NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Tobacco Use and Dependence Guideline Panel. Treating Tobacco Use and Dependence: 2008 Update. Rockville (MD): US Department of Health and Human Services; 2008 May.

Cover of Treating Tobacco Use and Dependence: 2008 Update

Treating Tobacco Use and Dependence: 2008 Update.

Show details

1Overview and Methods


Tobacco use has been cited as the chief avoidable cause of illness and death in our society and accounts for more than 435,000 deaths each year in the United States.37,38 Smoking is a known cause of multiple cancers, heart disease, stroke, complications of pregnancy, chronic obstructive pulmonary disease (COPD), and many other diseases.4 In addition, recent research has documented the substantial health dangers of involuntary exposure to tobacco smoke.4 Despite these health dangers and the public's awareness of those dangers, tobacco use remains surprisingly prevalent. Recent estimates are that about 21 percent of adult Americans smoke,3 representing approximately 45 million current adult smokers.3,39 Moreover, tobacco use remains a pediatric disease.4042 Each day, about 4,000 youth ages 12 to 17 years smoke their first cigarette, and about 1,200 children and adolescents become daily cigarette smokers.4344 As a result, new generations of Americans are at risk for the extraordinarily harmful consequences of tobacco use.

Tobacco use exacts a heavy cost to society as well as to individuals. Smoking-attributable health care expenditures are estimated at $96 billion per year in direct medical expenses and $97 billion in lost productivity.28 It has been estimated that the per pack additional cost of smoking to society is approximately $7.18 per pack,45 and the combined cost of each pack to society and the individual smoker and family is nearly $40.46 If all smokers covered by state Medicaid programs quit, the annual savings to Medicaid would be $9.7 billion after 5 years.47

Despite the tragic consequences of tobacco use, clinicians and health care systems often fail to treat it consistently and effectively. For instance, in 1995, about the time of the release of the first clinical practice guideline, smoking status was identified in only about 65 percent of clinic visits, and smoking cessation counseling was provided in only 22 percent of smokers' clinic visits.48,49 Moreover, treatment typically was offered only to patients already suffering from tobacco-related diseases.48 This pattern gradually began to improve as of 2005, with up to 90 percent of smokers reporting they had been asked about smoking status and more than 70 percent reporting having received some counseling to quit.23,50,51 However, the failure to assess and intervene consistently with all tobacco users continues despite substantial evidence that even brief interventions can be effective among many different populations of smokers.5258 Also, the use of effective medications is low. Among current smokers who attempted to stop for at least 1 day in the past year, only 21.7 percent used cessation medication.33

This Guideline concludes that tobacco use presents a rare confluence of circumstances: (1) a highly significant health threat;4 (2) a lack of consistent intervention by clinicians; and (3) the presence of effective interventions. This last point is buttressed by evidence that tobacco use interventions, if delivered in a timely and effective manner, can rapidly reduce the risk of suffering from smoking-related disease.613 Indeed, it is difficult to identify any other condition that presents such a mix of lethality, prevalence, and neglect, despite effective and readily available interventions.

Significant barriers interfere with clinicians' assessment and treatment of smokers. Many clinicians lack knowledge about how to identify smokers quickly and easily, which treatments are effective, how such treatments can be delivered, and the relative effectiveness of different treatments.5962 Additionally, clinicians may fail to intervene because of inadequate clinic or institutional support for routine assessment and treatment of tobacco use48,60,63 and for other reasons such as time constraints, limited training in tobacco cessation interventions, a lack of insurance coverage for tobacco use treatment, or inadequate payment for treatment.6467

Rationale for Guideline Development and Periodic Updates

In the early 1990s, the Agency for Healthcare Policy and Research ([AHCPR] now the Agency for Healthcare Research and Quality [AHRQ]) convened an expert panel to develop the Smoking Cessation Clinical Practice Guideline (the “Guideline”), Number 18 in the AHCPR series of Clinical Practice Guidelines. The need for this Guideline was based on several factors, including tobacco use prevalence, related morbidity and mortality, the economic burden imposed by tobacco use, variation in clinical practice, availability of methods for improvement of care, and availability of data on which to base recommendations for care. More than 1 million copies of the 1996 Guideline and its affiliated products were disseminated. The original Guideline recommendations inspired changes in diverse health care settings such as managed care organizations and the Veterans Health Administration. The original Guideline also provided a framework for educating clinicians, administrators, and policymakers about the importance of tobacco dependence and its treatment. It stimulated discussions that addressed the development of tobacco dependence treatment programs at the Federal and State levels and by professional medical organizations.

Significant new research findings regarding tobacco use and its treatment led to the 2000 Guideline update, which was authored by the expert panel that developed the 1996 Guideline. The 2000 Guideline update was a product of the U. S. Public Health Service (PHS), sponsored by a consortium of private and public partners, including AHRQ; National Cancer Institute (NCI); National Heart, Lung, and Blood Institute (NHLBI); National Institute on Drug Abuse (NIDA); Centers for Disease Control and Prevention (CDC); Robert Wood Johnson Foundation (RWJF); and University of Wisconsin School of Medicine and Public Health Center for Tobacco Research and Intervention (UW-CTRI).

