Cost-Effectiveness of Community-Based Tobacco Dependence Treatment Interventions: Initial Findings of a Systematic Review

Introduction Scientific literature evaluating the cost-effectiveness of tobacco dependence treatment programs delivered in community-based settings is scant, which limits evidence-based tobacco control decisions. The aim of this review was to systematically assess the cost-effectiveness and quality of the economic evaluations of community-based tobacco dependence treatment interventions conducted as randomized controlled trials in the United States. Methods We searched 8 electronic databases and gray literature from their beginning to February 2018. Inclusion criteria were economic evaluations of community-based tobacco dependence treatments conducted as randomized controlled trials in the United States. Two independent researchers extracted data on study design and outcomes. Study quality was assessed by using Drummond and Jefferson’s economic evaluations checklist. Nine of 3,840 publications were eligible for inclusion. Heterogeneity precluded formal meta-analyses. We synthesized a qualitative narrative of outcomes. Results All 9 studies used cost-effectiveness analysis and a payer/provider/program perspective, but several study components, such as abstinence measures, were heterogeneous. Study participants were predominantly English speaking, middle aged, white, motivated to quit, and highly nicotine dependent. Overall, the economic evaluations met most of Drummond and Jefferson’s recommendations; however, some studies provided limited details. All studies had a cost per quit at or below $2,040 or an incremental cost-effectiveness ratio (ICER) at or below $3,781. When we considered biochemical verification, sensitivity analysis, and subgroups, the costs per quit were less than $2,050 or the ICERs were less than $6,800. Conclusion All community-based interventions included in this review were cost-effective. When economic evaluation results are extrapolated to future savings, the low cost per quit or ICER indicates that the cost-effectiveness of community-based tobacco dependence treatments is similar to the cost-effectiveness of clinic-based programs and that community-based interventions are a valuable approach to tobacco control. Additional research that more fully characterizes the cost-effectiveness of community-based tobacco dependence treatments is needed to inform future decisions in tobacco control policy.


Introduction
Scientific literature evaluating the cost-effectiveness of tobacco dependence treatment programs delivered in community-based settings is scant, which limits evidence-based tobacco control decisions. The aim of this review was to systematically assess the cost-effectiveness and quality of the economic evaluations of community-based tobacco dependence treatment interventions conducted as randomized controlled trials in the United States.

Methods
We searched 8 electronic databases and gray literature from their beginning to February 2018. Inclusion criteria were economic evaluations of community-based tobacco dependence treatments conducted as randomized controlled trials in the United States. Two independent researchers extracted data on study design and outcomes. Study quality was assessed by using Drummond and Jefferson's economic evaluations checklist. Nine of 3,840 publications were eligible for inclusion. Heterogeneity precluded formal meta-analyses. We synthesized a qualitative narrative of outcomes.

Introduction
In the United States, tobacco use, the leading cause of preventable morbidity and mortality (1), poses a substantial economic burden. Nationwide, smoking causes an estimated $332 billion in annual lost productivity and direct health care costs (1,2). Elimination of combustible tobacco use will dramatically reduce this burden (1).
Community-based cessation approaches, such as quitlines, internet, and self-help interventions, are important components of comprehensive tobacco control. Although community-based initiatives may have lower cessation rates (3) than clinic-based programs, they are effective as they engage more participants than do clinic-based programs, including underserved and hard-to-reach smokers (4-6). Because community-based interventions are often less expensive to deliver, they may be as cost-effective as clinicbased interventions. Furthermore, community-based programs offer an opportunity for tailored behavioral interventions to facilitate optimal outcomes and reduce tobacco use disparities (7).
Cessation interventions can immediately affect the economic and public health consequences of tobacco use (8). Yet comprehensive tobacco control efforts, which are typically organized at the state level and include administrative, surveillance, evaluation, and monitoring components, incur high implementation costs. These costs are allocated toward changing social norms, increasing cessation, and reducing exposure to secondhand smoke. It is essential that decision makers have tools such as economic evaluation to make informed decisions about how to allocate resources.
Few economic evaluations of controlled community-based trials in the United States have been published, and no comprehensive literature review exists. This void limits evidence-based decision making and resource allocation and negatively affects health outcomes. This review addresses the void by systematically assessing the cost-effectiveness and quality of economic evaluations of community-based tobacco dependence treatments conducted as randomized controlled trials in the United States. This report potentially informs tobacco control policy, contributes to populationlevel strategies, and improves future economic evaluations by illustrating best and current practices.

