Medication-assisted quit rates in participants with smoking-related diseases in EAGLES: Post hoc analyses of a double-blind, randomized, placebo-controlled clinical trial

INTRODUCTION Greater understanding is required of how smokers with smoking-related diseases respond to smoking cessation medications. This post hoc analysis of EAGLES data compared continuous abstinence rates (CARs) in smokers with/without smoking-related diseases and assessed participant demographic and baseline characteristics that may serve as predictors of continuous abstinence (CA). METHODS EAGLES was a 24-week (12-week treatment, 12-week follow-up), double-blind, active- (nicotine replacement therapy; patch) and placebo-controlled study in motivated-to-quit smokers with/without psychiatric disorders. This analysis assessed CARs at weeks 9-12 (CAR9-12) and 9-24 (CAR9-24) in participants with smoking-related diseases [asthma, chronic obstructive pulmonary disease (COPD), diabetes, and/or cardiovascular disease (n=1372)] versus controls without these comorbidities (n=6039). Participants received varenicline 1 mg twice daily, bupropion 150 mg twice daily, nicotine patches 21 mg/day with taper, or placebo for 12 weeks. Stepwise logistic modeling was also performed to analyze odds ratio (OR) for predictors of CA at weeks 9-12 (CA9-12) and 9-24 (CA9-24). RESULTS Smokers with smoking-related diseases were older, had a longer smoking history, more quit attempts, and were more likely to have a psychiatric disorder and reside in the US versus smokers without comorbidities. Fagerström Test for Cigarette Dependence scores and treatment adherence were comparable between cohorts. Smokers with smoking-related diseases had lower CARs versus controls (CAR9-12: 20.8% vs 24.0%; CAR9-24: 13.0% vs 16.9%). Use of smoking cessation medication was the strongest predictor of CA after control for demographics, smoking characteristics, and psychiatric disorder. By treatment, OR and CI were: varenicline CA9-12 (OR=3.82; 95% CI: 3.21-4.54) and CA9-24 (OR=2.92; 95% CI: 2.40-3.54); bupropion CA9-12 (OR=2.17; 95% CI: 1.81-2.60) and CA9-24 (OR=1.99; 95% CI: 1.63-2.44); nicotine patches CA9-12 (OR=2.23; 95% CI: 1.87-2.67) and CA9-24 (OR=1.86; 95% CI: 1.52-2.28). CONCLUSIONS Smokers with smoking-related diseases had lower quit rates than controls. Of the active treatments compared, varenicline was most effective in smokers with asthma, COPD, diabetes, or cardiovascular disease. TRIAL REGISTRATION NCT01456936 (https://clinicaltrials.gov/ct2/show/NCT01456936).


INTRODUCTION
Globally, smoking is the second leading risk factor for deaths and disability-adjusted life-years after elevated systolic blood pressure 1 and two-thirds of smokers die of a smoking-related disease 2 . Encouraging smokers to quit is critically important 3 , as smoking cessation is considered an effective intervention by which to lower the risk of death and disability associated with a range of smoking-related chronic diseases 4 .
Lung diseases are a major consequence of cigarette smoking 4 . Studies in smokers with chronic obstructive pulmonary disease (COPD) have demonstrated the effectiveness of behavioral treatment and pharmacotherapy in aiding smoking cessation 5 . However, up to 40% of smokers with COPD may continue to smoke, possibly due to a high degree of nicotine dependence 6 . Studies have found smoking rates in patients with asthma to be similar to those in the general population 7 . Control of asthma symptoms and bronchial reactivity tends to improve substantially in smokers with asthma who quit, while inhaled antiasthmatic medication is more efficacious in nonsmokers than in those who smoke 6 . Despite these potential benefits, smoking cessation trials specifically tailored to patients with asthma are lacking 6 .
