistical analysis was conducted using SAS, version 9.4 (SAS Institute Inc), and 2-sided P < .05 indicated significance. The study was powered to detect an end-of-treatment 21% abstinence rate in the varenicline group compared with 6% in the placebo group in an intention-to-treat sample, based on estimates from previous trials.^7,14,16,25 With a type I error rate of 5% and power (1 − β) of 80%, the study required randomization of 83 participants per treatment group (166 total randomized) using a 2-sided Pearson χ² test statistic. No interim efficacy or futility analyses were specified or conducted.

Standard descriptive statistics were used to summarize characteristics of the study sample. The primary hypothesis was that participants in the varenicline group (varenicline participants) would have a higher probability than those in the placebo participants of end-of-treatment cotinine-confirmed 7-day abstinence. A logistic regression model with a sandwich variance estimate was used to assess this primary efficacy outcome.²⁹ To assess the potential effect of missing outcome data on parameter estimates, sensitivity analyses were completed (1) with all available data, (2) with missing outcome data imputed as smoking (worst-case imputation), and (3) using methods of multiple imputation.³⁰ Imputation of missing outcome data as positive for smoking is often used in cessation trials because it does not necessitate the missing-at-random assumption and allows for a correlation between missing status and smoking status.³¹ However, when the data are missing at random, this method may add significant bias to the parameter estimates. Further, multiple imputation was implemented using fully conditional specification with a logistic regression approach. Owing to the high rate of missing data, 100 imputation data files were created to assure reasonable relative efficiency.³² Parameter estimates were recovered from logistic regression models noted above. Imputation of cotinine-confirmed smoking status at the end of treatment was based on model variables (randomized treatment assignment, baseline-reported CPD, and participant age). No independent variables from the models were missing in the data file. Imputed values for smoking status were included in a logistic regression model and combined to generate parameter estimates and 95% CIs.³³ Risk ratios (RRs) and asymptotic 95% CIs were computed for efficacy estimates.

Repeated-measures logistic regression models using the methods of generalized estimating equations³⁴ were constructed to assess self-reported and CO-confirmed abstinence at weekly treatment visits (weeks 1-12) and 7-day abstinence at posttreatment follow-up visits (weeks 18 and 26). Working correlation structures were independently compared, and the final model structure was chosen using the quasi-likelihood under the independence model criterion³⁵ and direct comparison of the robust and model-based variance-covariance matrices; specifically autoregressive, compound symmetric, and banded structures were compared. In self-reported and CO-confirmed use data, a first-order autoregressive structure provided a better model fit than a compound symmetric or a banded structure. Risk ratios and asymptotic 95% CIs were computed for all efficacy estimates. Sensitivity analysis was completed as described above. For models with repeated outcome measures on participants, fully conditional specification methods were implemented using a logistic structure, and additional variables to account for visit number and clustering of data on participants were added to the imputation process at each visit.³⁶ Time to first 7-day abstinence was assessed using log-rank test statistics and Cox proportional hazards regression models. All models were adjusted for baseline-reported CPD, participant age, and study visit when appropriate.

Treatment-emergent adverse event rates were compiled using all reported events after randomization, regardless of attributed medication relatedness. Negative binomial regression models were used to compare event rates (counts per participant) as well as to compare rates of participants experiencing events (any vs none). All adverse event rate comparisons used 2-sided statistical tests.

Participant demographics and baseline characteristics are presented in Table 1, and progression through trial procedures is summarized in Figure 1. Of 206 screened individuals, 157 were randomized (94 male [59.9%] and 63 female [40.1%]; mean [SD] age, 19.1 [1.5] years), 77 to varenicline and 80 to placebo. This randomized total fell slightly short of the goal of 166 participants. Although enrollment progressed at the goal rate for most of the rolling recruitment period, amid reduced rates of adolescent daily smoking it progressively declined toward the end, despite an extension of the enrollment period and redoubled recruitment efforts. This situation ultimately led the investigative team to conclude enrollment at 157 randomized participants. Of these, 146 (93.0%) attended at least 1 postrandomization visit (68 varenicline and 78 placebo participants). The end-of-treatment (week 12) visit was attended by 90 participants (57.3%) (45 varenicline and 45 placebo group participants). Participants who attended the week 12 visit, compared with those who did not, were not significantly different on any baseline measures (Table 1). Week 26 follow-up data were collected on 83 participants (52.9%). Enrolled participants were predominantly white (120 [76.4%]). At study entry, participants smoked a mean (SD) of 11.5 (6.8) CPD and had been smoking regularly for more than 4 years.

