Theories Underlying the Strategy

An understanding of the relationship between ill health, disease, and behavioral choices led early health communication researchers to create prescriptive messages urging people to make healthier choices. Messages focused more clearly on influencing attitudes and beliefs have traditionally been more effective than messages without these types of information (Hornik 2002).

Individual-based theories of behavior change provide a rationale for how public health messages may affect behavior by influencing knowledge, attitudes, and beliefs. Early models of behavior change focused on different aspects of eliciting behavior change. The Health Belief Model focused on susceptibility, perceived severity of consequences for a behavior, cost-benefit analysis, and health motivation (Rosenstock 1974). The Theory of Reasoned Action (TRA) (Fishbein and Ajzen 1975) and the Theory of Planned Behavior (TPB) (Ajzen 1991) focused on behavioral beliefs, norms, and control beliefs and their effect on intention to engage in a behavior. Another example is the Social Cognitive Theory, which (Bandura 1986, 2004) focuses on the relationship between personal factors, environmental factors, and behavior, which is often affected by modeling. TRA and TPB have been updated in the Integrated Model of Behavior Change (IMBC). In IMBC, a number of exogenous variables, including exposure to media and health interventions, contribute to beliefs about a particular health-related topic (Bleakley et al. 2011). Behavioral beliefs lead to attitudes, intentions, and finally to behaviors. In these models, an individual’s attitudes, beliefs, and environmental factors (such as perception of norms) are thought to be central to influencing intentions and ultimately behavior change.

A number of communication theories on persuasion add to this literature by providing guidance on how to change attitudes and beliefs. The Elaboration Likelihood Model of Persuasion (Petty and Cacioppo 1986) and the Heuristic-Systematic Model (Eagly and Chaiken 1993) propose two processing systems. One system involves “central” or “systematic” processing in which the message content is considered more carefully and is elaborated upon more fully. The other system is the “peripheral” or “heuristic” system that involves processing of cues such as source credibility to reject or accept the message. “Central” and “peripheral” systems can be activated individually or simultaneously at varying levels. The models suggest that lasting change and persuasion are most likely to occur when an individual has the motivation and ability to process a message centrally if the argument contained in the message is presented well. However, if the argument in the message is poor, peripheral processing may produce more desirable effects, depending on the peripheral cues.

Many theorists also emphasize the importance of emotion for message processing and behavior change (Cohen 1990; Eagly and Chaiken 1993; Forgas 1995; Escalas et al. 2004; Dillard and Nabi 2006; Lang 2006; Baumeister et al. 2007). Public health messages that activate emotion systems may increase personal perceptions of vulnerability to a health risk by producing a mental shortcut through increases in emotional associations with actions, images, or ideas (Damasio 1994; Finucane et al. 2000) that a person may use when making decisions or judgments (Slovic 2001). Emotional information may function by increasing resources allocated to processing until information overload occurs, that is, until the number of resources required to process the message becomes more than the resources allocated to processing (Lang 2006). There are two basic parts of emotional activation: arousal, which is related to how much activation is occurring unrelated to the type of emotion being experienced, and valence. Valence can be divided into appetitive (positive) and aversive (negative) activation (Cacioppo and Gardner 1999; Lang 2006) or into discrete emotions such as happiness, sadness, or hopefulness (Nabi 2010).

There has also been increasing work in health communication on using narratives and exemplars to decrease processing defensiveness and thereby increase persuasiveness of health communication messages. Dunlop and colleagues (2010) found that greater levels of transportability (which is associated with becoming absorbed in the message’s narrative) were associated with greater intention to quit smoking. Furthermore, Moyer-Guse and Nabi (2010) found that narratives reduce reactance, thus increasing persuasion of messages that were high in narrativity.

Mass media messages may also exert influence through indirect interpersonal or social influences pathways (Rogers 1995b; Ball-Rokeach 1998; Yanovitzky and Stryker 2001). People obtain information about how best to respond to a health threat not only through direct exposure to campaign messages but also from social networks when the message is shared or discussed with others. For example, discussion among peers of antismoking messages is associated with increased perceptions of personal risk in adolescents (Hafstad and Aarø 1997; Morton and Duck 2001), and so, the social context in which a message is received and interpreted may influence the effects of that message (Lazarsfeld et al. 1944; Katz and Lazarsfeld 1955; McCombs and Shaw 1972).

Review Methodology

Many previous reviews have focused specifically on the effects of mass media antismoking messages on youth (USDHHS 1994; Pechmann 1997, 2001; Sowden 1998; Pechmann and Reibling 2000a; Farrelly et al. 2003a; Wakefield et al. 2003b,c). Other reviews have examined the broad impact of antismoking campaigns on both adults and youth (Flay 1987; Friend and Levy 2002; Siegel 2002; Jepson et al. 2006; Schar et al. 2006; NCI 2008) and the effects of campaigns on youth within the context of other strategies to prevent youth smoking (Lantz et al. 2000; Richardson et al. 2007).

This chapter examines the conclusions from these previous reviews and describes in Tables 6.1 and 6.2 the published studies of the effects of mass media campaigns on youth addressed in the three most recent comprehensive reviews (Richardson et al. 2007; Angus et al. 2008; NCI 2008). In addition, a systematic literature review for articles published since the latest review (NCI 2008) from May 2007 to June 2008 was conducted using the same search terms. The focus in that review, and for this section, was on studies that assessed the influence of mass media interventions (e.g., television, radio, print, and outdoor advertising) alone or in combination with other interventions (e.g., school, community, policy) (NCI 2008). These newer studies on youth are included in Tables 6.1–6.3.

Table 6.1

Summaries of controlled field trials of community-based mass media programs, by review(s).

Table 6.2

Summary of longitudinal and cross-sectional population-based studies examining the effects on youth of mass media antismoking campaigns.

Table 6.3

Controlled exposure and naturalistic exposure studies examining the relative effectiveness of different advertising messages for youth.

Overall Effectiveness of Mass Media Campaigns in Preventing Youth Smoking

Controlled field trials. The NCI review (2008) of the media and tobacco use described above highlights the difficulty of evaluating the media components of several early quasi-experimental studies of community-based cardiovascular programs because the media elements were combined with other program elements (e.g., in the North Karelia Project and the Minnesota Heart Health Program). However, the evaluations of the overall effects of these programs indicate positive immediate and intermediate effects on smoking levels among youth (Vartiainen et al. 1986; Perry et al. 1992) and on long-term effects on initiation of smoking by youth at 8- and 15-year follow-ups (Vartiainen et al. 1990, 1998). In contrast, another cardiovascular program aimed primarily at adults, the Stanford Five-City Project, allowed for the examination of the media effects alone and did not show any differences between intervention and control communities in the prevalence of smoking that could be traced to the media component. There was evidence, however, of a strong secular trend that may have reduced the ability to detect effects (Winkleby et al. 1993).

Early reviews of the published literature focused heavily on the findings of some of the controlled field experiments on the effectiveness of community-based antismoking programs for youth. Some of these trials were able to randomize allocation to the media campaign (Bauman et al. 1991; Flay et al. 1995; Biglan et al. 2000a), and others used matched “unexposed” communities as controls (Flynn et al. 1992; Slater et al. 2006). These programs varied greatly in the length and intensity of exposure to the campaign message and the time to follow-up assessment.

Reviewing the available literature up to the early 1990s from controlled field trials and limited population-based evaluations, the 1994 Surgeon General’s report on preventing tobacco use among young people emphasized that the mass media campaigns to prevent smoking by youth conducted up to that point were “meager” compared with the highly coordinated and well-funded marketing activities of the tobacco industry (USDHHS 1994). State agencies and volunteer organizations had conducted only “short-term efforts that have had limited evaluations” (USDHHS 1994, p. 150), and evaluations were completed on only a handful of the campaigns described in the report. Of the few reviewed experimental studies of different media strategies that had been conducted, only one had found a significant reduction in smoking among adolescents (Flynn et al. 1992).

In that study, Flynn and colleagues (1992) examined the effects of a media (television and radio)-plus-school intervention (refusal skills, accurate social norms, positive views of nonsmoking) and of a school intervention alone that both ran over 4 years. Assessments at the end of the 4-year intervention and then at a 2-year follow-up (Flynn et al. 1994) found that those in the media-plus-school intervention had significantly lower smoking rates than those in the school-only intervention. The 1994 Surgeon General’s report (USDHHS 1994) concluded that mass media campaigns can be cost-effective but that messages should be pretested to avoid and test for unintended effects (Worden et al. 1988) and that these campaigns should be intense enough and sufficient in length to ensure impact.

A Cochrane review completed a few years later (Sowden 1998) included longer-term follow-up reports for some of the studies (Bauman et al. 1991; Flynn et al. 1994, 1997; Flay et al. 1995) reviewed in the 1994 Surgeon General’s report (USDHHS 1994) as well as a new study (Hafstad and Aarø 1997; Hafstad et al. 1997a) and concluded that there was some evidence, although it was not strong, that mass media can be effective in preventing the uptake of smoking in young people. As did the 1994 Surgeon General’s report (USDHHS 1994), the Cochrane review emphasized that the effective campaigns were based on theory, used formative research to develop messages, and had relatively intense and ongoing exposure of messages.

In reviews published after 2000, Pechmann (2001), Friend and Levy (2002), Farrelly and colleagues (2003a), Wakefield and colleagues (2003b,c), and the Task Force on Community Preventive Services (2005) all concluded that the findings from controlled experiments indicate that campaigns have the potential to decrease tobacco use among youth, with some evidence that campaigns are more likely to succeed when they are coordinated with school- or community-based programs. Wakefield and colleagues (2003a,c) also highlighted the idea that the effects seem to be more reliable when exposure occurs in preadolescence or early adolescence and when ads lead to emotional arousal. Consistent with theoretical models indicating that the effect of public health messages may be mediated through interpersonal communication (Flay and Burton 1990; Yanovitzky and Stryker 2001), Wakefield and colleagues (2003a,c) also suggested that the discussion of media campaigns may play an important role in either reinforcing or neutralizing the potential effects of antismoking advertising, as indicated by the findings from Hafstad and Aarø (1997).

Methodologic shortcomings highlighted by Hornik (2002) and NCI (2008) may explain some of the variation in findings from the controlled field trials. These problems have included: (1) difficulties in developing the televised components of the media exposure (Flay et al. 1988, 1995); (2) a low intensity of the media campaign or short duration of exposure to it (Bauman et al. 1991; Meshack et al. 2004); (3) insufficient control for baseline community characteristics and smoking-related risk factors and for prior and concurrent secular trends (Winkleby et al. 1993); and (4) differential attrition in longitudinal samples (Perry et al. 1992; Hafstad et al. 1997a; Vartiainen et al. 1998). Also, most analyses were not based on the primary sampling units considered as a whole that received the intervention (i.e., communities, schools). Rather, analyses were conducted on individuals within these sampling units, which can increase the chance of a Type 1 (false-positive) error due to an artificially inflated sample and failure to consider the effect on responses of shared experience within communities (see Hornik [2002] and NCI [2008] for further discussion of these issues).

In an analysis that considered the early cardiovascular programs of the 1970s and 1980s along with specific controlled field trials of youth media campaigns, NCI (2008) determined that media can “play an important role in affecting smoking behavior” (p. 508). Only one of the four reviewed studies that examined the effect of media alone found a positive effect (Hafstad et al. 1996, 1997a; Hafstad and Aarø 1997), however, the other three did not (Bauman et al. 1991; Winkleby et al. 1993; Flay et al. 1995). In comparison, five of six studies found evidence for an effect when the media was combined with a school-based intervention (Vartiainen et al. 1986; Perry et al. 1992; Flay et al. 1995; Flynn et al. 1997; Biglan et al. 2000a; Meshack et al. 2004).

Adding to this literature, a 2009 longitudinal controlled field trial by Solomon and colleagues included four matched pairs of media markets across four states randomly allocated to receive a 3-year television-and-radio intervention to increase smoking cessation and reduce smoking prevalence among adolescents. The media messages were based on social-cognitive theory. Although the authors did not find a significant time-by-condition interaction, significantly fewer participants in the intervention group were smoking in the past month at 3-year follow-up than in the control group after adjustment for baseline smoking status. Those in the intervention communities had greater cessation rates (an 18.1% 30-day point prevalence rate of quitting) than those in the control communities (14.8%) after the first year of the intervention, but no further gains were made up to 3 years, and light and occasional smokers were most likely to quit. The analyses used an intention to treat (ITT) method, assuming those who were lost at follow-up to have smoked at least one cigarette in the past 30 days, minimizing the possible effects of attrition bias. Unlike many others, this study used multilevel analytic techniques to account for similarities in reaction within individuals and similarities due to shared experience within matched media markets (Solomon et al. 2009).

Longitudinal population studies. Pechmann (2001) stated that there is limited direct evidence from controlled trials that media alone can influence youth smoking, but reported indirect evidence of the effects of stand-alone media campaigns from longitudinal population surveys of adolescents. These population surveys linked self-reported exposure to ads and reductions in smoking initiation (Siegel and Biener 2000; Sly et al. 2001b). Siegel and Biener (2000) examined the effect of the Massachusetts state campaign on smoking initiation by following 12- to 15-year-olds over 4 years and found that those who were 12 or 13 years of age and recalled campaign messages at baseline were less likely to start smoking than those who did not recall the messages. There were no effects for 14- and 15-year-olds and no effects on most knowledge and attitude measures. Similarly, Sly and colleagues (2001b, 2002) conducted longitudinal surveys to examine the effects of the Florida “truth” campaign on smoking initiation; they found that the number of advertisements recalled and campaign-related beliefs among youth at follow-up were associated with decreased smoking initiation.

Pechmann (2001) cautioned, however, that reverse causality cannot be ruled out with this type of evidence because adolescents who had strong antismoking beliefs at baseline and/or follow-up may have been more likely to pay attention to antismoking ads and also less likely to start smoking. However, Sly and associates (2001a,b) and Siegel and Biener (2000) minimized the likelihood of this possibility by controlling for baseline age, gender, prior smoking status, and the smoking status of friends and parents; Siegel and Biener (2000) also controlled for extent of television viewing. But as pointed out in the NCI review (2008) of the media and tobacco use, the studies by Sly and colleagues (2001b, 2002) measured recall at follow-up and the one by Siegel and Biener (2000) did not adjust for nonresponse at follow-up through weighting or analytic techniques. If those in the studies by Sly and colleagues who recalled the advertisements and those in the study by Siegel and Biener who completed the follow-up survey were relatively more likely to be nonsmokers, the possibility of finding an effect could well have been inflated.

Cross-sectional population studies. The 1967 ruling by the Federal Communications Commission that the Fairness Doctrine applied to cigarette advertising provided the first chance to examine the effects of antismoking messages on youth smoking. Much later, Lewit and colleagues (1981) associated various estimates of exposure to the antismoking advertisements with adolescent smoking behavior while controlling for a comprehensive range of covariates (Table 6.2). These authors found that the prevalence of smoking among youth was 3.0–3.4 percentage points lower during the Fairness Doctrine period than during the 16 months before it and that those who watched more television and were exposed to more anti-smoking messages were less likely to smoke. This study used measures of potential exposure based on hours of daily television watching reported by youth that were related by the authors to the number of antismoking advertisements aired during the Fairness Doctrine period in a given year. The NCI review of the media and tobacco use (2008) described this early study as making “significant strides in using more complex measures of exposure” (p. 518); more sophisticated measures than those used in the early days were not employed again until much later (Emery et al. 2005; Farrelly et al. 2005; Terry-McElrath et al. 2007), when campaign exposure was measured using gross rating points (GRPs). GRPs measure the relative reach and frequency of exposure to the campaign among the target audience within specific media markets. Emery and colleagues (2005) found that exposure to at least one U.S. state-funded antismoking advertisement in the prior 4 months was associated with lower perceived rates of friends’ smoking, greater perceived harm of smoking, stronger intentions not to smoke in the future, and lower likelihood of being a smoker. The variation in campaign exposure across different media markets in this study design provided natural comparison groups for examining the effects of campaigns and different intensities of exposure. These studies all used a comprehensive set of potential confounders, but only one (Terry-McElrath et al. 2007) also controlled for preexisting prevalence of youth smoking (in this case in 1995–1996) in different media markets to account for correlations between these rates and the frequency of antismoking advertisements aired in each market.

The findings from these and other cross-sectional, population-based evaluations of state and national anti-smoking campaigns developed by tobacco control programs can be more fully understood by examining the reported findings from 20 relevant papers cited in the three most recent comprehensive reviews (Richardson et al. 2007; Angus et al. 2008; NCI 2008). Of the 12 studies that examined attitudes or beliefs relating to smoking (Murray et al. 1994; Popham et al. 1994; Sly et al. 2001a, 2002; Farrelly et al. 2002; Hersey et al. 2003, 2005a,b; White et al. 2003; Niederdeppe et al. 2004; Emery et al. 2005; Terry-McElrath et al. 2007), all but 1 (Murray et al. 1994) found favorable changes associated with exposure to the campaign, and all 13 studies that examined intentions to smoke found favorable effects of such exposure (Popham et al. 1994; Seghers and Foland 1998; Bauer et al. 2000; Sly et al. 2001a,b, 2005; Farrelly et al. 2002; Niederdeppe et al. 2004; Emery et al. 2005; Hersey et al. 2005a,b; Johnston et al. 2005; Terry-McElrath et al. 2007; White et al. 2003). Fourteen of 16 cross-sectional population studies that examined smoking behavior (i.e., smoking prevalence, initiation of smoking, or quitting) associated with televised antismoking campaigns found a favorable change in the behavior (Lewit et al. 1981; Popham et al. 1994; Bauer et al. 2000; Siegel and Biener 2000; Sly at al. 2001a,b; White et al. 2003; Niederdeppe et al. 2004; Emery et al. 2005; Farrelly et al. 2005; Hersey et al. 2005a,b; Johnston et al. 2005; Terry-McElrath et al. 2007).