Smokers with cardiovascular disease (CVD) are most likely to quit following an acute event compared with smokers with other smoking-related diseases 8 . In the acute CVD setting and in smokers with stable CVD, smoking cessation pharmacotherapies (in particular, bupropion and varenicline) are shown to be effective 9 . However, psychosocial and other triggers appear to contribute to the resumption of smoking in up to 50% of quitters as the acute phase of the illness abates 10,11 .
Reported rates of smoking vary among patients with type 2 diabetes 8,12 and studies of the effectiveness of smoking cessation medications are limited. Smoking contributes to abnormal fat distribution, adversely affects insulin sensitivity, and reduces glucose disposal 13 ; however, cessation may be hindered by fear of weight gain and loss of glycemic control. Although some studies show worsening of diabetes post cessation, this phase seems to normalize with longer follow-up 14,15 . Furthermore, weight gain after smoking cessation does not negate the beneficial cardiovascular effects of quitting 15,16 .
To help increase quit rates in smokers with smoking-related diseases, greater understanding is needed of whether these smokers differ in their smoking characteristics compared with smokers without comorbid conditions, and whether the response to smoking cessation pharmacotherapies and behavioral support differs between these populations. EAGLES (Evaluating Adverse Events in a Global Smoking Cessation Study) was a large (n=8144), phase IV, smoking cessation, randomized clinical trial (RCT) that captured detailed individual participant data with regard to smoking-related diseases 17 . The objectives of this post hoc analysis of EAGLES efficacy and safety data were to compare carbon monoxide (CO)-confirmed continuous abstinence rates (CARs) in smokers with asthma, COPD, diabetes, and/or CVD versus smokers without these comorbidities, and to assess participant demographic and baseline characteristics that may serve as predictors of continuous abstinence (CA). A plain language summary is available for this article in Supplementary Material.

Study design and population
EAGLES (ClinicalTrials.gov: NCT01456936) was a 24-week (12-week treatment, 12-week nontreatment follow-up), double-blind, active-[nicotine replacement therapy (NRT): nicotine patch 21 mg/ day with taper] and placebo-controlled, multicenter, parallel-group, triple-dummy RCT designed to assess the safety and efficacy of varenicline 1 mg twice daily and bupropion 150 mg twice daily for smoking cessation in smokers with and without pre-specified psychiatric diagnoses.
For this post hoc analysis, EAGLES participants were stratified into the smoking-related-disease cohort (those with a self-reported history of asthma, COPD, diabetes, or CVD), or the control cohort (without these comorbidities) (Supplementary file Figure S1). Because of the heterogeneity of cancer diagnoses, smokers who reported present, active cancer (n=2: one diagnosis each of colon cancer and non-Hodgkin lymphoma) were excluded. Likewise, since alcoholism may confound the relationship between smoking and smoking-related diseases and possibly lower overall adherence, those who reported past or present alcohol dependence were also excluded. Diagnoses included alcohol abuse (n=553), alcohol dependence syndrome (n=584), and alcoholism (n=64); participants may have one or more diagnosis (cumulative frequency: n=715). Participants with past or present schizophrenia (n=18) were excluded because of possible bias in the analysis due to poor medicine compliance or co-abuse of other substances that can contribute to lower quit rates in this population.

Statistical analysis
Demographic and other baseline characteristics were summarized by cohort and by smoking-relateddisease subcohort; participants with more than one smoking-related disease could be assigned to multiple subcohorts. CO-confirmed CAR9-12 and CAR9-24 were presented by: 1) cohort and by smokingrelated-disease subcohort; and 2) treatment and cohort/smoking-related-disease subcohort. The incidence of NPSAEs and study-drug exposure (days) were summarized by treatment and cohort/ smoking-related-disease subcohort. Common adverse events (AEs) (≥5% of participants in any treatment by cohort) and change in body weight from baseline were summarized by treatment and cohort. Descriptive statistics included the number of observations, percentage, mean, and standard deviation (SD).
