Varenicline versus nicotine patch with brief advice for smokers with substance use disorders with or without depression: effects on smoking, substance use and depressive symptoms

Abstract

Aims

Varenicline was compared with transdermal nicotine (NRT) for smokers with current substance use disorders (SUD) for effects on 3-month smoking abstinence (primary outcome) and, secondarily, on 3- and 6 month abstinence while adjusting for medication adherence, and on additional smoking and substance use outcomes. Moderation by major depressive disorder history (MDD) and adherence were investigated.

Design

Double-blind double-placebo-controlled randomized design, stratifying by MDD, gender and nicotine dependence, with 3 and 6 months follow-up.

Setting

University offices in Rhode Island, USA.

Participants

Adult smokers (n = 137), in SUD treatment, substance abstinent <12 months (n = 77 varenicline, 60 NRT).

Intervention and comparator

Twelve weeks of varenicline (2 mg/day, after 1-week dose run-up) or NRT (21 mg/day decreasing to 7 mg/day).

Measurements

Primary: point-prevalence smoking abstinence (7-day, confirmed) at 3 months. Secondary: point-prevalence abstinence at 6 months, quantity and frequency of smoking and substance use at 3 and 6 months, and within-treatment abstinence, medication adherence and depressive symptoms. Smoking outcome analyses were repeated controlling for adherence and investigating adherence as a moderator.

Findings

Effects on 3-month abstinence were P < 0.065 without a covariate (Bayes factor 3.35, supporting the effect strongly) and differed significantly when controlling for baseline smoking [varenicline: 13%, NRT: 3%; odds ratio (OR) = 4.81, 95% confidence interval (CI) 1.00, 23.13, P < 0.05]. The threefold difference at 6 months was not significant. Medication effect on abstinence across time was significant (P < 0.05) covarying adherence and baseline smoking (OR = 6.40, 95% CI = 1.00, 40.93). Medication differences in 3-month abstinence occurred among participants with ≥ 77% adherence (P < 0.02). No significant medication effects on heavy drinking, drug use or depressive symptoms were found.

Conclusions

Varenicline appears to improve the chances of achieving at least 3 months of smoking abstinence in smokers with substance use disorders trying to stop, compared with transdermal nicotine patches, the effect being independent of history of depressive disorder.

INTRODUCTION

Smokers with substance use disorders (SUD) are more likely to die from smoking than from alcohol use [1–3] and have limited success quitting smoking early in sobriety [4]. SUD treatment provides an opportunity to intervene with smoking, but pharmacotherapies for smoking cessation have rarely been investigated among people with non-alcohol SUDs and rarely early in recovery. Transdermal nicotine replacement (NRT) with counseling results in only 3.5% abstinence at 12 months on average with recently or long-term sober smokers with alcohol use disorders [5,6]. Smokers with current alcohol use disorders did not benefit from transdermal NRT [7,8].

Low motivation to quit smoking in this population [9–11] correlates significantly with perceiving more barriers to quitting smoking [9]. Expecting to be unable to tolerate the discomforts of tobacco abstinence predicts less tobacco abstinence 3 months later [12], so reducing withdrawal distress may be essential. Another barrier is concern that smoking cessation will make it harder to stay sober [13]. As continued smoking undermines long-term sobriety [14,15], these smokers need corrective information.

Varenicline is more effective for smoking than other medications except combination nicotine replacement in the general population [16,17], even among those who do not complete it [18], and in smokers with low motivation to quit [19]. Varenicline reduces tobacco withdrawal [20,21], smoking pleasure and craving (e.g. [22,23]). Furthermore, in heavy drinkers and in one study with people with alcohol use disorders (e.g. [24,25]), varenicline reduced drinks per day and alcohol craving.

Varenicline increased smoking abstinence compared to placebo in methadone patients [26], but not compared to NRT [27]. However, varenicline has not been tested with smokers with a range of SUDs, nor combined with counseling designed to increase motivation to quit and address perceived barriers [28], and most studies [16] excluded major depressive disorder (MDD). Given frequent MDD in patients in SUD treatment [29], we stratified on MDD history and investigated it as a moderator of outcome.

We hypothesized that 12 weeks of varenicline versus NRT would be more effective in producing (Aim 1) 7-day point-prevalence abstinence at 3 months (primary outcome) and 6 months (secondary outcome) and (Aim 2) across 3 and 6 months when controlling for baseline smoking and degree of medication adherence (a secondary outcome) for smokers in SUD treatment receiving brief advice [30,31] to motivate smoking cessation. We considered 3 months as primary for determining the extent of a treatment effect by the end of treatment, while maintenance of treatment effects over time we considered as secondary as only relevant if initial treatment effects occur.¹ We also examined effects on (Aim 3) quantity and frequency of smoking, (Aim 4) on within-treatment smoking abstinence and depressive symptoms and (Aim 5) on substance use. Finally, we examined (Aim 6) potential moderators (MDD, medication adherence, nicotine dependence and gender) of medication effects on outcome.