New studies published since these reviews further support these findings, indicating that well-funded state and national antismoking campaigns can reduce smoking among youth (Davis et al. 2007a; Evans et al. 2007; Tangari et al. 2007; Niederdeppe et al. 2008; White et al. 2008b; Farrelly et al. 2009). For example, Niederdeppe and colleagues (2008) surveyed 5,010 12- to 18-year-olds for their recall of Florida’s “truth” campaign ads, anti-industry beliefs, and nonsmoking intentions from April 1998 to May 2000. Rates of change were examined using interrupted time series techniques before and after budget cuts by the Florida Tobacco Control Program that took place between May 1999 and September 1999. After controlling for demographics, smoking in the home, degree of parental smoking, and parental monitoring, the study found that upward trends in recall of the Florida “truth” campaign weakened and nonsmoking intentions became relatively less prevalent following the budget cuts to the campaign.

As outlined in a number of reviews (Pechmann 2001; Jepson et al. 2006; NCI 2008), there are methodological issues with cross-sectional population studies to consider in determining the relative strength of those findings that linked media campaigns with preventing smoking among youth. Some of the cross-sectional studies used post-only (White et al. 2003) or single pre-post surveys (Seghers and Foland 1998; Bauer et al. 2000); these designs make it difficult to gauge whether any changes found were due to the media campaign or to secular trends in the exposed community and/or other events and activities unrelated to the media exposure. Use of a comparison group (Murray et al. 1994; Sly et al. 2001a; Niederdeppe et al. 2004), along with a comprehensive set of controls for preexisting demographic characteristics and levels of smoking in the community, may help to increase confidence that the observed effects are due to campaign exposure rather than preexisting baseline factors or secular trends (Farrelly et al. 2002, 2005; Emery et al. 2005; Terry-McElrath et al. 2007). Studies that provide measures at multiple baselines (e.g., Farrelly et al. 2002) can also help establish prior secular trends. Use of multiple measures during and after the campaign (Popham et al. 1994; Bauer et al. 2000; Niederdeppe et al. 2004; Emery et al. 2005; Johnston et al. 2005) and observation of changes in factors thought to be mediators of the effect of campaigns, such as certain beliefs and attitudes (Sly et al. 2002; Hersey et al. 2003, 2005a,b; Evans et al. 2004a), can also help increase confidence that any observed changes in smoking behaviors are the result of campaign activity rather than alternate trends or concurrent events.

Still, a key difficulty in attempting to assess the specific media effects of statewide and national media campaigns is the fact that most were developed and run within the context of broader tobacco control programs and activities, such as tax increases (Friend and Levy 2002; Farrelly et al. 2003a). Regardless, some authorities suggest that integrating media campaigns within a broader tobacco control program is important to their effectiveness (Schar et al. 2006; Angus et al. 2008; NCI 2008), and thus, considerations of precisely determining the effects of the media campaigns, while important, perhaps need to be seen as less compelling than meeting the goal of offering a program that produces positive changes. Schar and colleagues (2006) point to the success of mass media campaigns in Finland as well as in California, Massachusetts, and a number of other states that have implemented youth tobacco campaigns that included other program elements (see “Comprehensive State-Level Tobacco Control Programs” later in this chapter for more detail); these included such initiatives as a school curriculum, cessation programs, and policy changes that increased cigarette taxes and smoke-free environments and strengthened laws restricting youth access. Schar and colleagues (2006) conclude that “a key contributor to successful mass media campaigns is the synergy resulting from the different program elements working together to change society’s prevailing attitudes about tobacco use” (p. 5). Finally, Richardson and colleagues (2007) indicate that campaigns are likely to “work best when combined with broader tobacco control initiatives produced by tobacco control bodies” (p. 4).

The consistent positive findings across a variety of study designs provide convincing evidence that anti-smoking media campaigns can be effective in reducing youth smoking but that certain factors and conditions are required for their success. There is broad consensus that these factors include the use of formative research in the development of messages and, for campaign messages, sufficient intensity and duration of exposure (USDHHS 1994; Sowden 1998; Pechmann and Reibling 2000b; Siegel 2002; Farrelly et al. 2003a). Recent research and reviews have begun to focus more heavily on which message characteristics work best, what the ideal level of exposure is, and which types of youth are most or least affected by mass media campaigns against smoking.

Factors That May Optimize the Effectiveness of Mass Media Campaigns

Mass media campaigns against smoking, especially those with televised components, require considerable investment, making it particularly important to understand the factors and strategies that optimize their effectiveness. This section summarizes conclusions from various reviews and new research (Pechmann 2001; Siegel 2002; Farrelly et al. 2003a; Wakefield et al. 2003b,c; Schar et al. 2006; Richardson et al. 2007; Angus et al. 2008; NCI 2008) on the effects of different types of messages, the optimum intensity and duration of exposure to messages, and how messages may influence different youth (i.e., classified by gender, age, race/ethnicity, socioeconomic status, risk status).

Theme, emotional tone, format, and characteristics of execution. Studies to assess differences in the responses of youth to various types of ads have usually used controlled exposures; less often, they have employed naturalistic exposures. In controlled-exposure studies, youth typically view a series of messages and then either discuss their reactions to them (often in focus groups) or complete an experimental study. In experimental studies youth may rate ads in terms of their emotional impact, liking, or other features thought to be associated with increased antismoking attitudes and behaviors, or are asked about these attitudes and behaviors directly. It is also possible that youth will complete cognitive processing tasks (Shen et al. 2009), have physiological data recorded such as heart rate (Leshner et al. 2011), or complete memory questions on viewed messages (Leshner et al. 2010) among many possible experimental approaches aimed at better understanding the processes behind mediated message effects for youth.

The limitations of these controlled-exposure methods are that the exposure does not mimic real-world viewing contexts and that one cannot examine the effects of multiple exposures occurring over months and years. The advantages of naturalistic studies are that the effects of different types of messages can be examined in a real-world setting: messages are viewed within a crowded media environment, often within a person’s home; there are a myriad of distractions; and the effects of exposure over weeks, months, or years can be studied. The limitations of these naturalistic-exposure studies are that they rely on self-reported recall of messages, which may be correlated with smoking intentions and behaviors, and they cannot rule out other factors that may influence outcomes, such as policy changes and geographic or historic differences in exposure to different types of messages.

Pechmann’s (2001) review highlighted the mixed findings from the early controlled-exposure studies that compared different ad themes (Goldman and Glantz 1998; Teenage Research Unlimited 1999). For example, one study that used 20 focus groups indicated that ads showing the serious physical consequences of smoking—portrayed either graphically, dramatically, or emotionally—performed well (Teenage Research Unlimited 1999), while another study, summarizing the findings of 186 focus groups, indicated that ads about secondhand smoke or about industry manipulation rated best (Goldman and Glantz 1998). And in a copy-test study (representative populations view ads and answer survey questions afterwards), Pechmann and colleagues (2003) found that ads depicting the impact of smoking on infants and children, those showing that smoking is socially unacceptable, and ads indicating that nonsmoking is the norm significantly decreased youth’s reported intentions to smoke.

Siegel (2002) suggested that the mixed findings from early studies may be explained by the fact that the studies considered only differences in the messages’ themes (Goldman and Glantz 1998; Pechmann et al. 2003) and not their emotional content. Subsequent reviews (Farrelly et al. 2003a; Wakefield et al. 2003b; Schar et al. 2006; NCI 2008) have considered both the theme and emotional tone of advertisements and have examined findings of more recent naturalistic studies as well as controlled-exposure studies. In support of theories of persuasion that emphasize emotion (Cohen 1990; Eagly and Chaiken 1993; Forgas 1995; Escalas et al. 2004; Baumeister et al. 2007), these reviews concluded that there is consistent evidence that ads eliciting strong emotional responses (such as disgust, loss, sadness, dread, and anger) through personal testimonials and visceral imagery of the health effects of smoking, or that portray deception on the part of the tobacco industry, can increase attention, generate greater recall and appeal, and affect young audiences’ smoking-related beliefs and intentions to smoke. However, exposure to high levels of negative emotion may actually hinder persuasiveness and elicit undesirable negative consequences depending on the stimulus itself (Erceg-Hurn and Steed 2011). This makes message testing extremely important. Ads featuring harm to appearance, addiction, and decreased athletic performance are concluded to be less effective than those about health effects or the tobacco industry’s deceptive practices (Goldman and Glantz 1998; Pechmann et al. 2003; Smith and Stutts 2006). The NCI review (2008) of the media and tobacco use noted that some themes (e.g., those on health effects) lend themselves to the elicitation of negative emotions more readily than others, while the “encouragement to quit” theme is often more upbeat and positive. Copy-test studies have shown that when the message’s theme and executional style have been controlled, it is the negative emotional elements that are independently related to more encouraging appraisals of the message (Terry-McElrath et al. 2005; Wakefield et al. 2005a).

The American Legacy Foundation “truth” campaign used a mix of serious and sarcastic ads to get the overall message across to youth that tobacco companies are deceptive and misleading; the intent was to elicit outrage and spur young people to resist tobacco use. Recent population-based research on the effects of the first 3 years of the “truth” campaign (Davis et al. 2007a; Farrelly et al. 2009) indicated that confirmed exposure to the campaign was associated with stronger antitobacco attitudes and intentions not to smoke in the future. Reviews caution, however, that ads that use humor have been found to be less effective than those that evoke negative emotions (Schar et al. 2006; NCI 2008). It is not known whether “truth” ads that evoke negative emotions differ in effectiveness from those that use humorous techniques in terms of creating the observed effects. Also, reviews have suggested the need for repeated exposure over time to several different types of ads that deal with the industry’s manipulations and deceptive practices to educate audiences about these topics, as the ads may be misunderstood at first (Wakefield et al. 2003b; Schar et al. 2006). The reviews also caution that research into the effectiveness of the counter-industry ads (typically those used in the “truth” campaign that highlight the deceptive practices of the industry) has been limited to the United States, and the findings may have limited transferability to countries where the tobacco industry has a lower profile. Indeed, a recent focus group conducted in the United Kingdom (Devlin et al. 2007) found that industry-manipulation ads provided new information that led to greater interest among adolescents, but comprehension was a barrier with many youth needing the ideas explained.

One review (Schar et al. 2006) summarized findings from controlled field trials, controlled-exposure studies, and focus groups and suggested that ads about the social consequences of smoking and about refusal skills can be effective (Flynn et al. 1992, 2007; Biglan et al. 2000a; Pechmann et al. 2003; Devlin et al. 2007). In addition, a recent longitudinal controlled field study conducted in four media markets within each of four states, detailed earlier in this chapter, provided some modest support for the ability of ads about social norms to influence smoking by youth (Solomon et al. 2009). A new series of controlled-exposure studies added to this literature (Zhao and Pechmann 2007) by examining four versions of the same basic social-disapproval antismoking message (depicting a gathering of young college students) that varied along two dimensions (positive vs. negative frame, promotion-vs. prevention-focused message) that were presented to adolescents categorized as either promotion focused (motivated by achievements and advancement) or prevention focused (motivated to avoid threats to security and safety). The study found that promotion-focused, positively framed messages were most effective at persuading promotion-focused adolescents not to smoke and that prevention-focused, negatively framed messages were most effective for prevention-focused adolescents. Most of these studies examining the influence of these types of themes have been conducted using controlled exposure to ads; one population-based study that specifically used these message themes found no effects on antismoking attitudes or smoking behavior (Murray et al. 1994). Therefore, the extent to which these messages would be effective at the level of a broad population-based mass media campaign is unclear.

As discussed in Chapter 5, ads developed by the tobacco industry that counsel youth not to smoke and emphasize personal choice, such as the “Think. Don’t Smoke” ads developed by Phillip Morris, generally had the lowest ratings and effects on smoking intentions or behavior among all ads that were viewed (Teenage Research Unlimited 1999; Biener 2002; Niederdeppe et al. 2005; Wakefield et al. 2005a; Henriksen et al. 2006; Pechmann and Reibling 2006; Farrelly et al. 2008). Angus and colleagues (2008) reported that four of five studies reviewed found that industry campaigns performed poorly compared with tobacco control campaigns. One of these studies showed that youth who recalled the industry campaigns were significantly more likely than their unexposed peers to have intentions to smoke in the future (Farrelly et al. 2002). Another study (Wakefield et al. 2006) found that greater exposure to industry ads directed at youth was associated with stronger intentions to smoke among younger survey participants, and that exposure to industry ads directed at parents was associated with several undesirable outcomes, including stronger approval of smoking and stronger intentions to smoke, for older survey participants. Supporting this research, a new study by Farrelly and colleagues (2009) found that at 3-year follow-up, exposure to the Philip Morris campaign was associated with more favorable beliefs and attitudes toward tobacco companies and a trend for weaker intentions not to smoke.

The NCI review (2008) of the role of the media and tobacco use pointed out that structural features, such as pacing, use of loud music, and cuts or edits of advertisements, may be important in that they can increase the “message sensation value,” which has been associated with greater processing of the message (Niederdeppe et. al 2007). Niederdeppe and colleagues (2007) examined 32,977 adolescents from seven cross-sectional waves of the American Legacy Foundation’s Media Tracking Surveys, which assessed these youth for confirmed recall of television ads from the “truth” campaign and their need for sensation. After controlling for a comprehensive set of ad features, demographics, and “truth” campaign GRPs, the odds that the messages were recalled increased with more frequent edits and unrelated cuts, intense imagery, sound saturation, loud and fast music, and second-half punch (surprising or shocking ending).

Despite the common use of television, radio, and outdoor advertising in many state and national antismoking campaigns, few studies have examined the relative effectiveness of these different formats, although commercial information suggests that television has the broadest reach. In a cross-sectional study, Seghers and Foland (1998) found that television ads were associated with greater recall than were other formats, and in a controlled-exposure study, Flynn and colleagues (2007) found that televised messages generally received higher ratings than did radio messages. In a recent controlled field trial (Solomon et al. 2009), no differences in smoking outcomes were found by format for those in the exposed group who had heard at least one radio message, but those who had reported seeing at least one television message were less likely to have smoked in the past 30 days than were those who had not seen any messages (54% vs. 62.6%). In a longitudinal study (Siegel and Biener 2000), neither radio nor outdoor advertising was associated with reduced initiation of smoking at 4-year follow-up, but recall of a television message was associated with reduced initiation in 12- and 13-year-olds. It is unclear whether the lack of success of these radio campaigns was due to the format, the messages typically broadcast on the radio stations, or the lower population reached by radio.

In recent years, antismoking messages have increasingly been presented via antitobacco Web sites. A study of differences between design elements, persuasive strategies, and information content across the Web sites of youth antitobacco organizations (which also included the areas for prevention of youth smoking on tobacco industry Web sites) indicated that the industry sites provided the weakest persuasive messages; grassroots (costkids.org [2012]) and government sites provided the strongest messages; and medical sites provided mostly scientific information for specialists (Lin and Hullman 2005). Delivering a message through the Internet can encourage changes in smoking behavior through interactive communication; interactivity can range from quizzes, contests, and games to connecting to campaign Web sites and other users through sites such as Facebook and MySpace. Antismoking campaigns may be able to increase their reach and persuasive impact by using these social networking sites, given a survey indicating that over one-half of U.S. youth who use the Internet have accessed these sites (Lenhart and Madden 2007). For example, the American Legacy Foundation launched the truth profile pages (INFECT truth) on a range of social networking sites. Preliminary results indicate that the addition of these profile pages was associated with an estimated increase of 20,000 unique visitors a week to the truth Web site (2010) in a comparison with traffic during typical campaigns that do not involve social networking sites (Vallone 2007). The video-sharing Web site YouTube provides another modality through which youth may be exposed to both traditional and innovative antitobacco messages from antitobacco organizations and motivated individuals (e.g., “Thanks Tobacco: You Killed My Mom” posted on April 13, 2007 [YouTube 2007]). YouTube also allows viewers to post comments about videos and send links to others. Determining the impact of messages conveyed through this medium is a fertile area for new research. The effects of antismoking messages delivered via text messaging and the use of this technology as a way for smokers to seek help for quitting smoking after exposure to antismoking messages is another important area for research.

Intensity and duration. Despite the conclusion of most reviews that campaign funding and exposure need to be “sufficient” to ensure effects, there is little research as to what levels of intensity and duration might be “sufficient.” Nevertheless, studies indicate that increased exposure to antismoking messages over time results in a greater likelihood of having beliefs consistent with the campaign against smoking, decreased youth smoking, a lower intent to smoke, and less initiation of smoking than in those not exposed (Emery et al. 2005; Farrelly et al. 2005; Johnston et al. 2005; Terry-McElrath et al. 2007).

Sly and colleagues (2002) found a dose-response effect of Florida’s antismoking advertising in its “truth” campaign, with greater numbers of different ads recalled at follow-up (but not greater overall exposure) associated with greater odds of remaining a nonsmoker during a 22-month period. Later, Emery and colleagues (2005) reported that if the average exposure among youth was less than one state-sponsored antismoking ad over a 4-month period, there were no discernible effects. Exposure to one or more ads for the same period was associated with lower odds of being a smoker. Elsewhere, Farrelly and colleagues (2005) found dose-response effects of the American Legacy Foundation “truth” campaign for up to an average of four ads per month (average cumulative 10,000 GRPs over a 2-year period), after which there were diminishing returns. This suggests that in efforts to reduce youth smoking, there is a threshold of exposure below which antitobacco advertising may not have an influence, and effects increase with increasing exposure up to four ads per month (CDC 2007b). Terry-McElrath and colleagues (2007) used the same study design as Emery and colleagues (2005), but with more years of data from state anti-tobacco campaigns, and also found a dose-response effect with no point of diminishing returns. It should be noted, however, that state tobacco control campaigns that aired during the 1999–2003 period of this study may not have been broadcast at a level sufficient to detect the point of wear out (among 12- to 17-year-olds the average was just 1.08 target rating points [TRPs] per month) (Wakefield et al. 2005b). Only Arizona in 1999 and 2000, Florida in 1999, Minnesota in 2001, and Utah in 2001–2003 averaged more than four exposures per month to state antitobacco ads among 12- to 17-year-olds (Szczypka et al. 2005).

