A Randomized Trial of Contingency Management for Smoking Cessation in the Homeless

Abstract

Smoking cessation services are an unmet need among the homeless, who smoke at rates more than 4 times the national estimate. Successful interventions have high potential for improving tobacco-related health disparities among homeless smokers. Contingency management (CM) is a behavioral intervention with efficacy in a number of substance use disorder populations, including smokers. However, no randomized studies have evaluated the effect of CM in homeless smokers. We examined smoking-related outcomes in homeless smokers (N = 70) randomized to standard care smoking cessation involving transdermal nicotine replacement therapy (NRT), standard counseling, and carbon monoxide (CO) monitoring or the same standard care plus prize CM for negative CO sample submissions. Participants randomized to CM achieved significantly longer durations of consecutive abstinence and submitted a significantly higher proportion of CO-negative samples relative to standard care participants. At 4 weeks post-quit day, 22% were abstinent in the CM condition and 9% were abstinent in the standard care condition. At the 6-month follow-up, about 10% of smokers in both conditions were abstinent. This study demonstrates that CM is an efficacious option to increase initial quit rates in homeless smokers, but methods to extend effects are needed.

Up to 81% of homeless persons smoke (Baggett & Rigotti, 2010; Connor, Cook, Herbert, Neal, & Williams, 2002; Torchalla et al., 2011; Tsai & Rosenheck, 2012), rendering this population in great need of effective smoking cessation services. Furthermore, homeless persons experience tobacco-related mortality rates 3 to 5 times that of non-homeless persons (Baggett et al., 2015). Smoking-related diseases (e.g., cancers, cardiovascular, respiratory diseases) disproportionately impact the homeless population due to the high rates of smoking and barriers to health care, poor nutrition, crowding, and exposure that exacerbate illness (Hwang, Wilkins, Tjepkema, O’Campo, & Dunn, 2009; Jones et al., 2009; Wiersma et al., 2010). Effective smoking cessation treatments for this population would improve individual health and reduce costs to the public health care system.

A majority (up to 84%) of homeless smokers are interested in quitting smoking (Arnsten, Reid, Bierer, & Rigotti, 2004; Baggett, Lebrun-Harris, & Rigotti, 2013; Connor et al., 2002). Despite this interest in quitting and the potential positive health impact, few studies have examined smoking cessation treatments in the homeless. A large clinical trial (Okuyemi et al., 2013) of 430 homeless smokers randomized participants to 8 weeks of transdermal nicotine replacement therapy (NRT) or NRT plus 6 sessions of motivational interviewing. There were no differences between groups at the end of treatment or the week 26 follow-up, with about 9% and 7% verified abstinent at each time point, respectively. Other studies (Burling, Burling, & Latini, 2001; Okuyemi et al., 2006; Segan, Maddox, & Borland, 2015; Shelley, Cantrell, Wong, & Warn, 2010) using counseling plus pharmacotherapy find carbon-monoxide (CO) verified quit rates at 6 months of 4-14%, which are lower than those found among non-homeless community smokers using NRT (~21%) (Eisenberg et al., 2008). Given their challenging life circumstances and comparably lower quit rates, homeless smokers may need different and more intensive treatment options.

Contingency management (CM) interventions for smoking cessation may benefit homeless smokers by providing an additional level of support for abstinence. CM is based on behavioral operant learning principles and provides frequent, tangible rewards for verified abstinence. Meta-analyses (Benishek et al., 2014; Lussier, Heil, Mongeon, Badger, & Higgins, 2006; Prendergast, Podus, Finney, Greenwell, & Roll, 2006) find that CM is an efficacious adjunct to treatment for substance use disorders, including nicotine dependence. CM for smoking abstinence in the homeless has been explored in two pilot studies. The first study (Businelle et al., 2014) compared 10 homeless smokers who received CM to a historical control group of 58 homeless smokers who received the usual smoking cessation services at the shelter. Usual services included 6 weekly smoking cessation groups and access to smoking cessation medication. The CM participants earned gift cards starting at $20 and escalating by $5 for weekly self-reported smoking abstinence verified by CO samples of <8 ppm. CM participants (30%) were more likely to be abstinent at the 4-week post-treatment assessment compared to the historical control participants (2%), but no follow-up data were presented.