Methods
We searched 8 databases (MEDLINE/PubMed, Tufts Cost-Effectiveness Analysis Registry, EMBASE, CINHAL, Scopus, Web of Science, Global Health, and the NHS Economic Evaluation Data-base) by using medical subject headings and synonyms for tobacco dependence treatment, economic evaluation, and randomized controlled trial. We searched databases from their beginning to February 2018.

Study selection
Selection criteria were structured by population, intervention, comparator, outcome, and study design (Box). We included economic evaluations that randomized individuals or communities to intervention or control groups. We excluded review articles but examined them for studies that met criteria. All outcome time points, perspectives (ie, societal, payer/provider/program, or individual), and types of economic evaluations were included. We also included studies without biochemical verification. Our review focused on community-based tobacco dependence treatments, such as those delivered through quitlines/telephone counseling, in-person counseling, postal mail, and the internet, and included alternatives to traditional, medically trained clinical interventions. Although "community-based" has several definitions, we defined it as programs that 1) referred to a community as a setting or a geographic location where interventions were implemented, 2) included a variety of approaches, levels, and locations, and 3) focused on changing individuals' behavior as a method of reducing population risk (9). Locations included state quitlines and local and population-based residents throughout the United States (reached by postal mail, internet, telephone counseling, and in-person counseling). Only studies with adult smokers and abstinenceframed measures such as cost per quit and incremental cost-effectiveness ratio (ICER) (ie, attributable cost per quit) were included. We excluded special or targeted populations such as adolescents, employees, insured, and hospitalized patients; studies conducted outside of the United States; and studies solely reporting outcomes in terms of enrollees, patients, or recruitment. We removed duplicates, and 2 researchers (S.A.R., S.K.) independently performed an initial screening of titles and abstracts. Publications remaining were reviewed in entirety to determine eligibility. References in included studies were reviewed. We contacted authors for additional information when necessary. The same 2 researchers independently abstracted data and assessed quality. Any disagreement or uncertainty was resolved through discussion or a third independent researcher (M.E.W.).
Initial searches returned 3,480 articles. After removing 993 duplicates, we screened titles and abstracts of 2,487 records. Of these, 2,419 were excluded, leaving 68 full-text articles that were systematically examined for the following criteria in the following order: presence of an economic evaluation, randomized controlled trial, community-based, and set in the United States. Of these 68 studies, 9 that were published in peer-reviewed journals met inclusion criteria (Figure).

Figure.
Article search and selection process using PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses). "Records screened" are titles and abstracts.

Data extraction
Two researchers (S.A.R., S.K.) independently abstracted data for 1) research question, objective, or hypothesis, 2) location or setting, 3) population description, 4) study design and intervention description, 5) intention-to-treat analyses, 6) summary effects, 7) type of economic evaluation, 8) perspective, 9) the analytic horizon, or the duration of the study time frame including outcome assessment (22), 10) type of abstinence measure used for the economic evaluation, 11) source of the valuation of resources (eg, invoices, contracts, website data) used to determine costs, or how components of intervention were valued, 12) sensitivity analyses, 13) economic evaluation results, 14) subgroup analyses, 15) generalizability, and 16) limitations. We assessed the quality of each study by using Drummond and Jefferson's economic evaluation checklist (10), which consists of 35 questions that use 4 response options (yes, no, not clear, and not appropriate) to assess study design, sources and quality of data collected, data analysis, and interpretation of results. We noted any items not explicitly addressed but inferred from the text. We synthesized a qualitative narrative of outcomes. We assessed inter-rater reliability by calculating percentage agreement, expected agreement, and the Cohen κ statistic in Stata version 12 (StataCorp LLC). Cohen κ, values for which range from −1 to +1, represents the proportion of agree-ment after chance is excluded; larger values indicate better reliability. The statistic accounts for the possibility of guessing, but the assumptions of rater independence and other factors may lower the estimate of agreement, and interpretation and accepted values can vary by discipline (11,12).

