Smokers with bipolar disorder, other affective disorders, and no mental health conditions: Comparison of baseline characteristics and success at quitting in a large 12-month behavioral intervention randomized trial

Abstract

Background:

The extent to which smokers with bipolar disorder (BD) differ from other smokers on cessation-related characteristics and outcomes is unknown and could improve knowledge of treatment needs for this group. These analyses compared smokers with BD versus smokers with other affective disorders (ADs; anxiety and unipolar depression) and smokers with no mental health conditions (MHCs).

Method:

Participants (n=2,570) were a subsample of those enrolled in a smoking cessation trial comparing two web-delivered intervention approaches: acceptance and commitment therapy (ACT) and cognitive behavioral therapy. Those included in this analysis self-reported having BD (n=221), other ADs (n=783) or no major MHCs (n=1,566). Surveys assessed baseline characteristics and self-reported abstinence at 3, 6, and 12-months post-randomization. Treatment utilization was tracked via page views.

Results:

Smokers with BD were distinct from both AD and no MHC smokers on the majority of baseline characteristics. At 12-months, quit rates were lower for smokers with BD (20%) than no MHCs (29%; p =.01), but no different than other ADs (20%; p= .467). Interactions between treatment assignment and diagnostic group were non-significant for cessation outcome. The number of logins was higher for smokers with BD than AD in the ACT arm only (p=.001), but this finding was not replicated across other utilization indicators.

Conclusions:

Smokers with BD and other ADs had similar long-term quit rates despite numerous differences in baseline characteristics. Despite being lower than for smokers without MHCs, long-term quit rates from web-based treatment are promising for smokers with BD as well as other ADs.

1. Introduction

In epidemiologic studies, bipolar disorder (BD) is associated with high rates of ever-smoking and a low proportion of successful quitters (Lasser et al., 2000). Consistent with the high prevalence of smoking and low rate of quitting, the prevalence of tobacco-related disease in adults with BD--including cardiovascular disease, diabetes, and respiratory disease--exceeds that of individuals without mental health conditions (MHCs) (Carney and Jones, 2006; Crump et al., 2013). As a result, life expectancy is several decades shorter for people with BD than for those without MHCs (Colton and Manderscheid, 2006).

Despite these known disparities in tobacco smoking and its health consequences for people with BD, there is limited research on the treatment needs and outcomes of quit attempts for this group of smokers. This stands in stark contrast with the much broader literature on smoking cessation among those with psychotic disorders, unipolar depression, and anxiety (Ziedonis et al., 2008). Six of eight published studies focused on smoking cessation for people with BD have sample sizes of twenty or less (Brunette et al., 2018; Chengappa et al., 2014; Evins et al., 2014; Frye et al., 2013; Heffner et al., 2013; Heffner et al., 2015; Weinberger et al., 2008; Wu et al., 2012). Although epidemiologic data suggest that quit rates may be lower for smokers with BD than for smokers without MHCs (17% of ever smokers with BD report being former smokers vs. 43% of ever smokers without MHC) (Lasser et al., 2000), this has not been tested directly in a prospective cessation study. This leaves open the possibility that differential quit rates could be due to a lower likelihood of making a quit attempt, lower likelihood of being successful on any given quit attempt, or a combination of the two. Although depression has been linked to lower utilization of some cessation treatments, including both behavioral (Zeng et al., 2015) and pharmacotherapy (Hood et al., 2013) interventions, there has not been any examination of whether smokers with BD demonstrate lower utilization of cessation interventions such that the likelihood of quitting is reduced. Finally, there have been no comparisons of how treatment-seeking smokers with BD may differ from smokers without MHCs or smokers with other ADs (including unipolar depression and anxiety disorders) on demographic, smoking, and other characteristics relevant to cessation. These comparisons would provide important information about potential mechanisms underlying differential smoking prevalence in BD and whether smokers with BD may have different treatment needs. Such findings could inform the development of targeted treatments to improve quit rates for smokers with BD.

