A comprehensive description of study methods can be found in the main outcomes paper [21]. In brief, participants were 2512 of the 2637 adult smokers from all 50 states enrolled in the parent RCT, which compared the effectiveness of two Web-based smoking cessation programs [NCT01812278]. For the purposes of these analyses, the analytic sample was limited to participants who had complete data on at least one screening measure assessing affective disorder symptomology (see below for more details on these measures). To be eligible for the parent trial, participants had to (1) be 18 years of age or older; (2) smoke 5 or more cigarettes per day for the last year; (3) express a desire to quit smoking within 30 days; (4) have access to high-speed internet, including access to email; (5) not be participating in other smoking interventions or treatments; (6) never have used the control intervention, Smokefree.gov; (7) never have participated in one of our previous smoking studies; (8) not have another member in their household participating in the study; (9) express a willingness to be randomized to treatment, complete 3 follow-up surveys, and provide contact information for themselves and two relatives; (10) reside in the United States; and (11) be able to read in English. After completing the informed consent and baseline survey online, participants were randomized to one of two Web-based programs based either on acceptance and commitment therapy (ACT) or standard care (the National Cancer Institute’s Smokefree.gov website). The ACT-based website presented content in a sequential manner through four phases and prompted users to track their smoking each day. The Smokefree.gov website had three main sections and users could navigate through all pages of the website freely with no restrictions on the order in which content had to be viewed. Participants had access to their assigned interventions for 12 months and received up to 4 messages per day for 28 days (via text or email) designed to encourage website use. Outcome surveys were completed at 3, 6, and 12 months postrandomization. The 12-month data retention rate was 88%, and neither treatment arm [21] nor mental health status predicted retention [22]. Further details on the recruitment and retention methods can be found elsewhere [22].

For descriptive purposes, baseline characteristic variables of participants with ADS (each diagnostic group considered separately) were examined and compared against the no AD group using chi-square tests for categorical variables and t tests for continuous variables. Logistic regression models were used to evaluate differences in treatment satisfaction and cessation outcomes between each diagnostic group and the no AD group. Negative binomial models were used to evaluate differences in website use between groups. Models included variables used in stratified randomization for the main trial (ie, gender, education, and smoking more than 20 cigarettes per day), treatment arm, and treatment arm by diagnostic group interactions were considered. Nonsignificant interaction terms were dropped from final models. Analyses were completed using R version 3.4.2 (The R Foundation) and R package MASS [33], and all statistical tests were two-sided with alpha=.05.

Of the 2512 participants who provided sufficient data on the screening measures to establish whether the screen was positive or negative, the majority (1938/2512, 77.15%) screened positive for at least one affective disorder at baseline. Of those, the majority screened positive for depression (1470/1938, 75.85%), followed by PTSD (1383/1938, 71.36%), PD (1145/1938, 59.08%), GAD (903/1938, 46.59%), and SAD (797/1938, 41.25%). Additionally, as can be seen in Table 1, most of these participants screened positive for more than one affective disorder. Overall, 15.18% (355/2339) screened positive for one affective disorder, 13.59% (318/2339) screened positive for two, 16.76% (392/2339) screened positive for three, 15.01% (351/2339) screened positive for four, and 14.92% (349/2339) screened positive for all five affective disorders assessed in the study, suggesting high levels of elevated affective symptomology overall.

The baseline characteristics of all groups can be found in Table 2. Participants with ADS differed from those without ADS in a number of ways on demographic and smoking variables. Compared to smokers without ADS, all ADS subgroups were significantly younger, less educated, and more likely to identify as a sexual minority (ie, lesbian, gay, or bisexual). Some groups (GAD, SAD) comprised a greater proportion of women, and all ADS subgroups except the PD subgroup were less likely to be married. No differences were found between groups regarding race or ethnicity. Relative to the group without ADS, all ADS subgroups had significantly greater levels of nicotine dependence (all P<.001), avoidance of physical sensations that cue smoking (eg, urges) (all P<.001), and lower commitment to quitting scores (all P<.001).

