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

Patnode CD, O'Connor E, Whitlock EP, et al. Primary Care Relevant Interventions for Tobacco Use Prevention and Cessation in Children and Adolescents: A Systematic Evidence Review for the U.S. Preventive Services Task Force [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2012 Dec. (Evidence Syntheses, No. 97.)

Cover of Primary Care Relevant Interventions for Tobacco Use Prevention and Cessation in Children and Adolescents

Primary Care Relevant Interventions for Tobacco Use Prevention and Cessation in Children and Adolescents: A Systematic Evidence Review for the U.S. Preventive Services Task Force [Internet].

Show details


Key Questions and Analytic Framework

Using the methods of the USPSTF,71 we developed an analytic framework (Figure 1) and key questions (KQs) to guide our literature search, in consultation with liaisons from the USPSTF. This review examined the benefits and harms of primary care relevant interventions designed to both prevent tobacco use in children and adolescents and help child and adolescent tobacco users stop using tobacco. The KQs we examined were:

Figure 1 is an analytic framework for the key questions of this report that depicts the events that children and adolescents (ages ≤18 years) could experience while undergoing assessment and interventions for the prevention and cessation of tobacco use. In general, the figure illustrates how interventions designed for tobacco prevention or cessation could lead to no tobacco use as an intermediate outcome. For longer-term health outcomes, no tobacco use may lead to improved respiratory and dental/oral health and decreased amounts of adult smokers. The figure also depicts the possibility of harms or adverse events occurring as a result of the interventions.

Figure 1

Analytic Framework. Key Questions (KQs) KQ 1. Do interventions in primary care designed to prevent tobacco use or improve tobacco cessation rates in children and adolescents improve health outcomes in children and adolescents (i.e., respiratory health, (more...)

KQ 1. Do interventions in primary care designed to prevent tobacco use or improve tobacco cessation rates in children and adolescents improve health outcomes in children and adolescents (i.e., respiratory health, dental/oral health) and reduce the likelihood of adult smoking?

KQ 2. Do interventions in primary care prevent tobacco use in children and adolescents or improve tobacco cessation rates in children and adolescents who use tobacco? What are elements of efficacious interventions? Are there differences in outcomes in different subgroups, as defined by age, sex, race, socioeconomic status, type or pattern of tobacco use, urban versus rural, depressed versus nondepressed?

KQ 3. What adverse effects are associated with interventions to improve tobacco cessation rates or prevent tobacco use in children and adolescents?

Data Sources and Searches

The previous USPSTF recommendation was based on a 2000 report from the PHS.2 This report was subsequently updated in 2008.72 As such, we began by identifying and evaluating all trials included in the 2008 updated report for possible inclusion in the current review. Additionally, we evaluated all trials that were included or excluded (where available) in three previous reviews.39, 43,73 These reviews addressed issues that were applicable to the KQs in our review and had inclusion/exclusion criteria consistent with (or broader than) the current review. We also judged the search methods employed in these three reviews to be acceptable. For tobacco prevention, we identified two relevant systematic reviews that collectively covered the literature relevant to our inclusion/exclusion criteria through July 2002. For tobacco cessation, we identified a Cochrane review that searched for cessation trials through August 2009. We then searched MEDLINE, PsycINFO, the Cochrane Central Register of Controlled Trials, and the Database of Abstracts of Reviews of Effects for trials of tobacco use prevention starting in January 2002, and for trials of tobacco cessation starting in January 2009, ending all searches on September 14, 2012. See Appendix A for a sample search strategy. We also searched bibliographies of 20 additional relevant reviews;47-58,60-62,74-78 solicited expert input; searched Web sites of government agencies such as Agency for Healthcare Quality and Research (AHRQ), Institute of Medicine, Office of the Surgeon General, FDA, and National Institute of Clinical Excellence for relevant grey literature (February to September 2011); searched bibliographies of other relevant publications; and used news and table-of-contents alerts beginning in January 2011 to help us identify potentially eligible trials.

Study Selection

Two investigators independently reviewed 2,453 abstracts against prespecified inclusion and exclusion criteria; of those, 111 articles were subsequently evaluated for inclusion (Appendix B). Articles that were excluded are listed in Appendix C, along with their reason for exclusion. Disagreements were resolved by discussion or consultation with the larger project team. Detailed inclusion and exclusion criteria are provided in Appendix D.

We examined trials of interventions designed to prevent tobacco use in children or adolescents, or trials that promoted the cessation of tobacco use (with or without the adjunctive use of medication) published in or after 1980. Included interventions were targeted at children or adolescents (either tobacco users or nonusers) or their parents and were delivered individually or in small groups in a health care or comparable setting. Included trials had control arms that offered minimal or no treatment, or an attention control arm, and had to report tobacco use prevalence or a comparable outcome at least 6 months after the baseline assessment. We only considered controlled trials for questions related to benefits of treatment (KQs 1-2). We considered controlled trials and comparative observational studies for harms of pharmacotherapy (KQ 3).