Addressing these gaps in the literature is a necessary first step toward decreasing the high prevalence of smoking among adults with BD. To begin addressing the gaps, we conducted post hoc analyses using data from a large, randomized controlled trial (n=2,637) comparing the effectiveness of two web-delivered treatments for smoking cessation (Bricker et al., 2018). The aims of the present study were (1) to determine to what extent treatment-seeking smokers with BD differed from smokers with other ADs and from those with no MHCs on demographic characteristics, smoking behavior, and the presence and severity of a range of mental health symptoms, and (2) to compare treatment utilization and long-term outcomes for smokers with BD to smokers with other ADs and no MHCs. We expected that the smokers with BD would differ from both smokers with other ADs and smokers with no MHCs on baseline characteristics and would demonstrate lower rates of treatment utilization and quitting than both comparison groups, regardless of which treatment they were assigned to receive.

2. Material and Methods

2.1. Participants

Participants (n=2,570) were a subsample of those enrolled in the web-delivered cessation intervention trial (n=2,637). Participants included in this analysis self-reported having BD (with or without comorbid anxiety disorder; n=221), other ADs (n=783) or no major MHC (n=1,566). In order to be eligible for the trial, participants had to: (1) be age 18 or older, (2) smoke at least 5 cigarettes per day for the last 12-months, (3) be motivated to quit in the next 30 days, (4) reside in the United States, (5) be able to read in English, (6) have Internet and email access, (7) have never used Smokefree.gov, the publicly available control condition, or the experimental web intervention via participation in a previous study by the research team, (8) not currently be enrolled in any cessation interventions, (9) not have another household member participating, and (10) be willing to be randomized and complete all study procedures. Targeted recruitment strategies were employed to achieve at least 25% representation of racial/ethnic minorities.

2.2. Procedures

A complete description of study procedures is provided in the report of the main outcome study (Bricker et al., 2018). Briefly, participants were recruited for the trial via Facebook, Google and Craigslist advertisements, an online survey panel, earned traditional media, family and friend referrals, and organic search engine results. Interested individuals were directed to a recruitment web page for further information about the study. Those who were interested completed a screening survey to establish eligibility. Those who were eligible then received an emailed link to enter the study website to provide informed consent, complete a baseline survey and contact form, and be randomized to one of two web-based cessation interventions based either on a novel Acceptance and Commitment Therapy (ACT) program or standard care (the National Cancer Institute’s Smokefree.gov website). Web interventions were available to participants for 12-months following randomization. All participants also received up to four text messages per day for 28-days to support quitting and use of the assigned intervention website, with the content of the messages differing by arm.

Outcome surveys were completed at 3, 6, and 12-months post-randomization. To maximize response rate, the study used a multimodal survey protocol beginning with web-based surveys and offering telephone, mailed, and postcard-only options sequentially as needed until a response was received or the protocol was completed, and the participant was classified as a non-responder. Compensation for completing each outcome survey was $25, with a $10 extra incentive for responding within 24 hours to the emailed survey link. Follow-up rates were 88% (2274/2570) at 3 months, 89% (2283/2570) at 6 months, and 88% (2249/2570) at 12-months.

2.3. Measures

2.3.1. Self-Reported Mental Health Conditions.

The baseline survey included an item developed by the North American Quitline Consortium (Consortium, 2012; Tedeschi et al., 2016) to assess self-reported mental health conditions (“Do you have any of the following mental health conditions?”). Response choices included anxiety disorder, depression disorder, bipolar disorder, schizophrenia, alcohol abuse, drug abuse, or none of the above. Taking into account diagnostic hierarchies and potential comorbidities, participants were first grouped according to whether they had bipolar disorder, regardless of whether they reported any co-occurring conditions. The remaining participants who did not report bipolar disorder were then classified as reporting other affective disorders (i.e., anxiety disorder or depression disorder) or reporting no mental health conditions. Participants who reported having only schizophrenia (1% n=17), alcohol abuse (5%; n=119) or drug abuse (3%; n=67) were not included in this subgroup analysis because of the low proportions of participants with these disorders and resultant concerns about the generalizability of findings.

