Efficacy and safety of pharmacotherapies for smoking cessation in anxiety disorders: Subgroup analysis of the randomized, active‐ and placebo‐controlled EAGLES trial

Abstract Background Smoking rates are high in adults with anxiety disorders (ADs), yet little is known about the safety and efficacy of smoking‐cessation pharmacotherapies in this group. Methods Post hoc analyses in 712 smokers with AD (posttraumatic stress disorder [PTSD], n = 192; generalized anxiety disorder [GAD], n = 243; panic disorder [PD], n = 277) and in a nonpsychiatric cohort (NPC; n = 4,028). Participants were randomly assigned to varenicline, bupropion, nicotine‐replacement therapy (NRT), or placebo plus weekly smoking‐cessation counseling for 12 weeks, with 12 weeks follow‐up. General linear models were used to test the effects of treatment group, cohort, and their interaction on neuropsychiatric adverse events (NPSAEs), and continuous abstinence weeks 9–12 (treatment) and 9–24 (follow‐up). Results NPSAE incidence for PTSD (6.9%), GAD (5.4%), and PD (6.2%) was higher versus NPC (2.1%), regardless of treatment. Across all treatments, smokers with PTSD (odds ratio [OR] = 0.58), GAD (OR = 0.72), and PD (OR = 0.53) had lower continuous abstinence rates weeks 9–12 (CAR9–12) versus NPC. Varenicline demonstrated superior efficacy to placebo in smokers with GAD and PD, respectively (OR = 4.53; 95% confidence interval [CI] = 1.20–17.10; and OR = 8.49; 95% CI = 1.57–45.78); NRT was superior to placebo in smokers with PD (OR = 7.42; 95% CI = 1.37–40.35). While there was no statistically significant effect of any treatment on CAR9–12 for smokers with PTSD, varenicline improved 7‐day point prevalence abstinence at end of treatment in this subcohort. Conclusion Individuals with ADs were more likely than those without psychiatric illness to experience moderate to severe NPSAEs during smoking‐cessation attempts, regardless of treatment. While the study was not powered to evaluate abstinence outcomes with these subgroups of smokers with ADs, varenicline provided significant benefit for cessation in those with GAD and PD, while NRT provided significant benefit for those with PD.


| INTRODUCTION
Individuals with anxiety disorders (ADs) smoke tobacco at rates 2-3 fold higher than those without mental health conditions (Cougle, Zvolensky, Fitch, & Sachs-Ericsson, 2010;Lasser et al., 2000;McCabe et al., 2004). Like other smokers with mental health conditions (Evins, Cather, & Laffer, 2015), smokers with AD typically smoke more heavily (Kelly, Jensen, & Sofuoglu, 2015a), are more severely nicotine-dependent (Okoli, Otachi, Manuel, & Woods, 2018), and experience earlier and more severe nicotine-withdrawal symptoms (Piper, Cook, Schlam, Jorenby, & Baker, 2011) than smokers without mental health conditions, making quitting smoking more challenging (Kelly et al., 2015b). Despite this high rate of co-occurrence and greater difficulty quitting, relatively few randomized controlled trials (RCTs) have evaluated the safety and efficacy of the front-line smoking cessation pharmacotherapies in smokers with AD. Moreover, most RCTs conducted to date of smoking cessation interventions for smokers with AD included no placebo control or comparison group of smokers without mental health conditions, focused on one AD condition at a time, and tested only one of the US Food and Drug Administration (FDA)-approved medications in that AD subcohort.
Thus, there are insufficient data to draw comparisons across different diagnostic subgroups of smokers with AD, control smokers, and smoking cessation medications.
A notable exception was the retrospective analysis of a placebocontrolled RCT of nicotine-replacement therapy (NRT) and bupropion in a subcohort of smokers who endorsed prior panic attacks (not necessarily meeting criteria for panic disorder [PD]), social phobia, or generalized anxiety disorder (GAD; Piper et al., 2011). Smokers with AD in this study had lower abstinence rates than smokers without mental health conditions and appeared to derive no added benefit from bupropion, NRT, or their combination over placebo plus counseling. While there was some specificity between the subtype of ADs, the results were more similar than different. To our knowledge, no placebo-controlled RCTs have examined the efficacy of varenicline in smokers with ADs, most likely reflecting a reluctance to prescribe this effective medication to individuals with ADs because of concerns raised about its neuropsychiatric safety (Anthenelli et al., 2016). For example, a retrospective chart review of 78 smokers with posttraumatic stress disorder (PTSD) found an increased number of mental health encounters while subjects were taking varenicline compared with a pretreatment baseline period (Campbell & Anderson, 2010) examining varenicline in the largest randomized trial to date will provide critical safety information to guide policies and practices.
Given the paucity of safety and efficacy data from methodologically rigorous, placebo-controlled RCTs of pharmacotherapeutic cessation aids, clinicians have had little guidance on how to advise patients with specific AD diagnoses about which cessation treatment might work best for them. It is also not clear whether cessation obstacles equally cut across this heterogeneous group of mental health conditions (i.e., PTSD, GAD, and PD with/without agoraphobia) that form the broader AD category.
The multinational EAGLES (Evaluating Adverse Events in a Global Smoking Cessation Study; Anthenelli et al., 2016), which compared the non-nicotine medications varenicline and bupropion with an active comparator (transdermal nicotine patch, NRT) and placebo in smokers with/without mental health conditions, provides a unique opportunity to address these knowledge gaps. With approximately 20% (n = 787) of the EAGLES psychiatric cohort having a primary AD diagnosis, it is the largest sample of smokers with AD ever enrolled in a placebo-and active-controlled RCT for smoking cessation treatments, and the first to compare all three first-line smoking cessation aids head-to-head in such smokers. In this planned secondary analysis we compare the safety and efficacy of varenicline, bupropion, NRT, and placebo across AD subcohorts (PTSD, GAD, and PD), and compare clinical characteristics and rates of clinically significant neuropsychiatric adverse events (NPSAEs) and cessation in smokers with AD versus a cohort of smokers without psychiatric disorders.

