Adherence to Pharmacological Smoking Cessation Interventions: A Literature Review and Synthesis of Correlates and Barriers

Abstract

Introduction

Efficacious pharmacological interventions for smoking cessation are available, but poor adherence to these treatments may limit these interventions overall impact. To improve adherence to smoking cessation interventions, it is first necessary to identify and understand smoker-level characteristics that drive nonadherence (ie, nonconformance with a provider’s recommendation of timing, dosage, or frequency of medication-taking during the prescribed length of time).

Methods

We present a literature review of studies examining correlates of, or self-reported reasons for, nonadherence to smoking cessation pharmacotherapies. Studies were identified through PubMed—using MeSH terms, Embase—using Emtree terms, and ISI Web of Science.

Results and Conclusions

This literature review included 50 studies that examined nonpreventable (eg, sociodemographics) and preventable (eg, forgetfulness) factors associated with adherence to smoking cessation medication and suggestions for overcoming some of the identified barriers. Systematic study of this topic would be facilitated by consistent reporting of adherence and correlates thereof in the literature, development of consistent definitions of medication adherence across studies, utilization of more objective measures of adherence (eg, blood plasma levels vs. self-report) in addition to reliance on self-reported adherence.

Implications

This article provides the most comprehensive review to date on correlates of adherence to pharmacological smoking cessation interventions. Challenges and specific gaps in the literature that should be a priority for future research are discussed. Future priorities include additional research, particularly among vulnerable populations of smokers, developing standardized definitions of adherence and methods for measuring adherence, regular assessment of cessation pharmacotherapy adherence in the context of research and clinical practice, and development of novel treatments aimed at preventable barriers to medication adherence.

Introduction

Efficacious Pharmacological Smoking Cessation Interventions

A range of Food and Drug Administration (FDA)-approved over-the-counter and prescription pharmacological treatments for smoking cessation are available to smokers interested in quitting. Findings from a recent meta-analysis indicated that nicotine replacement therapy (NRT), bupropion (ie, Wellbutrin/Zyban), and varenicline (ie, Chantix) all increased the odds of smoking cessation versus placebo.¹ Specifically, NRT and bupropion increased the odds of cessation nearly twofold (odds ratio [OR] = 1.84 and 1.82, respectively), while varenicline increased the likelihood of cessation nearly threefold (OR = 2.88) versus placebo.² These therapies were not associated with an increased rate of serious adverse events.¹ However, the absolute proportions of smokers able to quit using these methods were as follows: NRT (17.6%), bupropion (19.1%), and varenicline (27.6%), as compared to placebo (10.6%).² Thus, while efficacious pharmacological methods for smoking cessation exist, the absolute proportion of smokers able to quit using these methods remains low.

Medication Adherence as a Barrier to Cessation

Medication nonadherence is a challenge in clinical practice and research that contributes to lower smoking cessation rates.³ Vrijens et al.⁴ define medication adherence as “the process by which patients take their medication as prescribed.” They posit that medication adherence is composed of three quantifiable phases: initiation: when patients take their first dose of a prescribed medication, discontinuation: when patients stop taking their prescribed medication, and implementation: the extent to which patients’ actual dosing corresponds to the prescribed regimen, from initiation until the last dose.⁴

Estimates indicate that 20% of smokers who receive prescriptions for smoking cessation medication never fill that prescription.^5,6 Individuals receiving pharmacotherapies for tobacco dependence also often use them at a lower dose and for less time than evidence suggests is optimal.^7–10 For instance, adherence with the recommended length of treatment occurs among 50% or fewer NRT users.^11–14 This lack of compliance with the recommended treatment length is significant because substantial evidence—both observational and within the context of clinical trials—indicates that greater medication adherence is associated with greater abstinence from smoking,^3,8,15–18 even after controlling for potential bias due to reverse causation.³

