Behavioral Counseling and Pharmacotherapy Interventions for Tobacco Cessation in Adults, Including Pregnant Women: A Review of Reviews for the U.S. Preventive Services Task Force

Cigarette smoking and exposure to smoke result in more than 480 000 premature deaths in the United States every year, along with substantial illness (1, 2). Despite considerable progress in tobacco control over the past 50 years, in 2013, an estimated 17.8% of U.S. adults (3) and 15.9% of pregnant women aged 15 to 44 years were current cigarette smokers (4). Many tools are available to help smokers quit, including counseling by health care providers, telephone- and print-based interventions, computer and text-messaging interventions, and pharmacologic agents (that is, nicotine replacement therapy [NRT], bupropion hydrochloride sustained release [bupropion], and varenicline).

In 2009, the U.S. Preventive Services Task Force (USPSTF) reaffirmed its 2003 recommendation that clinicians ask all adults about tobacco use and provide interventions for cessation for those who use tobacco products (grade A recommendation) (5). The original USPSTF recommendation (2003) and reaffirmation (2009) were based on the Public Health Service's clinical practice guidelines on treating tobacco use and dependence (6, 7). Because there were no plans to update the Public Health Service report, we undertook the current review to assess the benefits and harms of behavioral and pharmacologic interventions for tobacco cessation in adults, including pregnant women, to assist the USPSTF in updating its 2009 recommendation. Because of the rapid increase in the use of electronic nicotine delivery systems (ENDS) and the vigorous debate about the public health effect of these devices and their role in smoking cessation (8–13), our review also synthesized the primary trial evidence on the efficacy and safety related to this technology as a means for quitting conventional smoking.

Methods

We relied primarily on a review of reviews method for this update. We did not replicate quality rating or data abstraction for original studies or replicate review-specific analyses. However, we decided a priori to conduct a de novo search for primary evidence related to the effectiveness and safety of ENDS. In addition, we did a bridge search for evidence related to pharmacotherapy interventions among pregnant women because of the limited number of studies included in the available systematic reviews and the length of time that had elapsed since their last search dates.

We developed an analytic framework and 3 key questions with input from the USPSTF (Appendix Figure 1). The final version of the framework and key questions reflects both USPSTF and public input. The full report provides detailed methods (14).

• Download figure
• Download PowerPoint

Data Sources and Searches

We searched the following databases for relevant reviews from January 2009 to 1 August 2014: PubMed, PsycInfo, Cochrane Database of Systematic Reviews, Health Technology Assessment database, and Database of Abstracts of Reviews of Effects of the Centre for Reviews and Dissemination. We also searched the following organizational Web sites: the Agency for Healthcare Research and Quality, the British Medical Journal Clinical Evidence (through 7 August 2013), the Canadian Agency for Drugs and Technologies in Health, Guide to Community Preventive Services, the Institute of Medicine, the National Institute for Health and Clinical Excellence, the National Health Service Health Technology Assessment Programme, and the Surgeon General. We supplemented our searches with suggestions from experts. We searched PubMed for primary evidence related to ENDS through 1 March 2015 and for pharmacotherapy interventions among pregnant women through 15 August 2014 (the full report outlines the search strategies for these 2 searches [14]).

Study Selection

Two investigators independently reviewed all identified abstracts and dually reviewed full-text articles against prespecified eligibility criteria (14). We resolved disagreements through discussion. We included systematic reviews—with or without meta-analysis—that examined the effectiveness of interventions for tobacco cessation for adults, including pregnant women, and were linked to primary care or took place in a general adult population. We excluded nonsystematic meta-analyses and narrative reviews. We also excluded reviews that focused on reduction of tobacco harms, interventions for relapse prevention, or cessation medications that were not approved by the U.S. Food and Drug Administration as first-line medications for cessation (such as nortriptyline). We included only the most recent version of updated reviews. We outlined separate selection criteria when considering primary evidence related to ENDS and pharmacotherapy among pregnant women, as described in the full report (14).

Data Extraction and Quality Assessment

At least 2 independent reviewers rated the quality of all included systematic reviews using a slightly modified version of the Assessment of Multiple Systematic Reviews tool (15, 16) (see the full report for modifications and methods for determining the overall quality rating of individual reviews [14]). We excluded all poor-quality studies (17). One reviewer completed primary data abstraction, and a secondary reviewer checked all data for accuracy and completeness.