A more recent study by Biener and colleagues (2008) provides strong support for the relative utility of emotionally evocative advertising as well as an idea of how its effects relate to broadcast intensity (broadcast volume, i.e., media weight or rating points in reaching targeted audiences). The authors assessed confirmed recall of nine specific antitobacco ads in a sample of 3,332 12- to 17-year-old adolescents from January 2001 to June 2002. The intensity and duration of the broadcast of the nine ads were estimated from adolescent TRPs, and each ad was given an emotional intensity score based on a previous study of ad ratings with adolescents. The analyses controlled for demographics, household education level, TV-watching frequency, and smoking status; the findings indicated that the level of the ads’ emotional intensity was a significant predictor of recall. As emotional intensity increased from the lowest to the highest level, the odds of recall rose by more than a factor of three. The authors also found that the broadcast volume (media weight) was a significant predictor of recall: as the TRPs increased from the lowest to the highest level, the odds of recall more than doubled. In addition, TRPs were a significantly stronger predictor of recall of the two ads low in emotional intensity (odds ratio [OR] = 2.68) than of the two ads high in emotional intensity (OR = 1.36). These findings indicate that for ads high in emotional intensity, less media weight is required to generate recall than for those that are low in emotional intensity.

Higher recall does not necessarily equate to the effectiveness of an ad and, ultimately, to changes in behavior. However, population-based research indicates that recall of campaign messages has been associated with reduced smoking behavior in youth (Siegel and Biener 2000; Sly et al. 2002). Other research indicates that emotionally evocative messages are perceived as more effective (Pechmann and Reibling 2006), even after controlling for recall (Biener et al. 2000; Biener 2002).

Research linking cuts in the funding for antitobacco campaigns to the halting of declines in youth smoking or even to increases in youth smoking (Sly et al. 2005; Niederdeppe et al. 2008; White et al. 2008b) indicates that optimal implementation for campaigns would involve ongoing exposure at regular intervals. This conclusion highlights the notion, widely acknowledged in advertising literature, that media campaigns influence behavior while they are on air but that their effect diminishes very quickly once they are removed from the air (Tellis 2004).

Context. There is a need not only to identify the characteristics of messages and the level of exposure most likely to change attitudes and behavior among youth about smoking, but also to understand the influence of the circumstances surrounding exposure to messages. Evidence from the broader public health and advertising domains indicates that the contexts in which ads are viewed (Goldberg and Gorn 1987; Sharma 2000) and the extent and type of discussion that ads generate (Morton and Duck 2001) can influence the processing and impact of the messages they impart. Research into the effect of these factors on the responses of adults to antismoking campaigns has shown that messages may be processed less effectively when they are aired during programs that transport viewers into the story (e.g., drama and soap operas [Durkin and Wakefield 2006, 2008]) rather than during lighter entertainment (e.g., comedy). Other research has found that engagement in ad-related discussions can enhance the impact of antismoking messages on both intentions to quit and attempts to quit by adolescents (Hafstad et al. 1996, 1997a; Hafstad and Aarø 1997) as well as adults (Durkin and Wakefield 2006; Dunlop et al. 2008). Several studies (Hafstad et al. 1996, 1997a; Hafstad and Aarø 1997) found that in adolescents the most important predictor of positive behavioral reactions was campaign-stimulated discussion with peers. In a more recent study, adults were most likely to discuss advertising that contained information about the negative health consequences of smoking presented through graphic images or simulations of bodily processes (Dunlop et al. 2008). This result is consistent with the observation that interpersonal discussion can bring antismoking messages into an immediate social environment that may lead to either the extension or reduction of a message’s impact (Flay and Burton 1990; Southwell and Yzer 2008).

Audience segmentation. Tailoring the message’s content to specific audience subgroups (defined, for example, by age, gender, race/ethnicity, a desire for sensation, or socioeconomic status) has the potential advantage of increasing a message’s relevance and ability to persuade. However, tailoring the ad’s message, settings, and actors to specific population subgroups requires funding multiple campaigns to convey a variety of messages or tailored versions of a key message rather than simply producing general campaigns to convey messages likely to resonate with all population groups. Also, given the finite resources of most public health campaigns, this type of tailoring may result in having a lower proportion of funds available to broadcast these ads, resulting in lower rates of exposure to the messages. The extent of tailoring and segmentation, therefore, needs to be weighed carefully against goals of maximizing campaign exposure.

Youth- versus adult-targeted campaigns. Although most of the reviewed studies examined campaigns that were specifically targeted to youth, it is a matter of debate whether these campaigns are the best choice for reducing youth smoking (Hill 1999). Beaudoin (2002) found that many youth-targeted campaigns presented the short-term, social consequences of smoking and used humor, while ads targeted to adults more often highlighted the long-term consequences and evoked fear. A study by Flynn and colleagues (2007) that examined ratings for a series of messages on social norms (many of which used humor) indicated that it may be particularly difficult to design messages that appeal to older youth and found strong differences in ratings between age groups. Evidence that younger youth may be more likely than older youth to decrease their intentions to smoke in response to counter industry mass media campaigns (Sly et al. 2001b; Wakefield et al. 2003b; Farrelly et al. 2005) was interpreted in one review (Schar et al. 2006) as indicating that older adolescents may be better addressed by campaigns targeted to a general audience.

Evidence from studies that compared responses from younger and older youth to a range of youth- and adult-targeted messages (e.g., on cessation, secondhand smoke, family guidance, health benefits, health effects, industry manipulation, and smoking being “uncool”) found that youth responded as favorably to adult-targeted ads as to youth-targeted ads (Terry-McElrath et al. 2005; Wakefield et al. 2005a, 2006; NCI 2008). This finding is consistent with findings from adult-targeted mass media campaigns that have successfully reduced the initiation of smoking and of smoking behavior among youth (Lewit et al. 1981; Siegel and Biener 2000; White et al. 2003; Schar et al. 2006). In population studies of U.S. youth (Emery et al. 2005; Terry-McElrath et al. 2007), beneficial effects on youth smoking were found from exposure to the overall complement of state antitobacco campaign ads, not just youth-targeted campaigns, and a study by Emery and colleagues (2007) indicated that a majority of the state campaign GRPs came from adults rather than youth. The NCI review (2008) of the media and tobacco use proposed that the success of adult-targeted campaigns for adolescents may be due in part to changing the broader social norms about smoking. Further exploring this issue, Angus and colleagues (2008) suggested that using adult-focused campaigns for reducing smoking in youth may avoid the danger that “using youth targeted mass media campaigns in isolation may create the impression that, whilst children should avoid it, tobacco use is an acceptable adult behavior” (p. 16).

Gender, race/ethnicity, and socioeconomic status. The limited amount of research that has examined differences between youth subgroups in their appraisals of antitobacco ads has not yet found any systematic differences by gender, race/ethnicity, or nationality (Terry-McElrath et al. 2005, 2007; Wakefield et al. 2005a; Flynn et al. 2007). In fact, these studies indicate that the advertisement’s characteristics are much more important than the characteristics of the audience. Consistent with this research and with studies of adult responses to advertising against smoking (Siahpush et al. 2007), White and colleagues (2008b) found that across socioeconomic groups, 12- to 15-year-old adolescents showed parallel reductions in smoking behavior during the period of the well-funded Australian National Tobacco Campaign, which included emotionally evocative messages about the health effects of smoking. However, during periods of low funding, when adolescents were exposed to sparse, sporadic campaigns, smoking among 12- to 15-year-olds increased, and those from the lower socioeconomic groups had the greatest monthly and weekly increases. This study suggests that when well-funded campaigns are not on the air, it is youth from lower socioeconomic groups who are most negatively affected. This is consistent with research that suggests disparities in health knowledge may widen when there are only low or moderate levels of publicity about these campaigns (Viswanath et al. 2006).

High-sensation seekers and high-risk youth. Despite early indications that media interventions may be especially effective for high-risk youth (Flynn et al. 1994, 1997), subsequent studies have provided mixed results on this issue. For example, population-based studies have shown that the impact of the American Legacy Foundation’s national “truth” campaign on smoking by youth was similar among high- and low-sensation-seeking adolescents (Farrelly et al. 2005; Niederdeppe 2005; Thrasher et al. 2006). Niederdeppe and colleagues (2007) examined the structural elements of ads and found that results were nearly identical between youth with high needs to seek sensation and those with low needs, although the magnitude was somewhat higher among youth with a high need for sensation. Thrasher and colleagues (2006) also found that anti-industry attitudes were similar across sensation-seeking groups, but were lower among adolescents weakly bonded to social supports such as families, schools, and communities. However, the relationship between anti-industry attitudes and smoking was consistent across both risk groups (both sensation-seeking and weakly social-bonding risk groups).

In contrast to early predictions, Pechmann and Reibling (2006) found that youth with conduct disorders (who also are often high-sensation seekers) did not give a variety of antitobacco messages higher ratings than they gave to the control message, but for youth who did not have conduct disorders (81% of the sample), advertisements focusing on young victims suffering from serious smoking-related disease elicited disgust, enhanced anti-industry motivation, and reduced intentions to smoke. A study by Helme and colleagues (2007) randomly allocated middle school students to either a high- or low-sensation-value message. Students’ level of need for sensation seeking (high vs. low) was also measured. The authors found no differences between high- and low-sensation-value messages in changing antismoking attitudes, future intentions to smoke, self-efficacy not to smoke, perceived effectiveness of the message, and perceived risk for self and others. The authors found, however, that high-sensation seekers were more likely to show changes than were low-sensation seekers in antismoking attitudes, intentions not to smoke, self-efficacy not to smoke, perceived effectiveness of the message, and perceived risk from smoking. In assessing the importance of the effects of these campaigns on high-risk youth, however, it is important that the proportion of youth who fall into these categories (of high- and low-sensation seeking) be considered. A greater population effect on the prevalence of smoking among youth is likely to be achieved by focusing on what is effective for the majority of youth, and the proportion of youth who have high needs for sensation might not be large enough in some cases to make them a specific target group for interventions to prevent smoking.

Theoretical implications. Some support for models of health behavior change is provided by studies finding that exposure to antismoking messages leads to changes in, or increased salience of, attitudes, beliefs, and intentions relative to smoking as well as reduced smoking behavior (e.g., Popham et al. 1994; Sly et al. 2001b, 2005; Farrelly et al. 2002; White et al. 2003; Meshack et al. 2004; Niederdeppe et al. 2004; Emery et al. 2005). These cross-sectional studies could not, however, examine whether the changes in attitudes and beliefs preceded the changes in intentions and behavior. Controlled and longitudinal studies are better for testing these pathways. Some longitudinal studies have found changes in smoking intentions and behavior without concurrent changes in attitudes and beliefs (Siegel and Biener 2000; Solomon et al. 2009), and others have found that changes in these proposed mediators have occurred before the change in smoking behavior. Flynn and colleagues (1992, 1994) found support for social-cognitive theory, with differences between intervention and control groups on mediating variables (such as smoking norms, attitudes toward smoking, refusal skills) occurring before differences in smoking behavior. Further support for the idea of changes in health behaviors resulting from exposure to antismoking messages is afforded by a series of cross-sectional, population-based studies that surveyed youth in states with relatively higher exposure to the American Legacy Foundation “truth” campaign and found them to have greater agreement with campaign-relevant beliefs and lower rates of smoking initiation than youth from states with relatively lower exposure (Hersey et al. 2005a). Finally, Evans and colleagues (2004a) found that the perceptions of positive social images for not smoking among nonsmokers targeted by the “truth” campaign mediated the relationship between exposure to the campaign and smoking status.

Summary of the Current Evidence Base Regarding the Use of Mass Media

The power of the mass media to influence public perceptions of tobacco was first documented in the aftermath of the 1967 Fairness Doctrine ruling, when considerable reductions in youth smoking were shown to be associated with government-sponsored antismoking television messages. Reviews of early field trials provided some support for the effectiveness of media interventions combined with school programs within communities, but since then, a host of population-based investigations on mass media campaigns have provided convincing evidence that these campaigns, by themselves, can decrease youth smoking. The NCI review (2008) of the media and tobacco use concludes that: “Evidence from controlled field experiments and population studies conducted by many investigators in many countries shows that antitobacco mass media campaigns can reduce tobacco use” (NCI 2008, p. 537). More recent studies (Davis et al. 2007a; Farrelly et al. 2009; Solomon et al. 2009) provide further support for the utility of mass media campaigns to reduce youth smoking.

In summary, the evidence is sufficient to infer a causal relationship between adequately funded antismoking media campaigns and a reduced prevalence of smoking among youth. Evidence has been consistently strong across a wide range of longitudinal-cohort and cross-sectional, population-based studies that have controlled for a variety of potential confounders, have compared effects of exposure with less or no exposure, and have shown diminishing effects when exposure is reduced.

Evidence also suggests a dose-response relationship between exposure to antismoking media messages and reduced smoking behavior among youth, which is further evidence of the effectiveness of these messages. Very few studies, however, have explored the optimum level and duration of exposure to these messages for exerting effects on youth smoking. The few studies to examine this question suggest that levels between one ad per 4-month period and four exposures of the target audience per month are needed to observe an impact, with dose-response findings indicating closer to four exposures per month are needed to be more effective and one study indicating that emotionally evocative messages need less exposure than less emotional messages.

The research reviewed in this section also provides consistent, strong evidence through controlled-exposure and population-wide studies that media ads designed for adults decrease the prevalence of smoking among youth. This effect may be the result of changing the social norms of youth about smoking by altering their perceptions of smoking prevalence among adults as well as reduced exposure to adult smoking (NCI 2008). In addition, a number of population-based and controlled-exposure studies provide evidence that the characteristics of advertising messages seem to be more influential than the characteristics of the audience in terms of the results obtained, suggesting that messages developed for specific target groups may in fact translate successfully to broader audiences and that the expense of developing and airing many different ads for specific target groups may be able to be alleviated.

It is clear that not all campaigns will be equally effective, and recent research has focused on the factors that differentiate influential campaigns and messages from those that are less successful. The research provides consistent evidence from controlled-exposure studies that ads evoking strong negative emotions (including those about the health effects of smoking and exposure to secondhand smoke as well as those about the deceptiveness of the tobacco industry) show greater recall and are rated higher on measures of appeal and smoking-related beliefs and intentions not to smoke than are ads that do not evoke these kinds of emotions.

This review, then, provides important evidence on the efficacy of antismoking mass media campaigns and considerable direction on how those campaigns should be developed in content, tone, and intensity.

Prior Reviews

A Cochrane review of community-based interventions for preventing smoking in young people defined community interventions as coordinated, widespread programs in a particular geographic area or in groupings of people who share common interests or needs that support nonsmoking behavior (Sowden and Stead 2003). The review included 17 RCTs and non-RCTs published up to 2002 that assessed the effectiveness of multicomponent community interventions in comparisons with no intervention or with single-component interventions or school programs alone in young people under the age of 25 years. Four studies reported interventions aimed at preventing the uptake of smoking in the community among young people that were part of larger, community-wide programs to reduce cardiovascular disease in all age groups in specific areas: California (Winkleby et al. 1993); Minnesota (Perry et al. 1994); North Karelia, Finland (Vartiainen et al. 1998); and Rotherham, England (Baxter et al. 1997). One study evaluated an intervention targeted at cancer prevention in New South Wales, Australia (Hancock et al. 2001), and another examined a community-level intervention in Minnesota and Wisconsin that focused on deterring tobacco use via a public policy initiative (Murray et al. 1994). Five other interventions focused exclusively on preventing the uptake of smoking in young people in specific locations: Wensleydale, England (Davidson 1992); Chicago, Illinois (Kaufman et al. 1994); Cardiff, Wales (Gordon et al. 1997); Sydney, Australia (Tang et al. 1997); and Oregon (Biglan et al. 2000a). Six other interventions were aimed specifically at young people, with the focus on preventing or reducing the use of tobacco, alcohol, and drugs in certain locations: Kansas City, Kansas, and Kansas City, Missouri (Pentz et al. 1989b); Wisconsin (Piper et al. 2000); Boys & Girls Clubs of America across the United States (St. Pierre et al. 1992); New Jersey (Aguirre-Molina and Gorman 1995); California (Sussman et al. 1998); and American Indian reservations (Schinke et al. 2000).

All 17 studies in the Cochrane review used a controlled trial design, with 6 using random allocation of schools or communities. Of 12 studies that compared community interventions with no-intervention controls, 2 (both part of programs to prevent cardiovascular disease) reported a lower prevalence of smoking following the intervention (Perry et al. 1994; Vartiainen et al. 1998). Of four studies comparing community interventions with school-based programs, only one found differences in the reported prevalence of smoking (Biglan et al. 2000a), and samples of expired carbon monoxide detected no differences in smoking between groups. One study reported a lower rate of increase in the prevalence of smoking in a community receiving a multicomponent intervention than in a community exposed to a mass media campaign alone (Kaufman et al. 1994). Finally, one study reported a significantly lower prevalence of smoking among the group receiving media, school, and homework components than in the group receiving the media component only (Pentz et al. 1989b).

Overall, Sowden and Stead (2003) concluded that there was some support for the effectiveness of community interventions in preventing the uptake of smoking by young people. The reviewers found it was not possible to pool the results because the studies were heterogeneous in terms of interventions, communities, participants, and measurement of outcomes. Indeed, it could be argued that the very nature of a community intervention means that no two initiatives could ever be the same and, therefore, that their findings should not be aggregated. Furthermore, establishment of control groups in these kinds of studies is difficult and may require extensive negotiations or a “delayed” intervention condition. And because communities are assigned to intervention or control groups, the analysis of outcomes needs to be at the level of the community rather than the individual level. Furthermore, the large size of community interventions means that the measurement of their implementation can be difficult and expensive. Regardless, the studies included in the review represent the most methodologically rigorous set of studies available on the effectiveness of community interventions in influencing smoking among young people.

In their review, Sowden and Stead (2003) recommended several principles to be considered in planning future community interventions: building on the elements of existing programs shown to be effective rather than repeating methods with limited success, adapting program components to suit the community, pretesting and fine-tuning program messages and activities before full implementation, being guided by theoretical constructs of behavior change, and ensuring that activities reach the intended audience.