In addition to other key confounders,¹⁹ medication nonadherence also likely explains at least part of the discrepancy between efficacy observed in tightly managed clinical trials and real world clinical practice, where in the latter, nonadherence occurs at higher rates^20–22: NRT achieves quit rates of 18%–31% in clinical trials²³ but only 5%–15% in general population smokers using NRT as a quit aid.²⁴ Though it is difficult to make similar comparisons regarding bupropion and/or varenicline given that relatively less research has been published for these treatments in community-based settings, limited evidence exists to support relative worse adherence to varenicline in community-based settings (24%–28%)^17,25 versus clinical trials (56%–74%).^26–28

One key to reduce medication nonadherence is understanding the characteristics that drive nonadherence—both those that are nonmodifiable as well as those that can be modified via intervention. Identification and understanding of such characteristics can enable more precise thinking about targeted smoking cessation interventions. Despite discussion of the need to identify adherence factors within the literature,²⁹ there have been no reviews of characteristics associated with smoking cessation intervention adherence. To address this, we aim to evaluate characteristics associated with adherence to pharmacological smoking cessation interventions.

Methods

Inclusion Criteria

The aim of this article was to review studies that explored self-reported and statistical correlates of adherence to smoking cessation pharmacotherapies. Inclusion criteria were (1) articles focused on smoking cessation, (2) reporting use of pharmacotherapy for smoking cessation, (3) reporting self-reported reasons for and/or analysis of potential correlates of adherence to smoking cessation pharmacotherapy, (4) English-language articles, (5) analysis-driven articles (ie, review/commentary papers and conference abstracts were excluded), and (6) publication in peer-reviewed journals.

Search Strategy

In July 2017, we searched PubMed using MeSH terms (“Smoking Cessation” OR “Tobacco Use Cessation Products” OR “Varenicline” OR “Bupropion” AND “Medication Adherence”) and ISI Web of Science using the key words “smoking cessation” AND “medication adherence” to obtain all of the published English language articles on the topic. We also searched Embase using Emtree terms (“Smoking cessation” OR “nicotine replacement therapy” OR “varenicline” OR “bupropion” AND “medication compliance”). We reviewed reference sections of published articles to obtain additional references.

Data Synthesis

Consistent with the World Health Organization’s (WHO) approach to classifying types of nonadherence,³⁰ we distinguished between factors that are potentially preventable (eg, beliefs about the efficacy of smoking cessation medications), and therefore can serve as targets in interventions, as well as those that are nonpreventable (eg, race/ethnicity, comorbidities). We reported characteristics of identified studies in Table 1, and organized identified reasons for/correlates of medication adherence in such a way as to highlight the quality and support for the identified associations (Supplementary Table 1). We also organized the studies based on whether they reported self-reported reasons for (Supplementary Table 2) or statistical correlates of adherence to pharmacotherapy (Supplementary Table 3).

Table 1.

Characteristics of Studies Included in Qualitative Synthesis (n = 50)