The 2000 Guideline, titled Treating Tobacco Use and Dependence, comprised specific evidence-based recommendations to guide clinicians, tobacco treatment specialists, insurers, purchasers, and health care administrators in their efforts to develop and implement clinical and institutional changes that support the reliable identification, assessment, and treatment of patients who use tobacco. This title underscores three truths about tobacco use.68 First, all tobacco products—not just cigarettes—exact devastating costs on the Nation's health and welfare. Second, for most users, tobacco use results in true drug dependence, comparable to the dependence caused by opiates, amphetamines, and cocaine.6972 Third, both chronic tobacco use and dependence warrant clinical intervention and, as with other chronic disorders, these interventions may need to be repeated over time.73,74

The 2000 Treating Tobacco Use and Dependence document was the most widely disseminated Guideline ever released by AHRQ, with more than 5 million copies of the Guideline and related products distributed. Moreover, it has had an enormous influence on tobacco use treatment and policy worldwide, serving as the basis for Guidelines in Australia, Canada, Chile, Japan, Portugal, and Switzerland, among other countries.

The continued expansion of new scientific findings on the effective treatment of tobacco use led to calls for the current update, Treating Tobacco Use and Dependence: 2008 Update. The 2008 update reviewed scientific evidence from 1975 to 2007 on selected topics and in total reviewed more than 8,700 scientific publications. The result of this methodologically rigorous review is an updated set of recommendations on effective counseling and medication treatments and institutional policies that can guide clinicians, specialists, and health systems in intervening with tobacco users. Appendix D summarizes new recommendations and changes to the 2000 Guideline.

The clinician audience for this Guideline update is all professionals who provide health care to tobacco users. This includes: physicians, nurses, physician assistants, medical assistants, dentists, hygienists, respiratory therapists, psychologists, mental health counselors, pharmacists, and others. The ultimate beneficiaries of the Guideline are tobacco users and their families.

Most tobacco users in the United States are cigarette smokers. As a result, the majority of clinician attention and research in the field has focused on the treatment and assessment of smoking. Clinicians, however, should intervene with all tobacco users, not just with those who smoke cigarettes. To foster a broad implementation of this Guideline update, every effort has been made to describe interventions so that they are relevant to all forms of tobacco use. In some sections of this Guideline, the term “smoker” is used instead of “tobacco user.” The use of the term “smoker” means that all relevant evidence for a recommendation arises from studies of cigarette smokers. Additional discussion of noncigarette forms of tobacco use is found in Chapter 7.

The 2008 Guideline update generally is consistent with the findings of the 2000 Guideline (see Appendix D). It also is important to note that other Guidelines and analyses on the treatment of tobacco dependence have been published with essentially consistent findings, including those from the American Psychiatric Association,75,76 the American Medical Association,77 the American Dental Association,78 the American Nurses Association,79 the American College of Obstetricians and Gynecologists, the Institute of Medicine,80 the United Kingdom Guideline,81 and the Cochrane Collaboration ( Finally, throughout the Guideline update, the terms “tobacco use treatment” and “tobacco dependence treatment” will be used interchangeably to emphasize the fact that both chronic use and dependence merit clinical intervention.

Tobacco Dependence as a Chronic Disease

Tobacco dependence displays many features of a chronic disease. Only a minority of tobacco users achieve permanent abstinence in an initial quit attempt. The majority of users persist in tobacco use for many years and typically cycle through multiple periods of remission and relapse. A failure to appreciate the chronic nature of tobacco dependence may impede clinicians' consistent assessment and treatment of the tobacco user over time.

Epidemiologic data suggest that more than 70 percent of the 45 million smokers in the United States today report that they want to quit, and approximately 44 percent report that they try to quit each year.3 Unfortunately, most of these efforts are both unaided and unsuccessful. For example, among the 19 million adults who attempted to quit in 2005,39 only 4 to 7 percent were likely successful.82,83 These statistics may discourage both smokers and clinicians.

Modern approaches to treating tobacco use and dependence should reflect the chronicity of tobacco dependence. A chronic disease model recognizes the long-term nature of the disorder with an expectation that patients may have periods of relapse and remission. If tobacco dependence is recognized as a chronic disease, clinicians will better understand the relapsing nature of the condition and the requirement for ongoing, rather than just acute, care. The existence of numerous effective treatments gives the clinician and patient many options should repeated quit attempts be needed.

A chronic disease model emphasizes for clinicians the importance of continued patient education, counseling, and advice over time. Although most clinicians are comfortable in counseling their patients about other chronic diseases such as diabetes, hypertension, or hyperlipidemia, many believe that they are less effective in providing counseling to patients who use tobacco.84,85 As with these other chronic disorders, clinicians should be encouraged to provide tobacco-dependent patients with brief advice, counseling, and appropriate medication. It is important for clinicians to know that assessing and treating tobacco use generally leads to greater patient satisfaction with health care.23,50,8688 Moreover, policy changes (e.g., tax increases, smoke-free ordinances) often lead smokers to seek treatment for this chronic disease.