| Study design
EAGLES was a multinational RCT (ClinicalTrials.gov identifier NCT01456936) conducted at the request of the FDA and European Medicines Agency. It was an active treatment-(transdermal nicotine patch, NRT) and placebo-controlled trial of varenicline and bupropion for 12 weeks, with a 12-week nontreatment follow-up assessment.
Institutional review boards or ethics committees at participating institutions approved the study procedures. The study adhered to the Declaration of Helsinki and the International Conference on Harmonization Good Clinical Practice Guidelines. All patients signed informed consent. The main outcome paper provides details of the study design (Anthenelli et al., 2016). Here we present a post hoc subgroup analysis of the EAGLES safety and efficacy outcomes identical to the parent trial.

| Participants
Eligible participants were adults aged 18-75 years who smoked an average of ≥10 cigarettes/day with an exhaled carbon monoxide (CO) >10 parts per million (ppm) at screening and wanted to stop smoking.  First, Gibbon, Spitzer, Williams, & Benjamin, 1997). Smokers with a primary diagnosis of obsessive-compulsive disorder (n = 27) and social phobia disorder (n = 48) were also enrolled in the parent trial but not included in the present analysis because of the small sample size ( Figure S1).
Participants could have a comorbid psychiatric diagnosis, provided it did not prevent the participant from complying with study requirements. If the comorbid diagnosis was a substance-use disorder, the participant was required to be in sustained full remission to be eligible. AD participants must have been considered psychiatrically stable as evidenced by no exacerbations of their psychiatric condition in the prior 6 months and if on pharmacotherapy, to have been on a stable dose for ≥3 months. They could not be considered at high risk of self-injurious or suicidal behavior. Nonpsychiatric cohort (NPC) participants were confirmed via SCID to have no major psychiatric disorders (n = 4,028).

| Randomization and treatment
In EAGLES, participants were stratified into one of four psychiatric subcohorts that included those individuals with AD, or the NPC, while also stratifying into one of four geographic regions. Subsequently, participants were randomized to study treatment in a 1:1:1:1 ratio using a computer-generated schedule. They were not randomized based on their specific AD diagnosis but stratified by whether they had any AD. Participants received either varenicline 1 mg twice daily, bupropion 150 mg twice daily, transdermal nicotine patch (NRT) 21 mg/day with taper, or placebo for 12 weeks. Treatment was administered in a double-blind, triple-dummy fashion (i.e., participants received one active and two placebo medications, or three placebo medications), and all assessments were blinded throughout the study. All participants received brief weekly smoking cessation counseling (e.g., managing withdrawal and cravings) that lasted <10 min and was based on US Agency for Healthcare Research and Quality Guidelines (Fiore et al., 2008). No routine mental health counseling was provided for mental health issues, including ADs.