First author, year	Cessation modality	Country	Population	Type of characteristics	Phase of adherence
Alterman, 1999	Nicotine patch	USA	General population smokers	C	Imp
Balmford, 2011	NRT, bupropion, varenicline	Canada, UK, USA, Australia	General population survey	SR; C	Imp; D
Ben Taleb, 2015	Nicotine patch	Syria	General population smokers	C	Imp
Berg, 2013	Nicotine patch	USA	Female prisoners	C	Imp
Browning, 2016	Varenicline, NRT	USA	HIV-positive smokers	C	Imp
Burns, 2008	NRT	USA	General population smokers	SR; C	D
Carroll, 2017	Varenicline	USA	Smokers with cancer	C	Imp
Catz, 2011	Varenicline	USA	General population smokers	SR; C	Imp, D
Cooper, 2004	Nicotine patch	USA	General population smokers	C	Imp
Cropsey, 2015	Bupropion	USA	Smokers under community corrections supervision	C	Imp
Cropsey, 2017	Bupropion	USA	Smokers under community corrections supervision	C	Imp
de Dios, 2012	Varenicline, nicotine patch	USA	HIV-positive smokers	SR	Imp
de Dios, 2015	Nicotine patch	USA	HIV-positive smokers	SR; C	Imp
Epperson, 2010	Naltrexone, nicotine patch	USA	General population smokers	C	Imp
Fish, 2009	Nicotine patch, gum, lozenge	USA	Pregnant smokers	SR; C	Imp; D
Fucito, 2009	Bupropion	USA	General population smokers	C	Imp
Fucito, 2016	Naltrexone, nicotine patch	USA	General population smokers	C	Imp
Gariti, 2009	Nicotine patch, bupropion	USA	Light smokers	C	Imp
Grandi, 2016	Bupropion	Canada	Myocardial infarction patients	C	Imp
Harrison-Woolrych, 2010	Varenicline	New Zealand	General population smokers	SR	Imp
Hawk, 2012	Varenicline	USA	Not reported	C	Imp
Hays, 2010	Varenicline, bupropion	USA	General population smokers	C	Imp
Hollands, 2013	Nicotine patch, inhaler, lozenge, gum, sublingual tablet	UK	Primary care patients	C	Imp
Hood, 2013	Nicotine patch	USA	Female smokers	SR; C	Imp
Killen, 2010	Transdermal selegiline	USA	Adult smokers	C	Imp
Kushnir, 2017	Nicotine patch	Canada	Adult smokers	SR; C	Imp
Lam, 2005	Nicotine patch, gum, inhaler	China	Chinese smokers	C	Imp
Leischow, 2015	Bupropion	USA	Adolescent smokers	C	Imp
Liberman, 2013	Varenicline	USA	General population smokers	C	Init
Nollen, 2013	Bupropion	USA	African American light smokers	C	Imp
Okuyemi, 2010	Nicotine gum	USA	African American light smokers	C	Imp
Rojewski, 2016	Nicotine gum, patch, lozenge	USA	Hazardous drinkers	C	Imp
Scherphof, 2014	Nicotine patch	Netherlands	Adolescent smokers	SR; C	Imp
Schmitz, 2005	Bupropion	USA	Female smokers	C	Imp
Shadel, 2016	Nicotine patch	USA	HIV-positive smokers	SR	Imp
Shelley, 2015	Varenicline	USA	HIV-positive smokers	C	Imp
Shiffman, 2008	Nicotine patch	USA	General population smokers	C	Imp
Solberg, 2010	Any prescription medication	USA	General population smokers	C	Init
Suehs, 2014	Varenicline	USA	MADP members	C	Imp
Swan, 2005	Bupropion	USA	Not reported	C	D
Swan, 2010	Varenicline	USA	Treatment-seeking smokers	SR	D
Tague, 2017	Nicotine patch, bupropion, varenicline	USA	Hospital inpatients	C	Imp
Tseng, 2017	Varenicline	USA	HIV-positive smokers	C	Imp
Vaz, 2016	Nicotine patch	UK	Pregnant smokers	C	Imp
van Boven, 2015	Varenicline, bupropion	Netherlands	General population smokers	C	Imp
Voci, 2016	Nicotine patch, gum, inhaler	Canada	Adult smokers	SR; C	Imp
Ware, 2015	Nicotine patch	UK	Adult daily smokers	C	Imp
Wiggers, 2016	Nicotine patch	Netherlands	Cardiovascular patients	SR; C	Imp
Yingst, 2015	Nicotine patch	USA	General population smokers	SR; C	Imp
Zeng, 2011	Varenicline, bupropion, NRT	USA	General population smokers	C	Init

C, correlates; SR, self-report; Init, initiation; Imp, implementation; D, discontinuation.

Results

Figure 1 presents a flow diagram of the process used to obtain the studies using the search and screening process. The initial keyword search returned 932 articles (PubMed: 215, ISI Web of Science: 299, and Embase: 418). Review of these abstracts revealed 50 studies that met inclusion criteria. All included studies were published 1999–2017. Author LRP conducted the initial search and authors FJM and HBB reviewed the extracted information for appropriateness and identified additional studies for inclusion.

Figure 1.

PRISMA flow diagram.