2.3.2. Baseline Demographics, Smoking and Mental Health Symptoms.

The baseline survey included questions covering demographics such as age, gender, race and ethnicity, marital status, educational attainment, and sexual minority status. To assess the smoking history and current smoking-related characteristics, participants completed the Fagerström Test for Nicotine Dependence (Heatherton et al., 1991) as a measure of nicotine dependence severity. To assess the theory-based targets of the ACT intervention, participants completed the Commitment to Quitting Smoking Scale (CQSS) (Kahler et al., 2007) as a measure of commitment to achieving abstinence and the Avoidance and Inflexibility Scale (AIS) (Gifford et al., 2004) as a measure of smoking-specific experiential avoidance (i.e., efforts to reduce aversive internal states, such as craving, by smoking). The survey also included questions about the current use of e-cigarettes and number of quit attempts in the past 12-months. Validated mental health screeners administered at baseline covered depression [CES-D (Radloff 1977); positive screen is ≥16], generalized anxiety [GAD-7 (Spitzer et al., 2006); positive screen is ≥ 10], panic disorder [ANSQ (Stein et al., 1999)], post-traumatic stress disorder (PTSD) [PCL-6 (Lang and Stein, 2005); positive screen is ≥ 14], social anxiety [mini-SPIN (Connor et al., 2001); positive screen is ≥ 6], and at-risk drinking [AUDIT-C] (Babor et al., 1992). Heavy drinking on the AUDIT-C was defined as drinking 4 or more (for women) or 5 or more (for men) drinks on a typical drinking day (Sanchez-Craig et al., 1995). On the ANSQ, which is scored qualitatively rather than quantitatively, a positive screen for panic disorder required: (a) indication of a panic attack in the past month, and (b) at least one attack in the past month occurring in a situation in which they were not in danger or not the center of attention (Stein et al., 1999).

2.3.3. Cessation outcomes.

Cessation outcomes were calculated according to self-reported time since last cigarette (even a puff). As in the parent trial, the primary cessation outcome was 30-day point prevalence abstinence (PPA) at 12-months using complete-case analysis. To promote comparability with other trials, we also included as secondary outcome measures the complete-case 7-day and 30-day PPA at 3- and 6-month follow-ups as well as missing=smoking 7-day and 30-day PPA at 3-, 6-, and 12-month follow-ups. Self-reported cessation outcomes are recommended for large population-based cessation trials where there is no face-to-face contact, demand characteristics are minimal, and biochemical confirmation is not feasible due to sample size (SRNT Subcommittee on Biochemical Verification, 2002).

2.3.4. Treatment Utilization.

Server login data of recorded page views were used to calculate the number of logins to the assigned program over the 12-month treatment period. Page views separated by at least 15 minutes were recorded as separate logins to account for participants who leave the program open between sessions.

2.4. Statistical Analysis

Comparison of baseline characteristics was conducted using chi-square tests for categorical variables and analysis of variance (ANOVA) for continuous variables, with follow-up pairwise comparisons conducted using a p-value adjustment controlling for the false discovery rate (FDR) (Benjamini and Hochberg, 1995) for categorical variables and Tukey’s HSD for continuous variables. Adjusted logistic regression and negative binomial models were used to evaluate differences in cessation outcomes and treatment utilization, respectively, by diagnostic group. Models included terms for treatment group assignment and considered treatment by diagnostic group interactions. Stratification variables for the larger trial (gender, education, and smoking more than 1-pack per day) were also included. Interaction terms that were not significant were dropped from the final models. Baseline demographic and smoking variables were included in the models as covariates if they differed by diagnostic group and were associated with the outcome (i.e., cessation or treatment utilization). Due to the post hoc, exploratory nature of the analyses, we did not control for multiple testing with the exception of controlling for familywise error rate as described previously. All tests used a 2-sided significance level of p< .05. Analyses were conducted using R, v. 3.4.2 (R Core Team, 2017) and R package ‘MASS’(Venables and Ripley, 2002).