The second pilot study (Carpenter et al., 2015) to examine CM incorporated mobile technology and reinforced smoking abstinence remotely. In this single group design, 20 homeless smokers received 4 counseling sessions, 6 weeks of bupropion SR and NRT, and mobile CM. For the mobile CM, participants were trained to video record themselves taking their own CO reading twice daily using a mobile phone and upload this video to a website for review. Participants earned $1 for the first negative CO sample (< 6 ppm), and incentives increased by $0.25 for each consecutive negative sample. CO-verified quit rates at the end of treatment and at 6 months were 50% and 45%, respectively.

Building on the promising findings of these early studies (Businelle et al., 2014; Carpenter et al., 2015) using CM with homeless smokers, we conducted a randomized controlled trial comparing standard smoking treatment to the same plus CM. Standard treatment involved NRT, counseling, and CO monitoring. Participants in the CM group received the same standard treatment plus the opportunity to earn prizes for evidence of smoking abstinence. We hypothesized better during-treatment smoking outcomes for CM relative to standard care (SC) participants, and we examined post-treatment smoking abstinence through 24-weeks post-quit day.

Method

Participants and setting

Participants (N = 70) were recruited from local facilities that serve the homeless population using flyers, referrals and word-of-mouth efforts. Those meeting the following criteria were invited to participate in the study: (1) 18 years or older; (2) receiving services at a homeless facility or otherwise meeting the federal definition of homeless and intending to stay in the area for at least 4 weeks; (3) average past month smoking of ≥ 5 cigarettes per day (CPD); (4) smoking at least 1 year; (5) CO reading of ≥ 6 parts per million [ppm] and urine cotinine reading consistent with > 100 ng/ml; (6) willing to abstain from all forms of tobacco; (7) self-reported interest in quitting, and (8) willing to return for study visits and to provide locators. Individuals were excluded if they (1) had uncontrolled, severe psychopathology and/or severe cognitive impairments; (2) were non-English speaking; (3) had medical contra-indications of NRT use; (4) reported current or past month treatment for smoking cessation; or (5) were in recovery for pathological gambling. The university Institutional Review Board approved study procedures.

Study recruitment began in October 2012, and follow-ups were completed in July 2016. Study visits, including treatment sessions and follow-up assessments, occurred at a local homeless facility in the northeast. The facility offered several programs, including a soup kitchen, an emergency shelter, and transitional housing, but did not offer smoking cessation services.

Of 345 individuals screened for the study, 115 were ineligible based on the screening and an additional 96 were not consented largely due to lack of interest or failure to attend the scheduled meeting. Of 134 consented participants, 21 were determined to be ineligible at intake and 43 were excluded mostly due to failure to complete the intake process. The remaining 70 participants were randomized to SC (n = 33) or SC+CM (n = 37) conditions. See Figure 1.

Figure 1.

The flow of participants through the study

Measures

Research assistants interviewed participants and entered questionnaire responses directly into the electronic data capture tool, REDCap (Harris et al., 2009). The Modified Housing Inventory (Aidala & The Center for Homelessness Prevention Studies, 2012) assessed current housing status and housing history, use of homeless services and patterns of homelessness, financial support, health status, and employment. A Smoking History Form collected current and past smoking habits, including prior quit attempts. The Fagerstrom Test of Nicotine Dependence (FTND) (Heatherton, Kozlowski, Frecker, & Fagerstrom, 1991) assessed nicotine dependence severity, with scores of 6 or greater indicating severe dependence. To assess problems related to alcohol and drug use, we administered the 10-item Alcohol Use Disorder Identification Test (AUDIT) (World Health Organization, 2001) and the 11-item Drug Use Disorder Identification Test (DUDIT) (Berman, Bergman, Palmstierna, & Schlyter, 2002). AUDIT or DUDIT scores of 8 or higher indicate hazardous drinking or drug use, respectively (Voluse et al., 2012; World Health Organization, 2001). The Timeline Follow-back (TLFB) (Sobell & Sobell, 1992) retrospectively assessed cigarette smoking and NRT compliance (during applicable 8 week period) from 60 days pretreatment through the end of the 6-month follow-up period.