Data Synthesis and Analysis

When we found several fair- and good-quality reviews that met the inclusion criteria in a given population and intervention subgroup, we applied criteria (Appendix Table 1) to identify 1 or more reviews that represented the most current and applicable evidence to serve as the basis for the main findings (called “primary reviews”). We reviewed the remaining reviews for complementary or discordant findings. When we encountered discordant bodies of evidence, we sought explanations for these differences by examining the eligibility criteria and included studies within each review.

We used the pooled point estimates presented in the included reviews when appropriate. We did not reanalyze any of the individual study evidence. We evaluated the appropriateness of meta-analytic procedures and used our technical judgment to interpret pooled analyses accounting for limitations or concerns around heterogeneity, statistical approaches (18, 19), and other factors.

Role of the Funding Source

This review was funded by the Agency for Healthcare Research and Quality. Agency staff provided technical oversight for the project. Liaisons from the USPSTF helped resolve issues around the review's scope but were not involved in its conduct.

Results

We reviewed 638 abstracts and 114 full-text reviews for possible inclusion (Appendix Figure 2). We identified 54 systematic reviews that met our eligibility criteria (20–73), and 22 of these served as the basis for the primary findings (Table 1). In general, results across all included reviews were consistent within each population and intervention grouping. Our results are organized by outcomes and subcategories by population and interventions. Eleven of the 54 included reviews synthesized evidence on interventions among specific subpopulations of adults (such as persons with depression and young adults) that are not included here but appear in detail in the full report (14).

• Download figure
• Download PowerPoint

Behavioral Interventions Among Adults

Eleven reviews served as primary reviews examining the effects of behavioral interventions for smoking cessation among the general adult population (Table 1) (21, 22, 31, 37, 55, 58, 60, 61, 67, 71, 78).

Health and Cessation Outcomes

Data on health outcomes after behavioral interventions were limited to 1 study (79) that was reported in 1 review (58) (Table 2). This study reported no statistically significant differences in rates of total mortality, coronary disease mortality, and lung cancer incidence and mortality at 20-year follow-up among men at high risk for cardiorespiratory disease (n = 1445) (80). However, at 33-year follow-up, there were significantly fewer deaths from respiratory illnesses among participants who received an intervention than control participants (58).

Several behavioral interventions increased smoking cessation at 6 months or more, including physician- (58) and nurse-delivered (55) counseling interventions, tailored self-help print materials (37), and telephone counseling (60), when compared with minimal intervention or usual care (Table 2 and Appendix Table 2). Smokers who were offered cessation advice by a physician, for example, were 76% more likely to have quit at 6 months or more than those who received no advice or usual care (risk ratio [RR], 1.76 [95% CI, 1.58 to 1.96]; I² = 40%; 28 trials; n = 22 239) (58). Both minimal and intensive advice (>20 minutes, additional materials beyond a brochure, or >1 follow-up visit) showed statistically significant increases in cessation rates when compared with control participants who did not receive advice. Direct comparisons between intensive and minimal advice in 15 trials suggested that more intensive advice offered a significant advantage (RR, 1.37 [CI, 1.20 to 1.56]; I² = 32%; 15 trials; n = 9775) (58).

A separate meta-analysis of 38 randomized, controlled trials (RCTs) done among more than 15 000 smokers found a small relative benefit of adjunctive behavioral support to pharmacotherapy when compared with pharmacotherapy alone (RR, 1.16 [CI, 1.09 to 1.24]) (61). Cessation rates were relatively high in both the intervention (21.4%) and control (18.3%) groups because both groups received pharmacotherapy (Appendix Table 2).

There was mixed evidence of improved tobacco cessation for the following interventions: nontailored self-help materials (37), interactive or tailored Internet or computer programs (31), mobile telephones (71), biomedical risk assessment (22), exercise (67), acupuncture (70), and hypnotherapy (21) (Appendix Table 2).

Adverse Events

One review reported minor adverse events related to ear acupuncture, ear acupressure, and other auriculotherapy (33). No other reviews found or reported adverse events related to other behavioral or complementary and alternative therapies (Table 2).

Pharmacotherapy Interventions Among Adults

Six reviews served as primary reviews on the effectiveness or harms of NRT, bupropion, or varenicline among current adult tobacco users (Table 1) (40, 47, 49, 54, 59, 73).