Characteristics of Included Studies

Study Populations

Most studies (n = 24; 48%) focused on general population cigarette smokers (Table 1). Twenty-four percent (n = 12) of study populations were selected based on medical comorbidity (eg, HIV-positive smokers) or utilization of medical services (eg, hospital inpatients). Samples from three studies (6%) were selected based on smoking intensity (eg, light smokers) and samples from an additional three studies (6%) were based on racial/ethnic group membership. The remainder were identified based on being under community corrections supervision (n = 3; 6%); sex (n = 3; 6%); age (n = 2; 4%); hazardous alcohol consumption (n = 1; 2%); and two were not reported. Note that percentages do not sum exactly to 100% because some studies fell into multiple categories (eg, female prisoners).³¹

Geographic Regions

Thirty-seven studies (74%) were conducted in the United States, 3 (6%) in the United Kingdom, 3 (6%) in Canada, 3 (6%) in the Netherlands, 1 (2%) in China, 1 (2%) in Syria, 1 (2%) in New Zealand, and 1 study (2%) was an international collaboration (ie, United States, United Kingdom, Canada, and Australia).

Smoking Cessation Pharmacotherapies and Adherence Characteristics

Regarding pharmacotherapy, 21 (42%) studies reported on NRT alone, 9 (18%) reported on varenicline alone, and 8 studies (16%) reported on bupropion alone. The remaining studies reported on combinations of pharmacotherapy: varenicline and bupropion (4%); varenicline and NRT (4%); bupropion and NRT (2%); NRT, varenicline, and bupropion (6%); naltrexone and NRT (4%); transdermal selegiline (2%); and any smoking cessation medication (2%). Most studies (n = 35; 70%) reported correlates of adherence alone, 8% reported self-reported reasons alone, and 22% of studies reported a combination of correlates of and self-reported reasons for adherence. A majority (n = 41) of the studies reported on self-reported reasons for and correlates of medication implementation alone (82%), followed by medication discontinuation alone (6%), a combination of implementation and discontinuation (6%), and medication initiation alone (6%).

Nonpreventable Characteristics

Sociodemographic Characteristics

Sociodemographic characteristics were among the most consistently identified characteristics correlated with adherence. Moreover, the directionality of associations between demographic characteristics and adherence were largely consistent. With two exceptions,^32,33 male sex^{5,6,13,14,34,35} was associated with better adherence to smoking cessation pharmacotherapies. Additionally, older age^{5–7,13,15,27} (though see refs.^35,36) greater educational attainment,^13,28,37 and Caucasian race^{5,27,28,38–40} (with one exception³⁹) were associated with better adherence across studies that have examined these factors. Poverty¹¹ and current employment status³⁶ were also associated with poorer adherence. These findings are largely substantiated by findings from a systematic review of the larger medication adherence literature.⁴¹

Comorbidity and Medical Factors

Mental and physical health comorbidities also emerged as correlates,^6,34,42–44 though the comorbidities identified were varied, and the direction of associations was often inconsistent from study to study. For instance, symptoms of depression were found to be associated with poorer adherence in one study,⁴³ but associated with a greater likelihood of adherence in another.⁶

Tobacco Use Characteristics

Tobacco use characteristics of prior quit attempts—having prior experience with a specific cessation modality,^{6,13,17,36,45,46} having made a prior quit attempt,⁴⁷ and/or having made a greater number of quit attempts in the past^31,37,44—were consistently associated with a greater likelihood of medication adherence. Conversely, relapse to smoking was consistently associated with poor adherence.^{11,34,36–38,43,48} Greater nicotine dependence,^12,16,48,49 frequency/history of smoking,^12,16,35,50 and higher baseline biochemical measures of nicotine/tobacco exposure (eg, cotinine)⁴⁹ were largely associated with poorer medication adherence, though some discrepancies were observed.^7,31,43,44 Greater motivation to quit⁵¹ and being in Contemplation/Preparation Stages of Change³⁶ were associated with a greater likelihood of adherence, while not being ready to quit/still wanting to smoke^34,36,47,52 was associated with a decreased likelihood of adherence.