Cotinine-confirmed 7-day abstinence at the end of treatment was not statistically significantly different between randomized groups (varenicline group, 4 of 45 [8.9%]; placebo group, 4 of 45 [8.9%]; adjusted RR, 0.97; 95% CI, 0.29-2.99; P = .96). For secondary efficacy outcomes, of a possible 1884 weekly treatment visits for the 157 randomized participants, 1281 (68.0%) had associated self-reported smoking and CO measures available; 609 of 924 (65.9%) were available in the varenicline group and 672 of 960 (70.0%) in the placebo group. Abstinence rates are presented in Table 2 (with available data and with imputed missing data) and illustrated over time in Figure 2 (with available data). In the analyses using available data, compared with the placebo participants, the varenicline participants had nearly twice the probability of any self-reported abstinence during treatment (measured at weekly visits; RR, 1.81; 95% CI, 1.09-2.99; P = .02). Similarly, varenicline participants were more likely than placebo participants to report 7-day abstinence at either of the 2 posttreatment follow-up visits (RR, 1.82; 95% CI, 1.01-3.28; P = .02). Incorporating breath CO confirmation of weekly self-reported abstinence yielded similar patterns of between-group differences in efficacy outcomes during study treatment but attenuated estimates at the posttreatment follow-up visit (Table 2); however, the prevalence of cannabis use among the study sample potentially confounds interpretability of CO findings (93 participants [59.2%] had positive results of urine cannabinoid tests at baseline). When imputation methods were implemented, resulting parameter estimates were reasonably consistent with those found using available data.

Of the 157 randomized participants, 55 (35.0%) reported at least 7 consecutive abstinent days at any time during treatment (varenicline group, 31 [40.3%]; placebo group, 24 [30.0%]; log-rank P = .02). In Cox proportional hazard regression models (adjusted for baseline CPD and age), varenicline participants were nearly twice as likely as placebo participants to report at least 7 days of abstinence during treatment (hazard ratio, 1.91; 95% CI, 1.12-3.27). In those who reported at least 7 consecutive days of abstinence during treatment, the median time to the first 7 or more days of abstinence was 39 days (interquartile range, 26-59 days) in varenicline participants and 59 days (interquartile range, 45-72 days) in placebo participants.

Of the 157 randomized participants, 115 (73.2%) experienced at least 1 adverse event during the study (varenicline group, 55 [71.4%]; placebo group, 60 [75.0%]; RR, 0.95; 95% CI, 0.79-1.15; P = .61). Of 391 total adverse events, 214 were reported in the varenicline group and 177 in the placebo group; most were unrelated to study treatment. Varenicline participants reported a mean (SD) of 2.8 (2.8) adverse events; placebo participants, 2.2 (2.2) adverse events (RR, 1.26; 95% CI, 0.90-1.75; P = .18). The most commonly reported categories were gastrointestinal tract and neuropsychiatric, with gastrointestinal tract events occurring in 53 participants (33.8%) (varenicline group, 29 [37.7%]; placebo group, 24 [30.0%]; P = .31) and neuropsychiatric events occurring in 50 participants (31.8%) (varenicline group, 26 [33.8%]; placebo group, 24 [30.0%]; P = .61). No medication-related serious adverse events occurred in either treatment group.

Treatment adherence, calculated from daily medication diaries and weekly pill counts as doses taken compared with doses dispensed, was 98% in the varenicline group and 96% in the placebo group (P = .24). Similarly, 49 varenicline participants (63.6%) and 51 placebo participants (63.8%) took 100% of dispensed doses.

Findings indicate that varenicline, when added to brief smoking cessation counseling, is adequately tolerated but does not promote end-of-treatment abstinence compared with placebo among adolescent and young adult cigarette smokers. Secondary findings suggest varenicline may hasten abstinence during treatment, allowing for continued reinforcement of abstinence and relapse prevention in the context of a structured cessation counseling program, which may translate into longer-term posttreatment sustained abstinence.³⁷

In light of postmarketing concerns regarding neuropsychiatric adverse events with varenicline, it is encouraging that treatment-emergent adverse events, in general and specific to neuropsychiatric events, did not differ between the varenicline and placebo groups. This finding, however, occurred in the context of a trial with strict exclusion criteria (ie, history of mood or psychotic disorders, suicidality, homicidality, or significant hostility/aggression history) that reduced the likelihood of neuropsychiatric events. How the safety or efficacy findings may translate to adolescents with co-occurring psychiatric disorders, who are disproportionately represented among adolescent smokers, is not clear.³⁸