METHODS

Participants

Smokers were recruited via advertisements in Providence, RI, from 2009 to 2014. Inclusionary criteria were: (a) SUD diagnosis; (b) enrollment in any out-patient SUD treatment; (c) 10+ cigarettes per day for the past 6 months; and (d) aged 18–75 years. Exclusionary criteria were: (a) evidence of hallucinations or delusions, (b) current smoking cessation treatment, (c) contraindications for either medication (such as pregnancy, uncontrolled hypertension or severe renal impairment), (d) using medications affected by smoking cessation (antipsychotics, warfarin, theophylline and insulin), (e) suicidal plan or attempts in past 5 years, (f) not willing to try to quit smoking and (g) substance use reported on the day of or before recruitment or positive breath alcohol at recruitment (this could impair accurate answers and ability to consent; they could reschedule once to avoid this rule).Participants did not need to be motivated for smoking cessation treatment; this is a study of cessation induction regardless of motivation.

Recruitment

Brief screening and informed consent by research interviewers was followed by physical examination, laboratory tests and screening interviews. Participants designated someone as a locator and informant.

Overview of design, randomization and procedures

Using a double-blind double-placebo design, participants were randomized to 12 weeks of varenicline plus placebo NRT after a 1-week dose run-up, or 12 weeks of transdermal NRT plus placebo capsules after a 1-week placebo capsule lead-in. All were asked to quit smoking at the end of the 7-day capsule lead-in period and attend 10 brief advice sessions. Within-treatment assessments were conducted 2, 5 and 9 weeks after randomization, and follow-up assessments at 3 months and 6 months after randomization. All procedures were conducted at university offices downtown and approved by the university Institutional Review Board. The trial was registered with Clinical trials.gov, with identifier NCT00756275.

Urn randomization [32] to medication condition (1 : 1 assignment) was stratified by gender, median score on the Fagerström Test for Nicotine Dependence [33] from a previous study [28] and history of MDD. The randomization program was run by the Project Manager, who concealed the assignment from all other staff. Conditions did not have exactly the same n by chance due to randomization. Participants receiving capsules on randomization day were included in intent-to-treat analyses.

Intervention conditions

Varenicline and placebo capsules

Varenicline and matching placebo capsules were prepared by a compounding pharmacy. Participants received 0.5 mg/day for 3 mornings and 1 mg/day (0.5 mg 2×/day) for the next 4 days (dose run-up), and 2 mg/day (1.0 mg 2×/day) thereafter. Placebo capsules were on the same schedule. Written instructions and a physician contact card were provided.

NRT and placebo patch

NRT followed clinical practice guidelines [30], modified for 12 weeks [34]: 4 weeks each of 21 mg/day, 14 mg/day and 7 mg/day. Placebo patches (Rejuvenation Labs, Inc., Midvale, UT, USA) matched active ones. All were instructed to start using patches on quit day after the 7-day capsule dose run-up.

Counseling for both medication conditions

Fully manualized brief advice [31,35] included a 20-minute session at the start of the dose run-up week, a 30-minute session on quit day a week later, then eight weekly 5–10-minute sessions. Assistance involved advice about cognitive–behavioral methods [35], a self-help handbook [36] and problem-solving around barriers [13] and medication issues. Concerns about effects on sobriety were countered with corrective information [28]. The two research therapists (masters’ level, bachelors’ level) had 10 hours of training. Session audiotapes were reviewed in weekly supervision with a doctoral level clinical psychologist, and rated for competence and adherence with immediate feedback.

Assessments

Assessment procedures

Research interviewers blind to condition conducted assessments at baseline, quit day and 2 weeks, 5 weeks, 9 weeks, 3 months and 6 months after the start of dose run-up. At 3 and 6 months, participants were first tested (Alco Sensor IV by Intoximeters, Totnes, UK) to insure a breath alcohol ≤ 0.02 g/dl [37]. Participants who withdrew from treatment were followed for assessments. Participants received $185 for completing all assessments.

Individual difference measures

Diagnoses were made using the Structured Clinical Interview for DSM-IV–patient version [38]. A 6-month time-line follow-back (TLFB) interview [39–41] at baseline was scored for number of days of drug use, number of heavy drinking days (at least six/five drinks/day for men/women [42]), and number of cigarettes per day. The Fagerström Test for Nicotine Dependence was given. Breath carbon monoxide (CO) was collected using an EC50 Micro III Smokerlyzer^® (Bedfont Scientific Ltd, Kent, UK). Barriers to Quitting Smoking in Substance Abuse Treatment (BQS-SAT [13]) identified barriers for corrective feedback in counseling.

