Effect of varenicline directly observed therapy versus varenicline self-administered therapy on varenicline adherence and smoking cessation in methadone-maintained smokers: a randomized controlled trial

Abstract

Background and Aims:

Level of adherence to tobacco cessation medication regimens is believed to be causally related to medication effectiveness. This study aimed to evaluate the efficacy of varenicline directly observed therapy (DOT) on varenicline adherence and smoking cessation rates among smokers with opioid use disorder (OUD) receiving methadone treatment.

Design:

Multicenter, parallel-group two-arm randomized controlled trial.

Setting:

Urban opioid treatment program (OTP) in the Bronx, New York, USA.

Participants:

Daily smokers of ≥5 cigarettes/day, interested in quitting (ladder of change score 6–8), in methadone treatment for ≥3 months, attending OTP ≥3 days/week. Participants’ mean age was 49 years, 56% were male, 44% Latino, 30% Black, and they smoked a median of 10 cigarettes/day.

Interventions:

Individual, block, random assignment to 12 weeks of varenicline, either directly observed with methadone (DOT, n=50) or via unsupervised self-administration (SAT, n=50).

Measurements:

The primary outcome was adherence measured by pill count. The secondary outcome was 7-day point prevalence tobacco abstinence verified by expired CO<8 parts per million.

Findings:

Retention at 24 weeks was 92%. Mean adherence was 78.5% (95% confidence interval (CI) 71.8–85.2%) in the DOT group versus 61.8% in the SAT group (95% CI 55.0–68.6%); differences were driven by DOT effects in the first 6 weeks. CO-verified abstinence did not differ between groups during the intervention (p=0.26), but was higher in the DOT than the SAT group at intervention end (DOT 18% versus SAT 10%, difference 8%, 95% CI −13, 28); this difference was not significant (p=0.39) and was not sustained at 24-week follow up.

Conclusions:

Among daily smokers attending opioid treatment programs, opioid treatment program-based varenicline directly observed therapy was associated with early increases in varenicline adherence compared with self-administered treatment, but findings were inconclusive as to whether directly observed therapy was associated with a difference in tobacco abstinence.

Introduction

Over three-quarters of persons with opioid use disorder (OUD) smoke cigarettes (1). Among smokers with OUD, cessation rates with evidence-based smoking cessation medications are less than half the rates observed in smokers without OUD (2–6). Poor medication adherence, a major determinant of limited treatment efficacy, is consistently associated with active drug use (7–9). However, even smokers in treatment for OUD experience challenges adhering to smoking cessation pharmacotherapy, including: comorbid disease (10), limited social support (11,12), poor self-efficacy (13,14), and risk perceptions that prioritize other medical concerns over tobacco-related risks (15). Consequently, smokers with OUD often fail to initiate or adhere to prescribed nicotine replacement therapy (NRT), bupropion, or varenicline (4,16–22).

Adherence to NRT, bupropion, or varenicline is associated with improvement in smoking cessation outcomes (16–18,23–33), with a dose-dependent relationship between adherence and abstinence (19,25,32). In large smoking cessation trials among smokers with OUD, adherence to cessation medications was associated with improved cessation outcomes (16,17,21,23,34); in one trial there was a seven-fold odds of abstinence on days on which participants used (vs. did not use) NRT (16). These results highlight the critical importance of addressing adherence to achieve tobacco abstinence among smokers with OUD.

Despite robust associations between medication adherence and smoking cessation, few studies have evaluated efficacy of interventions to improve adherence to smoking cessation pharmacotherapy (35). Interventions to promote medication adherence in other chronic diseases have been evaluated, including reminder systems, counseling, contingent reinforcement, and directly observed therapy (DOT), but few studies have tested interventions to improve adherence with smoking cessation medications (35). In trials among smokers without OUD, contingent reinforcement and medication electronic monitoring system feedback increased adherence to NRT and bupropion (36–38), while adherence counseling did not (38–42).