Genetic and Personality Factors

Genetic factors, such as having a copy of rs1051730 allele⁵³ or having one or more A1 allele of the dopamine DRD2 gene⁵⁴ were associated with poorer adherence. Personality characteristics like conscientiousness and agreeableness were associated with an increased likelihood of adherence.³⁷

Preventable Characteristics

Smokers’ Perceptions, Beliefs, and Knowledge

Smokers’ perceptions—specifically perceived lack of need for a pharmacological cessation aid^{15,34,38,43,48,54,55} and the belief that the cessation aids do not help with cessation^{11,15,34,37,38,47,52,54,56}—were among the most commonly cited reasons, and most consistently associated with medication nonadherence. Smokers expressed fear of becoming dependent by using the nicotine patch⁵² and a desire to test one’s ability to remain quit without pharmacological aid,^34,36,47 both of which were associated with poor adherence. Lack of knowledge regarding the effects/adverse effects of³⁶ and how to correctly use smoking cessation pharmacotherapy³⁴ were associated with poorer medication adherence while using pharmacotherapy correctly was associated with better adherence.^47,52 Additionally, both perceived stress⁴⁴ and perceived social support⁵⁷ were associated with a greater likelihood of adherence.

Health Care Characteristics

Advice from a doctor to discontinue smoking cessation pharmacotherapy was associated with poorer adherence,³⁶ while attending a greater number of health care visits⁵ and having greater health-related quality of life⁵⁸ were associated with better adherence.

Adverse Effects

The experience of^{11,15,34,37,47,55,56} and concern about experiencing adverse effects^43,52,55 were commonly identified self-reported reasons for poor adherence to smoking cessation interventions.

Psychosocial Characteristics

Life events and stress were commonly reported as being detrimental to medication adherence,^34,36,48,55 though one study⁴⁴ reported positive associations between stress and adherence to nicotine gum. Self-efficacy for adherence^28,59,60 and motivation^12,48 were consistently associated with an increased likelihood of adherence. Consistent with correlates identified within the larger adherence literature,^34,61–65 forgetfulness was often cited as a reason for nonadherence to NRT^{34,36–38,55} and varenicline or placebo pills.⁵⁵ Social environments that included other smokers^42,45 and heavier alcohol consumption^34,35 were both negatively associated with smoking cessation medication adherence.

Financial Characteristics

Financial and insurance-related characteristics were associated with adherence. Financial concerns—including cost prohibitive prices,^11,38,56 lack of insurance,^38,66,67 higher copays,⁶ and lack of willingness to pay for pharmacotherapies¹³—were identified as barriers to adherence.

Intervention-Specific Characteristics

Characteristics of the interventions themselves were found to be associated with adherence. For instance, characteristics of adjunctive treatments (eg, psychosocial counseling) that were coupled with smoking cessation pharmacotherapy were influential of medication adherence.^{12,14,39,45,68} Two studies found that attending a greater number of adjunctive counseling sessions increased the likelihood of adherence to NRT³⁵ and bupropion.⁵⁹

Discussion

A variety of factors associated with adherence to smoking cessation pharmacotherapies were identified from the literature; the quality and support for the associations is varied. Below, we discuss potential explanations for observed associations between identified correlates and poor smoking cessation medication adherence and suggest potential solutions to these barriers.

Facilitating Medication Adherence in Light of Identified Correlates

Nonpreventable Characteristics

Despite being classified as “nonpreventable,” it is possible to tailor interventions to individuals who, because of these characteristics, are at increased risk for nonadherence. For instance, individuals with characteristics associated with poorer adherence may benefit from higher intensity, longer duration, or adjunctive interventions to improve smoking cessation pharmacotherapy adherence. Additionally, employing qualitative and mixed-methods research can be advantageous and effective when designing interventions for specific populations. Shadel et al.,⁵¹ for instance, conducted two phases of qualitative research to assess barriers to and facilitators of adherence to the nicotine patch in order to develop an intervention to improve patch adherence among Latino smokers living with HIV/AIDS. Preliminary results from this intervention indicate superior adherence with the nicotine patch and smoking abstinence among individuals in the adherence intervention condition compared to the standard condition.⁶⁹ To effectively inform the development of targeted interventions for populations of smokers presenting with nonpreventable characteristics that adversely influence medication adherence, more research is needed to assess which factors mediate the association between these characteristics and medication adherence.