Newer Studies

Several studies published since the Cochrane review by Sowden and Stead (2003) also suggest modest support for community interventions. One such study involved an evaluation of the effects on youth of the NCI-funded Community Intervention Trial for Smoking Cessation (COMMIT), a multicomponent, community-based intervention designed to decrease the prevalence of smoking among adults and increase quitting among adult smokers (Journal of the National Cancer Institute 1991). In addition to its components for adults, COMMIT (Lichtenstein et al. 1994) included youth-oriented activities in four principal areas: school-based education programs, smoking policies in schools, legislative activities related to youth smoking, and participation by students and teachers in other COMMIT activities. The evaluation, which was reported by Bowen and colleagues (2003), involved a two-group pretest/posttest with matched communities randomly assigned to either control or intervention; the ninth-grade classroom (students 14 and 15 years of age) was the unit of assessment. Bowen and coworkers (2003) found no differences in changes in smoking over time between youth in the intervention and control communities.

Full Court Press (FCP), a multifaceted community intervention to change social norms about tobacco use, was intended to reduce the uptake of smoking among youth in Tucson, Arizona. The program included media advocacy, mobilization of youth to build a network of young people committed to reducing tobacco use and advocating for policy change, improvements in the enforcement of laws governing youth access, and development of cessation services (Ross et al. 2006). Results indicated that the prevalence of youth smoking declined 27% between 1996 and 2000 in Tucson during the FCP intervention period, which was larger than changes observed in national and statewide trends for prevalence after accounting for gender and racial/ethnic differences. A subsequent study of FCP that adjusted for other changes in the sociodemographic and economic environment (e.g., increases in cigarette prices) also found beneficial effects on the prevalence of smoking (Ross et al. 2006).

Summary Regarding Community-Level Programs

Coordinated, multicomponent community programs may be able to reduce smoking among young people, and they do so more effectively than can single strategies. Results are likely to depend upon the mix of strategies chosen and the reach of the program’s efforts into communities. The most effective components should form the basis for future community interventions (Sowden and Stead 2003).

Prior Reviews

Several reviews have examined the effectiveness of statewide tobacco control programs on reducing smoking by youth. Wakefield and Chaloupka (2000), who reviewed published literature, reports of program evaluations, and working papers about the effects of state programs in Arizona, California, Florida, Massachusetts, and Oregon, found youth in these states to have high levels of recall of the state’s mass media campaigns and generally positive improvements in tobacco-related beliefs and attitudes. In addition, the combination of program activity and increases in tobacco taxes was found to reduce cigarette consumption more than would be expected from price increases alone. Reviews of programs in California and Massachusetts documented beneficial effects on the prevalence of adolescent smoking compared with other states (Briton et al. 1997; CDC 1999a; Independent Evaluation Consortium 2002), and Florida had promising indications of reduced smoking when its program was reviewed (CDC 1999b). Siegel (2000) reviewed these three state programs, as well as those of Arizona and Oregon, commenting that the extent of the tobacco industry’s attempts to undermine the programs was a good indicator of the programs’ effectiveness. Responses by the tobacco industry had been aggressive, including more intensive tobacco marketing; increased lobbying at the state and local levels; attempts to limit the tobacco control programs’ funding, scope, and messages (Ibraham et. al. 2004; Ibraham and Glantz 2006, 2007); promotion of preemption legislation to allow state laws to override more stringent local laws; and funding of local groups to fight against ordinances mandating clean indoor air (Siegel 2000). A later review by Pierce (2007), with the benefit of more recent data from states, rated the evidence as strong that state programs reduced tobacco use, including among youth (Sly et al. 2001a; Rigotti et al. 2002; Soldz et al. 2002; Niederdeppe et al. 2004; Pierce et al. 2005). Similarly, Bonnie and colleagues (2007) found “compelling” evidence that comprehensive state tobacco control programs can achieve substantial reductions in tobacco use. Such reductions, however, could well rely on the extent to which strategies are comprehensive and integrated. To be effective, they must also be consistent, and budget cuts in many states’ tobacco control programs have threatened that consistency. Thus, a report by the Institute of Medicine recommended that all states maintain funding for their tobacco control activities at the level suggested by CDC—about $15 to $20 per capita, depending on the state’s population, demography, and smoking rate (Bonnie et al. 2007). The President’s Cancer Panel’s report made the same recommendation in 2007 (NCI 2007).

A challenge for evaluating these state programs is that, by definition, they have multiple components, making it difficult to assess the relative contribution of each one. Still, several studies have attempted to quantify the relative amounts of effort expended by state tobacco control programs. For example, Schmitt and colleagues (2007) surveyed partners in state tobacco control—including the state health department, voluntary health agencies, and tobacco control coalitions—to assess the strength of tobacco control in various states by determining the proportion of partners working on interventions recommended by the Task Force on Community Preventive Services (2001). In addition, the community guide recommended a standardized approach, but this study found great variation between states in their overall levels of effort and in the relative degree of effort apportioned to media campaigns, tax increases, legislation on clean indoor air, supporting cessation assistance for smokers wanting to quit, and quitline services. Regardless, the strength of state-based tobacco control measures has not been the subject of studies to determine whether it is related to change in youth smoking.

Several studies, however, have focused on the overall level of tobacco control efforts within the states in an attempt to determine their impact on youth (and adult) smoking (Farrelly et al. 2003b, 2008; Tauras et al. 2005a). For example, Tauras and colleagues (2005a) related annual inflation-adjusted per capita expenditures on tobacco control to annual survey data for 8th-, 10th-, and 12th-grade students completing Monitoring the Future (MTF) surveys from 1991 to 2000. State expenditures were summed from (1) real per capita state-specific excise tax funding and other state-appropriated funds earmarked for tobacco control programs; (2) real per capita nongovernmental state-level expenditures on tobacco control; and (3) per capita tobacco control expenditures from ASSIST, IMPACT, SmokeLess States, and the National Tobacco Control Program (Tauras et al. 2005a). After adjusting for cigarette prices; the strength of laws on clean indoor air; laws on youth access; possession, use, and purchase (PUP) laws; and a range of individual characteristics associated with smoking, real per capita tobacco control expenditures had a statistically significant negative relationship with the prevalence of student smoking and the amount smoked by students. If states had spent the minimum amount of funding recommended by CDC, the relative prevalence of student smoking would have been between 3.3% and 13.5% lower than was observed over this period (Tauras et al. 2005a). Reduced prevalence was not observed in all states, however, as documented by Alesci and colleagues (2009) in Minnesota.

Fichtenberg and Glantz (2000) investigated the effects of the California Tobacco Control Program, implemented in 1989, on cigarette consumption and age-adjusted death rates from heart disease. Between 1989 and 1992, the rates of decline in per capita cigarette consumption and mortality from heart disease in California, relative to the rest of the United States, were significantly greater than the pre-1989 rates (p <0.001). These rates of decline were reduced significantly when the program was cut back beginning in 1992. The researchers estimated that the program was associated with 59,000 fewer deaths from heart disease between 1989 and 1997 than would have been expected if the earlier trend in heart disease mortality had continued.

Lightwood and colleagues (2008) modeled the dynamic relationships between per capita tobacco control expenditures, per capita cigarette consumption, and health care expenditures in California, showing $86 billion in reduced personal health care expenditures between 1989 and 2004 than would have been expected absent the state’s tobacco control program. Lightwood and Glantz (2011) used a similar approach to investigate the relationship between per capita tobacco control expenditures, cigarette consumption, and health care expenditures in Arizona, which employed a youth-focused tobacco control program. The state’s tobacco control expenditures were associated with reduced cigarette consumption and with reductions in health care expenditures amounting to about 10 times the cost of the program through 2004.

Previous reports have reviewed the programmatic components and outcomes of state tobacco control programs, especially states that adopted these programs during the 1990s (USDHHS 1994, 2000b; Siegel 2000; Wakefield and Chaloupka 2000; Bonnie et al. 2007; NCI 2008). The next section outlines the comprehensive tobacco control program in New York state that began in 2000, with information provided as well on separate programmatic efforts in New York City from 2002, and the positive effects of these efforts on smoking among youth. Taken together, results from statewide comprehensive tobacco control programs provide strong evidence that they reduce the prevalence of smoking by youth. To maintain their effectiveness, such programs need to be funded according to CDC recommendations in a sustained manner and include policy change, such as creation of smoke-free environments that reinforce a nonsmoking norm.

Case Study: New York Statewide Program

In 2000, New York state began implementing a comprehensive tobacco control program with funds from the Master Settlement Agreement and revenue from the state’s cigarette tax. The New York Tobacco Control Program (NYTCP) implements three key strategies: taking community action, producing and disseminating public health communications, and carrying out interventions to promote cessation. The program, whose components are supported by surveillance, evaluation, and statewide coordination, has attempted to reduce smoking among youth by working to change adult smoking norms and behaviors. From 2000 to 2005, funding for the program was one-half of what CDC recommended as a minimum (RTI International 2004), and in the first independent evaluation, which covered 2000–2003, NYTCP was found not to have expended all available funds in any year since the program had begun and thus did not have a fully implemented program (RTI International 2004). Bureaucratic procedures prevented NYTCP from fully implementing its strategic plan, especially a countermarketing campaign, and from establishing contracts with partners and contractors in a timely fashion (RTI International 2004). However, in 2002, New York increased its tobacco tax, and this produced reductions in smoking (RTI International 2004). Unfortunately, the program missed an opportunity to have a large impact on its intended outcomes by failing to implement media campaigns consistently with messages that elicited strong emotional responses among the target audiences and by not timing its media to coincide with the implementation of the Clean Indoor Air Act (2003) (RTI International 2004).

During 2004–2005, NYTCP began to broadcast more ads with high emotional impact, but there was a 6-month period when no media messages at all were broadcast. The program also established 19 centers focused on increasing the number of health care organizations with systems in place that supported smoking cessation, more actively promoted a fax-based quitline referral system to health care providers, distributed free starter kits of nicotine replacement therapy to eligible quitline callers, and implemented a new statewide initiative to combat the influence of tobacco advertising, sponsorships, and promotions. In 2004, the Fire Safety Standards for Cigarettes Act was implemented, requiring manufacturers to certify that all cigarettes offered for sale in New York met a specific standard for propensity to ignite. Cigarette-caused fires and deaths caused by cigarette fires both declined following implementation of the law (New York State Department of Health 2009). A 2005 evaluation by RTI International found that the program was having an impact on tobacco use and that rates of decline in New York had outpaced rates of decline in the rest of the country (RTI International 2005). However, tax evasion (i.e., purchasing cigarettes from low-tax or untaxed sources) reduced the effect of the increases in cigarette excise taxes by negatively affecting outcomes for smoking cessation (RTI International 2005).

The Clean Indoor Air Act (2003) noted above was associated with reductions in exposure to secondhand smoking among both youth and adults in New York state (RTI International 2005). During 2004–2005, the budget for NYTCP doubled from $44 million to $85 million (the latter around 90% of CDC’s minimum recommended level), and by 2007, the program had significantly expanded its media campaign efforts, promotion of quit-lines, and partnerships. In 2006, the prevalence of smoking among youth and adults declined faster in New York than in the United States as a whole, and the use of other tobacco products by youth and adults also declined (RTI International 2007). Between 2000 and 2006, smoking among middle school students in the state declined from 10.5% to 4.1% (RI = 61%); among high school students it declined from 27.1% to 16.3% (relative improvement [RI] = 40%) (RTI International 2007).

Alongside efforts at the state level, New York City began implementing its own five-point tobacco control program in 2002 with increased taxation to a level greater than the New York state tax, then continued in 2003 with the establishment of smoke-free workplaces (including restaurants and bars), education of the public and of health care providers, cessation services, and rigorous evaluation of its program. The latter included annual, cross-sectional, citywide telephone surveys using the same measures as CDC’s state-based Behavioral Risk Factor Surveillance System (CDC 2007c). Starting in 2006, New York City implemented an extensive, television-based, antitobacco media campaign using graphic images of the health effects of smoking, a campaign that was aired simultaneously with the New York state anti- tobacco media campaign. Declines in the prevalence of adult smoking were observed during 2002–2004 (Frieden et al. 2005; CDC 2007c), coinciding with the tax increase and smoke-free laws, and in 2006 among men and Hispanics, coinciding with the first year of the city’s media campaign (CDC 2007c). From 2003 to 2005, smoking among high school youth in New York City decreased substantially, from 14.8% to 11.2% (RI = 24%), while the rate nationally remained unchanged at approximately 23% (CDC 2007c).

Summary Regarding State-Level Programs

The total weight of evidence from the consistent findings of cross-sectional studies that have controlled for differences between exposed and unexposed populations, combined with high theoretical plausibility and coherence, is sufficient to infer a causal relationship between exposure to comprehensive state-level tobacco control programs and reduced prevalence of smoking among youth.

Federal Taxes

As part of the Balanced Budget Act of 1997, Congress passed a two-stage increase in the federal tax: the first stage increased the federal excise tax from $0.24 per pack to $0.34 per pack on January 1, 2000, and the second increased it from $0.34 to $0.39 per pack on January 1, 2002. These were the first changes to federal excise taxes on cigarettes since January 1, 1993 (Table 6.5). Moreover, the Balanced Budget Act of 1997 increased the excise tax rates on all other tobacco products in two stages and established an excise tax rate for roll-your-own tobacco (Table 6.6). On April 1, 2009, the federal excise tax on cigarettes was increased from $0.39 to $1.01 per pack (Table 6.5), and federal excise taxes on other tobacco products were also increased. Revenue generated from the 2009 tobacco excise tax hikes was used to fund an expansion of the Children’s Health Insurance Program.

Table 6.5

Federal cigarette excise taxes, selected dates, 1993–2009.

Table 6.6

Federal tax rates on other tobacco products, selected dates, 1993–2009.

State and Local Taxes

All 50 states and the District of Columbia currently impose an excise tax on cigarettes. As of August 1, 2011, the rates ranged from $0.17 per pack in Missouri to $4.35 per pack in New York (Table 6.7). State excise taxes have increased considerably in recent years. Since January 1, 2002, 47 states, the District of Columbia, and several U.S. territories have increased their cigarette excise taxes 105 times. Even Kentucky, North Carolina, and Tennessee—tobacco-producing states that have traditionally resisted raising tobacco taxes—have increased their tax rates on cigarettes. Moreover, hundreds of municipalities impose taxes on cigarettes, but the rates are generally relatively small when compared with state taxes. However, in recent years, several cities and counties have implemented large increases. For example, in 2002, New York City increased its tax on cigarettes from $0.08 per pack to $1.50 per pack. Similarly, both the city of Chicago and Cook County, Illinois (Cook County includes Chicago as well as many other jurisdictions), raised taxes on cigarettes. Combining federal, state, and local taxes, individuals purchasing cigarettes in New York City and Anchorage, Alaska, pay the highest cigarette excise taxes in the country at $5.85 and $4.20 per pack, respectively, as of October 7, 2011 (Campaign for Tobacco-Free Kids 2011b).

Table 6.7

State cigarette excise taxes (dollars per pack) and sales tax rate applied to cigarettes.

Another kind of tax, the general sales tax, is also quite common. In 2010, 45 states and the District of Columbia imposed general sales taxes on cigarettes (Table 6.7; Orzechowski and Walker 2010); as of November 1, 2010, these taxes added between $0.14 and $0.43 to the price of a pack of cigarettes. In addition, 49 states currently apply excise taxes on tobacco products other than cigarettes; these taxes are predominantly ad valorem. Finally, in most states the general sales tax is applied to other tobacco products as well as to cigarettes.

Cigarette Taxes and Cigarette Prices

Increases in taxes on cigarettes and other tobacco products increase their purchase price. Excise taxes are per unit taxes, but unless they are increased regularly, the inflation-adjusted value of the tax will fall over time. Given the importance of taxes in determining the price of cigarettes, increasing them only infrequently will likely result in declines in the inflation-adjusted price for cigarettes.

The years 1997–2002 witnessed some of the most dramatic increases in the inflation-adjusted retail price of cigarettes in the United States; during this period the inflation-adjusted price increased by 71.1% (Figure 6.2). This large increase was partly the result of the two federal tax increases mentioned earlier and the numerous increases in state excise taxes, and it also reflected the significant increases in the wholesale price of cigarettes. In fact, between 1998 and 2003, wholesale prices for cigarettes increased 122% (Capehart 2004), largely as a result of increased costs associated with expenses for individual state tobacco settlements and expenses related to the Master Settlement Agreement.

Figure 6.2

Cigarette prices and prevalence of smoking among youth, 1975–2011. Source: Cigarette prices from Orzechowski and Walker 2011; 30-day smoking prevalence data for students in grades 8, 10, and 12 from Monitoring the Future 2011, University of Michigan (more...)

Effects of Price on the Demand for Tobacco Products

One of the fundamental principles of economics is that as the real price of a good increases, consumption of that good falls (the downward slope of demand). Some researchers once believed that because of the addictive properties of nicotine, tobacco products might be an exception to this fundamental principle, but numerous econometric studies conducted over the past four decades, including several studies that explicitly modeled the addictive nature of cigarettes, have confirmed that an inverse relationship indeed exists between the prices of cigarettes and their consumption. Because increases in tobacco taxes have the potential to increase the real price of tobacco, increasing those taxes can be an effective policy lever for decreasing tobacco consumption.

Economists measure how responsive tobacco consumption is to changes in the real price of tobacco with a concept known as the “price elasticity of demand.” Formally, this is the percentage change in the amount of tobacco consumed that results from a 1% increase in the price of tobacco. For example, a price elasticity of −0.4 implies that a 10% increase in price will decrease consumption by 4%.

The two most recent comprehensive reviews of the literature on the impact of price on tobacco consumption include the International Agency for Research on Cancer (IARC) Handbooks of Cancer Prevention in Tobacco Control Volume 14 (IARC 2011) and a summary of key findings by Chaloupka and colleagues (2011). A few conclusions can be drawn from these reviews. First, increases in cigarette prices lead to substantial reductions in cigarette smoking. The consensus estimate from the two reviews is that a 10% increase in cigarette price will result in a 3–5% reduction in overall cigarettes consumed. Second, increases in cigarette prices will decrease not only the prevalence of smoking but also the average number of cigarettes smoked by smokers. Third, a majority of the previous research on cigarette consumption among youth suggests that both youth and young adults are more responsive than adults to changes in cigarette prices, with several studies finding youth and young adults to be two to three times as responsive to changes in price as adults. Finally, mixed results have been found in the relatively few studies that have examined the impact of cigarette prices on the initiation of smoking among adolescents.