Medication adherence

NRT use and capsule adherence were assessed through counts of returned used patches and MEMSCaps (TM; Aardex, Sion, Switzerland) data.

Outcome measures

The primary outcome for Aim 1 was self-reported 7-day point-prevalence abstinence confirmed with a CO level ≤ 4 parts per million (p.p.m.) and salivary cotinine level ≤ 15 ng/ml [43–45] (cotinine not used if using NRT at the time) at 3 months (to detect maximum medication effect after all medication was completed). Secondary measures included 7-day confirmed point-prevalence abstinence at 6 months (Aims 1 and 2; to detect maintenance of effects) and 3- and 6-month average number of cigarettes per day (including abstinent days as zero) and percentage of smoking days (including abstinent days as zero) based on TLFB data (Aim 3). Other secondary measures included within-treatment (5 and 9 weeks after start of dosing) point-prevalence smoking abstinence confirmed by CO ≤ 4 p.p.m. (to see time–course of medication effects) (Aim 4), length of longest continuous abstinence from weeks 9–12 in days (including non-abstainers as zero) (Aim 4), and drug and alcohol use at 3 and 6 months (Aim 5; any heavy drinking, any drug use, number of heavy drinking days and number of drug use days). To count as abstinent from drugs, both self-report and urine drug screen (On Trak^® test cups [Roche, Indianapolis, IN, USA) for screening confirmed with the enzyme multiplied immunoassay technique (EMIT), gas chromatography and mass spectrometry] must have been negative. A significant other was interviewed to encourage honest reporting, but as such reports do not increase validity, they were not used as data [46].

Safety monitoring measures

The Beck Depression Inventory–II [47] (BDI) was administered at baseline and every counseling session to monitor depressive symptoms (Aim 5). At each session a checklist assessed adverse effects.

Statistical analysis approach

Power

In our Data and Safety Monitoring Plan we estimated that detecting a difference of 12 versus 4% abstinent would require n = 274, while detecting a difference of 15 versus 4% would require a total n = 162 with 80% power and alpha of 0.05. When funding concluded we had attained n = 137 randomized, with 62% power to detect the actual primary outcome, 32% power for the actual 6-month abstinence outcome.

Preliminary analyses

All variables were checked for assumptions of normality. Number of heavy drinking days, number of drug use days and percentage of smoking days at follow-up were skewed, so were log-transformed for analyses, with untransformed values presented for ease of interpretation. Medication condition differences in baseline characteristics, medication adherence, number of counseling sessions and follow-up rates were examined with one-way analyses of variance (ANOVA) for continuous variables and χ² for dichotomous variables.

Handling missing data

At follow-up, participants with a CO > 4 p.p.m., cotinine >15 ng/ml, missing CO or cotinine data (except if using NRT) or self-reported smoking were coded as having smoked. Interviewees with a positive, missing or contaminated drug screen were coded as having used drugs for that follow-up interval. Positive imputation for smoking research is most likely to reflect the true values [48]. However, analyses of smoking outcomes, drug use days and heavy drinking days were re-run using multiple imputation [48,49] (using MIANALYZE procedures [50]) as sensitivity analyses [51] for those missing verified abstinence, cigarettes per day, percentage of smoking days or number of days that any drugs or heavy drinking occurred at each follow-up [49,52]. As none of these results differed in significance level from analyses without multiple imputation, analyses showing multiple imputation of the smoking outcomes are presented only in a supplement (Supporting information, Appendix S1).

Post-treatment outcome analyses

For Aim 1, the primary and secondary outcomes (3- and 6-month confirmed 7-day point-prevalence abstinence) were each analyzed with two-group logistic regression, with and without covarying baseline number of cigarettes per day, as this covariate controls for variance due to individual differences unrelated to treatment assignment. For abstinence outcomes close to significant without using the covariate, a Bayes factor (http://www.lifesci.sussex.ac.uk/home/Zoltan_Dienes/inference/Bayes.htm) was calculated to determine whether the results were likely to support the hypothesis in comparison to the expected effect size [53]. For Aim 2, as clinicians want to know if a medication works better when patients have greater adherence, we next entered 3- and 6-month confirmed point-prevalence abstinence into a medication × time analysis while covarying percentage of capsules taken and pre-treatment cigarettes per day. Generalized estimating equations (GEE [54]) models included terms for time and medication × time interaction while being insensitive to missingness. For Aim 3, the average number of cigarettes per day and percentage of smoking days were also analyzed with GEE, entering as covariates the baseline value of the same variable and percentage of capsules taken. The Wald statistic was used for significance testing for dichotomous outcomes and effect size d for continuous outcomes.