Preventable Characteristics

The perceived lack of need for cessation aids^{15,34,38,43,48,54,55} and beliefs that cessation aids do not help with cessation^{11,15,34,37,38,47,52,54,56} were commonly cited as reasons for and were consistently associated with medication nonadherence. These findings are potentially indicative of inadequate knowledge regarding nicotine—specifically in the form of NRT⁷⁰—and other smoking cessation pharmacotherapies. Given that educational interventions show some promise for improving adherence to antiretroviral medications among HIV-positive patients,^71–73 these findings suggest the need for the systematic assessment of patient understanding regarding smoking cessation pharmacotherapies, and education to combat misperceptions and knowledge gaps, both within the context of clinical care and broader educational campaigns.

Though adverse effects often occur during a medication regimen, communication regarding accurate information about what patients and participants can expect, and provision of strategies for coping with adverse effects may help to decrease fears about and severity of adverse effects. Education regarding expected adverse effects is particularly important when considering that patients may confuse adverse effects of pharmacotherapy with routine nicotine withdrawal symptoms and needlessly discontinue medication. Additional clinical strategies in response to adverse effects that may improve adherence include modification of the prescribed dose and changing medications.⁷⁴

Regarding forgetfulness^{34,36–38,55} as a barrier to adherence, one possible solution includes mobile phone application- or text message-based reminder systems that prompt patients to take medication and use NRT as directed, which have been shown to be effective, at least in the short-term.⁷⁵ Smoking cessation pharmacotherapies are often time-limited, particularly in the case of varenicline and bupropion (ie, standard courses of treatment for both drugs is 12 weeks); as a result, electronic reminders may be particularly well-suited to these cessation modalities.

Additionally, regarding financial barriers to medication adherence, though NRT and bupropion or varenicline are often provided at no cost within the context of a scientific study, the cost of these pharmacotherapies remains a concern in real-life settings. Within clinical settings, a number of options to ameliorate concerns regarding cost are available, including: selection of a different medication or generic version of the treatment in question—when available; identification of a payment program for nongeneric drugs; and identification of a local low-cost drug program.⁷⁴

Challenges and Future Directions

Beyond the findings detailed above, despite the importance of this issue, little is known about the reasons for and correlates of nonadherence due to (1) inconsistent definitions of adherence, (2) lack of assessment of correlates of adherence, and (3) understudied factors that influence adherence.

Inconsistent Definitions and Measurement of Adherence

Inconsistent definitions of adherence and processes by which adherence is measured can lead to complications: Dunbar-Jacob and Rohay⁷⁶ summarize evidence indicating that measurement methods display a systematic bias, with self-report methods tending to reflect higher levels of adherence than other methods. The authors note that different measurement methods yield different predictors of adherence. The matter is further complicated by the definition of adherence varying from pharmacotherapy to pharmacotherapy, and oftentimes varies within studies of a single treatment modality. Within the present review, nicotine patch adherence has been defined as use of all patches distributed,¹⁴ the proportion of patches used and dichotomized to evaluate a cutoff (ie, 80% of patches used),^31,44 or the total number of patches used, evaluated as a continuous variable.⁴⁷ Adherence to bupropion and varenicline is often reported with an 80% adherence cutoff imposed.^27,28,67 Adherence to varenicline has also been quantified based on “purposeful nonadherence” (ie, not adhering to the medication regimen because the patient feels better or experiences adverse effects),¹⁵ and adherence to bupropion has also been measured based on full, partial, and nonadherence during the past 7 days.⁴² Hollands et al.^16,77 have suggested the following: (1) making a distinction between overall pharmacotherapy consumption/use and adherence to a prescribed regimen—and reporting on both outcomes, as a prescribed regimen may not be optimal; and (2) utilizing a continuous outcome (eg, proportion of medication consumed) versus a dichotomous variable indicating a satisfactory level of adherence (eg, 80% of pills taken).