3. Results

3.1. Baseline Comparisons

As shown in Table 1, the baseline characteristics of the subgroup of smokers with BD were distinct in several ways from both the AD and the no MHC groups. Relative to both comparison groups, smokers with BD were younger and more likely to be unmarried, not working, and identify as lesbian, gay, or bisexual (LGB). They were also more likely to screen positive for current mental health conditions than both comparison groups, with the exception of a similar likelihood of panic disorder in BD and AD. On smoking-related characteristics, the smokers with BD reported more severe nicotine dependence than either comparison group despite reporting a similar prevalence of heavy smoking. Regarding friend and family smoking, they reported generally greater exposure to other smokers (i.e., close friends, adult smokers in home) than smokers with no MHCs, but they were similar to the AD group on living with a smoker—either a partner or other adult in the household. On one of the psychological predictors of quitting—acceptance of internal smoking cues—the smokers with BD scored lower than the no MHC group but were not significantly different from the AD group.

Table 1.

Comparison of demographic, mental health, and smoking characteristics by diagnostic group.

Variable	Total (n=1,787)	No MHC (n=1,566)	BD (n=221)	AD (n=783)	Overall p-value (Baseline)	Bipolar vs No MHC	Bipolar vs AD
Demographics
Age	46.2 (13.3)	47.0 (13.4)	40.3 (12.6)	46.0 (13.0)	<0.0001	<0.0001	<0.0001
Male	517 (20%)	362 (23%)	37 (17%)	118 (15%)	<0.0001	0.062	0.616
Caucasian	1865 (73%)	1115 (71%)	144 (65%)	606 (77%)	0.0002	0.078	0.001
Hispanic	217 (8%)	130 (8%)	25 (11%)	62 (8%)	0.263
Married	967 (38%)	628 (40%)	61 (28%)	278 (36%)	0.001	0.001	0.035
Working /	1342 (52%)	926 (59%)	78 (35%)	338 (43%)	<0.0001	<0.0001	0.043
HS or less education	716 (28%)	415 (27%)	75 (34%)	226 (29%)	0.053
LGB	244 (9%)	122 (8%)	47 (21%)	75 (10%)	<0.0001	<0.0001	<0.0001
Mental Health
Current depression symptoms (CES-D ≥ 16)	1435 (56%), N=2556	666 (43%), n=1555	190 (86%), n=220	579 (74%), n=781	<0.0001	<0.0001	<0.001
Current anxiety symptoms (GAD ≥ 10)	889 (35%), n=2557	373 (24%), n=1559	145 (66%), n=219	371 (48%), n=779	<0.0001	<0.0001	<0.0001
Current panic disorder symptoms (positive screen on ANSQ)	1125 (49%), n=2307	507 (37%), n=1375	144 (71%), n=204	474 (65%), n=728	<0.0001	<0.0001	0.168
Current PTSD symptoms (PCL-6 ≥ 14)	1347 (53%), n=2561	620 (40%), n=1559	183 (83%), n=220	544 (70%), n=782	<0.0001	<0.0001	<0.001
Current social anxiety symptoms (mini-SPIN ≥ 6)	781 (30%), n=2563	342 (22%), n=1561	111 (50%), n=221	328 (42%), n=781	<0.0001	<0.0001	0.036
Heavy drinker	263 (10%), n=2511	153 (10%), n=1526	23 (11%), n=215	87 (11%), n=770	0.639
Smoking Behavior
Nicotine dependence severity (FTND score)	5.6 (2.2)	5.5 (2.2)	6.3 (2.1)	5.7 (2.2)	<0.0001	<0.0001	0.001
Smokes more than one pack per day	851 (33%)	505 (32%)	84 (38%)	262 (33%)	0.227
Used e-cigarettes at least once in past month	888 (35%)	523 (33%)	82 (37%)	283 (36%)	0.296
Quit attempts in past 12 months	1.7 (5.0), n=2450	1.7 (5.3), n=1490	1.3 (3.0), n=212	1.7 (5.0), n=748	0.609
Commitment to quitting (CQSS score)	4.0 (0.8), n=2562	4.1 (0.7), n=1561	4.0 (0.9), n=219	3.9 (0.8), n=782	<0.0001	0.380	0.280
Acceptance of physical triggers (AIS score)	2.9 (0.5), n=2536	3.0 (0.5), n=1541	2.8 (0.4), n=218	2.9 (0.5), n=777	<0.0001	<0.0001	0.088
Friend & Partner Smoking
Close friends who smoke	2.2 (1.6)	2.0 (1.6)	2.7 (1.8)	2.3 (1.7)	<0.0001	<0.0001	0.010
Number of adults in home who smoke	1.5 (0.8)	1.4 (0.8)	1.6 (0.9)	1.5 (0.8)	0.022	0.018	0.114
Living with partner who smokes	757 (29%)	464 (30%)	75 (34%)	218 (28%)	0.208