DOT improves TB and HIV medication adherence and health outcomes among people who use drugs (43–45). Opioid treatment programs (OTPs) provide an ideal setting for DOT because regulations require up to daily visits. In the U.S., there are more than 1,300 OTPs and nearly 500,000 persons receiving medication treatment of OUD (46). Though smoking cessation medications have been delivered via DOT in OTPs in some clinical trials (47–49), DOT effects were not evaluated. To address this, we conducted a parallel-design, randomized, controlled trial to test the efficacy of varenicline DOT delivered at an OTP, compared to self-administered treatment (SAT), for promoting adherence and tobacco abstinence among smokers with OUD. We hypothesized that at the end of the 12-week intervention, rates of adherence and tobacco abstinence would be higher among participants randomized to DOT versus SAT.

Methods

Design overview

We have previously described our study design (50). Between December 2011 and April 2015, we conducted a two-arm, individually randomized, parallel group, multi-center, randomized controlled trial with a 1:1 allocation using block randomization comparing DOT to SAT. Smokers with OUD who were enrolled in methadone treatment were randomly assigned to receive varenicline by DOT or SAT for 12 weeks. Research assessments occurred at baseline, and at weeks 1, 2, 3, 6, 9, 12, and 24; reimbursement was up to $110 for completion of all measures. The protocol was approved by a Data Safety and Monitoring Committee and the Einstein Institutional Review Board. We obtained a Certificate of Confidentiality from the National Institutes of Health.

Study setting and participants

Participants were recruited from the Einstein/Montefiore Division of Substance Abuse OTPs, which offer integrated opioid agonist treatment and primary medical care services in three locations in the Bronx, NY, USA. Research assistants recruited participants in OTP waiting areas, by word-of-mouth, fliers, and clinic counseling and medical staff referral. All participants provided written informed consent.

We have previously described eligibility criteria and screening procedures (50). In brief, eligible persons were: ≥18 years; smoked five or more cigarettes per day; interested in quitting smoking; did not take varenicline in the prior 30 days; enrolled in methadone treatment for at least three months, received methadone at the OTP three to six times per week with no more than two absences in the prior two weeks; English speaking; able to provide informed consent; and not pregnant, breastfeeding, or trying to conceive. To enhance recruitment, we broadened eligibility criteria after six months to include persons receiving methadone at the OTP ≥3 times per week (from ≥4 times per week). We excluded persons with unstable medical or psychiatric illness.

Sample Size Determination

We determined our sample size based on our prior trial of OTP-based antiretroviral DOT (45), in which we observed a 13% difference in adherence between DOT and SAT groups, and on a 44% rate of NRT adherence in another prior trial among smokers with OUD (16). We thus estimated adherence rates of 57% among DOT and 44% among SAT participants. At a two-sided alpha=0.05, and with an intraclass coefficient=0.3, we determined that a sample size of 17 per group would provide >90% power to detect this effect on our primary outcome of adherence.

To test potential group differences in abstinence (our secondary outcome), we decided to increase our sample size. We first estimated that abstinence over the 12-week intervention among SAT participants would be 22%, half that observed in varenicline phase 3 trials, and consistent with results from studies of NRT and bupropion among smokers with OUD (3,20,53) conducted prior to the conclusion of any trials of varenicline in smokers with OUD (4,5). We then estimated that increased adherence due to DOT would double the likelihood of participants achieving cessation (17,25,26,54), yielding an abstinence rate of 44% in the DOT group. We estimated that a sample size of 38 per group was required to detect this effect in mixed effects models with a within-subject intraclass coefficient=0.3 and a two-sided alpha=0.05. Estimating 25% attrition at 12 weeks, we planned a priori to recruit a total of 100 participants (50 per group). We did not perform, or base recruitment decisions on, interim analyses.

Randomization

Participants were randomized 1:1 to DOT or SAT in variably-sized blocks of 2–8 participants via computer-generated randomization. Randomization was stratified by clinical site and, to account for potential differences in experience with medication-taking and OTP-based DOT, HIV status. To ensure that the allocation sequence was concealed, a centrally-located data manager, not involved in eligibility determination, generated the sequence, stored it in a password-protected file, and assigned participants to groups. Because the intervention was not blinded, we varied block size to prevent anticipation of group assignment.