Most of the research published since 2000 supports the conclusion of previous reviews that an inverse relationship exists between age and responsiveness to changes in cigarette prices. Drawing the conclusion that youth will be the most responsive to price, however, does not settle things in terms of calculating demand among younger people. For example, a central issue when estimating equations for cigarette demand among youth (or any other sector of the population) is how to account for anti-tobacco sentiment in different states. This is important because during a particular period it may be sentiment against tobacco that is driving both changes in cigarette smoking and changes in cigarette excise taxes. Not controlling for antitobacco sentiment may result in bias from omitting a variable, thereby producing a spurious negative relationship between price and smoking and resulting in estimated price elasticities biased away from zero. Several strategies have been suggested to account for antismoking sentiment in equations on youth smoking, including controlling for state tobacco control policies that affect primarily adults and controlling for whether the respondent resides in a tobacco-producing state. To the extent that the enactment of tobacco control policies that affect adults (and have little impact on smoking by adolescents, such as worksite restrictions on smoking) and residing in a tobacco-producing state can serve as proxies for anti-smoking sentiment, the inclusion of these variables in the regression model will mitigate some of the bias from omitted variables on the price estimates (Tauras et al. 2005a).

Another approach is to approximate the magnitude of antitobacco sentiment within states by using the attitudes of individuals toward smoking and beliefs about tobacco policies obtained from survey data. Still another approach is to eliminate state-level heterogeneity that is time invariant (such as types of housing) and unobserved through the use of state-level fixed effects. To the extent that sentiment toward tobacco within states is time invariant during the period under investigation, the inclusion of state-level fixed effects will eliminate the bias from an omitted variable on the price estimates. The use of state-level fixed effects relies on within-state variation in cigarette prices or taxes over time (as opposed to interstate differences in prices and taxes) to quantify the effect of price on consumption. In essence, the use of state-level fixed effects in conjunction with year-level fixed effects compares the effect of tax (or price) on smoking for individuals who reside in states in which taxes (or prices) changed with the effects of tax (or price) on smoking for individuals who reside in states that did not observe a change in tax (or price) in that year. For the state-level fixed-effects approach to be viable, researchers must use multiple years of state data; 1 year of cross-sectional data would result in perfect multicollinearity between the state-specific taxes (or prices) and the dichotomous state indicators. Moreover, even if multiple years of state data are used, there must be reasonable variation in tax (or price) over time within states to avoid collinearity issues with the tax (or price) variable.

Prevalence of smoking and average smoking among youth. As can be seen in Figure 6.2, an inverse relationship exists between prevalence rates for smoking among young people and the inflation-adjusted price of cigarettes in the United States. Most of the research conducted during the past decade that has controlled for a host of other factors thought likely to affect youth smoking, including antitobacco sentiment in the state, supports the conclusion of previous reviews that an inverse relationship exists between smoking among youth and cigarette prices.

For example, using 1 year of cross-sectional data collected in 1996 for the Study of Smoking and Tobacco Use Among Young People, Ross and Chaloupka (2003) examined the effect of cigarette prices on smoking among high school students in the United States. Although they controlled for both state-level laws on smoke-free air and youth-access laws, the authors assessed the use of several alternative measures of cigarette prices in their analysis, including average state prices and perceived prices among the students. In their preferred specifications, they estimated total price elasticities of demand of −0.67 and −1.02 when using average state prices and perceived prices among youth, respectively. The price elasticity estimates were confirmed in a subsequent analysis by Ross and Chaloupka (2004) that also explicitly controlled for compliance with youth-access laws. The estimates from these studies suggest that adolescents are considerably more responsive to price changes than are adults on the basis of the consensus estimate for the latter population (Chaloupka and Warner 2000).

Using the same cross-sectional data as Ross and Chaloupka (2003), Powell and colleagues (2005) reexamined the determinants of smoking prevalence among high school students, incorporating the importance of peer effects in their analyses. Specifically, Powell and colleagues allowed cigarette prices to have both a direct and an indirect effect, via a social multiplier effect (i.e., the influence of peer interactions), on the prevalence of smoking among youth. They estimated the price elasticity of smoking prevalence among youth to be −0.50, with the peer effect playing a significant role in the prevalence of smoking by high school students. Specifically, the aforementioned price elasticity comprised a direct-prevalence price elasticity of −0.32 and an indirect-prevalence price elasticity (measuring the social multiplier effect) of −0.18. These estimates are consistent with Ross and Chaloupka (2003) and suggest a rather large social multiplier effect with respect to price changes and participation among youth in smoking.

Katzman and associates (2007) extracted data from the 1995–2001 national Youth Risk Behavior Surveys (YRBSs) to estimate equations for cigarette demand among individuals in grades 9–12. In this study, the authors took into account the manner in which the adolescents acquired their cigarettes, distinguishing between those who primarily bought their own and those who primarily “borrowed” them. In their analyses, the researchers controlled for whether the adolescents resided in tobacco-producing states, for laws banning smoking in private worksites, and for PUP laws. Although they allowed changes in cigarette prices to affect both the probability of being a buyer and borrower and the quantity smoked, given group membership, the authors concluded that the total price elasticity of cigarette demand among adolescents ranged from −0.556 to −0.857. Again, these results imply that high school students respond more to price changes than do adults.

Earlier, Gruber and Zinman (2001) controlled for both state and year fixed effects in their analyses of smoking by youth. These researchers used three data sets from the 1990s in their analyses: MTF surveys of 8th-, 10th-, and 12th-grade students, YRBSs of 9th- to 12th-grade students, and the Vital Statistics Natality detail files of mothers during pregnancy. The authors concluded that price had a sizable and significant impact on smoking by high school seniors, with prevalence-price elasticities ranging from −0.38 in the Natality data to −1.5 in the YRBS data, with the most reliable estimate of −0.66 coming from the MTF data. Moreover, they concluded that younger adolescents are less responsive to price changes than are high school seniors.

Tauras and colleagues (2005b) investigated the impact of cigarette prices and tobacco control policies on propensity to smoke and intensity of smoking among youth and young adults during the late 1990s through the early 2000s, a period characterized by dramatic increases in cigarette prices and taxes. These investigators used the first five waves of data (1997–2001) from the National Longitudinal Survey of Youth 1997 cohort (NLSY97). Using a two-way fixed-effects technique that controls for unobserved individual-level heterogeneity and individual-invariant year-specific unobserved heterogeneity, they found a strong negative impact of cigarette prices and taxes on propensity to smoke and intensity of smoking among youth and young adults and estimated the total price elasticity of cigarette demand to be −0.827. These authors separately considered the impact of price and tax on the probability of smoking and on the average number of cigarettes smoked by smokers, estimating smoking prevalence-price elasticity of demand and the conditional price elasticity of demand to be −0.311 and −0.516, respectively. These estimates imply that a 10% increase in the real price of cigarettes would decrease the number of adolescent and young adult smokers by approximately 3.1% and reduce the average number of cigarettes smoked by adolescent and young adult smokers by 5.2%. The estimated total price elasticity was twice as large (in absolute value) as the consensus estimate for adults (0.4) and is consistent with the notion that an inverse relationship exists between age and the price elasticity of cigarette demand (USDHHS 2000b; Chaloupka and Warner 2000).

Sloan and Trogdon (2004) used Behavioral Risk Factor Surveillance data from the 1990s and early 2000s to estimate equations for smoking prevalence among young adults (18–20 years of age) and older adults (21 years of age and older). Using both state and year fixed effects, the authors concluded that propensity to smoke among young adults was the most responsive to cigarette prices, with an estimated smoking prevalence elasticity of demand of −0.27. In addition, the authors found evidence that the absolute value of the price elasticity of smoking participation declined monotonically with age until 65 years of age.

More recently, DeCicca and colleagues (2008a) developed a direct measure of state-specific antismoking sentiment with a factor analysis technique using data extracted from the Tobacco Use Supplements to the Current Population Surveys during the 1990s. Employing data from the 1992 and 2000 waves of the National Education Longitudinal Study (NELS), they found that price had a strong negative (and significant) impact on the prevalence of smoking and on average consumption for youth and young adult smokers. The estimated price elasticities of smoking prevalence and average consumption by smokers ranged from −0.59 to −0.76 and from −0.3 to −0.66, respectively. Moreover, price was found to have a strong negative influence on average smoking by youth smokers in the 2000 cross-section even after controlling for the new measure of antismoking sentiment. However, when smoking sentiment was included in equations for smoking prevalence, the price effects lost statistical significance. Using the 2000 wave of data, the authors tested models that employed the newly developed direct measure of antismoking sentiment and compared it with models using alternative approaches to dealing with such sentiment. The strong negative impact of price on average smoking was robust to all the methods of dealing with unobserved state-level sentiment toward tobacco. Moreover, in all the models except the model that included the new measure of sentiment, price was found to have a significant negative impact, reducing smoking prevalence among youth. Given the findings when the direct measure of antismoking sentiment was included in the models, DeCicca and colleagues questioned the adequacy of using proxies to control for antismoking sentiment. However, some caution should be used in interpreting models that include a direct measure of antismoking sentiment in that reverse causality is likely in this type of estimation strategy. That is, the amount of smoking within a state is likely to have an impact on the level of antismoking sentiment within that state, resulting in simultaneity bias.

Carpenter and Cook (2008) addressed the concerns of DeCicca and colleagues (2008a) in a recent paper that used national, state, and local YRBSs from 1991 to 2005; they tested three alternative methods of dealing with anti-smoking sentiment. First, they estimated a cross-sectional model that relied on intrastate variation in cigarette taxes to identify the impact of price on youth smoking. Second, they estimated a two-way fixed-effects model that controlled for area and year fixed effects. Finally, they employed the same direct measure of antismoking sentiment used by DeCicca and colleagues (2008a). Carpenter and Cook found consistent evidence of a significant negative effect of cigarette taxes on smoking prevalence in the cross-sectional and fixed-effects approaches. Moreover, using the new direct measure of antismoking sentiment, they found a strong negative effect of taxes on the prevalence of smoking among youth, alleviating the concerns raised by DeCicca and colleagues. Using the tax effects from the national and state samples, Carpenter and Cook estimated price elasticities for the prevalence of smoking among youth of −0.56 and −0.25, respectively.

Effects of cigarette prices on smoking transitions. Many researchers examining the influence of price on the prevalence of smoking among youth have assumed that much of the effect of price reflects its impact on the initiation of smoking, while the effects of price on young adults and adults are thought to be dominated by its effects on escalation of smoking and on cessation. Whether these judgments are true or not, several recent studies have attempted to directly quantify the impact of price on initiation among youth and the effects of price on escalation and cessation among young adults. Most of the recent studies have used longitudinal data on smoking behavior and other determinants of smoking over time.

Initiation of smoking. Tauras and colleagues (2001) were the first to examine the impact of price on initiation of smoking among youth by using longitudinal data, in this case from three cohorts of students enrolled in the 8th and 10th grades in 1991–1993 who were part of the longitudinal component of the MTF. The authors examined three alternative measures of the smoking process over time, including a transition from not smoking to smoking any amount; daily smoking, defined as smoking at least one to five cigarettes per day on average; and heavy daily smoking, defined as smoking at least one-half pack per day on average. After controlling for youth-access laws and regional fixed effects, the average price elasticity estimates for smoking cigarettes for (1) any smoking, (2) smoking at least one to five cigarettes per day on average, and (3) smoking at least one-half pack per day on average were −0.271, −0.811, and 0.955, respectively. These estimates imply that the process of smoking uptake among youth responds to changes in cigarette prices.

Cawley and associates (2004), who investigated the determinants of smoking initiation among youth by using more recent data from the first four waves (1997–2000) of NLSY97, looked at two alternative measures of smoking initiation. The first was a transition from nonsmoker to smoking any quantity of cigarettes (termed “less stringent initiation”), and the second (“more stringent initiation”) was the transition from nonsmoker to frequent smoker, as measured by having smoked on at least 15 of the past 30 days. Although they controlled for smoke-free air laws, youth-access laws, and residence in tobacco-producing states, the authors concluded that initiation of smoking among male adolescents was very responsive to changes in cigarette prices, with the average price elasticities estimated to be −0.86 for less stringent initiation and −1.49 for more stringent initiation. Initiation of smoking among female adolescents was not significantly related to cigarette prices but was very responsive to concerns about body weight.

A follow-up paper on the initiation of smoking among youth by Cawley and associates (2006) found results very similar to the earlier paper by Cawley and colleagues (2004) despite the use of a longitudinal data set that spanned a longer period: the data were from 1988 to 2000 and were taken from the children’s cohort of NLSY79. After controlling for smoke-free air laws and youth-access laws, researchers found cigarette prices to have a negative impact on the initiation of smoking in all the models that were estimated; however, the price coefficients differed significantly from zero in only the male equations. Specifically, the price elasticity of smoking initiation among males on the basis of any cigarettes consumed was estimated to be −1.20.

In a series of papers, DeCicca and colleagues (2000, 2008a,b) examined the influence of price and tax on the initiation of smoking among youth and young adults. In one of the papers, DeCicca and associates (2008a) used data from the 1988, 1990, 1992, and 2000 waves of NELS to examine the influence of cigarette prices on decisions about smoking among adolescents and young adults. The authors found price to have a strong and significant negative influence on initiation when state fixed effects were omitted from the model. However, when state fixed effects were included in the regression analyses, price failed to reach significance at conventional levels. These researchers concluded that unobserved state-level heterogeneity (possibly in the form of differential antismoking sentiment), not price, drives decisions to smoke among youth and young adults. In a different paper, DeCicca and colleagues (2008b) used data from the 1992 and 2000 waves of NELS to examine the influence of cigarette excise taxes on initiation of smoking among young adults (18–26 years of age). These authors used three identification strategies in their equations: First, they used intrastate variation in cigarette excise taxes to identify the impact of price on initiation of smoking. Second, they included the direct measure of antismoking sentiment developed by DeCicca and colleagues (2008a) in their equations for initiation. Finally, they used variation in cigarette taxes faced by young adults who moved across state lines between 1992 and 2000 versus young adults who remained in the same state in these two specific years.

In this paper (DeCicca et al. 2008b), cigarette taxes were found to have a significant negative impact on the initiation of smoking among young adults for only those who remained in the same state (the third identification strategy). The authors concluded that cigarette prices have little impact on the initiation of smoking, but these results should be viewed with caution. First, the study was conducted on a sample of individuals who initiated smoking later in life (they were nonsmokers in high school but smokers by a modal age of 26 years). Most adults who have ever smoked initiate smoking before the age range investigated by DeCicca and colleagues, and the decisions on initiation of an older cohort may be quite different from those of younger individuals. Second, as discussed above, antismoking sentiment may be an endogenous variable that is determined simultaneously with smoking. Third, in the models that relied solely on intrastate variation in taxes, the authors found only weak evidence of a negative effect of taxes on the prevalence of smoking (the price effect failed to reach 5% significance levels in a two-tailed test). Finally, in an earlier study, DeCicca and colleagues (2000) examined the determinants of initiation among individuals of different races and ethnicities with data extracted from the 1988–1992 NELS. After controlling for state and year fixed effects, they found price to have a dramatic negative impact on decisions to initiate smoking among Hispanics and African Americans, but price had no influence on decisions to initiate smoking by Whites. The authors estimated that a price increase of $1.50 would decrease rates of initiation among Hispanics and African Americans to approximately 1%. However, the authors cautioned that the prediction for African Americans was based on a statistically insignificant estimate of the price coefficient. Regardless, the results of this earlier study (DeCicca et al. 2000) indicate that conclusions about the relationship between initiation and cigarette taxes may well need to consider race or ethnicity rather than being simply drawn for the population as a whole.

Dinno and Glantz (2009) used the February 2002 panel of the Tobacco Use Supplement of the Current Population Survey (54,024 individuals representing the U.S. population aged 15–80 years) to study the independent association of cigarette prices and state or local strong clean indoor air laws with current smoker status and consumption in a multilevel framework, including interactions with educational attainment, household income, and race/ethnicity. They found nonlinear relationships between price and smoking status and per smoker consumption, with no effect at higher prices. Below $3.28 per pack (in 2002), the OR for smoking, given a 10-cent increase in price, was 0.95 (95% confidence interval [CI] 0.93–0.97); this relationship ended above that price. The association of cigarette price with smoker status did not change with educational attainment, household income, or race/ethnicity. There was no interaction between clean indoor air coverage and cigarette price. There was no interaction between cigarette price (or strong clean indoor air laws) and educational attainment, household income, or race/ethnicity. Price increases (and clean indoor air laws) appear to benefit all socioeconomic and racial/ethnic groups in the study equally in terms of reducing smoking participation and consumption.

Smoking cessation. A few studies have examined the impact of price on the decisions of adolescents and young adults to quit smoking. Tauras and Chaloupka (2001) were the first to model decisions on cessation with longitudinal data that tracked individuals’ smoking behavior over time. In particular, these researchers used the longitudinal component of MTF surveys and a semiparametric Cox regression to assess the probability that smokers would make a transition from smoking to nonsmoking. The authors concluded that the likelihood of making an attempt to quit among both men and women increases significantly as cigarette prices rise. Their estimated price elasticities for smoking cessation ranged between 0.27 and 0.92 for men and 0.34 to 0.71 for women, implying that a 10% increase in price raises the probability of making a cessation attempt by as much as 10% for men.

Expanding on the original study, Tauras (2004) used the longitudinal component of MTF surveys and employed a stratified Cox regression to model multiple attempts to quit among young adults. His findings confirmed a positive relationship between cigarette prices and smoking cessation, with a 10% increase in the price of cigarettes increasing successful cessation by young adults by an estimated 3.5%.