Health and Cessation Outcomes

None of the reviews reported the effects of medications for smoking cessation on mortality, morbidity, or other health outcomes. For cessation outcomes, NRT, bupropion, and varenicline all improved rates of smoking cessation in adults at 6-month follow-up or longer (Table 2). Nicotine replacement therapy was effective in all forms and increased relative cessation rates by 53% to 68% when compared with placebo or no NRT (RR, 1.60 [CI, 1.53 to 1.68]; 117 trials; I² = 30%; n = 51 265) (Appendix Table 2) (59). No differences were found among NRT products (such as patch, gum, and lozenge) (59). Combining 2 types of NRT was found to be superior to a single form in 9 direct comparisons (RR, 1.34 [CI, 1.18 to 1.51]; 9 trials; I² = 34%; n = 4664) (59). A pooled analysis of 44 trials, including 13 728 smokers, found that bupropion increased relative cessation rates by roughly 62% at 6 to 12 months (RR, 1.62 [CI, 1.49 to 1.76]) (40). A smaller body of evidence (14 trials; n = 6166) compared varenicline with placebo and found relatively larger effects on smoking cessation (RR, 2.27 [CI, 2.02 to 2.55]), which was stringently defined as biochemically verified continuous abstinence (73) (Appendix Table 2).

Adverse Events

Pooled results suggested no serious harms from NRT (47, 49) or bupropion (40, 49). Nicotine replacement therapy was associated with an increased risk for any cardiovascular event, driven predominantly by minor cardiovascular events, such as tachycardia and arrhythmia (49). Although 2 reviews found no evidence of an increased risk for any or major cardiovascular adverse events for varenicline (49, 54), a separate meta-analysis of 17 trials found an increased risk for 1 or more serious adverse events among participants who received it (RR, 1.36 [CI, 1.03 to 1.81]; I² = 0%; 17 trials; n = 7725) (73).

Combined Behavioral and Pharmacotherapy Interventions Among Adults

A meta-analysis of 40 trials found a statistically significant benefit of combined pharmacotherapy (primarily NRT or bupropion) and behavioral interventions on smoking cessation at 6 months or more when compared with controls (RR, 1.82 [CI, 1.66 to 2.00]; I² = 40%; n = 15 021) (57) (Table 2 and Appendix Table 2).

Electronic Nicotine Delivery Systems

On the basis of our search for primary evidence and a review of 25 full-text articles published through 1 March 2015, we identified 2 RCTs that evaluated the effectiveness of ENDS (specifically electronic cigarettes [e-cigarettes]) to help current conventional smokers stop or reduce smoking (Table 2 and Appendix Table 3). In the largest trial, which we rated as fair quality, Bullen and colleagues (74) randomly assigned 657 smokers interested in quitting to a 16-mg nicotine e-cigarette, a 21-mg nicotine patch, or a placebo e-cigarette. All participants were also offered telephone-based support via a smoking quit line. At 6 months, this trial reported no statistically significant differences in biochemically verified continuous smoking abstinence between groups. Smoking cessation was generally low in all 3 groups: 7.3% of participants who received e-cigarettes, 5.8% of those who received nicotine patches, and 4.1% of those who received placebo e-cigarettes. Although more serious adverse events occurred in the nicotine e-cigarette group (27 events [19.7%]) than in the patch group (14 events [11.8%]), the difference was not significant.

Another fair-quality RCT done in Italy by Caponnetto and colleagues (75) randomly assigned 300 conventional smokers who did not intend to quit smoking to 1 of the 3 following regimens using e-cigarette nicotine cartridges: 7.2 mg for 12 weeks, 7.2 mg for 6 weeks followed by 5.4 mg for 6 weeks, or cartridges with no nicotine. Cartridge appearance was identical, but it is unclear whether allocation was concealed. At 52 weeks, biochemically verified cessation rates were borderline significantly different (P = 0.04) between participants in both nicotine groups (11%) and those who received the placebo cartridges (4%). The trial did not report comparisons between the individual treatment groups and placebo and reported no difference in the frequency of adverse events among study groups at 12 and 52 weeks. There was substantial loss to follow-up: 36% of participants who received one of the nicotine-containing cartridges and 45% of those who received nonnicotine cartridges did not provide 12-month follow-up data (75).