Interventions

All participants received 12 weeks of varenicline at standard doses: 0.5 mg once daily for three days, 0.5 mg twice per day for four days, and 1 mg twice per day for the remaining 11 weeks. While most participants had public insurance that covered the cost of varenicline, we used research funds to cover any co-payments and to purchase varenicline from a community pharmacy for uninsured participants.

Directly Observed Therapy (DOT) Intervention Condition.

Participants in the DOT group received varenicline at the same time as their methadone dose at their OTP, three to six times per week; they also received take-home varenicline doses packaged in individual pill boxes for self-administration on days off from clinic, evenings, and holidays. Participants were instructed to return pillboxes to nurses at their next OTP visit, whether or not they had taken the varenicline pills.

Self-Administered Therapy (SAT) Control Condition.

When a participant initiated his/her baseline research visit, the study physician called in their varenicline prescription to a community pharmacy of the participant’s choosing. Participants picked up a four-week supply and subsequent refills from this pharmacy, and self-administered all varenicline doses.

Counseling and medication education

All participants received a single session of brief (≤ 10 minute) physician smoking cessation counseling, based on Public Health Service guidelines. Brief advice has been shown to increase the likelihood of smoking cessation (55), is associated with comparable outcomes to motivational counseling (3,40), and is easily implemented in clinical settings. Participants also received a brochure promoting medication adherence, and instructions on management of anticipated varenicline side effects. Over the intervention period, participants who reported treatment-emergent symptoms met with the study physician for assessment and brief counseling on symptom management.

Visit schedule and measures

We conducted research visits in a private office on-site at participants’ OTPs at baseline and weeks 1, 2, 3, 6, 9, 12 and 24. We collected survey data using Audio Computer-Assisted Self-Interview (ACASI), which allows participants to listen to questions via headphones while simultaneously reading them on a computer screen, and has been shown to increase reporting of stigmatized behaviors (56).

Adherence measures.

Our primary outcome was adherence, assessed by pill-count. Pill count is an objective measure of adherence that correlates with electronic monitors and varenicline plasma levels (57). For DOT participants, we counted varenicline pills remaining in pill trays weekly. These included unconsumed pills that were not dispensed due to clinic misses, participant refusal, or nurse error, and those returned to nurses in pill boxes. We assumed that unobserved doses in unreturned pill boxes were not consumed.

For SAT participants, pill counts occurred at all study visits during the intervention period. To optimize pill bottle return and ensure pill count adherence measurement, we reimbursed participants partially for completion of research visit measures with an additional $5 for completion of pill counts.

Tobacco use measures.

We measured self-reported abstinence at each follow-up visit by asking “Have you smoked at least part of a cigarette in the past seven days, even a puff?” and verified abstinence with CO<8 p.p.m. (Micro Smokerlyzer®; Bedfont Scientific, Maidstone, UK). We selected the 7-day point prevalence abstinence measure and <8 p.p.m. threshold, and tested abstinence at 12 weeks (intervention end), to allow comparison to prior trials among smokers with OUD (3–5,53). We also evaluated 7-day point prevalence abstinence over the intervention period (weeks 1, 2, 3, 6, 9 and 12) and at 24 weeks. Given recent recommendations (58), we evaluated abstinence using a <5 p.p.m. threshold in post-hoc analyses.

Treatment emergent symptoms.

At weeks 1, 2, 3, 6, 9, and 12, we used both a structured questionnaire and open-ended questions to evaluate symptoms that emerged or increased in intensity during treatment. We also assessed whether symptoms prompted missed medication doses or medication discontinuation. We used the Mini-International Neuropsychiatric Interview 6.0.0. to assess incident psychiatric illness; and the Columbia Suicide Severity Rating Scale to assess suicidality (51,52).

Statistical analysis

Varenicline adherence (primary outcome).

With an intent-to-treat approach, we examined the effect of DOT on the continuous measure of pill count adherence at weeks 1, 2, 3, 4–6, 7–9, and 10–12. Adherence was calculated as the amount of medication taken as a percentage of the target dose.