Within-treatment (Aim 4)

GEE analyses of confirmed 7-day point-prevalence abstinence at weeks 5 and 9 covaried pre-treatment cigarettes smoked per day. The length of longest continuous abstinence from weeks 9 to 13 was analyzed with regression, entering pre-treatment cigarettes per day as a covariate.

Moderator analyses (Aim 6)

Potential moderators of 3-month point-prevalence abstinence were MDD, percentage of medication capsules taken, gender and Fagerström Test for Nicotine Dependence, investigated using medication by moderator path analysis (using PROCESS [55]). The significant interaction with capsule adherence was followed with simple slopes analyses of medication effects on abstinence at three levels of adherence (the moderator): low [≤ 1 standard deviation (SD) below the mean], average (from > −1 SD to < +1 SD) and high (≥ 1 SD above the mean).

Safety and substance use

BDI total score (Aim 4) was analyzed with group × time GEE of data at baseline and weeks 2, 5, 9 and 3 months, with post-hoc Tukey tests within time. Adverse events are handled descriptively due to low incidence. Substance use outcomes during each follow-up period (Aim 5) were analyzed using t-tests or χ² tests at each time-point and then across time with GEE, without covariates.

RESULTS

Participant characteristics, attrition and confirmation of abstinence

See Fig. 1 for recruitment figures and for data collected for the primary outcome. Follow-up interviews were completed by 89 of 137 (65%) at 3 months and 80 of 137 (58%) at 6 months, but point-prevalence abstinence was analyzed for all 137. Follow-up completion did not differ significantly by medication condition, race, gender, age, education, FTND, MDD or number of drinking days. CO and cotinine collection did not differ significantly by condition.

Figure 1.

Consolidated Standards of Reporting Trials (CONSORT) diagram of participant flow

See Table 1 for characteristics. No medication condition differences were significant for pre-treatment demographic, smoking or substance use variables.

Table 1.

Baseline participant characteristics and treatment compliance: mean (standard deviation) or percentage.

	Full sample (n = 137)	Varenicline (n= 77)	Nicotine replacement therapy (n = 60)
Male	53%	55%	51%
Race
White/Caucasian	83%	82%	83%
Black/African American	15%	16%	15%
Asian/Pacific Islander	<1%	1%	0%
Multi-racial	2%	1%	2%
Hispanic	5%	4%	7%
Married or cohabiting	13%	9%	18%
Annual household income
$0–9999	57%	61%	52%
$10000–29 999	28%	26%	30%
$30000–49 999	12%	10%	13%
$50000+	2%	0%	5%
Age	39.6 (10.0)	40.0 (10.5)	39.2 (9.7)
Years education	12.1 (2.20)	12.1 (2.3)	12.4 (2.1)
Employed	13%	12%	15%
Cigarettes per day	19.5 (7.4)	20.4 (11.1)	18.3 (9.3)
Fagerstrom Test for Nicotine Dependence	5.5 (1.9)	5.5 (1.9)	5.5 (1.9)
History of life-time major depression	28%	28%	29%
Beck Depression Inventory	11.3 (8.1)	11.9 (8.9)	10.5 (6.9)
No. of past substance abuse treatments	1.4 (1.27)	1.5 (1.2)	1.2 (1.1)
Alcohol use disorder	72%	73%	70%
Cocaine use disorder	61%	64%	57%
Opiate use disorder	34%	35%	34%
Marijuana use disorder	23%	35%	22%
Percentage of heavy drinking days	18.8 (27.8)	18.8 (27.3)	18.9 (28.6)
Percentage of drug use days	21.9 (29.3)	23.6 (30.2)	19.9 (28.4)
Counseling sessions completed	4.2 (2.5)	4.2 (2.4)	4.2 (2.6)
No. of weeks with 85% of capsules taken	3.7 (4.2)	3.4 (4.0)	4.2 (4.5)
Percentage of capsules taken	39.6 (37.1)	37.4 (35.1)	42.4 (39.7)
Percentage of patches used	40.9 (39.0)	40.0 (37.5)	43.3 (41.0)

Of participants reporting 7-day smoking abstinence, 95% (19 of 20) at week 5, 100% (of 17) at week 9, 86% (12 of 14) at 3 months, and 100% (of nine) at 6 months were confirmed abstinent. Of participants reporting abstinence from drugs at each follow-up, 91% (63 of 69) at 3 months and 95% (54 of 57) at 6 months were confirmed abstinent.

Compliance

There were no significant medication condition differences in number of counseling sessions attended or medication adherence (see Table 1). See Fig. 2 for weekly capsule use. (Patch adherence was collinear with capsule adherence, r = 0.89, so only capsule adherence was used.)

Figure 2.