Note: Values in table are M (SD) or n (%). P-values are based on chi-square tests for categorical variables and ANOVA for continuous variables. For pairwise comparisons between no MHC, BD, and other AD groups, we used Tukey’s HSD for continuous variables and p-value adjustment controlling the false discovery rate for categorical variables. MHC=mental health condition; BD=bipolar disorder; AD=(other) affective disorder; HS=high school; LGB=lesbian, gay, or bisexual; PTSD=posttraumatic stress disorder; FTND=Fagerström Test for Nicotine Dependence; CQSS=Commitment to Quitting Smoking Scale; AIS=Avoidance and Inflexibility Scale.

3.2. Comparisons of Treatment Outcomes.

On the primary outcome of complete-case 30-day PPA at 12-months, there was no significant treatment by diagnostic group interaction (p=.846), suggesting that the effects of the mental health diagnosis on smoking outcomes did not differ based on treatment group assignment. Combining across treatment arms, quit rates for the BD group (20%) were significantly lower than the no MHC group (29%; p=.007, OR=0.57, 95% CI=0.38-0-83), but no different than the AD group (20%; p=.467, OR=0.85, 95% CI=0.55–1.29). This finding remained when using the missing=smoking assumption for 30-day PPA as well as with 7-day PPA using both complete-case and missing=smoking methods (see results in Table 2 and Table S1¹).

Table 2.

Comparison of 30-day point prevalence smoking abstinence and web site utilization by diagnostic group.