Of 551 pill count adherence measures, 95.3% were between 0 and 100. As previously described (50,59), we modified measured pill count adherence according to the following parameters: we considered pill count adherence rates that were negative (0.2%) to represent 0% adherence. We considered adherence rates between 100% and 120% (2.5%; plausible if participants mistakenly took extra pills up to two times weekly) as 100% adherence. If pill count adherence rates were >120% (2%), we imputed new values for those data points by averaging five randomly generated imputed values based on the mean and standard deviation of other adherence rates from the same participant.

We fit the repeatedly measured outcome of pill count adherence using a mixed effects linear model, which included a fixed intervention effect (DOT versus SAT), time effect, intervention by time effect, and stratification variables (clinical site and HIV status). We used a two-tailed significance level of 0.05. To account for within-subject correlations, the covariance matrix of the error term was assumed to be unstructured. This modeling was repeated to include baseline variables (cannabis use, housing status) that differed between groups. Analyses were conducted using SAS, version 9.4 (SAS Institute, Cary, NC).

If a participant chose to discontinue medication, adherence was calculated as 0% from that date forward. However, if investigators discontinued varenicline, adherence was considered missing from the date of discontinuation. Missing pill count data were not included in analyses. In sensitivity analyses, we conducted multiple imputation modeling using Markov Chain Monte Carlo methods with five imputations, and fully conditional specification, for missing adherence data. Models included study arm, adherence, time, clinical site, HIV status, baseline cannabis use, and housing status.

Tobacco abstinence (secondary outcome).

With an intent to treat approach, we compared CO-verified abstinence between DOT and SAT groups at week 12 using chi-square tests. To account for missing abstinence data, we first considered missing to be equal to smoking (60). Given that the direction of potential bias with this approach is not known (61), we also conducted sensitivity analyses by constructing multiple imputation models to impute missing abstinence data. We used SAS PROC MI to construct fully conditional specification multiple imputation models; this approach can take into account uncertainties regarding the mechanism of missing data. We generated five imputed data sets, using the following predictors: age, sex, cigarettes per day, Brief Symptom Inventory Global Severity Index score, and methadone dose at baseline. We then applied a chi-square test to each imputed proportion of abstinent participants, and pooled the resulting proportions and p-values using Rubin’s rules (62) and the multiple parameter pooling method (63). We compared tobacco abstinence over the intervention period between the DOT and SAT groups using logistic mixed effects models that included study group, time, intervention by time, clinical site, HIV status, and baseline cannabis use and housing status. We also compared abstinence at week 12 using a complete-case approach, at week 12 using CO <5 p.p.m. threshold, and at week 24 (< 8 p.p.m.), using both missing=smoking and multiple imputation approaches.

Treatment emergent symptoms.

We compared the proportion of participants reporting treatment-emergent symptoms between groups using chi-square or Fisher’s exact tests.

Results

Baseline characteristics

We screened 1282 individuals, of whom 996 were ineligible, and 186 were unwilling to participate or did not complete screening procedures (Figure 1). A total of 100 participants were randomized (50 to DOT and 50 to SAT). Retention was 96% at 12 weeks and 92% at 24 weeks.

Figure 1.

Flow chart of participant screening, enrollment and follow up

Baseline demographic, smoking and clinical factors were similar between groups (Table 1). Participants’ median age was 49 years, 56% were male, 44% Latino, 30% Black, and 10% multiracial. They smoked a median of 10 cigarettes per day, had moderate nicotine dependence, and 88% smoked menthol cigarettes. Co-morbid psychiatric illness and substance use were common: 21% had a lifetime history of major depressive disorder, and 28% reported past-month heroin use.

Table 1.