As can be seen in Table 3, some differences emerged regarding subjective treatment acceptability ratings and objective website use data between those with and without ADS. Contrary to our predictions, relative to those who did not screen positive for any affective disorder, all ADS subgroups reported significantly greater satisfaction with their assigned website. Although differences in ratings of the other two indices of treatment acceptability (ie, usefulness of assigned website, whether or not they would recommend the website to a friend) were not significantly different, they were in the same direction such that those with ADS reported greater acceptability. No treatment by symptom group interaction was significant.

With regard to objective website use data, our hypotheses were partially supported. Relative to the no ADS group, the GAD group logged in to their assigned website significantly fewer times (P<.001). For all other symptom groups, there were significant treatment by symptom group interactions such that each ADS group logged in to the Smokefree.gov website significantly fewer times than the no ADS group. No significant differences in number of log-ins between groups were found in the ACT-based arm. All symptom subgroups spent significantly less time logged in to their assigned website compared with the no ADS group, and no treatment by symptom group interaction was significant.

Cessation rates for each ADS group are shown in Table 4. The 12-month complete-case quit rate among participants without ADS was descriptively higher than all other groups (153/520, 29.4%). The quit rate was statistically lower for some but not all ADS subgroups. Specifically, participants in the depression (306/1267, 24.15%; odds ratio [OR] 0.78, 95% CI 0.62-0.98; P=.03), PTSD (294/1215, 24.19%; OR 0.78, 95% CI 0.62-0.98; P=.03), PD (229/1003, 22.83%; OR 0.73, 95% CI 0.57-0.92; P=.009), and two or more ADS (323/1332, 24.25%; OR 0.78, 95% CI 0.62-0.98; P=.03) groups exhibited statistically lower quit rates of 20% to 25% relative magnitude than the no ADS group; those who screened positive for GAD (195/792, 24.6%; P=.08) and SAD (176/696, 25.3%; P=.13) did not. None of the treatment arm by symptom group interactions were significant. Models examining the secondary cessation outcome where missing values were imputed as smoking resulted in a similar pattern of results, with the exception that the difference for the GAD subgroup became significant.

To develop a better understanding of the factors contributing to lower quit rates among the symptom subgroups who were less likely to quit, post hoc analyses examined (1) whether each ADS group uniquely predicts abstinence after including baseline variables that differed between groups and were associated with cessation in the models and (2) whether website use (ie, log-ins, total time spent on website) mediated the relationship between ADS subgroup and abstinence.

As shown in Table 5, after adding the additional baseline variables to logistic regression models for each ADS subgroup, only screening positive for PD remained significantly negatively associated with cessation after controlling for baseline levels of nicotine dependence and commitment to quitting (OR 0.78, 95% CI 0.61-0.99; P=.04) in the analyses in which missing values were imputed as smoking. All other differences became nonsignificant and may represent variables that are part of a causal pathway between affective disorders and difficulty quitting. Mediation models tested the hypothesis that website use might indirectly explain the relationship between depression, PTSD, PD, and 2+ ADS subgroups (versus the no ADS group) and the primary cessation outcome [34]. However, none of the estimated indirect effects were significant, suggesting that fewer log-ins and less time spent on the websites did not help explain the observed lower quit rates for these ADS subgroups in this study (data not shown).

To inform future treatment development efforts for smokers with ADS, we conducted the first study comparing treatment acceptability and long-term smoking cessation outcomes between smokers who screened positive for five common affective disorders (depression, PTSD, PD, GAD, SAD, as well as those who screened positive for ≥2 diagnostic categories). All ADS subgroups differed from the no ADS group (those who did not screen positive for any affective disorder) on most demographic and smoking characteristics assessed at baseline, similar to other reports [19,35,36].

All ADS subgroups had higher subjective acceptability ratings of their assigned website compared with the no ADS group. It may be that, relative to smokers without affective disorder symptomology, smokers with ADS see greater potential for Web-based smoking interventions or find the Web-based format particularly helpful and convenient, leading to greater subjective acceptability among those groups. Although unanticipated, this finding is important and demonstrates that smokers who screen positive for affective disorders are open to this treatment modality and find them helpful. However, the higher subjective ratings did not translate into greater use of the websites as one might expect. Although most ADS subgroups logged in significantly less to the Smokefree.gov website (with no differential log-in rates for the ACT-based website) relative to the no ADS group, all ADS subgroups spent significantly less time on both websites overall, which is on par with our hypotheses. The differential log-in rates for the websites across groups may mean that smokers with ADS engage more with Web-based cessation programs (ie, find them more acceptable) when they are grounded in ACT, presented in a stepwise fashion, or when they have features that prompt tracking.