Stepwise logistic regression was performed to assess predictors of CA for weeks 9-12 (CA9-12) and weeks 9-24 (CA9-24). The model forced inclusion of treatment (varenicline, bupropion, NRT, or placebo), psychiatric cohort (psychiatric vs non-psychiatric), medical history cohort (smoking-related disease vs control), and region (non-USA vs USA) terms. Demographic and other baseline characteristic variables were considered as candidate terms, as was a term for treatment by medical history cohort interaction. Analyses were not pre-planned and should therefore be considered exploratory.

Study population
The mean (SD) age was 52.2 (11.8) years in smokers with smoking-related diseases (n=1372) versus 45.1 (12.2) years in controls (n=6039); the respective mean (SD) number of years smoked was 33.8 (12.5) versus 26.9 (12.0) years (Table 1). Baseline history of psychiatric disease was recorded in 705 (51.4%) smokers with smoking-related diseases versus 2693 (44.6%) in controls; the number of smokers who resided in the US was 745 (54.3%) versus 2974 (49.2%), respectively. Prior use of varenicline, bupropion, and NRT, and the number of previous quit attempts, was numerically higher in each of the smoking-related-disease subcohorts compared with controls. Fagerström Test for Cigarette Dependence (FTCD) mean (SD) score and baseline CO were the same in the smoking-related-disease cohort and controls [FTCD: 5.7 (2.0); CO: 21 ppm] and mean (SD) daily cigarette use over the previous month was also similar [21.3 (8.9) vs 20.5 (8.0)]. Mean (SD) body mass index was 29.5 (6.7) versus 27.7 (6.2) kg/m 2 in the smoking-related-disease cohort versus controls. Results were further stratified by active treatment (varenicline, bupropion, NRT) or placebo (Supplementary file Table S1). For smokers with 5 asthma or COPD, CAR9-12 and CAR9-24 were lower with all active treatments compared with controls. Conversely, for smokers with diabetes, CAR9-12 and CAR9-24 were similar or higher compared with the control cohort. For smokers with CVD, CAR9-12 was similar to the control cohort for all active treatments; however, CAR9-24 was lower compared with controls. In the smoking-related-disease and control cohorts CAR9-12 and CAR9-24 were higher in smokers treated with varenicline compared with bupropion or NRT. Following treatment with placebo, markedly lower CAR9-12 and CAR9-24 were observed for smokers with COPD or diabetes compared with the control group; however, quit rates were similar in smokers with asthma or CVD compared with controls.
Predictors of CA at weeks 9-12 were generally also predictors of CA at weeks 9-24; however, ORs for CA with varenicline, bupropion, and NRT treatment were lower at weeks 9-24 (Table 2). Comorbid smokingrelated disease (OR=0.705; 95% CI: 0.59-0.85) and psychiatric diagnosis (OR=0.720; 95% CI: 0.63-0.82) were inversely associated with CA at weeks 9-24. Previous attempts to quit did not predict CA at weeks 9-24; however, higher body mass index was predictive of CA over this follow-up period.
Exploratory logistic regression analyses were conducted to explore the effect of each medication versus placebo across each smoking-related-disease subcohort. For CA at both weeks 9-12 and 9-24, results showed varenicline to be superior to placebo in the COPD   Table S3). Similar results were observed for bupropion and NRT versus placebo at weeks 9-12; however, at weeks 9-24, superiority versus placebo was observed only in the diabetes subcohort and in controls (Supplementary file Table S3).

Treatment adherence
Overall, the mean number of days (SD) for which smokers in the smoking-related-disease cohort adhered to their assigned study medication were 73.0 (24.1) versus 74.4 (23.1) for controls ( Table  2). Regardless of treatment, mean adherence between smoking-related-disease subcohorts ranged between 72.3 and 75.2 days. Mean adherence was also similar between treatment groups overall, and when stratified by smokingrelated-disease subcohort ( Table 2).