We assessed smoking status using CO breath samples collected with PicoSmokerlyzer CO monitors (Bedfont, UK). CO samples ≤ 4 ppm were considered negative for smoking. Although many studies used higher cut-offs (≤ 6 or 8 or 10 ppm) for abstinence, such liberal values may allow for continued low-level smoking. Low cut-offs (≤ 4 ppm) have been used successfully by our group and others (Alessi, Badger, & Higgins, 2004; Irons & Correia, 2008; Lamb et al., 2007; Lamb, Kirby, Morral, Galbicka, & Iguchi, 2010) and limit the number of false negatives. At baseline and following the NRT period, participants also provided urine samples tested for cotinine using Accutest NicAlert strips (JANT Pharmacal Corp, Encino, CA). Readings consistent with concentrations of ≤ 100 ng/mL were considered negative for smoking.

Procedure

Individuals likely to meet criteria based on initial screening were invited to attend a baseline assessment at a local homeless facility. Research assistants obtained written informed consent under the supervision of the study investigator (CJR). Participants completed assessments and submitted breath and urine samples, and research staff assessed for medical contraindications to NRT use (e.g., prior allergic reaction to adhesive, acute and recent cardiovascular complications). Following the intake assessment, participants attended two 20-minute pre-quit counseling sessions with research staff. At the first quit prep session, the CDC patient handout, “You Can Quit Smoking: Consumer Guide” was discussed, and quit history and reasons for quitting were explored. Participants selected a quit date and developed a quit plan. At the second quit prep session, research staff identified barriers to the quit plan, assessed quit readiness, and provided tips for craving control. Research staff also provided participants with their first NRT patch (to be applied the morning of their quit day) with instructions for use and information about side effects. Also at the second pre-quit session, research staff randomized participants to study condition using a computerized urn procedure (Charpentier, 2003). Randomization was stratified by whether pretreatment reductions in smoking occurred, operationalize as CO < 6 ppm at either quit prep session; 91% randomized to SC and 89% randomized to CM met this criterion pre-randomization.

Follow-ups occurred at weeks 4, 8, 12 and 24 post-quit day. Participants received $25 for baseline and $50 for each follow-up visit, and bus tokens were provided as needed for follow-up visits if the individual had left the facility. Participants in the SC condition received $2 in small items (e.g., toiletries, gum, snacks, etc.) per breath sample submitted non-contingent on results during the active counseling phase.

Treatments

Standard care

On their quit day, participants began treatment. Treatment involved two phases. The intensive period was frontloaded with all three of the treatment components including twice-daily CO monitoring, twice weekly counseling, and NRT for the first four weeks post-quit day. In the second phase (weeks 5-8), only NRT continued. In weeks 5-8, participants met with research staff weekly for provision of NRT, but no farther counseling was provided.

In weeks 1-4, participants submitted breath samples twice daily on weekdays and received the $2 in small items non-contingent on smoking status. Samples were required to be at least 4 hours apart, and the CO monitoring sessions were typically about 5 minutes. Twice weekly, CO sessions were extended to include a 10 minute counseling session covering topics such as the withdrawal syndrome, cognitive and behavioral coping strategies, high-risk situations, and relapse prevention strategies. If lapses or relapses occurred, research staff worked with the participant to re-establish abstinence. Transdermal NRT was available for 8 weeks total post-quit day (weeks 1-8). NRT dose was tailored to the participant’s current smoking level as recommended, and the dose was reduced over time (e.g., 21mg for 4 weeks, 14 mg for 2 weeks, 7 mg for 2 weeks) (Fiore et al., 2008). Research assistants distributed patches daily in weeks 1-4 and weekly in weeks 5-8.

Standard care plus CM

Participants in the CM condition received all components of standard care, including NRT, CO monitoring, and counseling according to the same schedule outlined above except the $2 non-contingent payments. Instead, CM participants earned draws from a prize bowl for each negative (CO ≤ 4 ppm) sample submitted. Draws started at one on the first session of the quit day and escalated by one up to a cap of four draws per session with consecutive negative samples submitted. Positive (CO > 4 ppm), refused, or unexcused missed samples resulted in no draws for that session, and the number of draws was reset back to one for the next negative sample and escalation continued up to the max of four draws per sample. Research staff informed participants of the number of draws possible at the next session for a negative sample at each session.

The prize bowl contained 500 cards; 60% were winning and the remainder were nonwinning (i.e., “Good Job”). The majority (259/500) of winning cards were small prizes (about $1; e.g., small toiletries, food items); 40 cards were large prizes (up to $20; e.g., small electronics, gift cards). One card was a jumbo prize (up to $100; e.g., MP3/Ipod players). CM participants earned up to 155 total draws, including a non-contingent priming draw that occurred at randomization.