The use of biomarkers to verify smoking abstinence is typically regarded as a design strength, but in the present study, it may have introduced limitations. During the study’s enrollment period, the US national prevalence of adolescent tobacco use declined while rates of cannabis use increased.³⁹ This finding was reflected in the high rates of cannabis use among enrolled participants (59.2% of participants had a positive finding of a baseline urine cannabinoid test, and 106 [67.5%] reported cannabis or other illicit substance use in the 30 days before study entry). Cannabis smoking increases breath CO level, potentially yielding a false-positive breathalyzer result⁴⁰; in addition, blunts, a common mode of cannabis use, often contain residual tobacco and may yield a positive urine cotinine result.⁴¹ The US national prevalence of adolescent e-cigarette use also increased during the enrollment period³⁹; although participants were advised not to use e-cigarettes during study participation, use may have occurred, potentially affecting urine cotinine values. Given the likelihood that bias in tobacco use self-report would be similar across treatment groups, we have secondarily presented self-report abstinence and smoking outcomes, but note that between-group comparisons incorporating biomarkers generally parallel self-report comparisons.⁴²

Even amid declining rates of cigarette smoking among adolescents in the United States, the clear public health implications of adolescent-onset smoking indicate that efficacious treatments are needed for this age group, especially given that cessation during adolescence would prevent most of the morbidity and mortality incurred with years of tobacco smoking into adulthood. The present findings add to a mixed literature on adolescent smoking cessation interventions, which have generally yielded discouragingly low rates of abstinence in psychosocial and pharmacological treatment trials. Although present findings indicate that abstinence did not differ between the varenicline and placebo groups at the end of treatment, the secondary findings of earlier and sustained posttreatment follow-up response to varenicline suggest that it may provide an advantage in yielding longer-term abstinence among adolescent smokers. Future work should examine potential methods, such as the addition of behavioral incentives,¹⁶ to maximize the magnitude of smoking cessation success with varenicline and other treatments to improve health outcomes for adolescent smokers, a highly vulnerable group in need of improved treatment options.

In this study, among treatment-seeking adolescent and young adult daily cigarette smokers, a 12-week course of varenicline, compared with placebo, when added to weekly cessation counseling, was well-tolerated but did not improve end-of-treatment abstinence. However, secondary findings suggest that varenicline may promote abstinence early in treatment and increase abstinence at posttreatment follow-up.

Accepted for Publication: June 26, 2019.

Corresponding Author: Kevin M. Gray, MD, Department of Psychiatry and Behavioral Sciences, Medical University of South Carolina, 67 President St, Mail Stop 861, Charleston, SC 29425 (graykm@musc.edu).

Published Online: October 14, 2019. doi:10.1001/jamapediatrics.2019.3553

Author Contributions: Dr Gray had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Gray, Baker, Carpenter.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Gray, Baker, McClure, Squeglia.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Baker.

Obtained funding: Gray, Baker, Saladin.

Administrative, technical, or material support: McClure, Tomko, Saladin.

Supervision: Gray, McClure, Saladin.

Conflict of Interest Disclosures: Dr Gray reported consulting for Pfizer, Inc, and receiving grant support from the National Institutes of Health (NIH). Mr Baker reported receiving grant support from the NIH. Dr McClure reported receiving grant support from the NIH. Dr Tomko reported receiving grant support from the NIH during the conduct of the study and outside the submitted work. Dr Squeglia reported receiving grant support from the NIH. Dr Saladin reported receiving grant support from the NIH. Dr Carpenter reported consulting for Pfizer, Inc, during the conduct of the study. No other disclosures were reported.

Funding/Support: This study was supported by grants U01 DA031779 (Dr Gray), K01 DA036739 (Dr McClure), K12 HD055885 (Dr Tomko), and K23 AA025399 (Dr Squeglia) from the NIH. Varenicline and placebo tablets were supplied at no cost by Pfizer, Inc.

Role of the Funder/Sponsor: The sponsor had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, and approval of the manuscript.

Disclaimer: The opinions in this report are those of the authors and do not represent the official position of the US government.

Data Sharing Statement: See Supplement 2.

Additional Contributions: The authors thank the volunteers who participated in the study. Lori Ann Ueberroth, BS, Department of Psychiatry and Behavioral Sciences, Medical University of South Carolina, with her project staff, served as program coordinator. The NIH funding supported the efforts of Ms Ueberroth and her staff.