In updating the Guideline, the Panel has presented evidence-based analytic findings in a format accessible and familiar to practicing clinicians. Although this should aid clinicians in the assessment and treatment of tobacco users, clinicians should remain cognizant that relapse is likely and that it reflects the chronic nature of dependence. Most smokers who ultimately quit smoking experience episodes of relapse on the way to success. Relapse should not discourage the clinician or the tobacco user from renewed quit attempts.

Coordination of Care: Institutionalizing the Treatment of Tobacco Dependence

Increasing evidence shows that the success of any tobacco dependence treatment strategy cannot be divorced from the health care system in which it is embedded. Data strongly indicate that the consistent and effective delivery of tobacco interventions requires coordinated interventions. Just as a clinician must intervene with his or her patient, so must the health care administrator, insurer, and purchaser ensure the provision of tobacco dependence treatment as an integral element of health care delivery. Health care purchasers and insurers should ensure that evidence-based tobacco dependence counseling and medications are a covered and available health insurance benefit for all enrollees and that enrollees are aware of such benefits. Health care administrators also should provide clinicians with the training and institutional support and systems to ensure consistent identification of and intervention with patients who use tobacco. Therefore, insurers, purchasers, and health care organizations should promote the utilization of covered treatments and assess usage and outcomes in performance measurement systems.89 Finally, increasing evidence shows that, for maximum public health benefit, access to effective treatments should be increased during and following the implementation of population-level tobacco control policies (i.e., tobacco tax increases and clean indoor air laws), which boost motivation and support for quitting efforts.90

Guideline Development Methodology


Panel recommendations are intended to provide clinicians with effective strategies for treating patients who use tobacco. Fundamentally, this document is a clinical practice guideline. Recommendations were influenced by two goals. The first was to identify effective treatment strategies. The second was to formulate and present recommendations that can be implemented easily across diverse clinical settings (e.g., primary care and specialty clinics; pharmacies; hospitals, including emergency departments; worksites; and school-based clinics) and patient populations.

The Guideline update is based on three systematic reviews of the available scientific literature. The first review occurred during the creation of the original Guideline published in 1996 and included literature published from 1975 through 1994. The second review was conducted for the 2000 Guideline and included literature from 1995 through January 1999. The third review was conducted on literature published from 1999 to June 2007. The three data sets were combined into a single database that was used for the 2008 analyses.

The Panel identified randomized placebo/comparison controlled trials as the strongest level of evidence for the evaluation of treatment effectiveness. Thus, evidence derived from randomized controlled trials serves as the basis for meta-analyses and for almost all of the recommendations contained in this Guideline. Questions have been raised about medication placebo controls because individuals sometimes guess their actual medication condition at greater than chance levels.91 It is possible, therefore, that the typical randomized control trial does not control completely for placebo effects. This should be borne in mind when appraising the results of the medication meta-analyses. Further, in studies of counseling, it often is not possible to control for a nonspecific placebo effect.

The Panel occasionally made recommendations in the absence of randomized controlled trials when faced with an important clinical practice issue for which other types of evidence existed. This Guideline clearly identifies the level or strength of evidence that serves as the basis for each of its recommendations.

Topics Included in the Guideline

The Panel identified tobacco use as the targeted behavior and tobacco users as the clinical population of interest. Tobacco dependence treatments were evaluated for effectiveness, as were interventions aimed at modifying both clinician and health care delivery system behavior. At the start of the 2008 update process, Guideline Panel members, outside experts, and consortium representatives were consulted to determine those aspects of the 2000 Guideline that required updating. These consultations resulted in the following chief recommendations that guided the update efforts: (1) to conduct new literature reviews and meta-analyses on topics distinguished by their public health importance and for which significant new evidence is available; (2) to review previous recommendations and to identify a subset of recommendations for which to review new data; special attention was paid to clinical situations for which the Panel had previously achieved consensus in the absence of relevant controlled trials (“C”-level recommendations) to ensure that these still warranted Guideline Panel support; (3) to consider anew the strategies that might be used in clinical settings to deliver brief tobacco dependence interventions (see Chapter 3); and (4) to identify important topics for future research. Eleven topics out of 64 considered were chosen by the Panel for updated meta-analysis (see Table 1.1).

Table 1.1. Topics chosen by the 2008 Guideline Panel for updated meta-analysis.

Table 1.1

Topics chosen by the 2008 Guideline Panel for updated meta-analysis.

This Guideline update was specifically intended to review the evidence regarding clinical treatment of tobacco dependence. Interventions for the primary prevention of tobacco use were not examined in detail, with the exception of interventions directly relevant to clinical practice. Readers also may refer to the 1994 Surgeon General's Report, Preventing Tobacco Use Among Young People41 and the 2000 Surgeon General's Report, Reducing Tobacco Use,6 for information on the primary prevention of tobacco use. Community-level interventions (e.g., mass media campaigns) that are not usually implemented in primary care practice settings were not addressed. For more information on community-based tobacco use prevention, refer to the Centers for Disease Control and Prevention Guide to Community Preventive Services.92 The Guideline update did not examine evidence regarding unaided quit attempts as this Guideline focused on clinical interventions. Finally, the use of exposure reduction strategies93 (strategies in which tobacco users alter, rather than eliminate, their use of nicotine or tobacco in an attempt to reduce or avoid its harmful consequences) were not considered due to a lack of data and the fact that they are beyond the scope of a clinical practice guideline focused on treating tobacco use and dependence. Current research does not offer answers to key questions regarding exposure reduction strategies: their population-wide impact on cessation and initiation of smoking, their long-term benefits as compared with those of a strategy focused on tobacco abstinence, and their success in reducing long-term exposure to tobacco toxins.