DeCicca and colleagues (2008b), in their study using data from the 1992 and 2000 waves of NELS, examined the influence of cigarette excise taxes on the decisions of young adults to quit smoking. When these authors used intrastate variation in cigarette excise taxes to identify the impact of taxes on smoking cessation, they found young adults to be very responsive to tax changes; indeed, the price elasticity of cessation was estimated to be 0.93. In a different specification, these investigators added the direct measure of antismoking sentiment developed by DeCicca and associates (2008a) and estimated the price elasticity of cessation to be 0.47, but here the parameter estimate for price was insignificant, indicating that the elasticity was substantially driven by variation in cigarette excise taxes and antismoking sentiment. Finally, as discussed within “Initiation of smoking” earlier in this section, they used variation in cigarette taxes faced by young adults who moved across state lines between 1992 and 2000 versus young adults who remained in the same state in 1992 and 2000. In this specification, cigarette taxes were found to have a positive impact on smoking cessation among young adults only for those who moved to a different state in those 2 years. The price elasticity of cessation among those who moved was relatively large (1.49), and the authors concluded that despite the lack of significance of price in this specification, most likely owing to the small sample (n = 321), price is likely to play a strong role in decisions to quit smoking among young adults.

Finally, using an experimental framework, Ross and colleagues (2005) examined the expected reaction to a future price increase among smokers in high school. The authors used cross-sectional data collected in 1996 for the Study of Smoking and Tobacco Use Among Young People, which contained information on individuals’ current smoking status and expected smoking behavior after a hypothetical change in cigarette price. After controlling for smoke-free air laws and youth access laws, the authors found increases in cigarette prices to have a strong positive impact on decisions by youth to quit smoking: the estimated price elasticity of cessation ranged from 0.895 to 0.930.

Other smoking transitions. In a study that looked at smoking transitions other than initiation or cessation, Tauras (2005) examined the impact of cigarette prices on such transitions among youth and young adults in the United States. This author examined the transition from nondaily to daily smoking and the transitions from light smoking intensity (defined as 1–5 cigarettes per day) and moderate smoking intensity (defined as 10 cigarettes per day on average) to higher intensities. Tauras (2005) employed baseline surveys from the 1976–1993 longitudinal component of MTF data along with follow-up surveys through 1995 in the analyses and controlled for antismoking sentiment with a variety of techniques. These included having separate indicators for whether the individual resided in a tobacco-producing state or resided in Utah, using U.S. Census Bureau division fixed effects to capture differences between these divisions in smoking sentiment, and estimating the smoking progression equations on a subsample of the respondents who did not reside in either a tobacco-producing state or in Utah during the time the survey was conducted. Cigarette prices were found to have a strong negative impact on all of the estimated smoking transitions. In particular, the estimated mean price elasticities of daily uptake, moderate uptake, and heavy uptake were −0.646, −0.576, and −0.412, respectively. These results indicate that a 10% increase in cigarette prices will decrease daily uptake, moderate uptake, and heavy uptake by an estimated 6.46%, 5.76%, and 4.12%, respectively. These findings clearly indicate that increases in cigarette prices will prevent many young adults from progressing into higher intensities of smoking.

Other tobacco products. Numerous studies on the economic determinants of demand for cigarettes among youth have been published during the past decade, but very few recent econometric studies have been published on the impact of taxes on other tobacco products.

In one such study, Tauras and colleagues (2007) used data extracted from the 1995–2001 national YRBSs to examine the impact of taxes on smokeless tobacco on use of this product among male high school students. The estimates developed clearly indicate that higher taxes on smokeless tobacco would significantly reduce the number of male students who use this product and the number of days they would use it. The estimated tax elasticities of the prevalence of smokeless tobacco ranged from −0.197 to −0.121, and the estimated tax elasticities of days using smokeless tobacco ranged from −0.085 to −0.044. The study also found that cigarette prices had a significant negative impact on both the prevalence of smokeless tobacco and the number of days that male high school students used smokeless tobacco. The estimated cross-price elasticity of the prevalence of smokeless tobacco was −0.715, and the cross-price elasticity of the number of days of use of smokeless tobacco was −0.413. These estimates indicate that a 10% increase in the price of cigarettes would decrease the prevalence of smokeless tobacco by an estimated 7% and would lower the number of days using smokeless tobacco by an estimated 4% among male high school students. Thus, the estimates indicate that smokeless tobacco products and cigarettes are economic complements in consumption for young males. These findings are particularly important in light of the fact that the cigarette companies have purchased smokeless tobacco companies and are now actively promoting dual use of cigarettes and smokeless tobacco with the same branding (e.g., Marlboro Snus and Camel Snus) (Mejia et al. 2010). (More data on the use of multiple tobacco products by young males can be found in Chapter 3.)

Finally, Ringel and colleagues (2005) used data from the 1999 and 2000 waves of the National Youth Tobacco Survey to estimate the impact of cigar prices on demand for cigars among adolescents in grades 6–12. After controlling for laws on smoke-free air and on youth access, the researchers found the price of cigars to be inversely related to the prevalence of cigar use among youth. Specifically, the price elasticity of the prevalence of cigar smoking among youth was estimated to be −0.34.

Tax Avoidance

A preponderance of the aforementioned studies on the effects of price on the demand for tobacco products among adolescents used individual-level survey data and state-level price data. Aside from the problem of intrastate variation in prices, using average prices within a state does not account for an individual’s opportunities to avoid taxes. For example, some individuals living near American Indian reservations or close to the border of a state with lower taxes on cigarettes will be able to pay less than the average price for cigarettes in their own state. Thus, when using individual-level data, this type of measurement error in the independent variable (i.e., price) will likely result in an underestimate of the true price elasticity of demand. There will be an underestimate of the response to price because some smokers will maintain their consumption after a tax increase by turning to cheaper (tax-avoided) cigarettes, making it look as though the tax increase had little or no impact on their consumption. Future studies on demand that account for a person’s opportunities for tax avoidance are warranted.

Summary Regarding Taxation and Pricing

A few general conclusions can be drawn from recent studies on the effects of taxes and prices on tobacco consumption among youth and young adults:

1. Most of the research over the past decade has concluded that increases in cigarette prices lead to reductions in the prevalence of smoking and its intensity among youth and young adults.
2. A majority of the existing research suggests that the effects of price on smoking prevalence involve both a decrease in initiation of smoking among youth and an increase in cessation among young adults.
3. Most of the recent research has concluded that adolescents and young adults are more responsive than adults to changes in cigarette prices.
4. Limited evidence suggests that higher cigarette prices will prevent young adults from progressing into higher intensities of smoking.
5. A few recent studies have found an inverse relationship among adolescents between product-specific tobacco taxes (or prices) and the propensity to use smokeless tobacco, the intensity of its use, and the prevalence of cigar smoking.
6. The magnitude of the impact of taxes (or prices) on the demand for cigarettes seems to depend on how the model controls for antismoking sentiment.

Future research that uses a large number of waves of longitudinal data on adolescents and young adults during a period of significant changes in tobacco taxes and prices should be helpful in obtaining the most precise estimates for the impact of price on the intensity, prevalence, initiation, and cessation of smoking, smokeless tobacco use, and on other tobacco use transitions.

Effects of Clean Indoor Air Laws on Smoking by Youth

The first evidence that laws and policies on clean indoor air could reduce adolescent smoking came from cross-sectional studies. Liang and colleagues (2003), who reviewed studies on the effects of tobacco control policies, including the effects of clean indoor air laws on youth smoking rates, found that restrictive laws and workplace policies were an effective tool for reducing smoking among youth. They also reviewed the evidence concerning smoking policies in households and found several studies showing a strong inverse relationship between the presence of such policies and the chances of trying smoking as well as experimentation (Liang et al. 2003). Since that review, McMullen and colleagues (2005) used data from both the YRBS and the National Survey on Drug Use and Health (NSDUH) to examine the relationship between the prevalence of youth smoking at the state level and the “state clean indoor air law score” as reported by the State Cancer Legislative Database. For both sets of data, the strength of laws on clean indoor air was inversely related to the prevalence of smoking among youth.

Using the longitudinal data on young adults from MTF, Tauras (2004) found that stronger restrictions on smoking in private worksites and public places increased the probability of smoking cessation among young adults. Later, Siegel and colleagues (2005, 2008) published two papers from a longitudinal study of adolescents (n = 3,834) in Massachusetts; comparing baseline figures and the 2-year follow-up surveys they reported that youth who lived in a town with a strong smoke-free ordinance for local restaurants were significantly less likely to progress to regular smoking than were youth in towns where such restrictions were either weak or of medium strength (Siegel et al. 2005). These researchers reported that at the 4-year follow-up, youth in the group with a strong ordinance on smoking in restaurants had reduced odds for both overall progress to established smoking and transition from experimentation to regular smoking (Siegel et al. 2008). More recently, Klein and colleagues (2009) reported a much smaller effect in a report from the Minnesota Adolescent Community Cohort Study, which included 4,233 Minnesota youth who were 11–16 years of age at baseline. Participants were interviewed every 6 months from 2000 to 2006. The authors found a 6% lower likelihood of monthly smoking and a 13% lower likelihood of weekly smoking if youth lived in areas with a strong policy on clean indoor air. The study also found a strong association between a household smoking ban and reduction in the likelihood of smoking by youth.

Prohibitions by colleges on smoking may have characteristics of worksite, school, and household smoking bans because they can affect one or more aspects of the students’ lives. As discussed in “School-Based Programs to Prevent Smoking” later in this chapter, the amount of research on the role of school policy in preventing youth smoking is surprisingly small and, similarly, there are few published reports on college policies regarding students’ smoking behavior. Using data from the 1999 survey of the Harvard School of Public Health College Alcohol Study, Wechsler and associates (2001) found that current smoking prevalence was lower among students living in smoke-free campus residences than among those living in unrestricted residences. In addition, smokers who started smoking in college reported smoking fewer cigarettes if they lived in smoke-free residences. Czart and associates (2001), who used 1997 survey data from the Harvard School of Public Health College Alcohol Study, found that complete smoke-free policies lowered the intensity of smoking and strong enforcement decreased participation in smoking, but both findings were of only marginal significance statistically.

Effects of Home Smoking Policies on Youth Smoking

Restrictions in the home may be a powerful tool to reduce smoking by youth. In a report on 1996 survey data for high school students across the United States, from the Study of Smoking and Tobacco Use Among Young People, Wakefield and colleagues (2000) found that a 100% smoke-free policy for everyone in the home was associated with a reduced likelihood that youth would advance from their current smoking stage for every stage from susceptible to established smoker. In addition, in a study of youth 15–17 years of age from the Current Population Surveys of 1992–1993 and 1995–1996, those who lived in smoke-free households were only 74% as likely to be smokers as those who lived in households with unrestricted smoking (Farkas et al. 2000), independent of the smoking status of individuals in the household. Furthermore, youth already smoking were more likely to quit. However, partial restrictions showed no effect on smoking. Later, analysis of the 1998–1999 Current Population Survey produced the same results and extended them to young adults living with parents (Clark et al. 2006). In both adolescents and young adults, complete bans on smoking were associated with never having been a regular smoker, not being a current smoker, and having quit smoking. The adjusted odds of being a current smoker (using never smoking as the referent) were about 50% lower in households with strict smoking rules than in those without rules on smoking.

At this point, more information is needed on how home smoking policies vary by sociodemographic characteristics. Some information is available, however, on American Indian youth. In a recent study of a convenience sample of 336 urban youth who were American Indian, 43% reported living in a household that banned smoking for everyone (Forster et al. 2008). Lifetime nonsmokers to date were significantly more likely to live in a completely restrictive household than those who had ever smoked, and bans on smoking were associated with level of smoking among these youth. There is also a positive effect of smokefree legislation that applies to workplaces and public places on the prevalence of voluntary home smokefree policies (Cheng et al. 2011; Mills et al. 2011; Hovell et al. 2011). Cheng and colleagues (2011) found that living in a county fully covered by a 100% clean indoor air law in workplaces, restaurants, or bars is associated with an increased likelihood of having a voluntary 100% smoke-free home policy both for people living with smokers (OR = 7.76; 95% CI, 5.27–11.43) and not living with smokers (OR = 4.12; 95% CI, 3.28–5.16).

Effects of Home Smoking Policies on Exposure of Youth to Secondhand Smoke

In addition to reducing youth smoking, bans on smoking in the household have the potential to reduce youth’s exposure to secondhand smoke. Youth who reside in multiunit housing are particularly at risk of exposure, even if they do not live with a smoker, as smoke can travel through walls, air ducts, windows, and ventilation systems (Wilson et al. 2011). An analysis of NHANES data from 2001 to 2006 found that young people living in an apartment in which no one smoked had significantly higher cotinine levels (a biological measure of smoke exposure) than those living in a detached home in which no one smoked (Wilson et al. 2011). In 2009, the U.S. Department of Housing and Urban Development issued a memorandum strongly encouraging public housing authorities to implement nonsmoking policies in some or all of their public housing units (Winickoff et al. 2010).

Summary Regarding Policies on Clean Indoor Air

Laws and policies on clean indoor air support multilevel efforts that can be effective in reducing exposure to secondhand smoke and potentially youth smoking. This argues for a comprehensive approach to reducing smoking among youth.

Possible Strategies

The three possible strategies for encouraging compliance to age-of-sale laws are taking appropriate steps in the retail environment, educating merchants, and actively enforcing the laws. Taking appropriate steps in the retail environment includes requiring that tobacco products be located behind the counter, posting signage informing customers that it is illegal for minors to purchase tobacco, and banning vending machines and self-service sales (Forster and Wolfson 1998). Taking these steps reduces the likelihood that youth will obtain cigarettes even if the store’s clerk is inattentive. Education of merchants is an attempt to inform retailers of the laws; it is assumed that educated retailers would be less likely to sell cigarettes to minors (Rigotti 1999). “Self-enforcement” and education of merchants are not enough, however, to prevent minors from purchasing tobacco from commercial establishments (Feighery et al. 1991; DiFranza and Brown 1992; Landrine et al. 1996; Gemson et al. 1998; Altman et al. 1999; Rigotti 1999); penalties are needed. Penalties for selling tobacco to minors include revoking store licenses and fining merchants and clerks who sell to youth, both of which are usually done after a random compliance check.

The 1992 Synar Amendment can be seen as a supply-side strategy for limiting and controlling the supply of cigarettes. Its premise is that if youth-access policies are well enforced, they will lead to less youth smoking. This sentiment is echoed by CDC, which includes control of youth access in its Best Practices for Comprehensive Tobacco Control Programs Guide (CDC 2007b) as well as in Healthy People 2010, which specifies policy goals on youth access (USDHHS 2000a).

Another strategy is penalizing youth for possessing, using, or purchasing tobacco. The underlying theory behind PUP strategies is that these consequences will reduce demand among youth for tobacco. One potential downside of this approach, as discussed in Chapter 5 (Wakefield and Giovino 2003), is that punitive legal measures directed at youth may distract from focusing on the role of the tobacco industry or retailers.

Criteria for Evidence of Prevention

Of the two key criteria for evidence that strategies to limit access are effective, the first is that the supply of cigarettes available to youth is actually reduced; the second is that strategies affect the prevalence, intensity, initiation, and/or cessation of youth smoking. Rigorous evaluation of available strategies presents challenges, but such evaluations are necessary to determine whether these strategies meet the goals of prevention.

Effects of Interventions to Limit Youth Access: Prior Reviews

Several English-language systematic analyses have been conducted of interventions to limit the access of youth to tobacco, with the key paper a Cochrane review conducted by Stead and Lancaster (2005). These authors concluded that policies to limit youth access and enforcement of these policies can improve the compliance of retailers, and the prevalence of smoking will be affected if the commercial supply is sufficiently restricted through these means. The authors also concluded that enforcement had a greater effect than did the education of merchants, but as with all interventions, they noted that sustained compliance is a challenge.

The second review in this area was a meta-analysis of policy on youth access based on data from nine studies; the authors found no effect on smoking at any threshold of access control (Fichtenberg and Glantz 2002; Ling et al. 2002), although there have been some concerns about the methods used in this meta-analysis (DiFranza 2002; Jason et al. 2003). Levy and Friend (2002) also examined the empirical studies of policies on youth access and concluded that a comprehensive approach that includes active enforcement of laws, community mobilization, and training of merchants is the most promising way to control access. Even so, these reviewers found that past studies showed the effects of these policies on the prevalence of smoking among youth to be inconclusive (Levy and Friend 2002; Task Force on Community Preventive Services 2005). More recently, a 2009 study by DiFranza and colleagues examined the association between the compliance of merchants with youth access laws and current daily smoking while controlling for cigarette prices, restaurant smoking bans, media campaigns, and demographic variables. The study showed that the odds of daily smoking were reduced by 2% for each 1% increase in merchant compliance (DiFranza et al. 2009).

Wakefield and Giovino (2003) reviewed the empirical evidence for PUP laws and their enforcement and concluded that these laws were associated with reduced smoking among youth only for those young people who were unlikely to initiate smoking. Notably, both the existence of PUP laws and their enforcement have become extremely common in the United States.

Effects of Interventions to Limit Youth Access: Current Evidence Base

Critics of strategies that promote policies to limit youth access have argued that even if the commercial supply of cigarettes could be successfully reduced, the social supply of cigarettes would grow to fill the gap (Ling et al. 2002). Indeed, in communities where cigarettes have become relatively difficult for underage youth to purchase from commercial sources, adolescent smokers have increasingly relied on social sources (Forster et al. 1998; Altman et al. 1999; DiFranza and Coleman 2001). But this trend from relying on commercial sources to using social sources appears to be associated with less consumption of cigarettes among youth (DiFranza et al. 2009). Another finding of interest is that among adolescents who smoke, the heavier smokers are less likely to use social sources as their only source of cigarettes, although they are more likely to be a social source for other adolescents (Wolfson et al. 1997; Harrison et al. 2000; Forster et al. 2003). Finally, Widome and colleagues (2007) have demonstrated a trend in which a greater proportion of youth become heavy smokers in communities where more adolescent smokers exclusively use commercial sources, thus reinforcing the need for strong policies to restrict commercial access for young people.