Percentage of participants compliant with medication (> 85% of capsules) by assignment to varenicline versus placebo capsules after the week of dose run-up

Smoking outcomes

Smoking abstinence

For Aim 1, the primary outcome of smoking abstinence at 3 months just missed significance when analyzed without the covariate (13% for varenicline, 3% for NRT; odds ratio = 4.33, 95% confidence interval 0.91, 20.56, P = 0.065). A Bayes factor of 3.35 suggests that the data provide positive strong evidence in support of our hypothesis [53,56,57]. When including baseline smoking as a covariate, abstinence at 3 months differed significantly by medication condition. Treatment conditions did not differ significantly for smoking abstinence at 6 months with or without the covariate. A Bayes factor of 1.60 indicated only weak evidence for supporting the hypothesis [53,57], not enough evidence to distinguish our 6-month results from the null [56]. See Table 2 for results at each time-point and for statistics covarying pre-treatment smoking. For Aim 2, when also covarying medication adherence, GEE showed a significant medication effect on point-prevalence abstinence across both 3 and 6 months; those in varenicline were 6.4 times more likely to be abstinent during follow-up when covarying medication adherence (see Table 3). Main and interaction effects for time were not significant.

Table 2.

Outcomes by treatment condition: descriptive statistics and analyses at each time-point. Regressions were used for smoking variables with cigarettes per day included as a covariate; t-tests or χ² tests were used for substance use or depression.

Variable	Varenicline n (%) or mean (SD) (n = 77)	NRT n (%) or mean (SD) (n = 60)	n in analysis	Effect size d or OR	95% CI
Primary outcome: confirmed 7-day smoking abstinence at 3-month Follow-upa,b	10/77 (13)	2/60 (3)	137	4.81*	1.00, 23.13
Confirmed 7-day smoking abstinenceb
5 weeks	10/77 (13)	9/60 (15)	137	0.85	0.33, 2.26
9 weeks	10/77 (13)	7/60 (12)	137	1.25	0.44, 3.55
6-month follow-upa	7/77 (9)	2/60 (3)	137	3.17	0.63, 15.90
Average no. cigarettes per dayc
3-month follow-up	7.31 (7.35)	4.96 (5.54)	89	0.29	−0.13, 0.72
6-month follow-up	8.60 (7.88)	9.62 (9.85)	80	−.02	−.064, 0.26
Percentage of smoking daysc
3-month follow-upd	70.37 (38.79)	62.39 (38.84)	89	0.17	−0.26, 0.59
6-month follow upe	77.12 (37.61)	84.52 (31.25)	80	−0.25	−0.70, 0.20
Longest continuous smoking abstinence weeks 9–13 c	5.19 (9.96)	3.75 (8.34)	89	0.19	−0.15, 0.53
Relapse to any heavy drinkingb
3-month follow-up	9/52 (17)	7/37 (19)	89	0.82	0.28, 2.44
6-month follow-up	8/49 (16)	9/31 (29)	80	0.48	0.16, 1.41
Number of heavy drinking daysc
3-month follow-upf	3.18 (9.02)	1.16 (5.11)	89	0.19	−0.22, 0.61
6-month follow-upf	3.20 (10.26)	3.71 (14.22)	80	−0.07	−0.52, 0.38
Relapse to any drug useb
3-month follow-up	14/52 (27)	12/37 (32)	89	0.77	0.31, 1.93
6-month follow-up	18/49 (37)	8/31 (26)	80	1.67	0.62, 4.50
Number of drug use daysc
3-month follow-upf	2.25 (6.29)	2.89 (14.78)	89	0.15	−0.26, 0.57
6-month follow-upg	5.94 (17.87)	3.74 (14.32)	80	0.24	−0.21, 0.69
Beck Depression Inventorya
2 weeks	6.75 (6.79)	5.64 (6.05)	107	0.17	−0.02, 0.56
5 weeks	6.51 (8.22)	5.59 (7.42)	93	0.12	−0.29, 0.53
9 weeks	4.50 (6.06)	6.34 (7.97)	78	−0.26	−0.72, 0.19
3-month follow-up	5.75 (8.47)	5.19 (6.53)	89	0.07	−0.07, 0.52

^aAll participants were included in this analysis due to positive imputation.

^bDichotomous variable; effect size is expressed as an odds ratio.

^cContinuous variable; effect size is in standardized units of the dependent variable (d).

^dPercentage of smoking days 3-month follow-up, median = 88.33, interquartile range = 82.78. As significantly skewed, values were log-transformed prior to analysis.

^ePercentage of smoking days 6-month follow-up: median = 100, interquartile range = 24.17. As significantly skewed, values were log-transformed prior to analysis.