	Total (n=2,570)	No MHC (n=1,566)	BD (n=221)	AD (n=783)	FDR-adjusted overall p-value	BD vs AD OR (95% CI)	BD vs AD p-value	BD vs No MHC OR (95% CI)	BD vs No MHC p-value	Covariates (in addition to gender, education, smokes > 1 pack)
Cessation outcome
30-day PPA at 3-month follow-up, complete case, n (%)	313 (14%), n=2274	197 (14%), n=1388	23 (12%), n=189	93 (13%), n=697	0.588	--	--	--	--	Caucasian race, AA race, employed, CQSS
30-day PPA at 3-month follow-up, missing=smoking, n (%)	313 (12%)	197 (13%)	23 (10%)	93 (12%)	0.451	--	--	--	--	Caucasian race, AA race, employed, CQSS
30-day PPA at 6-month follow-up, complete case, n (%)	453 (20%), n=2283	304 (22%), n=1390	26 (14%), n=193	123 (18%), n=700	0.043	0.67 (0.41, 1.05)	0.090	0.56 (0.35, 0.87)	0.012	Caucasian race, AA race, CQSS, # of smokers in home
30-day PPA at 6-month follow-up, missing=smoking, n (%)	453 (18%)	304 (19%)	26 (12%)	123 (16%)	0.033	0.64 (0.39, 1.00)	0.060	0.55 (0.35, 0.84)	0.008	Caucasian race, AA race, CQSS
30-day PPA at 12-month follow-up, complete case, n (%)	569 (25%), n=2249	393 (29%), n=1370	36 (20%), n=184	140 (20%), n=695	<0.001	0.85 (0.55, 1.29)	0.467	0.58 (0.38, 0.85)	0.007	Caucasian race, AA race, CQSS, # of smokers in home
30-day PPA at 12-month follow-up, missing=smoking, n (%)	569 (22%)	393 (25%)	36 (16%)	140 (18%)	<0.001	0.81 (0.53, 1.21)	0.318	0.57 (0.38, 0.83)	0.005	Caucasian race, AA race, CQSS, # of smokers in home, FTND
Utilization
Number of logins by 12- month follow-up, mean (sd)										AA race, age
For ACT arm	9.2 (30.3), n=1279	9.6 (27.8), n=800	13.5 (68.1), n=107	7.1 (12.7), n=372	0.003	1.55 (117, 2.06)a	0.001	1.20 (0.93, 1.56)a	0.157
ACT Median (IQR)	3 (2–7)	3 (2–7)	2 (1–5)	3 (2–7)
For CBT arm	5.2 (12.0), n=1291	5.3 (14.0), n=766	5.2 (6.4), n=114	5.0 (8.6), n=411	0.666	--	--	--	--
CBT Median (IQR)	3 (1–5)	3 (1–5)	3 (2–7)	3 (1–5)
Time per login at the 12- month follow-up, minutes, mean (sd)	5.6 (6.1), n=2570	5.6 (6.0), n=1558	5.8 (5.7), n=220	5.6 (6.5), n=780	0.805					Age, Caucasian race, employed, # of friends who smoke, # of adults in home who smoke
Median (IQR)	3.9 (1.7–7.3)	3.8 (1.7–7.4)	4.1 (2.0–8.0)	3.9 (1.7–7.1)
Number of days logged in by 12-month follow-up, mean (sd)										Age, Caucasian race, AA race, married, employed, FTND, # of friends who smoke, AIS score
For ACT arm	7.3 (22.0), n=1279	8.0 (24.0), n=800	7.9 (33.8), n=107	5.7 (9.4), n=372	0.002	1.06 (0.82, 1.40)	0.635	0.79 (0.62, 1.02)	0.064
ACT Median (IQR)	3 (1–6)	3 (2–6)	2 (1–4)	3 (1–6)
For CBT arm	4.2 (7.1), n=1291	4.2 (8.0), n=766	4.6 (5.3), n=114	4.2 (5.6), n=411	0.707
CBT Median (IQR)	2 (1–5)	2 (1–5)	3 (1–6)	2 (1–4)

Note: MHC=mental health condition; BD=bipolar disorder; AD=(other) affective disorder; FDR=false discovery rate; PPA=point prevalence abstinence; ACT=acceptance and commitment therapy; CBT=cognitive-behavioral therapy; IQR=interquartile range AA=African American; CQSS=Commitment to Quitting Smoking Scale; FTND=Fagerström Test for Nicotine Dependence, AIS=Avoidance and Inflexibility Scale.

^aValues are incidence rate ratios (IRR) from a negative binomial model.

Cessation outcomes at the secondary time points of 3 and 6-months showed disparate results. Although the quit rates were descriptively lowest in the BD group across all outcome measures at the 3- and 6-month time points, there were no significant differences by the diagnostic group on any of the cessation outcomes at 3-months. At 6-months, the 30-day PPA rates were significantly lower in the BD group than the no MHC group in both complete-case (p=.012, OR=0.56, 95% CI=0.35–0.87) and missing=smoking (p=.008, OR=0.55, 95% CI=0.35–0.84) analyses, but they did not differ significantly from the AD group.

3.3. Comparisons of Treatment Utilization.

In the model evaluating diagnostic group as a predictor of the number of logins by 12-month follow-up, there was a significant interaction of diagnostic group by treatment assignment (p=.010). In the ACT arm, the BD group (M=13.5, SD=68.1) had a higher number of logins than the AD group (M=7.1, SD=12.7; IRR=1.55 (1.17, 2.06); p=.001), but did not differ significantly from the no MHC group (M=9.6, SD=27.8; IRR=1.20 (0.93, 1.56); p=.157). There were no differences in the number of logins to the Smokefree.gov website by diagnostic group (p=.666). On the secondary measures of treatment utilization—minutes per login and number of days logged in—the findings differed from the primary endpoint of the number of logins (see Table 2). On minutes per login, the differences among the three diagnostic groups were not significant (p=.805). On the number of days logged in, there was a significant interaction of diagnostic group by treatment arm (p=0.008). In the ACT arm, the AD group had a lower number of days logged in than the no MHC group (p<.001, IRR=0.74, 95% CI=0.64–0.87), but the BD group did not differ significantly from either the no MHC group (p=.064, IRR=0.79, 95% CI=0.62–1.02) or the AD group (p=.635, IRR=1.06, 95% CI=0.82–1.40). There were no differences in the number of days logged in to the Smokefree.gov website by diagnostic group (p=.655).