Baseline characteristics of the study population

	Total	DOT	SAT
	(n=100)	(n=50)	(n=50)
Sociodemographic characteristics
Age, median [interquartile range (IQR)]	49 (45,53)	49 (46,53)	49 (45,52)
Male sex, n (%)	56 (56)	26 (52)	30 (60)
Race/ethnicity, n (%)
Hispanic/Latino	44 (44)	22 (44)	22 (44)
Black	30 (30)	17 (34)	13 (26)
Multiracial/multiethnic	10 (10)	4 (8)	6 (12)
Non-Hispanic white	14 (14)	7 (14)	7 (14)
≤ High school education, n (%)	79 (79)	39 (78)	40 (80)
Married or living with partner, n (%)	45 (45)	26 (52)	19 (38)
Employed, n (%)	20 (20)	10 (20)	10 (20)
Lifetime history of incarceration, n (%)	75 (75)	36 (72)	39 (78)
Unstable housing, n (%)‡	64 (64)	37 (74)	27 (54)
Tobacco Use Characteristics
Cigarettes/day, median (IQR)	10 (8,20)	12 (8,20)	10 (8,15)
Carbon monoxide, median (IQR)	7 (5,10)	8 (5,10)	7 (5,10)
Fagerstrom Test of Nicotine Dependence score, median (IQR)	5 (4,6)	5 (4,7)	5 (4,6)
Ladder of change score, median (IQR)	8 (6,8)	8 (6,8)	8 (7,8)
Quit importance, median (IQR)	10 (9,10)	10 (9,10)	10 (9,10)
Quit confidence, median (IQR)	8 (5,10)	8 (6,10)	7 (5,10)
Any past quit attempts, n (%)	81 (81)	42 (84)	39 (78)
Median duration longest prior quit attempt,days (IQR) (n=74)	29.5 (3,365)	22 (5,365)	90 (2,365)
Any other household smoker, n (%) (n=98)†	55 (56)	28 (56)	27 (56)
Preferred brand menthol, n (%)	88 (88)	42 (84)	46 (92)
Psychiatric comorbidity
Lifetime major depressive episode, n (%)§	21 (21)	11 (22)	10 (20)
Lifetime psychotic disorder, n (%)§	10 (10)	4 (8)	6 (12)
Lifetime suicide attempt, n (%) ||	15 (15)	8 (16)	7 (14)
Severe global psychiatric symptoms, n (%)¶	17 (17)	9 (18)	8 (16)
Currently receiving psychiatric treatment, n (%)	37 (37)	18 (36)	19 (38)
Medical comorbidity
COPD/Asthma, n (%)	30 (30)	15 (30)	15 (30)
HIV/AIDS	20 (20)	11 (22)	9 (18)
Median weight, lbs (IQR) (n=99)†	184 (154,207)	183 (160,211)	186 (153,202)
Have a regular medical provider, n (%) (n=98)†	80 (82)	39 (78)	41 (85)
Substance use characteristics
Median duration methadone maintenance, years (IQR)	5.6 (3.3,8.9)	5.6 (3.3,8.9)	5.0 (3.3,8.9)
Median methadone dose, mg (IQR)	90 (61.5,112.5)	83.5 (70,110)	92 (60,120)
Self-reported use of illicit drugs in 30 days prior to baseline, n (%)
Heroin (n=97)†	27 (27.8)	11 (22.5)	16 (33.3)
Other opiates (n=98)†	29 (29.6)	11(22.5)	18 (36.7)
Cocaine (including crack) (n=98)†	24 (24.5)	9 (18.4)	15 (30.6)
Sedative/hypnotics (n=96)†	18 (18.8)	8 (16.3)	10 (21.3)
Marijuana (n=98)†	26 (26.5)	18 (36.7)	8 (16.3)
Hazardous alcohol use, n (%)** (n=99))†	24 (24.2)	13 (26)	11 (22.5)

^†n<100 due to missing data

^‡Living in a shelter, temporary housing, hotel/motel, or on the street

^§Assessed using the Mini-International Neuropsychiatric Interview 6.0.0

^||Assessed using the Columbia Suicide Severity Scale, a structured interview which assesses suicidal ideation, plans, intent, and behavior

^¶Assessed using the Brief Symptom Inventory Global Severity Index, with scores dichotomized at a T score ≥ 63

^**Assessed using the Alcohol Use Disorder Identification Test, with hazardous alcohol use defined as a score >= 4 for women or >=8 for men