This Guideline update is designed for two main audiences: first, clinicians; and second, health care administrators, insurers, and purchasers. It is designed to be used in a wide variety of clinical practice settings, including private medical practices; dental offices; pharmacies; academic health centers; mental health and substance abuse treatment clinics; telephone quitlines; managed care organizations; public health department clinics; hospitals, including emergency departments; and school or worksite clinics. The ultimate beneficiaries of the Guideline are tobacco users and their families.

Guideline Development Process

The 2008 Guideline update development process (see Figure 1.1) was initiated in mid-2006. The methodology was consistent with that followed by the 2000 Guideline except where specifically identified below.

Figure 1.1. 2008 Guideline development process.


Figure 1.1. 2008 Guideline development process.

Selection of Evidence

Published, peer-reviewed, randomized controlled studies were considered to constitute the strongest level of evidence in support of Guideline recommendations. This decision was based on the judgment that randomized controlled trials provide the clearest scientifically sound basis for judging comparative effectiveness. Most of these randomized trials, however, were conducted with individuals who proactively sought treatment and who volunteered to fulfill various research requirements. It is possible that these individuals were more highly motivated to quit smoking than the typical smoker encountered in a clinical practice setting. Thus, the percentage abstinent estimates supplied with the meta-analyses may overestimate the actual level of abstinence produced by some of the treatments in real-world settings. Analyses conducted for the previous Guideline editions, though, suggest that the treatment effect sizes (odds ratios or ORs) are relatively stable across individuals seeking treatment (“treatment seekers”) and those recruited via inclusive recruitment strategies (“all-comers”). Randomized controlled trials were exclusively used in meta-analyses. However, the Panel recognized that variations in study inclusion criteria sometimes were warranted. For instance, research on tobacco interventions in adolescents frequently assigns interventions on the basis of larger units, such as schools. These units, rather than individuals, were allowed to serve as units of analysis when analyzing interventions for adolescents. In such cases, studies were combined for inclusion in meta-analyses if the study satisfied other review criteria. A similar strategy was followed in the review of health systems research.

In certain areas, research other than randomized clinical trials was evaluated and considered to inform Panel opinion and judgment, though not submitted to meta-analysis. This occurred with topics such as tobacco dependence treatment in specific populations, tailoring interventions, and cost-effectiveness of tobacco dependence treatment.

Literature Review and Inclusion Criteria

Approximately 8,700 articles were screened to identify evaluable literature. This figure includes approximately 2,700 articles added to the literature since publication of the 2000 Guideline. These articles were obtained through searches of 11 electronic databases and reviews of published abstracts and bibliographies. An article was deemed appropriate for meta-analysis if it met the criteria for inclusion established a priori by the Panel. These criteria were that the article: (a) reported the results of a randomized, placebo/comparison controlled trial of a tobacco use treatment intervention randomized on the patient level (except as noted above); (b) provided followup results at least 5 months after the quit date (except in the case of studies evaluating tobacco dependence treatments for pregnant smokers); (c) was published in a peer-reviewed journal; (d) was published between January 1975 and June 2007; (e) was published in English; and (f) was one of the 11 topics chosen to be included in the 2008 update (see Table 1.1). It is important to note that the article-screening criteria were updated for the 2008 Guideline update. Additionally, articles were screened for relevance to safety, economic, or health systems issues. As a result of the original and update literature reviews, more than 300 articles were identified for possible inclusion in a meta-analysis, and more than 600 additional articles were examined in detail by the Panel. These latter articles were used in the formulation of Panel recommendations that were not supported by meta-analyses. The literature search for the update project was validated by comparing the results against a search conducted by the CDC and through review by the expert Panel.

When individual authors published multiple articles meeting the meta-analytic inclusion criteria, the articles were screened to determine whether they contained unique data. When two articles reported data from the same group of subjects, both articles were reviewed to ensure that complete data were obtained. The data were treated as arising from a single study in meta-analyses.

Preparation of Evidence Tables

Two Guideline staff reviewers independently read and coded each article that met inclusion criteria. The reviewers coded the treatment characteristics that were used in data analyses (see Tables 6.1 and 6.2 in Chapter 6). The same general coding procedure employed during the 2000 Guideline process was employed during the update. When adjustments to the coding process were made, articles coded with the original process were re-coded to reflect the changed coding (e.g., more refined coding criteria were used for the coding of treatment intensity).

Table 6.1. Topics meta-analyzed for the 2008 Guideline update.

Table 6.1

Topics meta-analyzed for the 2008 Guideline update.

Table 6.2. Topics meta-analyzed for the 1996 and 2000 Guidelines and included in the 2008 Guideline update (but not re-analyzed).

Table 6.2

Topics meta-analyzed for the 1996 and 2000 Guidelines and included in the 2008 Guideline update (but not re-analyzed).