Behavioral Interventions Among Pregnant Women

Health Outcomes

A meta-analysis of 19 trials found modestly higher mean birthweight among infants born to women who received a behavioral intervention for smoking cessation than those in the control group (40.78 g [CI, 18.45 to 63.10 g]; I² = 0%) (28) (Table 3 and Appendix Table 4). Evidence of beneficial health outcomes were also seen in the pooled analyses across all interventions and comparators for preterm birth and low birthweight, with an 18% risk reduction for preterm birth before 37 weeks (RR, 0.82 [CI, 0.70 to 0.96]; I² = 0%; 14 trials) and a similar significant estimate for low birthweight (28).

Cessation Outcomes

For smoking cessation, pooled analyses of all behavioral interventions among pregnant women (70 trials) indicated a significant effect during late pregnancy (RR, 1.45 [CI, 1.27 to 1.64]) and moderate to substantial heterogeneity of estimated effects (I² = 60%) (Table 3 and Appendix Table 5).

Adverse Events

None of the reviews on behavioral interventions among pregnant women reported adverse events related to the interventions.

Pharmacotherapy Interventions Among Pregnant Women

We included 1 additional fair-quality placebo-controlled trial of NRT (77) on the basis of our search and evaluation of primary evidence. Adding this trial to the review by Coleman and colleagues (32) left 7 trials that evaluated the effects of NRT among pregnant women (Table 3). No trials of bupropion or varenicline among pregnant women met our inclusion criteria.

Health Outcomes

Four NRT placebo-controlled trials reported on preterm birth (delivery at <37 weeks' gestation) (77, 81–83) (Table 3). All but the most recent study estimated effects in the direction of a reduced risk for preterm birth with NRT, including the smallest trial, which had a statistically significant result (RR, 0.41 [CI, 0.18 to 0.94]) (82). These 4 trials also reported birthweight outcomes, 2 of which found significantly higher birthweights among women allocated to the NRT group (82, 83). However, the largest trials (77, 81) did not find a birthweight benefit.

Cessation Outcomes

Meta-analysis of the 5 placebo-controlled efficacy trials among pregnant women (n = 1922) showed a nonsignificant pooled effect of NRT on biochemically validated smoking cessation (RR, 1.24 [CI, 0.95 to 1.64]) with low heterogeneity (I² = 0%) (Appendix Table 5). Adding the 2 other non–placebo-controlled trials to this analysis increased the estimate of the pooled effect but did not alter the statistical nonsignificance.

Adverse Events

We found no evidence of perinatal harms related to NRT use among pregnant women, but data for assessing rare harms were limited (Table 3). Although the largest trial (n = 1050) (81) reported a higher rate of cesarean sections in the NRT group (20% for NRT vs. 15% for placebo; odds ratio, 1.45 [CI, 1.05 to 2.01]), the most recent trial (n = 402) did not find a statistical difference (26% vs. 22%, respectively; odds ratio, 1.21 [CI, 0.76 to 1.91]) (77). Miscarriage rates did not differ statistically in the 3 studies included in pooled analyses (RR, 1.24 [CI, 0.37 to 4.17]; I² = 0%; n = 1407).

Discussion

We did this review of reviews to help the USPSTF update its 2009 recommendation on interventions for tobacco cessation among adults. The included reviews represented more than 800 RCTs, many of which were published since the last syntheses done as part of the Public Health Service guideline (which served as the basis for the 2009 USPSTF recommendation) (7, 84). The cumulative evidence suggests that behavioral, pharmacologic, and combined medication and behavioral interventions for smoking cessation that are readily available to primary care patients and clinicians can increase rates of smoking cessation in adults at 6-month follow-up or longer. Behavioral interventions, in particular, effectively help pregnant women stop smoking and improve perinatal health outcomes. Although evidence on the health outcomes of NRT during pregnancy was somewhat reassuring, it offered limited power to rule out rare potential harms.

Our updated findings are generally consistent with the 2008 Public Health Service guideline (7). We found similar evidence of effectiveness among the general adult population for physician advice to quit, varying formats of behavioral interventions (telephone counseling and individual and group counseling), and all 3 first-line medications approved by the U.S. Food and Drug Administration. We also found consistent evidence of effectiveness for behavioral interventions among pregnant women and limited data on the use of medications among pregnant women.