Varenicline adherence

In linear mixed effects models (Figure 2), estimated mean adherence was 78.5% (95% CI 71.8, 85.2%) in the DOT group and 61.8% (95% CI 55.0, 68.6%) in the SAT group. There was a significant time by treatment group interaction (p=0.01), with DOT effects on adherence most pronounced in the first six weeks of the 12-week intervention period. This effect persisted when adjusting for past-month cannabis use and housing status. Of a possible 600 pill count measures during the intervention period, there were 8 (DOT) and 41 (SAT) missing measures (p<0.01); our adherence estimates were unchanged in sensitivity analyses using multiple imputation modeling for missing data (not shown).

Figure 2. Adherence rates over time.

Adherence rates and their 95% confidence intervals are model-based estimates obtained from linear mixed effects models that adjust for time, clinical site, and HIV status. DOT = directly observed therapy; SAT = self-administered therapy

Tobacco use

At the end of the 12-week intervention period, CO-verified tobacco abstinence was 18% among DOT participants, compared to 10% among SAT participants (difference 8%, 95% CI −13, 28), p=0.39, relative risk of abstinence for DOT versus SAT: 1.8 (95% CI 0.65, 4.99). Over the intervention period, abstinence did not differ between groups in mixed effects models (OR = 1.12, 95% CI 0.91, 1.33, p=0.26). Abstinence was not maintained after active treatment ceased (Table 2). These estimates were unchanged when multiple imputation models were used for missing data.

Table 2.

Tobacco use outcomes

	DOT	SAT	p value	Percentage Point Difference (95% CI)
CO-verified seven-day point prevalence abstinence measure	n=50	n=50
Abstinence (CO < 8 p.p.m.) at 12 weeks, missing = smoking, % (n)*	18 (9)	10 (5)	0.39	8 (−13, 28)
Abstinence (CO < 8 p.p.m.) at 12 weeks, multiple imputation for missing data, % (n)†	20 (10)	10.8 (5.4)	0.33	9.2 (−11, 29)
Abstinence (CO < 8 p.p.m.) at 12 weeks, complete case, % (n)	18.8 (9)	10.4 (5)	0.39	8.4 (−13, 29)
Abstinence (CO < 5 p.p.m.) at 12 weeks, missing = smoking, % (n)*	12 (6)	8 (4)	0.74	4 (−16, 24)
Abstinence (CO < 8 p.p.m.) at 24 weeks, missing = smoking, % (n)	4 (2)	12 (6)	0.27	−8 (−28, 13)
Abstinence (CO < 8 p.p.m.) at 24 weeks, multiple imputation for missing data, % (n)†	4 (2)	12.4 (6.2)	0.25	− 8.4 (−28, 12)

^*7 day point prevalence abstinence at 12 weeks, biochemically verified with CO < 8 parts per million (p.p.m.), with missing considered smoking, compared between groups using chi-square tests

^†Sensitivity analyses of carbon monoxide-verified 7-day point prevalence abstinence outcomes using multiple imputation models with fully conditional specification for missing data

Adverse effects

A greater proportion of DOT than SAT group participants experienced vomiting (26% vs. 10%, p=0.04) and fatigue (24% vs. 8%, p=0.03). There were no differences between groups in the frequency of other treatment-emergent symptoms; symptoms prompting missed medications or medication discontinuation; incident psychiatric illness; or suicidality (Supplemental Table).

Discussion

In this trial of OTP-based varenicline DOT compared to self-administered treatment (SAT), we found that participants randomized to DOT had significantly higher levels of varenicline adherence (78.5% versus 61.8%) and achieved an 18% smoking cessation rate at the end of treatment. As in other trials of tobacco cessation treatment (3,5,18,53), cessation effects associated with DOT did not persist after the active DOT intervention ceased.