3.4. Post-Hoc Analyses: Potential Contributors to Similarities Between BD and AD Smokers.

Due to the unexpected similarity in quit rates in the BD and AD groups, both of which were lower than the no MHC group, we conducted several post-hoc tests to determine whether similarity in any specific baseline characteristics may have contributed to these findings. Three baseline variables on which smokers with BD were different from smokers with no MHC but similar to AD smokers were evaluated as potential mediators of the relationship between mental health condition and smoking abstinence: (1) positive screen for panic disorder, (2) experiential avoidance, as measured by the AIS, and (3) number of smokers living in the home. Using Baron and Kenny’s (Baron and Kenny, 1986) steps to preliminarily assess the conditions for mediation, one set of models tested the hypothesis that each of the three mediators indirectly explained the relationship between BD (vs. no MHC) and the primary outcome of 30-day PPA at 12-months. Another set of three models tested the hypothesis that each of the three mediators indirectly explained the relationship between AD (vs. no MHC) and the primary outcome of 30-day PPA at 12-months. None of the mediation analyses demonstrated statistically significant relationships among all three variables (predictor, mediator, and outcome) (see Supplementary Table S2²), providing no evidence that experiential avoidance, number of smokers in the home, or screening positive for panic disorder explain the similarly lower cessation rates for smokers with BD and AD relative to no MHC.

4. Discussion

In this study, we compared treatment-seeking smokers with BD to smokers with other ADs and to those with no MHCs on demographic, smoking, and current mental health symptoms, as well as on treatment utilization and cessation outcomes. Smokers with BD were distinct from both AD and no MHC smokers on the majority of baseline characteristics. At 12-months, quit rates for smokers with BD were significantly lower than for smokers without MHCs, but no different than those with ADs. The number of logins was higher for smokers with BD than AD in the ACT arm only, but this effect was not observed on the secondary measures of treatment utilization.

The numerous demographic and smoking characteristics that distinguish smokers with BD from the two comparison groups (e.g., greater severity of nicotine dependence, proximity to other smokers in the home or social context) could be expected to contribute to lower quit rates (Fiore et al., 2008) and potentially to lower treatment utilization (Christensen et al., 2009; Zeng et al., 2015) based on previous literature. Consistent with that prediction, the BD subgroup did have significantly lower quit rates than the no MHC group on the primary outcome of 30-day PPA at 12-months post-randomization (as well as all of the other cessation outcomes at 12-months). However, even with statistical control for baseline demographic and smoking characteristics that differed by diagnostic group and were associated with outcome, there was still a significant difference between smokers with BD and those with no MHC on cessation outcomes, indicating that the association between BD and cessation outcomes is independent of baseline differences in demographics and smoking. This finding raises the question of what other factors might be responsible for differences in cessation between smokers with BD and those without MHCs. Potential explanations for greater difficulty quitting among smokers with BD include more severe nicotine withdrawal during the quit attempt (Heffner et al., 2011); fear that quitting may jeopardize mental health (Prochaska et al., 2011); and social impediments to quitting, including high exposure to smokers, being encouraged to smoke either directly or indirectly, and feeling shamed or judged by others (Heffner et al., 2011; Heffner et al., 2018). A combination of cessation pharmacotherapy (to address their more severe nicotine withdrawal) as well as targeted behavioral support (to address problematic beliefs about quitting and social challenges such as high exposure to other smokers) may be needed to address these barriers and improve quit rates among smokers with BD (Heffner et al., 2011).