Results
For the initial review of titles and abstracts, the agreement rate was 98.6%, the expected agreement was 94.3%, and Cohen κ was 0.75. The agreement for full article review was 91.2%, the expected agreement was 72.7%, and Cohen κ was 0.68. The agreement for reasons of exclusion was 92.6%, the expected agreement was 76.3%, and Cohen κ was 0.69. The agreement for quality assessment was 93.9%, the expected agreement was 31.0%, and Cohen κ was 0.91.
Study populations. Study populations varied; generally, participants were English-speaking, middle aged, white, motivated to quit, and highly nicotine dependent. Key factors that differed were number of cigarettes consumed or level of nicotine dependence, smoking history, interest or willingness to receive or use NRT, willingness to quit smoking, and sociodemographic characteristics.
Costs. Overall, studies provided various levels of detail on costs (Table 2). Four studies indicated the resource that was used to determine how the component was valued (ie, source of the valuation) (13,14,19,21). When reported, information on costs was typically obtained from prevailing commercial costs, economies of scale, or study records. Most studies reported aggregate data with various levels of detail about which intervention components were included.
Analytic horizon. The analytic horizon was explicitly addressed by 1 study (14) and easily understood in all other studies. Horizons varied from 3 to 24 months; most were 6 or 7 months (13-16,18), and one was a single 24-month follow-up study (20). One study had multiple periods of assessment (3, 6, 12, and 18 months) (14).
Discounting. Discounting was not performed in any of the studies but was discussed in one (14).

Cost-effectiveness.
All studies conducted cost-effectiveness analyses, used cost per quit and/or ICER to present findings, and answered the study question posed. The combination of cost per quit and ICER was commonly used (14-16,18). Two studies calculated cost per quit only (13,21) and 3 studies reported ICER only (17,19,20). Studies compared relevant alternatives with various levels of detail (13)(14)(15)(16)(18)(19)(20)(21). Data and results were typic-ally presented in disaggregated and aggregated form to allow readers to calculate other ratios (14-21). Conclusions were provided for all but 1 study (17).
Despite various abstinence definitions, assessment time points, and scale of tobacco dependence treatments, the cost-effectiveness estimates were relatively similar (Table 3). Overall, when combining all studies, despite settings and methodology differences, cost per quit ranged from $5 (14) to $2,040 (16) and the ICER or cost per additional quit ranged from $357 (15) to $3,781 (14). When considering sensitivity analyses and the use of biochemically verified data, the ICER increased to $6,781 (19).
Cost per quit. When comparing combination therapy to monotherapy and duration of NRT therapy, using actual estimates only (excluding subgroup and sensitivity analyses), cost per quit ranged from $102.44 (18) (2 weeks of patch only) to $675 (15) (6 weeks of combination therapy). In studies examining duration of patchonly interventions, cost per quit ranged from $883 (4 weeks) (13) to $2,040 (8 weeks) (16). When examining dose or intensity of self-help interventions, cost per quit ranged from $5 for basic internet to $1,882 for enhanced internet plus phone (14). Conditions such as enhanced internet (14), mailed leaflets, leaflets plus maintenance manual, cessation manual, and cessation manual plus maintenance manual (21) were comparable.
ICER. Similar to cost per quit, when comparing combination therapy, monotherapy, and duration of NRT therapy, using actual estimates only, the ICER ranged from $357 (15) (2 weeks of combination therapy) to $3,131 (16) (8 weeks of patch only). When considering sensitivity analyses and the use of biochemically verified data, the range increased to $6,781 (19). When examining dose or intensity of self-help interventions, the ICER ranged from $361 (20) (intensive repeated mailings) to $3,781 (14) (enhanced internet plus phone). Compared with mailed self-help booklets alone, the cost per additional quit attributable to telephone counseling availability was approximately $1,300 (17).
Subgroup analysis. Three studies included subgroup analyses in the economic evaluation, including a comparison of combination therapy and monotherapy among uninsured participants (18), differences in abstinence by intervention use (14), and the effect of requesting 2 NRT shipments (13).
Generalizability. Two studies reported that the study design translated to real world or national samples (15,20). One study noted uncertainty about whether the sample would generalize to other smokers (14); others indicated the eligibility criteria (13,19) or intervention type (21) might limit the representativeness.