Our finding that DOT varenicline was associated with higher adherence than SAT extends limited evidence evaluating interventions to increase adherence to smoking cessation treatment. A meta-analysis of adherence interventions for tobacco cessation medications found modest improvements in adherence, but noted that evidence was limited in both quality and quantity (35). In one study, contingent reinforcement increased use of nicotine gum from a mean of six (with standard treatment) to ten pieces daily (36), while other studies showed that medication electronic monitoring with feedback increased adherence to NRT or bupropion by 11 – 25% compared to counseling or electronic monitoring alone (37,38,64). By contrast, adherence counseling has not increased adherence over control conditions (38–42,65,66). In this trial, we demonstrated robust adherence-enhancing effects of DOT varenicline, consistent with DOT effects in prior OTP-based trials of HIV and hepatitis C treatment (45,67,68). Because DOT both links varenicline adherence to an established behavioral skill (OTP attendance) and enhances social support through daily contact with nurses who administer varenicline, it mitigates against factors associated with non-adherence, such as active drug use (69). Unfortunately, we are unable to quantify which of the multiple components of DOT - direct observation, preparation of pill trays, nurse support - provides greatest effect. DOT effects are also consistent across disease states: the 16.7% increase in adherence we observed between DOT and SAT is on par with the 11–22% increase in adherence seen in prior trials of OTP- and community-based DOT for HIV and hepatitis C (45,67,68).

Though DOT increased adherence in our trial, it did not significantly increase smoking cessation. This finding is in contrast to the strong DOT effects on clinical outcomes such as HIV virologic suppression (45,68), but concordant with the modest impact of adherence enhancing interventions on smoking cessation (35). There are multiple potential reasons for this finding. First, some interventions increase adherence to smoking cessation medication at levels that are clinically modest (38); few interventions increase adherence to the ≥80% or ≥95% levels shown to be associated with cessation (30,42). Second, because smoking cessation pharmacotherapy lacks the potency of antiretroviral medication for HIV or hepatitis C, improving adherence alone may not be sufficient to yield tobacco abstinence. The complex relationship between pharmacotherapy adherence and smoking cessation is illustrated by an analysis of a prior clinical trial, in which groups with high probabilities of adherence had varying probabilities of tobacco abstinence (high, intermediate, and near zero) one month following a quit attempt (70). Finally, the causal relationship between adherence and abstinence is not linear, as successful tobacco cessation may further motivate medication adherence (30,38).

The difference in cessation observed at intervention end was not sustained at post-intervention follow up. This finding is concordant with other trials among smokers with OUD, in which interventions were associated with increased cessation over control conditions, but effects dissipated once active treatment ceased (5,18,47,53,71). Similarly, the effects of DOT on adherence and health outcomes are not maintained after DOT is stopped (72–74). Extended pharmacotherapy is a strategy that has been shown to help maintain abstinence and reduce relapse over time among smokers without OUD (75–78), but trials among smokers with OUD that have employed extended treatment have shown low rates of treatment adherence (4,22). As a next step in this line of research, our group is testing DOT and long-term treatment in an ongoing randomized 2 × 2 factorial trial.

This study has limitations. First, our sample size of 100 participants was set a priori prior to publication of varenicline trials among persons with OUD, and we overestimated varenicline efficacy in this group. This yielded limited power to detect significant cessation effects. Second, despite the internal validity of our eligibility criteria, it is possible that a broader population of smokers, such as those smoking <5 cigarettes per day, could benefit from DOT varenicline. Third, we recognize that reasons for medication non-adherence are multifaceted. Though a DOT strategy may not directly address potential barriers to adherence, such as poor psychological health, active substance use, or unstable housing, DOT interventions can provide adjunctive support to lessen their negative impact on adherence. Finally, our findings may not generalize to smokers not yet ready to quit, or not engaged in OUD treatment.

In conclusion, DOT varenicline administered in an OTP is associated with greater adherence than self-administered treatment. Given the marked concentration of tobacco use among smokers with OUD, scale up of DOT in OTPs has large potential public health impact. DOT can be feasibly implemented in other settings which provide regular medication support, such as residential drug treatment, supportive housing, or behavioral health programs, as well as in community-based settings (44,68,79–87). Given the marked burden of tobacco use and modest efficacy of other cessation approaches, OTP-based directly observed smoking cessation medication should be considered for smokers with OUD.