A third reviewer then examined the coding of both reviewers and adjudicated any differences. Discrepancies that could not be resolved through this process were adjudicated by the project manager, Panel chair, and/or the Panel's senior scientist. Finally, each article accepted for a meta-analysis had key fields reviewed by the project manager as a final quality check. The data then were compiled and used in relevant analyses and/or Panel deliberations. Analyses done for the 2000 Guideline revealed that intervention coding categories could be used reliably by independent raters.94

Outcome Data

Six-month followup after the quit date is a standard followup duration for reporting data from clinical trials. Therefore, focusing on a 6-month timepoint in meta-analyses allowed the investigators to capture the greatest number of studies for analysis. Also, research indicates that a high percentage of those who ultimately return to smoking will do so by 6 months.9598 Because a strict adherence to a 6-month timepoint would have eliminated a significant number of studies, a 1-month window was permitted such that studies with 5 months of followup data were included, but 6-month data were used if both 5- and 6-month data were available. When quit rates were provided for longer endpoints, outcome data from the endpoint closest to 6 months were used, so long as they did not exceed 3 years. Outcome data beyond 3 years rarely were available and were not included in the Guideline analyses. In the area of medication treatment, the inclusive meta-analysis reported in Table 6.26 was repeated with longer term outcome data (10–14 month postquit). This additional meta-analysis largely replicated the results of the meta-analysis based on a 6-month followup time frame. This suggested that the shorter, more inclusive, followup timepoint captured effect sizes that were similar to those yielded by the use of longer followup timepoints. There was one exception to the selection of followup data described above. In the case of pregnancy studies, both predelivery and postdelivery (5 months) outcomes were analyzed.

Table 6.26. Meta-analysis (2008): Effectiveness and abstinence rates for various medications and medication combinations compared to placebo at 6-months postquit (n = 83 studies)a.

Table 6.26

Meta-analysis (2008): Effectiveness and abstinence rates for various medications and medication combinations compared to placebo at 6-months postquit (n = 83 studies)a.

Panel staff also coded biochemical confirmation of self-reported tobacco use abstinence. Previous Guideline analyses show that studies with and without biochemical confirmation yield similar meta-analysis results. Therefore, meta-analyses presented in the Guideline reflect a pooling of these studies. If both biochemically confirmed and nonconfirmed data were available from the same study, however, the confirmed data were used in analyses. As in the 2000 Guideline, only studies that used biochemical verification were used in the meta-analyses of pregnant smokers because of the under-reporting of smoking status by pregnant women.

All of the new meta-analyses conducted for the 2008 Guideline were based exclusively on intent-to-treat data, in which the denominator was the number of participants randomized to treatment and the numerator was the number of abstinent participants contacted at followup. Some meta-analyses conducted for the 1996 and 2000 Guideline comprised a small number of studies in which the denominator consisted only of participants who completed treatment. The vast majority of studies across all analyses reported intent-to-treat data and these data were used if both types of data were available.

Studies were coded for how the outcome measures were reported—“point prevalence,” “continuous,” or “unknown/other.” If abstinence data were based on tobacco use occurrence within a set time period (usually 7 days) prior to a followup assessment, the outcome measure was coded as “point prevalence.” “Continuous” was used when a study reported abstinence based on whether study subjects were continuously abstinent from tobacco use since their quit day. “Unknown/other” was used when it was not possible to discern from the study report whether the authors used a point prevalence or continuous measure for abstinence or if abstinence was measured from some point other than the quit day.

As in the 1996 and 2000 Guidelines, a point prevalence outcome measure (7-day point prevalence, when available), rather than continuous abstinence, was used as the chief outcome variable. Point prevalence was preferred for several reasons. First, this was the modal reporting method among the analyzable studies. Second, continuous abstinence data may underestimate the percentage of individuals who are abstinent at particular followup timepoints, although some data suggest that these rates are similar.99 Finally, most relapse begins early in a quit attempt and persists.9597,100102 A point prevalence measure taken at 6 months certainly would capture the great majority of those relapse events. Therefore, whenever possible, 7-day point prevalence abstinence data were used. If point prevalence data were not available, the preferred alternative was continuous abstinence data.

Meta-Analytic Techniques

The principal analytic technique used in this Guideline update was meta-analysis. This statistical technique estimates the impact of a treatment or variable across a set of related investigations. The primary meta-analytic model used in this and the previous two Guidelines was logistic regression using random effects modeling. The modeling was performed at the level of the treatment arm, and study effects were treated as fixed. The panel methodologist chose to employ random effects modeling, assuming that both the subject populations and the treatment elements analyzed would vary from study to study (e.g., counseling might be done somewhat differently at two different sites). Random effects modeling is well suited to accommodate such variation among studies.103 The statistician used the EGRET Logistic Normal Model.104 A complete and detailed review of the meta-analytic methods used in the Guideline can be found in the Smoking Cessation Guideline Technical Report No. 18, available from AHRQ as AHCPR Publication No. 97-N004. The specific articles used in each meta-analysis included in the 2008 Guideline can be found at

In general, meta-analysis was used only with studies with randomization at the level of subject. In some areas (health systems changes, adolescents), however, studies often involved randomization at another level (e.g., clinician, clinic, etc.). Such studies were used in meta-analyses of a small number of topics when such studies occurred in sufficient numbers to permit inferences. Screening of such articles considered factors such as data nonindependence, the evaluation of pre-intervention or baseline status, and the number and types of higher level units.