^fNumber of heavy drinking days 3- and 6-month follow-up, number of drug use days 3-month follow-up: median = 0.00, interquartile range = 1.00. As significantly skewed, values were log-transformed prior to analysis. SD = standard deviation; NRT = nicotine replacement therapy; OR = odds ratio; CI = confidence interval.

^gNumber of drug use days 6 month follow up: median = 0.00, interquartile range = 0.00. As significantly skewed, values were log-transformed prior to analysis.

^*P < 0.05.

Table 3.

Analyses of smoking outcomes by medication and time [generalized estimating equations (GEE)] while covarying amount of capsules taken and baseline smoking.

Dependent variable	Predictor	Model coefficient	SE	Wald	Effect size d or OR	95% CI	P-value
3- and 6-month follow-up
7-day point-prevalence abstinencea,b
	Varenicline (versus NRT)	1.86	0.95	3.84	6.40	1.00, 40.93	0.05
	Pre # cigarettes/day	−0.07	0.03	5.19	0.93	0.87, 0.99	0.02
	% capsules taken	0.03	0.01	8.00	1.03	1.01, 1.05	0.01
	Time (centered)	0.00	0.76	0.00	1.00	0.22, 4.47	1.00
	Time × medication	−0.48	0.88	0.30	0.62	0.11, 3.47	0.58
Number of cigarettes per dayc
	Varenicline (versus NRT)	1.04	1.12	0.87	0.13	−0.15, 0.42	0.35
	Pre # cigarettes/day	0.39	0.08	22.15	0.05	0.03, 0.07	<0.01
	% capsules taken	−0.06	0.02	10.95	−0.01	−0.01, 0.00	<0.01
	Time (centered)	4.46	1.32	11.43	0.57	0.24, 0.90	<0.01
	Time × medicationd	−3.12	1.54	4.07	−0.79	−0.79, −0.01	0.04
Percentage of smoking daysc
	Varenicline (versus NRT)	0.03	0.12	0.05	0.05	−0.37, 0.46	0.82
	Pre # cigarettes/day	0.19	0.25	0.56	0.33	−0.54, 1.20	0.46
	% capsules taken	0.00	0.00	4.53	0.00	−0.01, 0.01	0.03
	Time (centered)	0.20	0.07	8.59	0.36	0.12, 0.60	<0.01
	Time × medicatione	−0.22	0.11	3.83	−0.40	−0.79, 0.00	0.05
Within-treatment: weeks 5 and 9
7-day point-prevalence abstinencea,b
	Varenicline (versus NRT)	−0.12	0.50	0.06	0.88	0.33, 2.36	0.81
	Pre # cigarettes/day	−0.03	0.02	3.55	0.97	0.94, 1.00	0.06
	Time (centered)	−0.29	0.35	0.68	0.75	0.37, 1.50	0.41
	Time × medicationc	0.29	0.42	0.48	1.34	0.58, 3.07	0.49

^aAll participants were included in this analysis due to positive imputation.

^bDichotomous variable; effect size is expressed as an odds ratio.

^cNormally distributed variable; effect size is in standardized units of the dependent variable (d).

^dSimple slopes test within nicotine replacement therapy (NRT), Wald ${\chi }_{(1)}^2=11.90$, P < 0.001.

^eSimple slopes test within NRT (nicotine replacement therapy), Wald ${\chi }_{(1)}^2=8.52$, P < 0.004. SE = standard error; OR = odds ratio; CI = confidence interval.

Secondary smoking variables

For Aim 3, for both cigarettes per day and percentage of smoking days, there was no main effect for medication condition but significant main and interaction effects for time (see Table 3), with a significant increase during the 6 month follow-up period in each smoking variable within the NRT condition but not within the varenicline condition (see Table 2 for means, Table 3 for simple slopes tests). For Aim 4, the longest continuous abstinence and within-treatment abstinence were not significantly different by medication group.

Moderator analyses (Aim 6)

Moderation of medication effects on 3-month point-prevalence abstinence by adherence showed a trend for significance, B = 0.04, standard error (SE) = 0.02, 95% confidence interval (CI) = −0.003, 0.09, P < 0.06. In simple slopes tests, abstinence at 3 months was significantly greater for varenicline versus NRT when adherence was ≥1 SD above the mean, B = 2.21, SE = 0.94, 95% CI = −363, 4.05, P < 0.02, but not when around the mean or below the mean. The predicted conditional probability of abstinence for participants with high adherence (77% or more of capsules taken) was 27% for varenicline versus 4% for NRT. There was no significant moderation by gender, MDD or nicotine dependence.

Safety and substance use

Adverse events

There were no medication-attributable serious adverse events. Twenty participants were discontinued from capsules due to non-serious adverse events, 11 on varenicline [one each for agitation, depression or seizure (with a history of seizures and her physician thought other changes were possible causes); other events were insomnia/nightmares, rash, upset stomach, anxiety)] and eight on NRT (one for agitation, three for depression; other events were insomnia/nightmares, rash, anxiety, tension and itch).