Our findings are also consistent with those of an “overview of reviews” done by Cahill and colleagues (85) on the effectiveness and safety of pharmacotherapies for smoking cessation. Both found that NRT, bupropion, and varenicline were superior to placebo for smoking cessation and that none seemed to have an adverse event risk that would negate their use among the general adult population. Our results also correspond with the results and synthesis of a 2013 review of reviews and recommendations for prevention of smoking during pregnancy by the World Health Organization (86).

Electronic nicotine delivery systems are relatively new technologies, and none of the specific products have been approved as cessation interventions by the U.S. Food and Drug Administration. Regardless, knowledge about these devices may be important for providers who wish to deliver comprehensive smoking-related counseling to their patients. On the basis of our primary review of 2 RCTs, we conclude that available data on the use of ENDS for smoking cessation are quite limited and suggest no benefit among smokers intending to quit. The most recent systematic review on this subject (87) included the same 2 trials that we summarized, and neither suggested a benefit on cessation rates at 6 months or more. In addition, neither of these trials nor the limited number of observational studies included in the recent review reported any serious adverse events considered to be plausibly related to ENDS use. The paucity of trial data on adverse events is part of the ongoing debate about the appropriateness of their use as a cessation tool.

Our review has several limitations, including our review of reviews approach, the methods and quality of the included reviews that synthesized the bodies of evidence, and the limitations of the primary studies themselves. The comprehensiveness of this review is inevitably limited by the comprehensiveness and quality of the source reviews. Although most of the primary reviews that served as the basis for the main results included evidence through at least 2012, there may be evidence on particular population and intervention subsets that have been published since then. Because of the consistency of the effects within each group over time, we expect that any new trials would have little bearing on the overall results of our synthesis, regardless of sample size or effect estimates.

By adopting a review of reviews approach, we relied on the data as described and assessed by the original reviewers. In doing this, we presumed that each review generally included the full available and eligible evidence base, that data abstraction was accurate, and that analyses were scientifically sound. We were cautious about reporting pooled results for small numbers of studies or highly heterogeneous bodies of evidence. Because the included reviews were not mutually exclusive in their eligibility criteria and, as a result, were not mutually exclusive in their included studies, some individual trials were represented in more than 1 review or meta-analysis. This is particularly true for trials related to behavioral interventions in adults. Although we could not address this overlap by recalculating all of the estimates reported in the reviews because of the effort involved, we do not expect that such adjustments would alter our conclusions. We likely mitigated this potential shortcoming by basing our estimates on primary reviews rather than reporting results from several reviews.

Our syntheses and source reviews identified many areas where more research is warranted. More research is needed on the different types of mobile telephone– and Internet-based behavioral interventions for smoking cessation, including text messaging and smartphone applications, which have high potential applicability to U.S. primary care. Two relatively large trials found favorable effects for personalized text messages (88, 89) and illustrate the particular promise for this new behavioral approach. Direct comparisons among combinations and classes of drugs would be informative (such as use of combinations of NRT and bupropion vs. placebo and NRT or bupropion vs. varenicline). The evidence base for varenicline, although consistent, is smaller than that for NRT and bupropion, and more trials (particularly those that closely monitor harms) would be useful. Further research on the benefit and safety of cessation medications among pregnant women is warranted, including assessment of optimal dosage and treatment timing. A recent pilot RCT on bupropion during pregnancy reported recruitment challenges and suggestions to inform future trials (90). Careful collection of adverse events and systems for deriving long-term consequences of exposure during pregnancy are also needed. Because of the variation and lack of regulatory oversight on the content of ENDS and the limited evidence available from well-designed studies, further research is clearly needed. We identified many current and planned clinical trials on the effectiveness and safety of e-cigarettes as an aid for smoking cessation that are referenced in the full report (14).

The extensive evidence on strategies to help persons stop smoking reviewed in this report confirms the effectiveness of a range of behavioral and pharmacologic interventions when used alone or combined. Clinicians may choose from an array of tools to aid their patients' efforts to quit smoking and can directly provide, refer, or prescribe those that patients find most acceptable, with informed consideration of the probable magnitude of benefits for 6-month cessation and beyond. Implementation of these evidence-based interventions for tobacco control and other comprehensive and systems-level interventions can help to end the burden of preventable disease and premature death.