The initial step in meta-analysis was the selection of studies that were relevant to the treatment characteristic being evaluated. After relevant studies were identified (i.e., those that contained a self-help intervention if self-help treatments were being evaluated), Panel staff reviewed the studies to ensure that they passed screening criteria. Some screening criteria were general (e.g., study presents greater than 5 months of followup data), whereas other criteria were specific to the type of treatment characteristic evaluated (i.e., in the analysis of quit lines, screening ensured that treatment arms were not confounded with differing intensities of in-person counseling).

The separate arms (treatment or control groups) in each study then were inspected to identify confounders that could compromise interpretation. Seriously confounded arms were excluded from analysis. Relevant characteristics of each arm were then coded to produce meaningful analytic comparisons. Criteria for performing a meta-analysis included: (1) the Guideline Panel judged the topic to be addressed in the meta-analysis as having substantial clinical significance; (2) at least two studies meeting selection criteria existed on the topic and the studies contained suitable within-study control or comparison conditions (e.g., each study had to contribute at least two arms that would permit the estimation of within-study effects); and (3) there was an acceptable level of interstudy homogeneity in the analyzed variable or treatment so as to permit meaningful inference (e.g., an analyzed treatment was sufficiently similar across various studies so that combining studies was meaningful).

Limitations of Meta-Analytic Techniques. Several factors can compromise the internal validity of meta-analyses. For example, publication biases (particularly the tendency to publish only those studies with positive findings) may result in biased summary statistics. The complement to publication bias is the “file-drawer effect,” in which negative or neutral findings are not submitted for publication. In addition, either the magnitude or the significance of the effects of meta-analyses may be influenced by factors such as the frequency with which treatments occurred in the data set and by the extent to which treatments co-occurred with other treatments. All else being equal, a treatment that occurs infrequently in the data set is less likely to be found significant than a more frequently occurring treatment. Also, when two treatments co-occur frequently in the same groups of subjects, it is difficult to apportion statistically the impact of each. In addition, comparability biases can exist when substantially different groups or treatments are coded as being the same (e.g., when treatments are similar only on a superficial attribute).

The generalizability of meta-analytic findings was evaluated for previous Guideline editions with respect to whether patients sought cessation treatment (“self-selected”) or whether treatment was delivered without the patient seeking it (“all-comers,” as when cessation treatment occurred as an integral part of health care). Conducting separate meta-analyses in these different subject populations yielded very similar findings across a variety of treatment dimensions (e.g., treatment format, treatment intensity). No other population characteristic (e.g., years smoked, severity of dependence) was explored in meta-analyses.

Interpretation of Meta-Analysis Results. The meta-analyses yielded logistic regression coefficients that were converted to odds ratios. The meaning or interpretation of an odds ratio can be seen most easily by means of an example depicted in a 2 × 2 table. Table 1.2 contains data showing the relation between maternal smoking and low birth-weight in infants. Data are extracted from Hosmer and Lemeshow, 2000.105 The odds of a low birthweight infant if the mother smokes are 30:44, or 0.68 to 1. The odds of a low birth-weight infant if the mother does not smoke are 29:86, or 0.34 to 1. The odds ratio may be estimated as (30/44)/(29/86) = 2.02 to 1. Therefore, the odds ratio can be seen roughly as the odds of an outcome on one variable, given a certain status on another variable(s). In the case above, the odds of a low birth-weight infant are about double for women who smoke compared with those who do not.

Table 1.2. Relation between maternal smoking and low birth-weight in infants.

Table 1.2

Relation between maternal smoking and low birth-weight in infants.

Once odds ratios were obtained from the meta-analyses, 95 percent confidence intervals (C.I.) were estimated around the odds ratios. An odds ratio is only an estimate of a relation between variables. The 95 percent confidence interval presents an estimate of the precision of the particular odds ratio obtained. If the 95 percent confidence interval for a given odds ratio does not include “1,” then the odds ratio represents a statistically significant difference between the evaluated treatment and the reference or control condition at the 0.05 level. The confidence intervals generally will not be perfectly symmetrical around an odds ratio because of the distributional properties of the odds ratio. The confidence intervals do not reveal whether active treatments differ significantly from one another, only whether they differ from the comparison condition (e.g., placebo medication, no contact). In the inclusive meta-analysis on medications, comparisons of an active medication versus the nicotine patch were accomplished via a posteriori contrasts, not on the basis of nonoverlapping confidence intervals.

After computing the odds ratios and their confidence intervals, the odds ratios were converted to abstinence percentages and their 95 percent confidence intervals (based on reference category abstinence rates). Abstinence percentages indicate the estimated long-term abstinence rate achieved under the tested treatment or treatment characteristic. The abstinence percentage results are approximate estimates derived from the odds ratio data. Therefore, they essentially duplicate the odds ratio results but are presented because their meaning may be clearer for some readers. Because the placebo/control abstinence percentage for a particular analysis is calculated exclusively from the studies included within that meta-analysis, these abstinence percentages vary across the different analyses. Therefore, the odds ratios and abstinence rates presented across the different tables are estimated relative to different placebo or control conditions.