| Safety
The primary safety outcome was the occurrence of any moderate to severe NPSAEs, which were operationally defined with guidance from the FDA (Anthenelli et al., 2016). Criteria for the primary neuropsychiatric endpoint required the four categories of NPSAEs (anxiety, depression, feeling abnormal, or hostility) most commonly associated with nicotine withdrawal to be rated as severe (i.e., significantly interfering with functioning). The other 12 categories (agitation, aggression, delusions, hallucinations, homicidal ideation, mania, panic, paranoia, psychosis, suicidal ideation, suicidal behavior, or completed suicide) were included in the primary composite safety endpoint if they were rated as either moderate (i.e., some interference with functioning) or severe in intensity.
Secondary safety endpoints included the incidence of each of the individual components and the subset of NPSAEs that were rated severe. We also examined general AEs of any intensity (mild, moderate, or severe) including those in the neuropsychiatric domain.

| Efficacy
The primary efficacy outcome measure was continuous abstinence rates weeks 9-12 (CAR9-12) and secondary efficacy was CARs weeks 9-24 (CAR9-24). Continuous abstinence was defined by selfreport of no smoking (even a puff), with an expired CO concentration of ≤10 ppm (Benowitz et al., 2002). Seven-day point prevalence abstinence (PPA) rates at week 12 (end of treatment) and week 24 (end of follow-up) were also evaluated as prespecified secondary endpoints.

| Assessments
Psychiatric diagnoses were evaluated using SCID-I and -II (First et al., 1997;First, Spitzer, Gibbon, & Williams, 2002). The severity of cigarette dependence was assessed at baseline with the Fagerström Test for Cigarette Dependence (FTCD; Fagerström, 2012). Anxiety and depression symptoms were assessed at baseline and serially over time using the Hospital Anxiety and Depression Scale (HADS; Zigmond & Snaith, 1983). The Buss-Perry Aggression Questionnaire (Buss & Perry, 1992) was used to assess trait aggression at baseline.

| Statistical analysis
Stepwise logistic regression (using a 5% level for entry/stay) was used to model the safety and efficacy endpoints of interest. Design terms for treatment, diagnostic subcohort, their interaction, and region (US or non-US), had forced inclusion (efficacy included cohort by region interaction as well). Several baseline characteristics, both numerical and categorical, formed the set of candidate covariate terms. Safety analyses included all participants who took ≥1 dose of randomized treatment. Efficacy analyses of CARs included all randomized participants. Consistent with the Russell standard (West, Hajek, Stead, & Stapleton, 2005), participants with missing smoking data were considered nonabstinent. The final model summaries for NPSAEs, CA9-12, CA9-24, and PPA at weeks 12 and 24 are provided in Table ST1.

| Baseline characteristics
There were 712 participants in the AD subcohort and 4,028 in the NPC. Baseline demographic, smoking, and psychiatric characteristics are displayed in Table 1 and demonstrate both AD subcohort and AD diagnostic subcohorts effects. For example, smokers with AD were more likely to be female, from the USA and of white race than NPC smokers. Smokers with PD had higher FTCD scores than NPC smokers. Smokers with AD had more psychiatric and substance-use comorbidity and greater lifetime suicidal ideation and behavior. In particular, at least descriptively, smokers with PTSD had higher rates of comorbid alcohol-use disorder history, other psychiatric disorders, prior suicidal ideation and/or behavior, and were less likely to be on psychotropic medication than smokers with GAD and PD. As expected, the HADS anxiety score was significantly higher among all AD diagnostic subcohorts versus NPC smokers, with smokers with GAD reporting the highest levels of anxiety. The HADS depression scores were higher in smokers with GAD and PTSD than in those with PD and NPC. Table 2 illustrates the incidence of moderate to severe treatmentemergent NPSAEs in each subcohort, the incidence of each individual component comprising the composite endpoint, and other prespecified NPSAE safety endpoints (e.g., serious NPSAEs). There was a significant AD subcohort effect (p = .0012), with smokers in all three AD subcohorts (PTSD, 6.9%; GAD, 5.4%; PD, 6.2%) having an increased incidence of moderate-to-severe NPSAEs versus NPC smokers (2.1%).

| Neuropsychiatric AEs
Across the AD subcohorts, the rates of treatment discontinuation due to NPSAEs were similar, ranging from 1.3% (GAD) to 2.1% (PTSD). Figure 1a displays the observed incidence of the primary composite safety endpoint for each subcohort as a function of treatment. Figure 1b depicts risk differences (RDs) and associated 95% confidence intervals (95% CIs) for contrasts comparing: (a) treatments overall; (b) subcohorts; and (c) treatments within each subcohort and the NPC. There were no significant differences in the incidence of moderate to severe NPSAEs by treatment group, nor were there any significant treatment-by-cohort interactions. Smokers with PD were significantly more prone to experience moderate to severe NPSAEs versus NPC smokers (RD = 4.0; 95% CI = 0.34-7.65).  95% CI = 1.20-17.10; and OR = 8.49; 95% CI = 1.57-45.78, respectively). Additionally, among smokers with PD, NRT significantly improved quit rates versus placebo (OR = 7.42; 95% CI = 1.37-40.35). Thus, although we found no frank interaction between primary AD diagnosis and treatment response, there appears to be beneficial effects on initiating smoking abstinence for varenicline and NRT in smokers with AD.