Comparisons between smokers with BD and other ADs yielded an unexpected finding: Baseline differences between BD and AD smokers did not translate into a difference in cessation outcomes. We hypothesized that this might relate to their similarly high (relative to the no MHC group): (1) prevalence of screening positive for current panic disorder symptoms, (2) scores on the measure of experiential avoidance (AIS), and (3) number of smokers in the home, all of which have been associated with cessation in previous studies (Charlotta et al., 2005; Farris et al., 2016; Farris et al., 2015; Piper et al., 2011). However, our post-hoc analyses to test these hypotheses suggested that these similarities did not represent common pathways to cessation difficulty, as none of the mediation pathways were fully supported in either the BD or AD group. It remains unclear, then, why the substantial differences observed at baseline between BD and AD did not translate into differential cessation rates. It is possible that these findings may be attributable to the unique characteristics of the study sample (e.g., predominantly women, who are more susceptible to affect-related cessation difficulties (Weinberger et al., 2017)) or to unmeasured variables (e.g., self-efficacy for quitting (Castro et al., 2014)) that are strong predictors of cessation and may be similar in BD and other ADs.

It is worth noting that differences in cessation outcomes between smokers with BD and no MHC were present at long-term follow-up, but not at the earliest follow-up of 3 months post-randomization. Although one might hypothesize that this effect would be attributable to higher relapse rates among the BD participants, the overall pattern of abstinence rates in all groups generally showed an increase in the proportion of successful quitters over time rather than a decrease, suggesting that relapse is not the primary driver of the emergence of differences at later follow-up points. Instead, it appears that the increased abstinence rates that occurred over time were more substantial in the no MHC group than in the BD group.

The primary limitation of the present study is that mental health conditions were self-reported. Although prior work suggests that a majority (~75%) of individuals who self-report affective disorders meet diagnostic criteria (Sanchez-Villegas et al., 2008), it is likely that some proportion of participants in the study who self-reported mental health disorders would not meet diagnostic criteria and that some of those who did not self-report a disorder would, in fact, meet the criteria. At the same time, the 8% (221/2637) prevalence of self-reported bipolar disorder among smokers in this sample is consistent with national-level data indicating that there are an estimated 3.5 million adults with bipolar disorder in the US who smoke (Lasser et al., 2000; Merikangas et al., 2007; US Census Bureau, 2012), which is approximately 9% of the current adult smokers in the US (US Department of Health and Human Services, 2014). Additionally, there was no biochemical verification of smoking abstinence. Generally, biochemical verification of smoking abstinence is considered unnecessary in population-level interventions with no face-to-face contact (SRNT Subcommittee on Biochemical Verification, 2002). However, our previous work demonstrated greater discrepancy between self-reported and cotinine-confirmed smoking abstinence among individuals with depressive symptoms (Heffner et al., 2017), indicating that self-report may overestimate abstinence among smokers with affective disorders. If this were the case, the differences in cessation outcomes between smokers with no MHC and those with BD or other ADs observed in this study would underestimate the true differences. Finally, given the exploratory nature of the study, we used a less conservative method of controlling for Type I error stemming from multiple testing (i.e., control for familywise Type I error via the false discovery rate, or FDR), and our conclusions are best characterized as preliminary.

4.1. Conclusions.

These analyses are the first to demonstrate that smokers with BD have greater difficulty quitting than smokers without MHCs in a prospective study with 12-month follow-up. This is also the first demonstration with any digital intervention for cessation that smokers with BD and those with other ADs have similar quit rates, despite smokers with BD differing from smokers with other ADs on a number of different demographics, smoking, and mental health characteristics at baseline. The number of participants with BD in this study is greater than any previously published smoking cessation study, which is a result of the large number of participants included in the parent trial (n=2,637) and the minimal exclusion criteria. Overall, this work provides novel insights into the cessation challenges of smokers with BD as well as those of smokers with other ADs. It also suggests that promising quit rates (i.e., 20% quit rates at one year) can be achieved in both smokers with BD and other ADs using web-based interventions, even if those rates are lower than for smokers without MHCs.

Highlights.

• Smokers with bipolar disorder differed on the majority of baseline characteristics

• Smokers with bipolar disorder and other affective disorders had lower quit rates

• Quit rates for web-based treatment are promising overall across diagnostic groups