Treatment-emergent AEs (reported in ≥5% of participants in any treatment) summarized by treatment and cohort are presented in Supplementary file  Smokers in EAGLES who had already developed a smoking-related disease, including asthma, COPD, diabetes, or CVD, had lower quit rates than smokers without these comorbid conditions. However, the efficacy of smoking cessation medications generally remained higher than placebo, and all three medications that were tested in the trial showed significant benefits at the end of treatment (week 12) that persisted up to the end of follow-up at week 24. Varenicline showed the greatest effect, with non-overlapping CIs at both 12-and 24-week time points between varenicline and bupropion or NRT. Given that EAGLES is the largest medication-assisted smoking cessation trial to date, these results may be important to clinical practice.
When summarized by smoking-related diseases, results suggest the lower quit rates in the smokingrelated disease cohort to be applicable to asthma, COPD, and (in a lesser sense) CVD only; smokers with diabetes had similar CARs at weeks 9-12 and 9-24 compared with controls receiving active treatment, but markedly lower quit rates with placebo.
Given the burden of smoking-related comorbidities, greater understanding is required of how smokers with smoking-related diseases respond to smoking cessation medications. Notably, they (overall, and in each smoking-related-disease subcohort) reported a higher number of previous serious quit attempts compared with controls, and a higher percentage prior use of varenicline, bupropion, and NRT. This suggests that lower quit rates in smokers with comorbidities are not due to an inherent lack of motivation to attempt to quit, nor a lack of willingness to use smoking cessation aids to do so. The study found varenicline was the most efficacious treatment, compared with bupropion and NRT, in both smokers with smokingrelated diseases and controls, and aligns with recent tobacco treatment guidelines which recommend treatment initiation with varenicline over bupropion or NRT 20 . It is noted that the estimation-oriented approach for treatment comparison by smokingrelated disease subcohort is suggestive in nature due to the lower subcohort sample sizes and resulting wider confidence intervals. We observed consistent mean study medication adherence rates among all smoking-related-disease subcohorts and controls (73 and 74 days out of 90 days, respectively). Our results suggest that neuropsychiatric safety is not a major concern when prescribing varenicline, bupropion, or NRT, to smokers with smoking-related diseases, and that commonly reported AEs were similar between smoking-related disease and control cohorts. Although the prospect of weight gain can be a concern to smokers wishing to quit, our results suggest no effect by cohort (smokingrelated disease vs control) or treatment. However, as the most effective drug seems to be varenicline, our results suggest that this should be the drug of choice.
Overall, the presence of a comorbid smokingrelated disease was identified as an independent predictor of significantly lower CA compared with controls at weeks 9-12 and 9-24, as was a psychiatric disease diagnosis. Smokers with smoking-related diseases were older and more likely to reside in the US compared with controls. Consistent with previous post hoc analyses of EAGLES data, older age was identified as a positive predictor of abstinence at weeks 9-12 and 9-24, whereas residing in the US was inversely associated with abstinence 21 . However, taken together, these differences do not explain the lower CARs in smokers with smoking-related diseases observed in the current analyses. Use of smoking cessation medication was by far the strongest positive predictor of abstinence.
Published data investigating quit rates in smokers with smoking-related diseases following treatment with smoking cessation pharmacotherapy are limited, but generally support the current analyses. In an RCT of 52 smokers with asthma who received varenicline or placebo, CARs at week 12 were significantly higher in the varenicline-treated group (69% vs 36%, respectively) 22 . However, between-treatment differences at 24 weeks were not significant due to a high relapse rate. In a recent open-label study 23 , 101 smokers who were hospitalized due to COPD, asthma, or pneumonia, self-selected to receive either varenicline plus intensive behavioral support versus no medicine and one consultation. Quit rates after 12 weeks were 55% versus 16% and after 1 year 52% versus 14% for the respective treatment regimens. A Cochrane meta-analysis that included 16 studies 5 found that varenicline, bupropion, and NRT, were all effective treatments in smokers with COPD; however, the analysis did not compare quit rates in smokers without COPD. In the current study, nicotine dependence did not differ between the overall smoking-related disease cohort and controls; however, smokers with COPD had slightly higher FTCD scores versus controls (6.0 vs 5.7), and a higher mean daily cigarette consumption (22.9 vs 20.5).