The impact of the Synar Amendment appears to have varied by sociodemographic characteristics, and there has been some research on how restrictions on access differentially affect youth from various demographic groups. In a Florida study, there was evidence that retailers in Hispanic neighborhoods in Miami (although not in the other cities studied) were more likely to sell tobacco to minors (Asumda and Jordan 2009). In contrast, stores in neighborhoods with a high percentage of Black residents were not more likely to sell tobacco to minors (Asumda and Jordan 2009). For individual youth, race/ethnicity may be associated with their chances of successfully purchasing tobacco. For example, a recently published study examined compliance checks in California from 1999 to 2003 and found more sales to Black and Asian underage decoys than to their White counterparts (Landrine et al. 2008). Earlier, Chaloupka and Pacula (1999) found that although restrictions on youth access had no impact on smoking rates among White youth, they were associated with a lower prevalence of smoking among Black youth.

Discussion Regarding Youth Access

Data on whether interventions to restrict access can lead to a reduction in the number of retailers selling tobacco to minors are mixed, although the Community Preventive Task Force concluded that community mobilization combined with additional interventions, such as stronger local laws directed at retailers, active enforcement of retailer sales laws, and retailer education with reinforcement are recommended (Task Force on Community Preventive Services 2005). A recent comprehensive review also supports the efficacy of enforced reductions in the sales of cigarettes to minors (DiFranza 2011).

Statistical Issues in Tobacco Advertising

Many empirical studies on the effects of cigarette advertising can be found in the academic literature that have used a variety of methodologies. Some have relied on small samples of data to address a specific question; for example, some small surveys have measured smoking behavior, exposure to advertising, receptivity or attitudes to tobacco advertising, or brand awareness during a baseline period and again during follow-up. Other studies have relied on large-scale data sets developed for public use, while some studies have used aggregated data at the national or international level. Advertising studies also can be divided into those using self-reported data on advertising, such as exposure or impact, and those containing market-level data. Studies have also addressed the impact of bans on advertising.

Regardless of the type of study, each raises statistical issues that researchers must consider carefully. These issues include the problems of dealing with measurement error, of properly adjusting for the effects of time by using a weighted average of current and past-period advertising, and the needs to specify an estimation equation, address the problem of uncontrolled individual heterogeneity, and deal with endogeneity, or reverse causality.

Measurement error is common in studies that rely on expenditures for advertising or on measurements of exposure to ads. The data here are either self-reported or are market-level data purchased from a firm specializing in advertising data. Measurement error will generally result in bias toward a finding of no effect of advertising. Self-reported advertising data contain measurement error because individuals who are considering use of a product, or who are current users, will generally be more aware of advertising for that product than other individuals will be. In the case of cigarettes, for example, individuals who are considering smoking, or who smoke, will usually report awareness of more tobacco ads than will other individuals. Thus, controlling for awareness levels will likely result in underestimating the impact of those most likely to smoke. Market-level data can be interpreted erroneously because everyone in the market is assigned the same value for assumed exposure to advertising, but not everyone in the market will actually have the same exposure. Thus, market-level data should preferably be evaluated by using a probability measure of exposure, since those most exposed are likely to be more strongly influenced by advertising and using a probability measure increases variability and the ability to detect a relationship between advertising and behavior.

The second issue, dealing with the effects of time, can also be challenging. For example, advertising in the current (present) period will have a lingering although smaller effect in the next period, but how much the effect declines over time remains unclear. In the case of cigarettes, research such as that by Boyd and Seldon (1990), found that the effects of advertising depreciate fully within a year. And yet, advertising has lingering effects as noted, and knowledge of these effects is the basis for a widely used technique known as “pulsing.” A pulse is a burst of advertising, in a specific market, that lasts for only a short time; after a period of time with no or minimal advertising, the market is exposed to another pulse. These pulses create variability in the amount of actual advertising from one period to the next, but because of lingering effects, a stock of advertising is created. To account for this stock of advertising, researchers should measure advertising as a weighted average of current and past-period advertising.

The third issue, specification of the estimation equation, is important because advertising has a diminishing marginal product. In brief, increments in advertising result in ever smaller increments in sales. That there is diminishing outcomes is a well-established tenet of economics and advertising; the important implication of this principle is that the functional relationship between advertising and sales should be specified as nonlinear.

The fourth issue, addressing the problem of not controlling for individual heterogeneity, can also be a vexing one. The ideal method for estimating the effects of advertising on smoking is a randomized trial, but ethical considerations prohibit experimentation with cigarette advertising. Without random selection, all individual characteristics that might influence smoking behavior must be controlled to ensure that the variation in advertising is the factor that causes the variation in smoking. This is not easily accomplished in standard regression models, however, and thus bias induced by heterogeneity is common. Fortunately, data sets from panels of individuals can be used to control for time-invariant individual characteristics, such as gender or race/ethnicity, and reduce this type of bias.

The fifth and final issue—endogeneity, known also as reverse causality—also creates bias; this is a problem in any study of advertising. Here, for example, rather than advertising driving revenues, revenues drive amounts of advertising; this may be particularly true for mature products. Thus, if smoking decreases, there may be less sales revenue to use for advertising, and advertising may decrease. The problem in this case is that lower advertising might be misunderstood as responsible for lower sales. This may also be a problem in studying the effects of advertising bans: a high level of smoking can lead to public pressure to legislate such bans and, for example, give the impression that such high levels are associated with bans. Endogeneity can be addressed with a well-identified structural model or a natural experiment that examines already existing data.

A partial ban on advertising may not reduce the total level of advertising but should reduce the effectiveness of the remaining media that are not banned (a ban on one or more media will generally result in substitution into the remaining media). This apparently counterintuitive phenomenon should be seen because each medium is subject to diminishing marginal product; the increased use of the nonbanned media will result in a lower average product for these media. Firms may try to compensate with more advertising, or they might increase the use of other marketing techniques, such as promotional allowances to retailers. From the research perspective, because bans on particular media result in cessation of advertising in those media, there are fewer issues overall with measurement error, diminishing marginal product, or lingering effects. Heterogeneity and reverse causality could still create problems for the investigator, however, depending on the nature of the data. In addition, researchers should be aware that there must be comprehensive bans in place to avoid substitution into other media. Finally, the researcher must control for other marketing activities. Data from a single country could reduce some problems caused by reverse causality in studies on bans, and longitudinal or aggregate data could reduce problems with heterogeneity.

Effects of Advertising Bans: A Prior Review

Lancaster and Lancaster (2003) reviewed 21 studies of advertising bans and found that 10 of these reported significant negative coefficients indicating that the bans on advertising were associated with decreased smoking or consumption. Of the 199 reported coefficients, 29.1% were negative and significant, 5.5% were positive and significant, and 65.3% were insignificant. Some of the coefficients may have been nonsignificant because the bans were limited to a few media, allowing substitution into other media. None of these studies accounted for the possibility of endogeneity (reverse causality).

Effects of Advertising Bans: Current Evidence Base

In a study of bans on advertising, Saffer and Chaloupka (2000) used an international data set from 22 countries that covered 1970 to 1992. Bans were considered weak if they were nonexistent or only one or two kinds of media, such as television and radio advertising, were banned; limited, if three or four media were banned; or comprehensive, if five, six, or seven media were banned. In an analysis limited to 1984 to 1992, they found that limited bans were not effective but that comprehensive bans were effective. Their results suggest that moving from a limited to a comprehensive set of bans can have a compounding effect, which is consistent with the theory that limited bans allow substitution of other media. The problem of endogeneity was not considered.

Iwasaki and colleagues (2006) found that advertising restrictions required by the 1998 Master Settlement Agreement decreased consumption of cigarettes in the United States. These restrictions included a ban on outdoor advertising and restrictions on youth-targeted advertising; in addition, the agreement provided funds for counter-advertising. Earlier, Chung and colleagues (2002) reported that the agreement’s restrictions on advertising to youth were easily avoided; they also noted that counter-advertising took a few years to initiate.

Iwasaki and associates (2006) constructed a time series data set from 1955 to 2002 for the United States in which the regression equations included interactions of advertising expenditures with dichotomous variables for four progressively more restrictive periods for advertising during the timeframe in question. These periods were 1955 to 1967, 1968 to 1971, 1971 to 1997, and 1998 to 2002. A break was seen between 1967 and 1968 because the first Surgeon General’s report on smoking was in 1964 (and related news on smoking causing lung cancer began in the 1950s and had substantial impact up to 1967). The 1971 break reflects the elimination of broadcast advertising, and the 1998 break reflects the Master Settlement Agreement. The coefficients from the first three periods were insignificant, perhaps because the United States did not have enough restrictions in place to prevent the substitution of television and radio advertising with other types of advertising and marketing activities. The coefficient from the final time period was both negative and significant, indicating that the agreement had reduced smoking. Thus, these data suggest that the most restrictive rules, including the ban on outdoor advertising, reduced smoking. Endogeneity was a problem, however, because, over time, sentiment against tobacco was increasing, and this sentiment would affect cigarette use as well as the passage of the Master Settlement Agreement. On the other hand, problems with controlling for heterogeneity of the population were reduced because aggregate data were used, but it should also be noted that there was no control for other forms of marketing. Data from the U.S. Federal Trade Commission (FTC 2011) indicated that other marketing expenditures increased dramatically after the Master Settlement Agreement.

Discussion Regarding Advertising Bans

According to FTC (2011), in 2008 more than $190 million was spent on cigarette advertising in the United States, but this represented only 1.9% of total monies spent for cigarette promotion (see Chapter 5, Table 5.3). Regardless, this amount of advertising constitutes a public health problem if it increases overall smoking or encourages youth to begin to smoke. The tobacco industry and associated researchers (e.g., Heckman et al., 2008) contend that there is no definitive research showing that advertising increases smoking, but this has now been countered with longitudinal research (see Chapter 5). Also, from a cost-benefit point of view, the potential public health advantage (such as in long-term worker productivity) of banning cigarette advertising is far greater than the private costs to tobacco companies and advertisers, and so a ban on such advertising makes sense from an economic perspective. As concluded in NCI Monograph 19: “The studies of tobacco advertising bans in various countries show that comprehensive bans reduce tobacco consumption. Noncomprehensive restrictions generally induce an increase in expenditures for advertising in ‘nonbanned’ media and for other marketing activities, which offset the effect of the partial ban so that any net change in consumption is minimal or undetectable” (NCI 2008, p. 281).

Effects on Youth of Current U.S. Health Warnings

A number of research studies indicate that the current U.S. text warnings have relatively little impact on youth smokers. Indeed, several studies of U.S. warnings suggest they are rarely noticed and suffer from low recall among youth, as illustrated by two studies that used eye-tracking equipment to examine attention paid to U.S. tobacco ads and recall of these warnings (Fischer et al. 1989; Krugman et al. 1994). The first study compared two existing U.S. health warnings in magazine ads with two “new” warnings and found that the “new” warnings were associated with more reading and attracted attention more quickly. However, relatively few respondents could accurately recall the wording or the general concepts of any of the four warnings. In the second study, adolescents were asked to view five tobacco ads that included a health warning. The average viewing time of the health warning was only 8% of the total time spent viewing the ads, and participants subsequently demonstrated a low recall of the warnings.

Brubaker and Mitby (1990), who conducted one of the few studies to examine U.S. text-based warnings on smokeless tobacco products, found results similar to those for Krugman and colleagues (1994): less than one-half of the persons (43%) exposed to the warnings recalled seeing them, and only one-third of those who recalled seeing them remembered the content of the message. Overall, the warning labels had no significant effect on whether adolescents would use the product.

More recent research suggests that although most youth report the U.S. cigarette health warnings to be “believable” (Cecil et al. 1996), few find them to be informative or relevant (Crawford et al. 2002). For example, in a series of focus groups conducted in 2001 among adolescents, most considered the warnings to be personally irrelevant and described the warnings as “vague,” “stale,” and “worn-out” (Crawford et al. 2002, p. 16).

In one of the few longitudinal studies of health warnings among youth, Robinson and Killen (1997) examined the association between adolescents’ knowledge of U.S. cigarette warning labels and subsequent smoking behavior by surveying 1,747 youth. At baseline, adolescent smokers were more familiar with the health warnings than were nonsmokers. When cigarette packages serve as the medium for health warnings, however, one would expect that consumption levels at baseline would be associated with knowledge of the warning labels.

Effects on Youth of the Size and Position of Health Warnings

Several studies demonstrate that an increase in the size of text warnings increases their impact (Environics Research Group 1999). For example, in studies in which Canadian youth were asked to rate the effectiveness of different health warnings, the largest warnings were most likely to be rated as effective (Environics Research Group 1999; Les Études de Marché Créatec 2008). These findings are consistent with research conducted among adults showing that smokers were more likely to recall larger warnings and often equated the size of the warning with the magnitude of the risk (Health Education Journal 1985; AGB Spectrum Research 1987; Cragg Ross and Dawson 1990; Centre for Behavioural Research in Cancer 1992; Action on Smoking and Health 1998; Strahan et al. 2002). Warnings that appear on the “front” or principal display area of packages are also likely to have greater impact. In one study, Rootman and Flay (1995) compared the effectiveness of U.S. and Canadian health warnings in 1995 among a youth sample. At the time, Canadian packages carried one of eight black-and-white text warnings on the front and back of packages, covering 25% of the display area on the package. Students were shown a package for 1 minute and then asked to recall everything they could about it. The most notable finding was that 83% of Canadian students mentioned the health warning on Canadian packs, a larger percentage than those who could recall the brand name. In contrast, health warnings on U.S. packs were recalled by only 6% of the U.S. students. A survey conducted with youth in The Netherlands also suggests that more prominent text warnings on the principal display area have relatively greater impact (Teeboom 2002). In addition, recent experimental research in Canada found that increasing the size of warnings from 50% to 75%, 90%, or 100% of the principal display area enhanced their impact among youth smokers and “vulnerable” youth nonsmokers (Les Études de Marché Créatec 2008).

Effects on Youth of Pictorial Health Warnings

In 2000, Canada became the first country in the world to introduce pictorial warnings on tobacco packages (Figure 6.6 provides an example). A series of focus groups and population-based surveys conducted among Canadian youth around this time suggested that large pictorial warnings were considerably more legible and credible and more likely to be noticed than were text warnings (Environics Research Group 1999, 2000; Nilsson 1999). A survey taken in 1999 in Canada, the year before the large pictorial health warnings on cigarette packages were introduced, found that youth in that country—both smokers and non-smokers—supported the use of pictorial health warnings on cigarette packages (Environics Research Group 1999). When shown health messages with and without pictures, 80% of youth reported that the message with the picture was more noticeable. Three years later, in a national survey of more than 19,000 Canadian youth between 11 and 15 years of age, the majority found the pictorial health warnings on cigarette packages to be believable and agreed that health-warning messages should be on cigarette packages (Chaiton et al. 2002). In a large national study conducted following implementation of the pictorial warnings, about 95% of Canadian youth reported that pictorial health warnings communicated the risks of smoking better than text-only warning labels (Bonnie et al. 2007, p. 294). Overall, the believability of the health warnings and the degree of endorsement were either similar to, or above, levels measured in 1994, 6 years before introduction of the large pictorial warnings. This research demonstrates that introducing large pictorial warnings does not decrease support or credibility among youth for messages about the health risks of cigarettes.

Figure 6.6

Pictorial warning on cigarette package in Canada. Source: Tobacco Labelling Resource Centre 2011a.

In addition, a series of 12 cross-sectional surveys were conducted with Canadian youth before and after the implementation of the large pictorial health warnings in 2000; these surveys showed significant increases in the frequency with which youth noticed, read, and thought about the health warnings after the pictorial messages were introduced. The most recent survey, in 2006, found that 86% of youth smokers reported the messages as effective in informing them about the health effects of smoking; 70% said that the messages had been effective in getting them to try to quit smoking; 66% reported that the messages had increased their desire to quit; and 56% said they smoked less around others as a result of the messages (Environics Research Group 2006).

Evidence from focus groups in Australia supports these findings. For example, although many Australian youth expressed a general lack of concern about the effects of smoking, they nevertheless reported being influenced by the health warnings (Elliott & Shanahan Research 2002). In particular, descriptive or emotive messages in the pictorial warnings had considerable impact, particularly those images portraying the external effects of smoking. Follow-up studies among Australian youth came to similar conclusions on the effectiveness of pictorial warnings (Elliott & Shanahan Research 2003; BRC Marketing & Social Research 2004). Evaluations have been conducted on pictorial warnings implemented in Australia (see Figure 6.7 for an example). A school-based study in western Australia found that students were more likely to report they had read, attended to, thought about, or talked about health warnings after the pictorial warnings were implemented in 2006 (White et al. 2008a). In addition, experimental and established smokers were more likely to have thought about quitting and forgoing cigarettes, and intention to smoke was lower among those students who had talked about the warning labels and had forgone cigarettes (White et al. 2008a).

Figure 6.7

Pictorial warning on cigarette package in Australia. Source: Tobacco Labelling Resource Centre 2011c.

In addition to increasing perceptions of risk, pictorial health warnings have been found to undermine the brand appeal of packages (Clemenger BBDO 2004; Thrasher et al. 2007; Les Études de Marché Créatec 2008; Stark et al. 2008). In addition, more than 80% of Canadian youth in a 2006 survey indicated that large pictorial health warning messages made smoking seem less attractive (Environics Research Group 2006). Overall, findings on the effectiveness of large pictorial warnings among youth are consistent with research conducted among adults, which has found associations between larger pictorial warnings and greater health knowledge, increased motivation to quit smoking, and greater attempts to quit (Hill 1988; Tandemar Research 1996; Borland and Hill 1997a,b; Liefeld 1999; Environics Research Group 2001; Portillo and Antoñanzas 2002; Willemsen et al. 2002; Cavalcante 2003; Hammond et al. 2003, 2004, 2006, 2007; Koval et al. 2005; Willemsen 2005; O’Hegarty et al. 2006; Ramesh 2006; UK Department of Health 2006; Quit Victoria 2007; Thrasher et al. 2007).