Data Analysis

We examined groups for baseline differences using t-tests and χ² tests as appropriate. Analyses of primary outcomes were based on the intent-to-treat sample, using all randomized participants. The majority of outcomes pertain to data collected during the intensive phase of treatment in the first four weeks post-quit day when participants met with research staff for smoking status verification via expired CO up to twice daily and counseling twice weekly. The exception is the NRT adherence outcome, which is reported for the 8 weeks that NRT was available.

We reported treatment adherence data for NRT use, CO session attendance, and counseling session attendance. NRT adherence represented the percentage of days the patch was worn out of 56 possible days (7 days a week for the 8 weeks of treatment); partial or full days of wear counted as adherent for the day. Following a similar approach as in NRT adherence, attendance to one or both CO sessions on a given day counted toward a day of CO monitoring adherence (out of 20 possible days). We calculated counseling adherence as the number of counseling sessions attended out of 10 possible sessions [2 pre-quit sessions plus twice weekly sessions for the first four weeks of treatment (8 sessions) for a total of 10 counseling sessions].

The two primary outcomes were the longest duration of abstinence (LDA) and percentage of negative CO samples submitted in Weeks 1 - 4. LDA was defined as the longest period of consecutive days of negative CO samples submitted during the first 4 weeks of treatment (possible range: 0 to 20 days, because samples were collected 5 days per week). To count toward LDA, at least one CO negative sample was required per day, with the second sample either negative or an excused absence (e.g., court appearance). If the second sample was positive, refused, or an unexcused absence, the LDA sequence was broken. Participants had a mean of five CO samples excused out of 40 possible samples, with no differences between conditions (p = .99). Percent negative CO samples was calculated with actual number of samples submitted in the denominator and the number of negative samples in the numerator (range 0-100%). These primary outcomes were available for 100% of cases. Both variables were non-normally distributed, and transformations failed to normalize the distributions. Thus, we examined group differences using directional, independent group t-tests with estimates derived from 1,000 bias-corrected bootstrapped samples. We also reported nonparametric test (i.e., Mann-Whitney U) results.

Secondary outcomes were past 7-day point-prevalence abstinence (PPA) assessed at weeks 4, 8, 12, and 24. PPA was defined as a self-report of no smoking in the past 7 days, confirmed by objective indicators of smoking status. For all time points, we report PPA defined as negative self-reported smoking status confirmed by negative CO (CO ≤ 4 ppm). At weeks 12 and 24 (post-NRT), we also report PPA based on self-reported 7-day smoking abstinence confirmed by negative urine cotinine (≤ 100 ng/mL) readings.

Changes in 7-day point-prevalence abstinence over time were examined using latent growth models with full information maximum likelihood estimation with robust standard errors (MLR). For this analysis, we used PPA derived from all available data. MLR uses all available data and does not exclude cases with missing data unless all time points were missing (7 cases excluded; 6 were lost to follow-up and an additional participant did not provide objective verification of smoking status at follow-up; see Figure 1). First, we modeled the growth in PPA at weeks 4, 8, 12, and 24 with fixed slope parameters. We then added the following additional covariates: treatment condition, LDA, NRT adherence, and counseling adherence. We included LDA because early treatment response may predict long-term outcomes. NRT and counseling adherence were included to assess whether these treatment components had an impact on post-treatment abstinence. The Likelihood Ratio Chi-Square evaluated the fit of the base growth model, with non-significance indicating a good fit of the model to the data. The sample size adjusted Bayesian Information Criteria (aBIC) was used to compare incremental fit for the models, with smaller aBIC values indicating better fit.

Analyses were conducted using SPSS version 24 and Mplus version 7.4 (Muthen & Muthen, 2015). Alpha was set at .05, and the sample size was powered to detect a treatment effect of d = 0.61 in the primary outcomes based on meta-analyses of studies using reinforcement for smoking abstinence (Lussier et al., 2006; Prendergast et al., 2006) and a study using similar design features as the present study with a disadvantaged smoker population (Alessi & Petry, 2014).