Discussion
Our review systematically assessed the cost-effectiveness and quality of the economic evaluations of community-based tobacco dependence treatment interventions conducted as randomized controlled trials in the United States. The studies reviewed addressed most of Drummond and Jefferson's recommendations for authors and peer reviewers of economic submissions; however, some studies provided limited details. Based on cost-effectiveness estimates reported among studies of predominantly middle-aged, white, motivated to quit, and highly nicotine dependent populations, basic internet had the lowest cost-effectiveness ratio. However, all interventions, even when considering biochemical verification, sensitivity analysis, and subgroup analysis, had a cost per quit of less than $2,050 or an ICER of less than $6,800. Thus, considering the most commonly accepted conservative threshold of $50,000 per quality-adjusted-life-year (QALY), all community-based interventions were cost-effective.
According to Drummond and Jefferson, cost-effectiveness comparisons among health care interventions should be made only when methods and settings are closely aligned (10). In our review, no studies used QALYs or a similar outcome, which would have allowed for direct comparison with studies on other health conditions. However, the cost-effectiveness estimates of communitybased interventions in our review (ie, cost per quit or ICER at or below $2,040 or $3,781, respectively) were similar to estimates in studies of clinical settings (ie, a few hundred to a few thousand dollars per quit [23][24][25][26][27][28][29]). Although community-based tobacco dependence treatment cost-effectiveness estimates in our review align with previous clinical findings and findings of nonrandomized controlled studies, more research is needed before definitive comparisons can be made.
Treatment for tobacco use is considered the gold standard of health care cost-effectiveness (30). Ranging from brief clinician advice to specialist-delivered intensive programs, including NRT or other medications, such treatment has been shown to be highly cost-effective (23). Population-wide policy, systems, and environ- mental changes, such as increases in the unit price of tobacco products, comprehensive smoke-free policies, and media campaigns increase cessation rates by motivating users to quit, increasing demand for tobacco dependence treatment, and making it easier to quit (7,8,(31)(32)(33). These approaches are most efficient and effective at reaching many people (7,31,33); however, policies and media campaigns require substantial resources, and like clinical interventions, they are often not tailored to specific populations in need, do little to directly address disparities in tobacco use and tobacco-attributable health outcomes, and fail to engage hard-toreach populations, such as those with social, economic, or geographic constraints.
In contrast, community-based programs are frequently tailored to targeted audiences. Tailored programs and messages are more effective than standardized, nontailored interventions (34), and have a greater effect on health behavior because they are often perceived as more relevant than generic communication (35-37). These benefits in reach and efficacy, which occur largely through engagement in evidence-based community interventions, can help reduce disparities in tobacco use (7). Furthermore, communitybased programs, such as those using community health workers, have strong potential for improving health outcomes (38). Although clinical and policy-based approaches are important, they are also labor intensive, costly, and do little to address subgroups of populations who are underserved. Community-based interventions are poised to focus on those gaps. This review shows community-based tobacco dependence treatment approaches are both effective and cost-effective and an important component of comprehensive tobacco control. With this additional information on the cost-effectiveness of community-based interventions, efforts should continue to promote such strategies.
Our study has several strengths. The systematic review used a comprehensive search strategy that was informed and reviewed by an academic research librarian. We used 2 independent researchers and a third to resolve conflicts with high interrater agreement. The search was performed in 8 databases and gray literature, which produced substantial overlap in returned results. The quality assessment used Drummond and Jefferson's established criteria and checklist, which is recommended in Cochrane reviews to inform appraisal of the methodological quality of economic evaluations (39). The study was limited in scope to include economic evaluations of community-based studies of tobacco dependence interventions conducted as randomized controlled trials in the United States. We identified a considerable number of studies outside the United States, but differences across countries' cost structures and health systems would have limited the validity of our review (10). Randomized controlled clinical trials were selected as the gold standard for estimating efficacy of treatment interventions and greatest internal validity.