There is great variability regarding how adherence is measured. Several studies reviewed here utilized pill or patch counts,^{16,28,31,42,43} while only two studies reviewed utilized biochemical verification (ie, blood plasma levels) of adherence to bupropion³² and naltrexone.⁷⁸ However, most studies utilized self-report as the primary measure of adherence. Though self-report is convenient and easy to administer⁷⁹—and one study suggests that pill counts are a valid and reliable measure of varenicline adherence⁸⁰—biological tests such as blood levels of medication metabolites are not subject to misreporting or response bias, and are therefore generally considered to be the most accurate measure of medication adherence.⁸¹ One caveat is that these measures are invasive and costly and may not be available for all types of smoking cessation aids (eg, NRT), limiting their practicality. As such, we recommend that objective methods (ie, blood plasma levels) for measuring medication adherence be used where feasible, and patch and pill counts be used in conjunction with self-reported adherence when biochemical measures are infeasible.

Assessment and Reporting of Correlates of Adherence

Correlates of adherence to pharmacological interventions are often unreported in the literature. For example, specific to medication discontinuation, publications resulting from clinical trials often include CONSORT diagrams that detail dropout at various stages of the study—and sometimes list reasons for dropout. However, predictors of discontinuation are not usually assessed. Furthermore, among studies that do report correlates of adherence, variability between studies in terms of design limit our ability to make direct comparisons between trials. Additionally, the populations of smokers in which correlates of adherence have been studied have been mixed. Limited amounts of research among a variety of populations does not lend itself well to draw conclusions regarding determinants of adherence, particularly when it is likely that these determinants vary based on the population under study.

Consideration of Other Factors that Influence Adherence

In addition to adherence measurement methods affecting which correlates of adherence are identified,⁷⁶ there is a great deal of heterogeneity across studies regarding which potential correlates are examined and frequency with which they are examined. Sociodemographic characteristics are considered as potential correlates of adherence in most of the identified studies, while genetic^53,54 and personality characteristics³⁷ were examined infrequently. Determinants of adherence to pharmacological smoking cessation interventions may differ based on the population being assessed. Vulnerable smokers (eg, substance users, individuals with chronic health conditions), for example, may face difficulties with adherence in addition to those faced by general population smokers. Individuals with physical health conditions that require medication management (eg, HIV) experience difficulties with medication adherence to begin with^82–84; putatively, the addition of another medication regimen (ie, for smoking cessation) further complicates medication self-management. The infrequent examination of medication adherence and correlates thereof among vulnerable smokers is problematic, given that vulnerable smokers often face a disproportionate burden in terms of overall smoking prevalence and smoking-related morbidity versus general population smokers.^85–88 Moreover, many trials utilizing medication for smoking cessation exclude participants with mental and physical health comorbidities.⁸⁹ As a result, identified correlates of adherence may not generalize to vulnerable populations of smokers. This highlights the need for additional research to investigate correlates of adherence among larger segments of the population, including vulnerable smokers, which will contribute to the development of interventions to increase adherence to smoking cessation interventions. Moreover, there is a need for the use of models and theoretical frameworks to drive the selection of factors relevant to medication adherence for study.

Additionally, current Clinical Practice Guidelines²³ for treating tobacco dependence indicate that the combination of counseling and smoking cessation medication is more effective for cessation than either approach alone. Given this, in addition to findings from the present review showing that characteristics of adjunctive treatments were found to be associated with adherence,^12,39,68 attention should be paid to the role of psychosocial intervention components, and how they may improve adherence to and effects of smoking cessation pharmacotherapies. Related to this, limited research has shown that greater counseling session attendance^35,59 and making a greater number of health care visits⁵ are associated with an increased likelihood of medication adherence. However, no additional studies included in this review reported on health care and health care professional-related factors (eg, providers’ attitudes and beliefs) and how these factors may contribute to adherence. Future work should examine the impact that these types of characteristics have on smoking cessation medication adherence.

Recommendations for Future Research and Clinical Practice

One of the most significant barriers to studying adherence to smoking cessation pharmacotherapies is the lack of consistent definitions regarding adequate adherence.⁷⁶ Within the tobacco control field, there is a need for consensus guidelines regarding standardized and gold standard definitions of adherence,^16,77 both within individual cessation modalities as well as across cessation pharmacotherapies. This will greatly increase the generalizability of findings and our ability to compare rates and correlates of adherence across studies in a meaningful way. Moreover, definitions of adherence should be based on theoretical considerations, including the mode of action of pharmacotherapies. Hollands et al.^16,77 have made suggestions regarding this topic (eg, utilizing continuous versus dichotomous measures of adherence), which represent an important starting point. However, in-depth discussion of this topic may merit the formation of a subcommittee via the Society for Research on Nicotine and Tobacco to develop guidelines on the measurement and reporting of adherence.