How To Read the Data Tables

Table 1.3 depicts results from one of the meta-analyses reported in this Guideline update. This table presents results from the analysis of the effects of proactive telephone counseling (see Formats of Psychosocial Treatments in Chapter 6). In this table, the comparison condition, or “reference group,” for determining the impact of different treatment options was smokers who received minimal or no counseling or self-help. The “Estimated odds ratio” column reveals that treatment conditions receiving proactive telephone counseling had an odds ratio of 1.6. The odds ratio indicates a statistically significant effect because the lower boundary of the confidence interval did not include “1.” This odds ratio means that when smokers receive proactive telephone counseling, they are more than one and one-half times more likely to remain abstinent than if they had received minimal or no counseling or self-help.

Table 1.3. Meta-analysis (2008): Effectiveness of and estimated abstinence rates for proactive telephone counseling compared to minimal interventions, self-help, or no counseling (n = 9 studies).

Table 1.3

Meta-analysis (2008): Effectiveness of and estimated abstinence rates for proactive telephone counseling compared to minimal interventions, self-help, or no counseling (n = 9 studies).

The column labeled “Estimated abstinence rate” shows the abstinence percentages for the two treatment conditions. For instance, the reference condition (minimal or no counseling) in the analyzed data set was associated with an abstinence rate of 10.5 percent. Consistent with the odds ratio data reviewed above, proactive telephone counseling produced modest increases in abstinence rates (15.5%).

The total number of studies included in each meta-analysis is provided within the title of the corresponding table. A list of published articles used in each meta-analysis can be found at: Finally, the 2008 Guideline update includes meta-analyses completed for the 1996, 2000, and 2008 Guidelines. In the title of each meta-analysis, the year in which it was first published is provided.

The column labeled “Number of arms” specifies the number of treatment groups across all analyzed studies that contributed data to the various treatment conditions (e.g., Quitline counseling was provided in 11 treatment arms). Therefore, this column depicts the number of treatment groups relevant to each analyzed category. Because a study may have multiple treatment groups, the number of treatment arms may exceed the number of studies included in a meta-analysis.

The outcome data in the tables may include findings from both studies with “all-comers” (individuals who did not seek a treatment intervention) and “self-selected” populations, studies using point-prevalence and continuous abstinence endpoints, and studies with and without biochemical confirmation, except where otherwise described. Some meta-analyses (such as those evaluating medications) included predominantly studies with “self-selected” populations who volunteered for intensive treatment. In addition, in medication studies, both experimental and control subjects typically received substantial counseling. Both of these factors might have produced higher abstinence rates in reference or placebo subjects than typically are observed among self-quitters. Finally, although there is an important scientific distinction between “efficacy” and “effectiveness,”106 this 2008 clinical update uses the term “effectiveness” exclusively, recognizing that the majority of the studies summarized here reflect efficacy research, which requires random assignment and a high degree of experimental control. This was done for purposes of clarity for the intended clinical audience.

Strength of Evidence

Every recommendation made by the Panel bears a strength-of-evidence rating that indicates the quality and quantity of empirical support for the recommendation. Each recommendation and its strength of evidence reflects consensus of the Guideline Panel.

The three strength-of-evidence ratings are described below:


Multiple well-designed randomized clinical trials, directly relevant to the recommendation, yielded a consistent pattern of findings.


Some evidence from randomized clinical trials supported the recommendation, but the scientific support was not optimal. For instance, few randomized trials existed, the trials that did exist were somewhat inconsistent, or the trials were not directly relevant to the recommendation.


Reserved for important clinical situations in which the Panel achieved consensus on the recommendation in the absence of relevant randomized controlled trials.

As noted previously, the Panel evaluated evidence from nonrandomized trials to inform members' understanding of certain topics (e.g., policy issues). If treatment recommendations were based primarily on such evidence, they were of the “C” level and depended on the consistency of findings across different studies. In some areas, the highest quality evidence does not depend on randomized trials (e.g., cost-effectiveness). In these areas, the strength-of-evidence rating depended on the number, quality, and consistency of the studies and evidence. Finally, the Panel declined to make recommendations when there was no relevant evidence or the evidence was too weak or inconsistent to support a recommendation.

Caveats Regarding Recommendations

The reader should note some caveats regarding Guideline recommendations. First, an absence of studies should not be confused with a proven lack of effectiveness. In certain situations, there was little direct evidence regarding the effectiveness of some treatments, and in these cases the Panel usually rendered no opinion. Second, even when there were enough studies to perform a meta-analysis, a nonsignificant result does not prove ineffectiveness. Rather, nonsignificance merely indicates that effectiveness was not demonstrated given the data available.