Our study has several limitations. We excluded several target populations (ie, adolescents, employees, insured, and patients) and other settings (ie, clinical settings and worksites) to increase the reliability of the findings. The only groups included a priori were low-resource groups, such as the uninsured. Socioeconomically disadvantaged populations underutilize smoking cessation treatments and are studied infrequently. Although some groups (eg, mentally ill, people who use drugs) could potentially be classified as low-resource, we determined that the underlying conditions would further complicate generalization. A broad body of research on those groups could be reviewed separately.
Cost-effectiveness ranges in our review were generalities; the ranges were not adjusted, because we lacked dollar-year data. However, when we used presumed dollar-years, adjusted cost-effectiveness estimates for the oldest study (21) and the study with the highest reported cost per quit (16) were below $2,700 in 2019 dollars. These approximations demonstrate that our findings would retain low cost-effectiveness ratios if dollar-year data were available to adjust costs to a single referent year. We could not directly compare the cost-effectiveness estimates among the 9 studies in our review because of heterogeneous study components and the lack of a common base case among study interventions, which complicate the comparison of studies and the ability to draw definitive conclusions. The heterogeneity precluded meta-analysis; rather, we performed a narrative synthesis and quality assessment. This approach is consistent with previous studies, and in practice, economic evaluations addressing a particular question do not generally present sufficient detail to permit adjustments required for a meta-analysis (40). Additionally, critical appraisal of the quality and reporting of health economic studies also vary. Because no minimum methodologic criteria or scoring systems exist, our evaluation of study quality was subjective.
Our review demonstrates that although smoking is an important public health issue, few economic evaluations of communitybased tobacco dependence treatment programs in the United States have been conducted. However, the cost-effectiveness estimates found in our review are a fraction of the generally accepted costeffectiveness threshold of $50,000-per-QALY. Given the degree to which tobacco influences numerous health outcomes and productivity, when the results are extrapolated to future savings, the low cost per quit or ICER is evidence that community-based cessation interventions are of great value and good investments for health.  (41). Funding for community-based programs is inadequate, and such limited support will continue to exacerbate smoking-related health care costs and disparities.
Our review revealed a lack of focus on economic evaluation of community-based tobacco dependence treatment programs in the United States. We demonstrated these programs are cost-effective. It is vital to expand this literature given the effect of tobacco on health and costs. To improve the state of the science and understanding of results, as well as identify potential opportunities for uptake of interventions, such studies should be conducted. Future studies should adhere to and explicitly address key components of Drummond and Jefferson's economic evaluation guidelines. Adding consistent base cases and valuation, subgroup analyses, cost-utility analyses, and standardizing approaches, especially related to measurement (eg, abstinence and costs) will improve comparability between studies and expand use in policy decisions. Future tobacco control economic evaluation research should also examine populations and areas with a high prevalence of smoking (eg, low-resource groups), in-person counseling, and mobile interventions. Tested the provision of 2 weeks of either combination therapy (nicotine patch and nicotine lozenge) or monotherapy (nicotine patch alone) to heavy smokers calling the quitline.
Adult residents of New York State (≥18 y); self-identified as current daily tobacco user; ≥20 CPD; contacted the New York State Smokers' Quitline; interested in using NRT; no known contraindications for NRT; and scored either 5 or 6 on the heaviness of smoking index (range 0-6).
Responder quit rates according to the protocol were higher for those given combination therapy than for those given monotherapy for 7-day (24% vs 21%) and 30day (21% vs 18%) point prevalence abstinence, although they were not significantly different (P > .05). Results did not differ between intention-to-treat and according-to-protocol analyses.
Smith et al (15), 2013 Type of NRT used and duration of use, quitline/ telephone counseling Abbreviations: CI, confidence interval; CPD, cigarettes per day; NRT, nicotine replacement therapy; OR, odds ratio. a 1,503 study participants were enrolled, but 8 were never sent NRT, resulting in 1,495 study participants. b A priori sample size was 600; 575 were randomized, but 7 people were ineligible because of medical contraindications after randomization and excluded from intention-to-treat analyses.
(continued on next page) Conducted an economic evaluation of The iQUITT study, a randomized trial comparing basic internet, enhanced internet, and enhanced internet + phone counseling at 3, 6, 12, and 18 months.