Depressive symptoms

For Aim 4, GEE was not significant for medication nor medication × time, but showed a significant main effect for time (see Fig. 3). Adding history of MDD to the GEE showed no significant main or interaction effects with MDD.

Figure 3.

Beck Depression Inventory score at baseline and weeks 2, 5, 9 and 13 after starting dose run-up. generalized estimating equations (GEE) showed a significant main effect for time, Wald χ² (d.f. = 1) = 13.43, P < 0.001, but not for medication or medication by time. Depression scores decreased from pre-treatment to week 2 [mean difference = −4.99, standard error (SE) = 0.97, 95% confidence interval (CI) = −7.65, −2.33), P < 0.01], with no significant change thereafter. NRT = nicotine replacement therapy; VAR = varenicline

Substance use

For Aim 5, GEEs (see Table 4) showed no significant effects. Fewer than 4% of days involved any heavy drinking and fewer than 7% of days involved drug use (Table 2).

Table 4.

Analyses of substance use at follow-up by medication and time [generalized estimating equations (GEE)] and covarying amount of capsules taken and baseline value of the substance use variable.

Dependent variable	Predictor	Model coefficient	SE	Wald	d or OR	95% CI	P-value
Relapse to any heavy drinkingb
	Varenicline (versus NRT)	−0.36	0.58	0.38	0.71	0.22, 2.18	0.54
	Pre-heavy drinking days	0.56	0.26	3.85	1.68	1.00, 2.81	0.05
	% capsules taken	−0.02	0.01	4.20	0.99	0.97, 1.00	0.04
	Time (centered)	0.58	0.33	3.16	1.79	0.94, 3.38	0.08
	Time × medication	−0.59	0.49	1.42	0.55	0.21, 1.45	0.23
Number of heavy drinking daysa
	Varenicline (versus NRT)	0.05	0.07	0.51	0.13	−0.22, 0.47	0.48
	Pre-heavy drinking days	0.09	0.04	5.36	0.22	0.03, 0.40	0.02
	% capsules taken	0.00	0.00	2.45	0.00	−0.01, 0.00	0.12
	Time (centered)	0.09	0.07	1.59	0.21	−0.12, 0.53	0.21
	Time × medication	−0.13	0.08	2.40	−0.30	−0.67, 0.08	0.12
Relapse to any drug useb
	Varenicline (versus NRT)	0.45	0.48	0.86	1.57	0.61, 3.99	0.35
	Pre # drug use days	−0.32	0.24	1.81	0.72	0.45, 1.16	0.18
	% capsules taken	0.01	0.01	3.07	1.01	0.10, 1.02	0.08
	Time (centered)	0.10	0.07	2.15	1.11	0.97, 1.27	0.14
	Time × medication	−0.60	0.32	3.31	0.55	0.29, 1.05	0.07
Number of drug use daysa
	Varenicline (versus NRT)	0.01	0.07	0.20	0.01	−0.13, 0.15	0.89
	Pre # drug use days	0.12	0.04	7.11	0.12	0.30, 0.21	0.46
	% capsules taken	0.00	0.00	7.65	0.00	−0.01, 0.00	0.01
	Time (centered)	0.06	0.08	0.44	0.06	−0.11, 0.22	0.51
	Time × medication	0.06	0.10	0.33	0.06	−0.14, 0.26	0.56

^aNormally distributed variable; effect size is in standardized units of the dependent variable (d).

^bDichotomous variable; effect size is expressed as an odds ratio. NRT = nicotine replacement therapy; SE = standard error; OR = odds ratio; CI = confidence interval.

DISCUSSION

Varenicline was more effective than NRT in producing point-prevalence smoking abstinence 3 months after medication started (the primary outcome) in smokers with SUD who had not sought smoking treatment, with four times greater odds observed with varenicline (13 versus 3% abstinent for NRT). When analyzed without covarying pre-treatment smoking levels, the evidence was strong that varenicline increases smoking abstinence at 3 months compared to NRT according to the Bayes factor, and the results were significant when controlling for pre-treatment levels of smoking. Differences were no longer significant at 6 months, when 9% of smokers were still abstinent on varenicline compared to 3% on NRT (comparable with the 2% abstinent at 3 and 6 months in our previous study with NRT [28]). This indicated that varenicline had less effect after 3 months off of it when people needed to stay abstinent without the help of a medication, consistent with other studies in this population [4]. However, clinicians also want to know how well a medication works for patients who take more of it. When accounting for amount of medication adherence in analyses, an average of 6.5 times as many people on varenicline compared to NRT had confirmed 7-day point-prevalence abstinence across 3 and 6 months (a significant difference) in analyses of medication × time effects, with no significant decrease over time. When looking at the participants who took at least 77% of the capsules they were given (1 SD above the mean), much larger differences in 3-month abstinence occurred: 27% for varenicline versus 4% for NRT. Thus, for these smokers with substance use disorders who took most of the medication, varenicline worked considerably better than NRT at 3 months and with no significant decrease in effects at 6 months.