Results

Baseline characteristics

Table 1 presents demographic and baseline characteristics by treatment condition. No significant differences were present between groups on baseline indices, including homelessness severity, smoking history, and alcohol/drug use problems. The sample was older and mostly male. The average age at the first episode of homelessness was early 30s. Participants reported several lifetime episodes of homelessness, and they were homeless about 15 months in the past 3 years. Average age of cigarette smoking onset was 15 years, and participants reported smoking about three quarters of a pack of cigarettes daily on average at study intake with low to moderate dependence levels. Mean scores on the AUDIT reflect few problems with hazardous alcohol use but drug use problems were evident, with mean DUDIT scores over 8 in both groups.

Table 1.

Demographic and baseline characteristics

Variable	SC	SC+CM	Statistical test, p
N	33	37
Age	45.5 (10.5)	44.7 (12.3)	t(68)=0.26, .79
Male, % (n)	75.8 (25)	73.0 (27)	χ2 (1) = 0.07, .79
Never married, % (n)	54.5 (18)	45.9 (17)	χ2 (1) = 0.52, .47
Years education	11.8 (1.3)	12.2 (1.9)	t(68)=−0.87, .39
Hispanic ethnicity, % (n)	27.3 (9)	16.2 (6)	χ2 (1) = 1.27, .26
Race, % (n)			χ2 (2) = 1.43, .49
White	45.5 (15)	51.4 (19)
Black or African American	33.3 (11)	37.8 (14)
Other	21.2 (7)	10.8 (4)
Age first homeless	34.7 (14.2)	30.7 (13.7)	t(68)=1.21, .23
Lifetime episodes of homelessness	4.2 (4.0)	6.8 (9.2)	t(68)=−1.50, .14
Months homeless in past 3 years	15.0 (14.7)	14.9 (13.7)	t(68)=0.04, .97
Age first smoked cigarettes	14.5 (5.1)	14.8 (4.2)	t(68)=−0.27, .79
Cigarettes per day	16.0 (8.5)	14.8 (8.4)	t(68)=0.57, .57
CO reading (ppm)	16.1 (8.9)	16.1 (10.6)	t(68)=−0.01, .99
Urine cotinine level*	6.0 (0.2)	5.9 (0.3)	t(68)=0.37, .71
FTND score	4.39 (2.2)	4.6 (2.0)	t(68)=−0.50, .62
AUDIT score	5.1 (7.0)	5.1 (9.1)	t(68)=0.01, 1.0
DUDIT score	10.7 (12.2)	9.6 (11.4)	t(68)=0.38, .70

Notes. Values are means (with standard deviations in parentheses) unless otherwise noted. AUDIT = Alcohol Use Disorder Identification Test. CM = contingency management. DUDIT = Drug Use Disorder Identification Test. FTND = Fagerstrom Test of Nicotine Dependence. SC = standard care.

^*Urine cotinine levels of 6 correspond to concentrations of >1000 ng/mL, consistent with concentrations expected by a tobacco user.

Treatment adherence

No differences in NRT patch adherence were present between CM participants (M = 42%, SD = 37) and SC participants (M = 51%, SD = 36) with both groups using patches about half of the 56-day NRT period, t(68) = 0.95, p = .35. Similarly, treatment conditions did not differ with respect to percentage of days of CO submission adherence, t(68) = 0.99, p = .32, or percentage of days of counseling attendance, t(68) = 0.93, p = .35. On average, SC participants submitted at least one CO sample on 59% (SD = 0.33) of 20 possible days and CM participants on 51% (SD = 0.35) of days. The total number of CO samples submitted (out of 40 possible) did not differ between groups (p = .70), with SC participants providing an average of 18.9 (SD = 11.6) samples and CM participants submitting an average of 17.7 (SD = 13.1) samples. Counseling adherence was 73% (SD = 0.29) in the SC group and 66% (SD = 0.29) in the CM group.

During-treatment effects of interventions

Table 2 presents primary outcomes by treatment condition. Participants in the CM condition achieved significantly longer periods of abstinence relative to those in the SC condition, t(56) = −2.33,p = .02; Mann-Whitney U = 757.5, p = .04. CM participants submitted a significantly higher percentage of negative CO samples relative to SC participants, t(68) = −1.85, p = .03; Mann-Whitney U = 750.0, p = .047. CM participants earned 44 (SD = 50.9; range = 1 to 147) draws on average, which represented an average cost per CM participant of $94.05 (SD = $126.25) in prizes.

Table 2.