The primary emphasis of this Guideline update is to identify effective interventions, not to rank-order interventions in terms of effectiveness. The most important goal of the analytic process is to identify effective interventions. Selection or use of particular intervention techniques or strategies usually is a function of practical factors: patient preference, time available, training of the clinician, cost, and so on. The Panel believes clinicians should choose the most appropriate intervention from among the effective interventions identified in this Guideline update, given clinical circumstances. An excessive emphasis on relative effectiveness might discourage clinicians from using interventions that have a small but reliable impact on quit rates. One meta-analysis that is new to this update does provide focused tests of the relative effectiveness of different interventions. Specifically, the inclusive meta-analysis of the tobacco use medications involved a posteriori tests of medication effectiveness versus the nicotine patch (Table 6.28). These tests of relative effectiveness were conducted on this topic because: (1) numerous treatments were available for comparison; (2) selection from among the various tobacco use medications has been noted as an important clinical concern;107109 and (3) the various interventions are somewhat interchangeable and widely available so that the clinician or patient might be able to select a medication based on effectiveness. Finally, the panel occasionally identified an intervention as superior to another in the absence of formal statistical contrasts; some interventions were so superior to control or no-treatment conditions that the Panel clearly identified them as superior to another intervention. For instance, although minimal person-to-person contact can increase smoking abstinence rates over no-treatment conditions, there is little doubt that longer person-to-person interventions have greater impact (see Chapter 6).

Table 6.28. Meta-analysis (2008): Effectiveness of and abstinence rates of medications relative to the nicotine patch (n = 83 studies)a.

Table 6.28

Meta-analysis (2008): Effectiveness of and abstinence rates of medications relative to the nicotine patch (n = 83 studies)a.

External Review of the Guideline

For the present update, the Panel and consortium members invited 106 reviewers to make comments. In addition, a draft of the Guideline was published in the Federal Register in September 2007 for public comment. A total of 81 invited reviewers and 15 members of the public supplied written comments. Peer reviewers included clinicians, health care administrators, social workers, counselors, health educators, researchers, consumers, key personnel at selected Federal agencies and State tobacco control programs, and others. All peer reviewers made financial disclosure statements, which were provided to the Panel. Reviewers were asked to evaluate the Guideline based on five criteria: validity, reliability, clarity, clinical applicability, and utility. Comments from the peer reviewers and public were incorporated into the Guideline when appropriate. Two individuals made oral presentations to the Guideline Panel during an advertised open presentation period.

Organization of the Guideline Update

This updated Guideline is divided into seven chapters that reflect the major components of tobacco dependence treatment (see Figure 1.2 for the treatment model):

Figure 1.2. Model for treatment of tobacco use and dependence.


Figure 1.2. Model for treatment of tobacco use and dependence.

Chapter 1, Overview and Methods, provides an overview and rationale for the updated Guideline, as well as a detailed description of the methodology used to review the scientific literature and develop the original and updated Guidelines.

Chapter 2, Assessment of Tobacco Use, establishes the importance of determining the tobacco use status of every patient at every visit.

Chapter 3, Clinical Interventions for Tobacco Use and Dependence, is intended to provide clinicians with guidance as they use brief interventions to treat tobacco users willing to quit, tobacco users unwilling to make a quit attempt at this time, and tobacco users who have recently quit.


For the Patient Willing To Quit, provides brief clinical approaches to assist patients in quit attempts.


For the Patient Unwilling To Quit, provides brief clinical approaches designed to motivate the patient to make a quit attempt.


For the Patient Who Has Recently Quit, provides clinicians with strategies designed to reinforce a former tobacco user's commitment to stay tobacco-free and assist patients who have relapsed.

Chapter 4, Intensive Interventions for Tobacco Use and Dependence, provides clinicians with more intensive strategies to treat tobacco users.

Chapter 5, Systems Interventions, targets health care administrators, insurers, purchasers, and other decisionmakers who can affect health care systems. This chapter provides these decisionmakers with strategies to modify health care systems to improve the delivery of tobacco treatment services.

Chapter 6, Evidence and Recommendations, presents the evidentiary basis for the updated Guideline recommendations.


Counseling and Psychosocial Evidence: Provides recommendations and analysis results regarding screening for tobacco use and specialized assessment, advice, intensity of clinical interventions, type of clinician, format, followup procedures, types of counseling and behavioral therapies, and the combination of counseling and medication.


Medication Evidence: Provides recommendations and analysis results regarding the seven first-line medications, combination medications, second-line medications, and other medication issues.


Systems Evidence: Provides recommendations and analysis results regarding systems changes, including provider training, cost-effectiveness, and health insurance coverage for tobacco use treatments.

Chapter 7, Specific Populations and Other Topics, provides information on specific populations, including HIV-positive smokers; hospitalized smokers; lesbian/gay/bisexual/transgender smokers; smokers with low SES/limited formal education; smokers with medical comorbidities; older smokers; smokers with psychiatric disorders, including substance use disorders; racial and ethnic minorities; women smokers; children and adolescents; light smokers; and noncigarette tobacco users. This chapter also presents information and recommendations relevant to weight gain after quitting smoking, with specific recommendations regarding future research on this topic.


Given the volume of literature referenced in this Guideline, references are listed at, rather than in this document. This was done to manage the length of this Clinical Guideline update and to facilitate electronic searches and manipulation of the references. Within this Web site, text references are numbered to match the numbers in this Guideline update. References to randomized control trials used in all of the meta-analyses (1996, 2000, 2008) are listed separately and by table number and title. The entire Guideline update, with and without references, can be downloaded from the site.


  • PubReader
  • Print View
  • Cite this Page

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...