In a review of smoking treatments for smokers in SUD treatment [4], only five of 17 studies found effects at 6 months (nicotine patch or gum, behavioral treatment or some combination), with four other studies finding effects at 3 months or less (behavioral treatment, patch or counseling plus exercise), with no effect of varenicline versus NRT in the only study to investigate it [27]. Thus, this is the first study to find more smoking cessation with varenicline than NRT in this population, although differences were only significant at 3 months and the study was underpowered to detect the differences that occurred at 6 months.

Other secondary analyses showed that while NRT and varenicline had nearly equivalent effects during the first 9 weeks of treatment, thereafter the benefits of NRT decreased markedly for abstinence, cigarettes per day and frequency of smoking, while the benefits of varenicline were mainly maintained. Our findings are notable, given several important aspects of this study: first, smokers did not need to be seeking smoking cessation to enroll; the study was designed to increase motivation while reducing barriers to cessation. Secondly, medication adherence was equivalently poor for both medications, with only 63% of participants taking most of their capsules during the first week after dose run-up, only 40–41% of all patches or capsules used and with on average 3.5–4-week adherence, so varenicline seemed to maintain effects after stopping it more effectively than did NRT. It is important to show that a medication can produce effects that persist after the medication is discontinued not just while taking it, as does this study, especially in smokers not very motivated to quit smoking. Thus, even in a SUD population that struggles with motivation and adherence, varenicline can improve smoking cessation 3–6 months after starting it in those who take most of it. Given that varenicline can be used with most smokers who rule out bupropion due to comorbid conditions or medications, this means varenicline with brief counseling may be a more effective treatment for smokers with SUD.

The lack of any significant increase in suicidality, depression or serious events despite 28% of participants having comorbid MDD is consistent with large randomized studies of smokers in the general population with or without psychiatric disorders [17]. The lack of any significant ill effects on substance recovery is also consistent with all other studies of voluntary smoking treatments [14,58]. Beneficial effects of varenicline on heavy drinking seen elsewhere [25] may have not been detectable, because all were alcohol-abstinent at treatment initiation, so little heavy drinking occurred. Thus, given the preponderance of evidence across studies, varenicline can apparently be used in early recovery from SUD even with patients with a history of MDD who are not currently suicidal.

Varenicline’s effects on smoking did not differ significantly by gender, MDD or degree of nicotine dependence. Women have 31% lower odds than men of quitting smoking without medications [46]. However, this study is consistent with others that found no significant differences in odds of quitting smoking for men versus women when using varenicline or NRT [46,59].

Limitations

There was insufficient power to detect the threefold difference in abstinence observed in 6-month outcomes. However, the lack of significant differences in substance use outcomes or depressive symptoms did not appear to reflect low power, given their small effect sizes. This study is limited to one low-income urban population with 17% minorities. Results may underestimate medication effectiveness among smokers with SUD who are seeking smoking treatment. We did not test varenicline against combination NRT, but in a study [28] where we offered combination NRT, only approximately 1% of smokers with SUD used anything but patches, suggesting low acceptance of combination NRT in smokers with SUD, at least in this community. Complier average causal effect modeling could improve the accuracy of evaluating treatment benefit for those adhering to medication, but we did not use this type of modeling.

Conclusions

There is strong evidence that varenicline increased smoking abstinence at 3 months compared to NRT for smokers with SUD regardless of MDD. Although the rates of abstinence were modest compared to results with smokers in the general population who had sought smoking treatment, abstinence results were favorable when compared to many other randomized trials of smokers in early SUD recovery, at least for short-term effects. Furthermore, effects of varenicline versus NRT at 6 months were approximately double those from a major meta-analysis [16], suggesting that low power prevented the 6-month results from reaching significance. By reducing the aversiveness of cessation [20,21], varenicline may make these smokers more willing to attempt cessation. However, ways to improve adherence to medication in this population are very much needed.

Supplementary Material

supplement

Table S1 Multiple Imputation Results of Medication Condition Effects on Smoking Outcomes: Descriptive statistics are the averaged results across the imputations without adjustment for the covariate. Effect size estimates covary baseline cigarettes per day.

Supporting Information

Additional Supporting Information may be found online in the supporting information tab for this article.

Appendix S1 Multiple Imputation Analysis.