| Efficacy
The secondary efficacy endpoint, weekly PPA at weeks 12 and 24, are presented in Figure 3. Consistent with some of the CAR9-12 results, these suggest the possibility of differential treatment responses to the first-line smoking cessation medications by primary AD diagnosis. For example, varenicline showed superior efficacy to placebo on 7-day PPA at week 12 in smokers with PTSD, GAD, and PD. This effect remained at week 24 among smokers with GAD. In addition, varenicline was superior to bupropion and NRT at week 12 among smokers with GAD. Among smokers with PD, NRT was superior to placebo for 7-day PPA at both weeks 12 and 24.
Observed CAR9-24 for varenicline, bupropion, NRT, and placebo ranged from 7 to 13% (PTSD), 6-22% (GAD), 5-21% (PD), and 11-26% (NPC), with placebo consistently the lowest ( Figure S2). Within AD subcohorts, estimated ORs versus placebo for CAR9-24 were >2 for all active treatments, with the exception of bupropion and varenicline in smokers with PTSD. However, the 95% CIs for all comparisons included one, with the exception of NRT in smokers with PD, which continued to demonstrate longer-term efficacy in that subcohort.   This finding is not surprising given the overlap of ADs and heightened anxiety sensitivity (Abrams et al., 2017) and distress intolerance (Farris et al., 2016a). Although the psychiatric NPSAE events that comprised the primary safety outcome had to be severe enough to interfere significantly with functioning, it is possible that hyperarousal to symptoms in smokers with AD contributed to a greater occurrence of NPSAEs. We also found that compared with smokers without mental health conditions, diagnostic subgroups of smokers with AD had higher baseline levels of state anxiety, histories of suicidal ideation/behavior, and were more likely to be white females -variables we have previously reported were independently associated with heightened NPSAE risk .
However, our data on the incidence of these events in subtypes of smokers with AD provides some reassurance that they occur relatively infrequently and are not related to any specific pharma-