PREVENTING CHRONIC DISEASE
Smokers recruited through active user interception sampling (entered terms "quit(ting) smoking," "stop(ping) smoking," or "smoking" in a major internet search engine and clicked on a link to www.quitnet.com); US residence; ≥5 CPD; ≥18 y; no prior use of QuitNet.
30-day point prevalence abstinence rates increased over time. Significant betweengroup differences in point prevalence were observed at 3, 6, and 12 months, but not at 18 months. Post hoc comparisons showed enhanced internet + phone outperforming the other 2 conditions at 3 and 6 months, and enhanced internet at 12 months (P < .003). The difference between enhanced internet and basic internet for point prevalence abstinence was not significant at any follow-up. 30-Day multiple point prevalence abstinence rates declined across groups over time. Significant betweengroup differences in multiple point prevalence were observed at all follow-ups with enhanced internet + phone outperforming other conditions. Difference between enhanced internet and basic internet for multiple point prevalence abstinence was not significant at any time. Abbreviations: CI, confidence interval; CPD, cigarettes per day; NRT, nicotine replacement therapy; OR, odds ratio. a 1,503 study participants were enrolled, but 8 were never sent NRT, resulting in 1,495 study participants. b A priori sample size was 600; 575 were randomized, but 7 people were ineligible because of medical contraindications after randomization and excluded from intention-to-treat analyses.
(continued on next page)  Maryland;and New York, New York;1979-1981 Examined long-term results of self-help smoking cessation programs involving no face-to-face contact during treatment.
Smokers responding to lung association announcements using standard newspaper advertisements provided by American Lung Associate, flyers, and media announcements.
Abbreviations: CI, confidence interval; CPD, cigarettes per day; NRT, nicotine replacement therapy; OR, odds ratio. a 1,503 study participants were enrolled, but 8 were never sent NRT, resulting in 1,495 study participants. b A priori sample size was 600; 575 were randomized, but 7 people were ineligible because of medical contraindications after randomization and excluded from intention-to-treat analyses.  Costs for recruitment, training of interviewers, and the telephone follow-up interviews. Costs were evenly allocated among all 4 experimental groups (leaflets; leaflets + maintenance manual; cessation manual; cessation + maintenance manuals). Costs: $12,451 recruitment; $4918.50 interview, assuming $4.50 hourly wage and 1,093 hours, utilities not included. Total direct costs for staff time were $1,700 with training costs prorated for 11 days using an assumed $20,000 annual salary.

PREVENTING CHRONIC DISEASE
Costs for the printing (not development) of the self-help materials, handling, and postage costs at the time of the study (1979)(1980)(1981), which were prorated to the number of participants in each group, but data/values were not provided. Costs of recruitment, available from 4 of the test sites, were statistically imputed for the fifth site (in Salinas, California) by prorating the average from the other 4 sites by the number of participants recruited at that site. a Resources (eg, invoices, contracts, website data) used to determine costs, or how components of intervention were valued.