While we sought to include trials that addressed both cigarettes and other forms of tobacco, all of the trials that met our inclusion criteria focused primarily or exclusively on cigarette smoking. We included trials conducted in, referred from, or potentially feasible for (or referable from) health care settings. We describe these collectively as “primary care relevant.” We excluded trials that were conducted in schools or other settings in which participants would be interacting with people in their existing social network as part of the intervention, because the social influence exerted by peers could not normally be replicated in a health care setting. Trials that recruited participants from schools but whose intervention was in a different setting could be included if participants were unlikely to be part of each other's pre-existing social networks (e.g., small numbers of students from multiple schools). In addition, a trial was included if it was conducted in a school setting but was feasible for primary care (e.g., school health nurse intervention or after school hours), provided we judged it unlikely that participants would be part of each other's pre-existing social networks. We excluded trials of broad community-based interventions (e.g., media campaigns, public policy changes, legislation). A comparison of the included trials in the current review with four previous systematic reviews39,43,61,73 is provided in Appendix E

Quality Assessment and Data Extraction

Two independent investigators conducted quality assessments of all trials meeting our inclusion criteria, resulting in a rating of “good,” “fair,” or “poor” (see Appendix F for quality criteria). Briefly, for benefits of treatment (KQs 1-2), we assessed the validity of the randomization and measurement procedures (including blinding and consistency between groups), comparability of the groups in baseline characteristics, overall and group-specific attrition, intervention fidelity, and statistical methods. Generally, good-quality trials blinded assessment and intake staff to participant group assignment, had followup data on 90 percent or more of participants, used reliable measures of tobacco use, reported group-specific followup with differences of less than 10 percentage points between groups, and used conservative data-substitution methods if missing data were imputed. Trials were rated as “poor” if attrition was greater than 40 percent, attrition in the treatment and control groups differed by more than 20 percentage points, or there were other important flaws. Poor-quality trials were excluded from the review. All trials meeting quality criteria for benefits of treatment (KQs 1-2) were also examined for harms of treatment (KQ 3). We did not require a minimum followup for harms of pharmacotherapy, since harms could potentially occur immediately after beginning use of a medication and may be a cause of high or differential attrition. Differences in quality ratings were resolved by discussion or consultation with the larger review team. One reviewer abstracted data from studies that were rated as “fair” or “good,” and this work was checked by another reviewer. Elements abstracted included information on study population, setting, recruitment methods, followup, intervention and control conditions, and outcomes.

Data Synthesis and Analysis

We conducted random effects meta-analyses to estimate the effect size of smoking prevention or cessation interventions on smoking status for trials reporting sufficient data. Our primary outcome was self-reported smoking status at followup. We chose self-reported smoking status rather than biochemically verified status because biochemical verification was not used consistently and is often not a reliable measure of smoking in adolescents due to sporadic tobacco use.23 Behavior-based and medication trials were analyzed separately.

While all of the medication trials were limited to smokers and targeted smoking cessation, the behavior-based trials varied in their target populations. Some behavior-based trials were limited to nonsmokers, focusing on primary prevention of smoking, and others were limited to smokers, addressing only smoking cessation (secondary prevention). Several trials included both smokers and nonsmokers at baseline and either delivered the same message to everyone (regardless of smoking status) or tailored the intervention to the smoking status of each participant. We examined smoking prevalence at followup separately for baseline smokers and baseline nonsmokers, and also analyzed combined samples that included both smokers and nonsmokers. Thus, some trials that reported outcomes for all three sets of participants (smokers only, nonsmokers only, and both groups combined) were included in all three analyses.

We entered the raw number of events (smokers, as defined by the study) in each group and the total number of participants in the analysis for each group at the pertinent followup into random effects meta-analysis to calculate pooled risk ratio estimates, using Stata 11.2 (StataCorp, College Station, TX). The meta-analysis was also adjusted for the cluster randomization of three trials79-81 by dividing the sample sizes in these studies by a design effect, which is based on average cluster size and the estimated intraclass correlation (ICC).82 We estimated the ICC to be 0.01, based on previously published literature.83 We generated forest plots that ordered the trials in alphabetical order by first author of the main outcomes publication within each of the three groups (combined, nonsmokers, smokers). We did not conduct statistical analyses for publication bias because we had fewer than 10 trials in all analyses. Statistical heterogeneity was assessed with the I2 statistic.84 We applied Cochrane Collaboration rules of thumb for interpreting I2: less than 40 percent likely represents unimportant heterogeneity, 30–65 percent represents moderate heterogeneity, 50–90 percent represents substantial heterogeneity, and more than 75 percent indicates considerable heterogeneity among the studies.82

Clinical and methodological heterogeneity were substantial, but were judged to be acceptable within the prevention, cessation, and combination subgroups to justify a meta-analysis of relative benefit (i.e., risk ratio). Because of the heterogeneity, however, we present qualitative synthesis and summary as well as quantitative, and view the quantitative pooling as adjunctive information.

There were too few trials and too much variability in a number of factors to statistically examine whether study or treatment characteristics influenced effect size in any of the analyses. Within each group, however, we did qualitatively explore patterns of association between effect size and the following factors: number of intervention sessions, time spent interacting with the interventionist, whether the intervention was tailored according to smoking status, whether there was a group component to the intervention, whether the intervention explicitly involved motivational interviewing, whether the primary treatment person targeted by the intervention was the youth, the parent, or both, theoretical basis, the measure of tobacco use, the type of control group used, study quality rating, average age of the participants, and sex distribution of the participants.

USPSTF Involvement

We worked with three USPSTF liaisons at key points throughout this review, particularly when developing the analytic framework, KQs, and scope of the review. AHRQ funded this review under a contract to support the work of the USPSTF. An AHRQ medical officer provided oversight of the project, reviewed the draft report, and assisted in the external review of the report.


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (1.9M)

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...