Treatment outcomes

Variables	Treatment Conditions		Test statistic, p	Effect size
	SC (n = 33)	CM (n = 37)
LDA, M (SD)	1.7 (3.0)	4.2 (5.6)	t(55) = −2.33, .02*	d = 0.55
% CO-negative, M (SD)	33.9 (36.7)	51.3 (41.6)	t(68) = −1.85, .03	d = 0.44
7-day PPA, % (n)
Week 4 (CO)	9.1 (3)	21.6 (8)	χ2 (1) = 2.07, .15	φ = .17
Week 8 (CO)	9.1 (3)	13.5 (5)	χ2 (1) = 0.34, .56	φ = .07
Week 12 (CO)	12.1 (4)	13.5 (5)	χ2 (1) = 0.03, .86	φ = .02
Week 12 (CO/cotinine)	9.1 (3)	8.1 (3)	χ2 (1) = 0.02, .88	φ = .02
Week 24 (CO)	12.1 (4)	8.1 (3)	χ2 (1) = 0.31, .58	φ = .07
Week 24 (CO/cotinine)	12.1 (4)	8.1 (3)	χ2 (1) = 0.31, .58	φ = .07

Notes. LDA = longest period of consecutive days negative for smoking (CO ≤ 4 ppm). PPA = point prevalence abstinence. PPA is defined as self-reported past 7-day abstinence from smoking confirmed by negative CO with missing considered not abstinent. For weeks 12 and 24, PPA was also confirmed by negative urine cotinine results.

^*df lower due to adjustments for unequal variances.

Point prevalence abstinence over time

Follow-up completion rates at weeks 4, 8, 12, and 24 were 84%, 74%, 81%, and 73%, respectively, with no differences by treatment condition (see Figure 1). Table 2 presents the 7-day PPA estimates at each follow-up, with missing considered not abstinent. No significant differences by group were present in PPA for any of the follow-up assessments. At the end of the CM and counseling phase (week 4), 22% (n = 8) of CM participants and 9% (n = 3) of SC participants reported smoking abstinence verified by CO. Over time, the percent abstinent decreased in both groups, with 11% (n = 8) abstinent at week 8 (verified by CO), 9% (n = 6) abstinent at week 12 (verified by CO and cotinine), and 10% (n = 7) abstinent at week 24 (verified by CO and cotinine).

The growth model of PPA from week 4 through 24 weeks post-quit day indicated a good fit of the data, LR χ² (10) = 8.53, p = 0.58. The slope was negative (M = −0.47, SE = 1.20) and stable over time, p = .69. The intercept and slope were not significantly correlated, p = .75. The addition of treatment condition, LDA, NRT adherence, and counseling adherence resulted in a better fitting model (growth only model: aBIC = 128.39, growth with predictors model: aBIC = 105.35). LDA was positively associated with starting values at the end of the intensive treatment phase (p = .015), but not with change over time (p = .55). Treatment condition, NRT adherence, and counseling adherence were not significantly associated with the intercept or slope (ps > .05).

Discussion

This study is the first randomized trial examining CM for smoking cessation in homeless smokers. We found significant effects of CM relative to SC in both primary outcomes: longest duration of consecutive abstinence and percentage of negative samples submitted during the intensive phase of the treatment period in weeks 1-4. CM doubled the number of consecutive days of abstinence, and these sustained periods of abstinence were significantly predictive of smoking status at the end of the intensive phase of treatment. This study provides evidence that CM is an efficacious intervention relative to the comparative standard treatment, increasing early smoking abstinence among homeless smokers.

Although we did not find group differences in PPA over time, 22% were abstinent in the CM condition and 9% were abstinent in the standard treatment condition 4 weeks post-quit day. Participants could continue on NRT for weeks 5-8, but abstinence decreased over this period to 14% in the CM group and 9% in the standard care condition. By month 6, about 10% were abstinent regardless of treatment assignment. This 24-week PPA is consistent with rates (4-14%) found in prior non-CM studies using pharmacotherapy conducted in the homeless (Burling et al., 2001; Okuyemi et al., 2006, 2013; Segan et al., 2015; Shelley et al., 2010), but lower than the 45% (12 out of 20 participants) abstinent in the Carpenter et al. (2015) CM study that used bupropion with NRT. CM in conjunction with bupropion, varenicline, and combination therapies are yet largely unexplored in this population, and these avenues may represent strategies to increase initial treatment response rates and promote long-term abstinence for these smokers.