In a related vein, including other methods of assessing adherence, in addition to self-report, is advantageous. Though gold standard methods, like biochemical verification (eg, blood plasma levels of medication) of adherence should be used whenever possible, pill and patch counts are often much more practical. In addition to practicality, research indicates that pill counts are reliable and valid measures of adherence;⁸⁰ as a result pill/patch counts should be used in addition to self-reported adherence.

The assessment and reporting of adherence—and its correlates—to smoking cessation pharmacotherapies, within the context of research studies and clinical practice, is essential. One strategy for accomplishing this—in both contexts—is through the use of brief self-report measures, such as the Adherence Starts with Knowledge questionnaire (ASK-20),⁹⁰ which allows researchers to collect information about adherence barriers with little additional burden on study/clinical staff or participants/patients. There is a need for assessment of all phases of medication adherence, as defined by Vrijens et al.,⁴ the majority of the studies—40 studies—evaluated by this review concerned the process of implementation, while 4 studies concerned discontinuation; 3 studies examined aspects of both processes. Only one study evaluated initiation of pharmacological smoking cessation interventions—albeit indirectly, via initial dispensing of cessation medications via electronic health records and pharmacy claims data.

Given the lack of research examining correlates of adherence to smoking cessation pharmacotherapies additional research is needed. In particular, research on this topic among vulnerable populations of cigarette smokers is lacking. This is significant given that vulnerable cigarette smokers bear a significant burden of smoking-related morbidity and mortality, and often face difficulties with adherence above and beyond those faced by smokers in the general population. Additional work in this area will lay the foundation for developing novel interventions to improve adherence to smoking cessation pharmacotherapies, which have the potential to increase the effectiveness of existing smoking cessation interventions.

Study Limitations and Strengths

There are some limitations to the current study. We cannot rule out the possibility for publication and selection biases, as the inclusion criteria for this review focused on studies published in English and excluded conference abstracts and off-line publications. Additionally, because we did not conduct a systematic review, it is possible that some relevant studies may have been excluded and that our findings may be susceptible to some biases. Though described separately in Supplementary Tables 2 and 3, to make a first pass at the literature, the present review gives equal weight to self-reported reasons and correlates of adherence—regardless of whether correlates were derived from multivariable or bivariable models. Future work may wish to distinguish between the two types of factors, and assign weight to factors based on whether they are derived from adjusted or unadjusted models, accordingly. Most papers came from studies conducted in the United States, with relatively few studies from Europe and Asia. Moreover, though this review utilizes a WHO classification system of nonadherence,⁴ it is worth noting that different approaches to classifying nonadherence have emerged.⁹¹ Additionally, this review yielded sometimes inconsistent findings regarding the directionality of associations between correlates and adherence; these inconsistencies may be, at least in part, due to a variety of study types (ie, both prospective and cross-sectional) being reviewed. Despite these limitations, to our knowledge, this review represents the first to examine and synthesize correlates of smoking cessation medication adherence.

Conclusion

In conclusion, we identified 50 studies that assessed self-reported reasons for and correlates of smoking cessation medication adherence. Within these studies, numerous preventable and nonpreventable characteristics associated with smoking cessation medication adherence were identified. The directionality of these associations is often inconsistent, though they have been summarized within this review. Challenges to this field include unstandardized definitions of adherence, reliance on self-report, and little research among vulnerable populations. Understanding characteristics that contribute to medication nonadherence is key to improve adherence and improve smoking cessation outcomes.

Supplementary Material

Supplementary data are available at Nicotine and Tobacco Research online.

Funding

This work was supported by the National Institutes of Health (T32 AI007392 and K01DA043413) and a Research Career Scientist Award from VA Health Service Research and Development (VA HSR&D 08-27 to HB).

Declaration of Interests

None declared.