Reasons for the discrepancy in the findings between the subjective and objective indices of acceptability are unclear. On the one hand, it may be that the cognitive and affective symptoms experienced by smokers with ADS impede them from using the websites as much as individuals without ADS. Alternatively, given that the websites in the parent trial were designed for the general population of smokers, it may also be that smokers with ADS found them less personally relevant, which could have contributed to them having lower levels of engagement [37-40].

The quit rates at 12-month follow-up for all symptom subgroups were promising (20% to 25%). However, in the primary analyses most symptom subgroups (those screening positive for depression, PTSD, PD, or ≥2 affective disorders) had significantly lower quit rates (23% to 24%) relative to the no ADS group (29%). Thus, there is some evidence of the possibility of differential outcomes being disorder-specific. However, all subgroups had descriptively lower quit rates than the no ADS group, suggesting that there is room to improve quit rates for all symptom subgroups and doing so would likely be significant at the population level considering the large proportion of smokers who screen positive for ADS.

Results from post hoc analyses suggest that the lower cessation rates among most groups may be influenced in part by their having greater levels of nicotine dependence and lower commitment to resist smoking in difficult situations at baseline. While greater levels of nicotine dependence among individuals with ADS is a robust finding in the literature [6,19,41], this is the first study to our knowledge demonstrating that these individuals also have significantly lower levels of commitment to resist smoking during difficult situations, which may be an important target to include in smoking interventions designed for these groups. However, unlike in previous studies of Web-based smoking interventions where greater use was associated with better cessation outcomes [42-45], the lower rates of cessation in the affective disorder subgroups were not mediated by website use. It may be that higher engagement only mediates cessation when the intervention adequately addresses the needs of the users. If the intervention does not address the user’s needs or is not perceived as relevant to the user, it follows that the user wouldn’t be as engaged and that engagement, per se, wouldn’t mediate cessation. Taken together, this suggests that increasing use alone among these groups may not be sufficient for improving cessation outcomes and that targeted interventions should also focus on addressing the unique, modifiable needs of smokers with affective conditions (eg, commitment to resist smoking in difficult situations) as a means of improving outcomes of these groups. Moreover, differences in nonmodifiable factors (eg, demographics) between these groups should also be considered when developing targeted interventions to increase their relevance for these groups. Future research is needed to determine if targeted Web-based interventions are associated with improved outcomes for these groups and if engagement mediates cessation outcomes in these cases.

There are numerous strengths to this study including a large, geographically diverse sample of adult smokers living in the United States, the use of validated self-report instruments to assess mental health conditions, and assessing long-term cessation outcomes at 12-months with a high rate of outcome data retention. Despite these strengths, the limitations of this study should be considered when interpreting the results. The primary limitation is the exploratory nature of the study as secondary analyses embedded in a larger clinical trial [21]; thus, our conclusions should be considered preliminary. A second limitation to the study is the use of self-report screening instruments to assess ADS. Individuals who screened positive for affective disorders may not all have the disorders they screened positive for and instead have elevated, subthreshold symptoms of the disorders. In line with previous research indicating that even individuals with subthreshold symptoms of affective disorders are at risk for continued smoking [3,46] and that screening instruments for affective disorders can be used to predict cessation [2], this study highlights that the screening instruments used in this study can be useful in identifying smokers who are less likely to quit even without a formal diagnostic assessment. Finally, we did not biochemically verify self-reported smoking abstinence at the 12-month follow-up. However, biochemical validation of abstinence is considered unnecessary for population-level intervention studies that otherwise have no face-to-face contact with participants because requiring this can bias study results [32].

This is the first study to compare treatment acceptability and long-term quit rates in response to Web-based smoking interventions between smokers who do versus do not screen positive for five common affective disorders. Findings suggest that while Web-based interventions are appealing to these groups, most individuals with ADS used their assigned website less and were less likely to quit smoking than their counterparts in response to the two interventions, each designed for the general population of smokers. Overall, these results support the need for developing targeted interventions for smokers with affective disorders and elevated affective symptomology.