In this study, participants reported wearing the patch on about half of treatment days and attended 69% of counseling sessions. Other reports of smoking cessation pharmacotherapy adherence have been variable. The Carpenter et al. (2015) study found that homeless smokers self-reported NRT use on 57% of treatment days and bupropion on 73% of days. The Shelley et al. (2010) study reported high rates of adherence for varenicline and bupropion, but only 2 to 3 weeks of NRT use on average. Physical ‘patch’ checks at study visits in the Okuyemi et al. (2013) study showed adherence declined over the 8-week treatment period from 53% adherent in the first week to 34% adherent in the final week. Counseling session attendance has also been variable across studies in the homeless population. Businelle et al. (2014) reported an average of 30% of sessions attended, whereas other reports (Okuyemi et al., 2006; Shelley et al., 2010) including this report find higher rates (~60%) of attendance to counseling sessions. This study suggested that homeless smokers were willing to engage in an intensive smoking cessation treatment that included daily CO monitoring, counseling, and NRT. Nonetheless, room for improvement exists, and targeting adherence via reinforcement protocols may be an alternate approach to pursue other than abstinence monitoring, which requires special equipment,.

Abstinence achieved during treatment was a significant predictor of smoking status at end-of-treatment above and beyond treatment condition and adherence factors. Abstinence during treatment has been identified as an important predictor of post-treatment abstinence in a number of CM studies (Higgins, Badger, & Budney, 2000; Higgins, Wong, Badger, Ogden, & Dantona, 2000; Petry et al., 2006; Petry, Alessi, Marx, Austin, & Tardif, 2005), and early abstinence appears to be important for smoking cessation (Heil, Alessi, Lussier, Badger, & Higgins, 2004; Higgins et al., 2006; Lussier, Higgins, & Badger, 2005; Romanowich & Lamb, 2010). Focusing on building these early periods of abstinence, using a combination of methods, may be a good path forward for increasing long-term abstinence in this population.

In this study, 30% of CM patients never left a negative sample, limiting the number of patients who accessed the reinforcement. This observation may be due in part to the abstinence requirement. Future studies might focus on interventions that target successive reductions in smoking over time, and CM protocols can be arranged to reinforce reductions from prior levels in addition to abstinence (Lamb et al., 2007; Lamb, Kirby, Morral, Galbicka, & Iguchi, 2004; Lamb et al., 2010; Lamb, Morral, Galbicka, Kirby, & Iguchi, 2005). We, consistent with other studies (Bonevski, Baker, Twyman, Paul, & Bryant, 2012; Okuyemi et al., 2013), found substantial smoking reductions, with more than half of this sample reducing their smoking rates by 50% or more. These rates suggest that a larger number of patients could meet target goals for smoking reduction, increasing the likelihood that participants are exposed to reinforcement, and thereby increasing the number of smokers who can potentially benefit from the CM program.

This study has a number of strengths, including that it is the first randomized trial of CM for smoking abstinence in homeless smokers. We used a strong multi-component active control consistent with current smoking cessation guidelines (Fiore et al., 2008), intent-to-treat analyses, and biochemical verification of smoking abstinence. The use of twice-daily CO measurement and a conservative CO cut-off decreased the chances of false negatives in smoking status determination. In addition, we included post-treatment smoking status assessments and achieved good retention rates for a transitory population.

Despite these strengths, results should be considered in light of study limitations. Sample size was relatively small. Although the study was powered for during-treatment effects, we had limited ability to detect smaller differences between groups in post-treatment analyses. Our sample was mostly male. Although these proportions are generally consistent with the state homeless population gender distribution (Housing and Urban Development, 2016), women are a fast growing segment of the homeless (National Council for the Homeless, 2009), and our results may not generalize to women-only shelter settings. In addition, inclusion criteria required participants to be interested in quitting smoking and to be willing to abstain from cigarettes and other tobacco products. Results may not generalize to less motivated individuals.

Overall, this study suggests that CM is efficacious in improving smoking abstinence in homeless smokers. The results of this study find short-term benefits from CM in reducing smoking in this health disparities population. Future studies should explore methods to extend smoking abstinence over time to improve the health outcomes in homeless smokers who suffer disproportionately from smoking-related morbidity and mortality.