(b)
F I G U R E 1 Incidence of neuropsychiatric adverse events (a) and risk differences (b) in the anxiety disorder subcohorts versus nonpsychiatric cohort. Period for the ascertainment of neuropsychiatric adverse events is during 12 weeks of treatment and ≤30 days after the last dose. The following variables were included in the risk-difference model: treatment group, anxiety disorder subcohort, treatment-by-subcohort interaction, and region (US or non-US). AD, anxiety disorder; CI, confidence interval; GAD, generalized anxiety disorder; NPC, nonpsychiatric cohort; NPSAE, neuropsychiatric adverse event; NRT, nicotine replacement therapy (i.e., transdermal nicotine patch); PD, panic disorder (with/ without agoraphobia); PTSD, posttraumatic stress disorder; RD, risk difference AYERS ET AL. are also consistent with nonplacebo-controlled trials of NRT (Hertzberg et al., 2013;Kelly et al., 2015a;Kelly et al., 2015b;McFall et al., 2010) showing beneficial effects in smokers with PTSD.
Among smokers with GAD, we observed a pattern of results similar to that found in smokers with heterogeneous mental health conditions that we described in our initial report ( Smokers with PD with/without agoraphobia appeared to respond similarly to NRT and varenicline. In fact, for CAR9-12 as well as 7day PPA at week 12, both medications were superior to placebo. This equivalent pattern of response, while similar to that for CAR9-12 in smokers with PTSD, is the first we have observed among any psychiatric subcohort of smokers we have studied to date in EAGLES, who typically show the pattern we described previously, wherein varenicline is better than bupropion and NRT, and all are better than placebo (Anthenelli et al., 2016). This novel finding warrants replication, but it is interesting to note that varenicline and NRT partially or fully agonize nicotinic acetylcholine receptors and ameliorate nicotine withdrawal, which has been found to be more severe in certain smokers with AD prone to negative reinforcementbased smoking (Farris et al., 2016b;Leventhal & Zvolensky, 2015).
To our knowledge, there has been only one other placebocontrolled trial of bupropion, NRT, and combinations thereof in smokers with AD to date. Piper and colleagues (Piper et al., 2011) retrospectively examined the relationship between ADs (including panic attack, social anxiety disorder, and GAD) and reported that neither bupropion, NRT, nor their combination was more effective than placebo and counseling. Our group found some differences among smokers with AD in that varenicline provided significant benefit for cessation in those with GAD and PD, while NRT provided T A B L E 3 Mild, moderate, or severe treatment-emergent adverse events a reported by ≥5% of participants in any subcohort Gastrointestinal motility and defecation conditions 15 (7.9) 9 (3. F I G U R E 2 Observed continuous abstinence rates (a) and odds ratios (b) for weeks 9-12. The following variables were included in the oddsratio model: treatment group, anxiety disorder subcohort, treatment-by-subcohort interaction, region (US or non-US), race, age, Fagerström Test for Cigarette Dependence, and cigarettes smoked/day in the past month. AD, anxiety disorder; CAR, continuous abstinence rate; CI, confidence interval; GAD, generalized anxiety disorder; NPC, nonpsychiatric cohort; NRT, nicotine replacement therapy (i.e., transdermal nicotine patch); OR, odds ratio; PD, panic disorder (with/without agoraphobia); PTSD, posttraumatic stress disorder (a) (b) F I G U R E 3 Observed 7-day point prevalence of abstinence during treatment and follow-up and odds ratios at week 12 and week 24 by cohort. The following variables were included in the odds-ratio model: treatment group, anxiety disorder subcohort, treatment-bysubcohort interaction, and region (US or non-US) and region-by-cohort interaction. CI, confidence interval; GAD, generalized anxiety disorder; NPC, nonpsychiatric cohort; NRT, nicotine replacement therapy (i.e., transdermal nicotine patch); OR, odds ratio; PD, panic disorder (with/without agoraphobia); PPA, point prevalence of abstinence; PTSD, posttraumatic stress disorder (c) (d) | 257 significant benefit for those with PD. The discrepancy with our results and Piper's investigation may have had to do with that group's smaller sample size, use of a structured interview that included lifetime diagnoses, and diagnostic differences (e.g., history of panic attacks vs. PD). Additional investigations (Gaspersz et al., 2017) examined anxiety in depressed patients and suggested that smokers with comorbid AD may be less able to deal with the stress and challenges of smoking cessation. While our results supported the idea that smokers with AD have lower rates of abstinence and higher rates of AEs during a cessation attempt than smokers with no psychiatric disorder, we also present evidence that first-line smoking cessation treatments do indeed work in those with ADs.
ORs for CAR9-12 for active treatments versus placebo ranged from 1.20 to 3.23 (PTSD), 2.43-4.53 (GAD), and 3.33-8.49 (PD). Thus, clinicians should utilize these treatments, as they produce clinical gains and do not increase the risk of moderate to severe NPSAEs, which occurred at low frequency overall and with no differences between treatment groups. Of note, there was lower efficacy of smoking cessation treatment on longer-term (CAR9-24) outcomes indicating smokers with ADs may require prolonged pharmacotherapy use or continued smoking cessation counseling.
The primary limitations of this study are related to generalizability. This study did not include individuals who were psychiatrically unstable or those with current substance-use disorders. The most recent DSM-5 no longer categorizes PTSD as an AD, but rather, a trauma-related disorder; nonetheless, there is substantial overlap in the symptoms of PTSD and ADs. Small samples of those with social phobia disorder and obsessive-compulsive disorder limited our ability to evaluate these diagnostic groups. Given the post hoc nature of this study, we did not adjust for multiple comparisons, which increases the probability of a Type I error. Finally, this study was not powered or stratified a priori for this subcohort analysis.
In conclusion, individuals with primary PTSD, GAD, and PD were approximately three times more likely to experience clinically significant NPSAEs during medication-assisted cessation attempts and were 28% (GAD) to 47% (PD) less likely to quit than smokers without mental health conditions. The incidence of moderate to severe

DATA AVAILABILITY STATEMENT
Upon request, and subject to certain criteria, conditions, and exceptions (see https://www.pfizer.com/science/clinical-trials/trialdata-and-results for more information), Pfizer will provide access to individual deidentified participant data from Pfizer-sponsored global interventional clinical studies conducted for medicines, vaccines, and medical devices: (a) for indications that have been approved in the USA and/or EU; or (b) in programs that have been terminated (i.e., development for all indications has been discontinued). Pfizer will also consider requests for the protocol, data dictionary, and statistical analysis plan. Data may be requested from Pfizer trials 24 months after study completion. The deidentified participant data will be made available to researchers whose proposals meet the research criteria and other conditions, and for which an exception does not apply, via a secure portal. To gain access, data requestors must enter into a data access agreement with Pfizer. Data subject to third party restrictions.