A retrospective pooled analysis of 15 Pfizersponsored studies in which smokers were treated with varenicline, reported that those with diabetes (n=323) had higher CAR9-12 (43.8%), CAR9-24 (27.5%), and CAR9-52 (18.4%) compared with patients who received placebo 24 . However, in contrast to the current analyses, CARs at each endpoint were lower compared with smokers without diabetes. In addition, observed quit rates with placebo were higher than those in the current analysis (CAR9-12: 24.8%; CAR9-24: 14.4%). Since these studies were not designed explicitly to explore quit rates in smokers with diabetes, an RCT is warranted to confirm whether quit rates in smokers with diabetes are indeed similar to those without comorbid conditions.
In an RCT of smokers hospitalized with acute coronary syndrome (n=302) CAR9-24 was higher following 12 weeks of treatment with varenicline versus placebo 25 . However, studies in which patients with acute coronary syndrome were treated with bupropion or NRT were inconclusive or failed to show improvement versus placebo, particularly in the longer term 26,27 . In an RCT of smokers with stable CVD (n=714), CAR9-12 and CAR9-52 were significantly higher following treatment with varenicline versus placebo 28 . In smokers with acute myocardial infarction (n=392) randomized to receive bupropion versus placebo for 9 weeks, CARs at week 52 were not significantly different between treatments (26.8% vs 22.2%) 29 . A recent meta-analysis of RCTs investigating the efficacy of smoking cessation interventions in smokers with CVD found evidence relating to efficacy of NRT to be inconclusive 9 .
Notably, comparative smoking cessation trials of smokers with smoking-related diseases versus otherwise healthy smokers are lacking. We compared the quit rates observed for smokers with asthma, COPD, diabetes, and CVD, in the current analyses with those from RCTs conducted by Pfizer in which patients with the above-mentioned smokingrelated diseases were enrolled 28,30 and with studies comprising a general population of smokers 17,31,32 (Supplementary file Figures S2 and S3). All studies included varenicline and placebo and were similarly designed. Taken together, the data support the findings presented in the current study: that quit rates are generally lower for patients with asthma, COPD, and CVD. The implications of lower quit rates in smokers with these smoking-related diseasesbut not for smokers with diabetes -suggest that intensification of smoking cessation interventions for these populations of smokers may be beneficial. For example, regular re-treatment of smokers who relapse was described previously in a study of smokers with mild COPD who were referred for smoking cessation treatment every 4 months during a 5-year period 33 .

Limitations
Our study has several limitations. First, EAGLES was an RCT and selection bias may have occurred because participants were a self-referred, motivatedto-quit population and not fully representative of the general population of smokers. Second, there may have been misclassification of participants to the smoking-related-disease subcohorts as medical diagnoses were self-reported and were not validated by medical records; therefore, the severity and the duration of these diseases were also not available. Third, although the overall smoking-related-disease cohort was relatively large (n=1372), the current analyses were not pre-planned and EAGLES was not powered to detect differences in quit rates in small subpopulations of smokers such as those stratified by smoking-related disease type.

CONCLUSIONS
In this post hoc analysis of EAGLES data, we report that smokers with asthma, COPD, and CVD, have lower CARs compared with otherwise healthy smokers, despite a similar degree of nicotine dependence and similar treatment adherence. Smokers with diabetes had a similar quit rate compared with controls when taking active medication, and further studies are required to confirm these findings. Smokers with asthma, COPD, and CVD, may therefore benefit from increased behavioral and pharmacologic support for smoking cessation, which may include re-treatment following a failed quit attempt(s) or relapse.