Evidence from numerous studies of adult populations indicates that health warnings are more likely to be effective if they elicit stong emotions, are larger and more visible (CDC 2011a). Although fewer studies examining the effects of pictorial warning labels have been conducted with youth than with adults, findings across countries show that the pictorial warnings better communicate the risks of smoking to young people than do text-only warnings.

Effects of “New” and Rotating Health Messages

Health warnings that are new or periodically updated are likely to have greater impact among youth than will “older” warnings, even in the absence of changes in the size and position of the messages. Indeed, youth commonly report on the stale or ineffective nature of “old” warnings that remain unchanged for more than several years (Environics Research Group 1999, 2000; Crawford et al. 2002). According to research findings from adults, health warnings have their greatest impact shortly after implementation and decline in effectiveness over time (Borland and Hill 1997b; Hammond et al. 2007). This is consistent with the basic principles of advertising and health communications, which suggest that the salience of a communication is greatest upon initial exposure and erodes thereafter (Bornstein 1989; Blair 2000).

Discussion Regarding Warning Labels

Research conducted to date demonstrates that the effectiveness of health warnings among youth increases with their size and placement as well as with the use of pictures. Small text-only warnings located in nonprominent locations, such as the side of the package in the United States, have relatively little impact. Furthermore, pictorial warnings that cover a significant proportion (e.g., 50% or more) of the package are associated with increases in health knowledge and motivation to quit smoking. Pictorial warnings also have the potential to reduce sociodemographic disparities in health knowledge and tobacco use among youth (CRÉATEC + Market Studies 2003). The existing text warnings in the United States require a college reading level, but pictorial warnings are easily understood by those with low literacy skills, including young children, youth with lower levels of education, and youth who may be literate but not in the language of the text warnings, such as young people in some immigrant families (Malouff et al. 1992).

The significant evidence base that has been developed since several countries implemented pictorial warning labels on cigarette packs clearly demonstrates that pictorial warning labels are an important component of tobacco control (Fong et al. 2009). The 2007 Institute of Medicine Report, “Ending the tobacco problem: A blueprint for the nation,” concluded that based on the available evidence, large, graphic warnings like those implemented in Canada, Brazil, and Thailand “…would promote greater public understanding of the risks of using tobacco products or reduce consumption” (Bonnie et al. 2007, p. 16). The report also recommended that FDA require pictorial and text-based warnings that cover 50% of the cigarette package (Bonnie et al. 2007). This requirement is currently subject to legal challenges. In June 2011, FDA announced it will require pictorial graphic warning labels on all packs of cigarettes sold in the United States (Figure 6.8) (USFDA 2011). One of nine pictures paired with one of nine text-based messages will be displayed on the top 50% of the front and back panels of each pack of cigarettes (USFDA 2011). These FDA requirements and related provisions of the Family Smoking Prevention and Tobacco Act are currently under judicial review. The evidence base is expected to increase in parallel with regulatory developments in tobacco labeling, which are rapidly progressing in response to the issuance of international standards through the World Health Organization’s (WHO’s) Framework Convention on Tobacco Control (FCTC) (WHO 2003). FCTC recommends warnings that cover 50% of the front and back panels but only requires warnings to cover 30%. Also, the treaty permits the use of pictures or graphics. More than 30 countries have either implemented or have committed to implementing large pictorial warnings that meet the recommended guidelines of FCTC.

Figure 6.8

Proposed pictorial warnings on cigarette packages in the United States. Source: U.S. Food and Drug Administration 2011.

Types of Parenting and Family-Focused Approaches

Several investigators have tried to classify the different types of family-focused approaches for prevention, but researchers in this field have not agreed on the definitions of the classifications. The review by the Center for Substance Abuse Prevention (CSAP 1998) of family-focused approaches defined eight approaches, but at that time, only four had sufficient validity to be considered evidence based: (1) cognitive-behavioral training for parents; (2) family skills training, including training of the parents, skills training of the children, and family practice time together; (3) family therapy (structural, functional, or behavioral); and (4) in-home family support or case management programs. Since the 1998 CSAP review, the very-low-cost strategy of involving parents with their children in homework assignments on the prevention of substance abuse has also shown promise as a cost-effective approach (Williams et al., 1995). In addition, cost-effective video, CD (compact disc), interactive DVD (digital video disc), and online versions of family programs have shown positive results (Gordon 2000; Schinke et al. 2000, 2004).

Theories Underlying the Strategy

The general logic of the family-based approach is that if parents learn and practice skills to become more effective at parenting and improve the parent-child relationship, learn how to be more effective in disciplining their children, and become better monitors, their children will have better developmental outcomes of all types, including those that relate to tobacco use. In addition, attention to the mechanisms of change has been identified as a crucial component for advancing theory in family-based treatment for substance use and ultimately for developing more effective prevention programs. For most family interventions, the underlying psychological theories are cognitive-behavioral, social learning, and/or family systems theory (Liddle et al. 2002). A key concept of many evidence-based programs is to reduce particular parent-child interactions that give rise to antisocial behavior and tobacco use, a process well documented by Patterson (1986) at the Oregon Social Learning Center. In general, the family systems approach uses reframing and cognitive restructuring methods to foster behavior change. Evidence-based interventions involve the whole family (rather than just the parents or the children) in processes that involve interaction, the building of skills, or behavior change rather than providing didactic educational lessons. These programs stress the importance of the engagement process and reducing barriers to attendance at program sessions, often through building relationships; extending personal invitations; providing meals, child care, and transportation; and sometimes by paying families for their time. Most begin with sessions designed to improve positive feelings in the family through positive reframing or through skills exercises that stress family strengths. Structured methods for communication and disciplinary techniques are also practiced once positive family feelings are established.

Systematic Reviews

For this Surgeon General’s report, two systematic Cochrane reviews of family-focused interventions in preventing tobacco use were identified (Thomas et al. 2003, 2007; Petrie et al. 2007); these reviews suggest that such interventions are effective.

In one of the Cochrane reviews, Thomas and associates (2007) assessed 20 RCTs of family-based interventions that included children or youth (5–18 years of age) plus family members and met their criteria for inclusion. Fourteen of the RCTs were conducted in the United States, two in Norway, and one each in Australia, Finland, India, and the United Kingdom. The studies reported on smoking status of children from baseline to at least 6 months from the start of the intervention; all 20 included at least a 1-year follow-up: 8 with 1 year; 1 with 20 months; 2 studies with 24 months; 6 with 36 months; and 1 each with 7, 15, 27, and 29 years, respectively, of follow-up data.

Of the 20 RCTs identified, 6 were classified by the Cochrane criteria for assessment of bias or quality of study (selection, performance, attrition, and detection) as Category 1, or of high quality with minimal risk of biased results (Bauman et al. 2001; Spoth et al. 2001, 2002; Storr et al. 2002; Curry et al. 2003; Schinke et al. 2004), 9 as Category 2, or medium risk of bias (Biglan et al. 1987; Ary et al. 1990; Nutbeam et al. 1993; Cullen and Cullen 1996; Elder et al. 1996; Jøsendal et al. 1998; Stevens et al. 2002; Wu et al. 2003; Jackson and Dickinson 2006), and 5 as Category 3, or high risk of bias. Studies in the last group were not included in the analysis.

Overall, the review by Thomas and colleagues (2007) found statistically significant results in 50% (three of six) of the Category 1 studies (Spoth et al. 2001; Storr et al. 2002; Schinke et al. 2004). In contrast, only 33% (three of nine) of the Category 2 studies (Jøsendal et al. 1998; Wu et al. 2003; Jackson and Dickinson 2006) found significant results for the interventions. The reviewers suggested potentially positive results for the family interventions when they were implemented with high-quality training and fidelity as was found in category 1 studies. In their review, the authors did not examine differential effectiveness by the major types of family interventions; many of the family interventions tested were minimal-contact, homework-based programs.

The second Cochrane review (Petrie et al. 2007) assessed 46 articles on 20 studies that met the authors’ review criteria for an RCT or were carried out as controlled before-and-after studies that focused on improving parenting skills. Although not mentioned as a criterion for inclusion, all the studies had at least 1 year of follow-up, with up to 6 years of follow-up for two studies. Of the 20 studies, 13 measured tobacco outcomes, of which 9 (representing 11 programs) resulted in significant positive reductions in tobacco use. Seven of the studies focused on the prevention of substance use in general (not tobacco specifically). Four of the nine programs found effective in this review were previously identified as effective by Thomas and colleagues (2003) in the protocol for the first Cochrane review of family-based smoking prevention programs.

The relative improvement (RI) rates calculated for the 11 effective programs are reported in the program descriptions below, and other details of the studies are shown in Table 6.8 (the programs that were not effective are discussed in Petrie et al. 2007). RI is the posttest difference between intervention (I) and control (C) groups minus the pretest difference between groups, divided by the control group posttest level: [(I% or mean –C% or mean) post – (I% or mean – C% or mean)pre]/ C% or mean post, expressed as a percentage. RI is similar to an effect size (ES) when the latter is defined as the posttest difference between groups divided by the pooled standard deviation (SD) at posttest: (I% or mean – C% or mean) post/SD (I + C) post.

Table 6.8

Descriptions and effect sizes (expressed as relative improvements) of parenting and family interventions for preventing use of tobacco among adolescents.

The Family-School Partnership intervention incorporated the Parents on Your Side program, which in this intervention included nine workshops for parents. In a 3-day workshop, teachers were trained to communicate better to parents. The parents also completed weekly parent-child homework assignments. Results for this family intervention (RI = 20.3%, relative risk [RR] = 0.69) were positive but almost the same as for a comparison group that received the classroom-based Good Behavior Game (RI = 22.2%, RR = 0.57) instead of the Parents on Your Side intervention (Storr et al. 2002).

Smoke-free Kids consisted of newsletters, six mailed tips on parenting, and gifts for participation. This program reduced initiation of smoking after 3 years to 11.9% of students, compared with 19.3% of minimal-contact controls, who had received five tobacco fact sheets (RI = 38%, RR = 8.4%, OR for not starting = 2.16) (Jackson and Dickinson 2003, 2006).

BE smokeFREE, a Norwegian school-based program reported by Jøsendal and associates (1998 (2005), found significant differences in number of cigarettes smoked per week (10 vs. 17 for controls, OR = 0.48, RI = 41%) at 6-month follow-up. A 3-year follow-up for 10th graders found reductions in lifetime (ever) use (31.5% vs. 41.7%, RI = 24%), weekly smoking (4.1% vs. 6.2% for controls, RI = 47%), and daily smoking (15.5% vs. 23.0%, RI = 28%) for the three-component intervention (a classroom program, involvement of parents, and teacher courses). The family component was not tested separately, but when the parenting intervention was dropped from the total intervention, the percentage of never smokers dropped 4.4 percentage points, from 41.7% to 37.3%, and RI dropped 6 percentage points, from 24% to 18%. However, the percentage of weekly smokers increased from 4.1% to 5.4%, and RI dropped 18 percentage points, from 47% to 29%. The percentage of daily smokers 3 years later among a group of 10th graders was 23% for controls, 15.5% for the model program, and 21.1% for the intervention minus the parenting component, for an RI of 13%, compared with an RI of 28% for the full intervention. Hence, the contribution of the parenting component appeared to be greater in the longer term for preventing daily smoking.

The Iowa Strengthening Families Program (ISFP) for youth aged 10–14 years (Kumpfer et al. 1996) is a seven-session family skills training program that was implemented during evenings for all sixth-grade students in randomly selected schools in an RCT in Iowa. Each session of ISFP involves parents and students in 1 hour of separate classes on parenting skills and on skills training for children followed by 1 hour of family practice time. The 4-year follow-up ITT analysis found a 32.6% rate of smoking initiation in the group receiving ISFP compared with 50% for the minimal-contact control group (RI = 34.8%) (Spoth et al. 2001).

Preparing for the Drug Free Years (PFDY), now called Guiding Good Choices, was tested in the same RCT as ISFP. PFDY is a five-session intervention that involves parents in five 1-hour parenting classes; the sixth-grade students had one session on peer-resistance skills. The same 4-year follow-up ITT analysis found a 44% rate of smoking initiation for the experimental group compared with 50% for the control group (RI = 12.5%). This comparative research suggests that ISFP was about three times as effective as PFDY in reducing the rate of initiation of cigarette use (Spoth et al. 2001).

Another study conducted in the Midwest (this time involving seventh graders) combined the seven-session ISFP and LifeSkills Training (LST), a school-based, youth-centered intervention that does not involve parents. Those who went through the combined program had a 12.1% rate for new use of cigarettes, compared with 16.7% for controls (RI = 27.5% reduction) and 13.9% for LST only (RI = 16.8% reduction) (Spoth et al. 2002). When the ISFP family program was replicated in a multicommunity RCT and combined with one of three youth-only programs (LST, All Stars, or Project ALERT), the percentage of new tobacco users dropped from 32% to 17% after 18 months (RI = 47%, Cohen’s d = 0.29) (Spoth et al. 2007).

Project STAR (Students Taught Awareness and Resistance), also known as the Midwestern Prevention Project (MPP) (Pentz et al. 1989b,c; Johnson et al. 1990), included homework assignments for the parents of youth who were engaged in a comprehensive prevention program that also featured a classroom curriculum and a mass media campaign. The family component (the homework assignments) was not tested separately. The 1-year RI was 41%, and the 3-year RI was 18% for reduction in tobacco use during a 30-day period.

Family Matters consisted of four brochures on parenting that were mailed to recruited parents and followed up by a call from a health educator. This minimal intervention was found by Bauman and colleagues (2001) to reduce the percentage of smokers from 55% to 48% at one-year follow-up, but at baseline the percentage of smokers was lower in the experimental group (24.5%) than in the control group (27.5%). The RI was 7.3%; the OR of 1.30 in the original analysis became 1.27 when Petrie and colleagues (2007) corrected for the design effect, producing a nonsignificant difference from the control group (p = 0.0595). This corrected result may explain why the Cochrane review conducted by Thomas and colleagues (2007) concluded that this program was not effective.

Another program reviewed by Thomas and colleagues (2007) was the intervention reported by Schinke and associates (2004), the CD-ROM (compact disc read-only memory) Youth and Parent program, a CD-ROM version of a youth and parenting program that was tested in an RCT. This program produced an RI of 31%.

Wu and associates (2003) tested Focus on Kids (FOK), an eight-session, small-group intervention providing training in social skills that is led by two older peers, both with and without a program called Informed Parents and Children Together (ImPACT), a 2-hour video on parenting skills plus two home visits by an instructor for practice sessions. The authors compared these two conditions with a third condition of both interventions plus two booster sessions. The study involved 817 high-risk Black youth 12–16 years of age in low-income communities in Baltimore, Maryland. At the 6-month follow-up, youth in families assigned to FOK plus ImPACT reported significantly lower rates of cigarette use than youth in families assigned to FOK only (RI = 20%). The booster sessions delivered at 7 and 10 months made no significant difference.

Elsewhere, a review of the D.A.R.E. (Drug Abuse Resistance Education) and D.A.R.E. Plus (Play and Learn Under Supervision) programs found significant reductions in smoking, alcohol use, and violence among boys but not among girls or for the total population when the D.A.R.E. Plus components (parent, peer, and extracurricular activities) were added to the junior high D.A.R.E. program (Perry et al. 2003).

A combined examination of the programs included in the two Cochrane reviews shows that the most effective family-focused program for preventing tobacco use by adolescents was a selective prevention program, Smoke-free Kids, that was targeted to high-risk children of smokers (RI = 38%). This program (Jackson and Dickinson 2003, 2006) was unusual in that it was a minimal-contact intervention. The next-best single intervention in terms of ES was ISFP, a purely family-focused intervention developed by Kumpfer and colleagues (1996). This program is of significantly greater dosage than others in its category because it involves the whole family in seven sessions of 2 hours of skills training (RI = 34.8%) (Spoth et al. 2001). Lowering the dosage and not including the children in the sessions seems to result in a lower ES. Generally, with higher-risk families, a higher dosage (or more time) is needed to produce effective behavioral changes. The CD-ROM version of Schinke and associates’ (2004) youth and parenting program also had a large RI (31%).

As a group, the family-involved programs targeting high-risk youth and their families had the largest ES, with a mean RI of 32.7%. The limited research reported here suggests that targeted selective prevention programs are likely to produce the largest ES in reducing tobacco use among adolescents. In the same RCT that included ISFP, the five-session PFDY parenting program (youth came for one session) had an RI of only 12.5% (Spoth et al. 2001). Adding the LST program to ISFP resulted in a lower RI (27.5%) than for the ISFP alone, but participants were in seventh grade rather than sixth, making a direct comparison difficult (Spoth et al. 2002).

Another group, not specifically targeting parents, the multicomponent school-based programs that consisted primarily of training in youth life skills with the added involvement of parents in homework assignments, averaged an RI of 28.3%. The largest ES in this category was for the Minnesota Heart Health Program and Class of 1989 Study, which indicated positive immediate and intermediate effects on smoking levels for youth smoking, with a large RI of 39.4% (Vartiainen et al. 1986; Perry et al. 1992).

The least effective type of family intervention, with an RI of 7.3%, was the universal application of a minimal intervention relying on mailings to parents followed up by calls from a health educator: the Family Matters program (Bauman et al. 2001). The base rates of smoking may have been too low in this universal sample, however, for a minimal-contact intervention like this one to produce significant changes compared with the no-treatment controls.

Ineffective Adolescent Tobacco Programs That Included Family Components

According to the Cochrane reviews, ineffective programs included (1) Kickbutts (Tang et al. 1997); (2) the South Carolina COPE program (Forman et al. 1990); (3) Biglan and colleagues’ (1987) training program in refusal skills; (4) Steering Clear clinical trial (Curry et al. 2003); (5) the Busselton Health Study (Cullen and Cullen 1996); and (6) one test of PFDY (Hawkins et al. 1999).

The ineffective programs were generally shorter (two to five sessions) than the effective ones, which were usually five to eight sessions plus two boosters or at least seven sessions. In their systematic review, Thomas and colleagues (2003) concluded that the ineffective programs had fewer training requirements for program delivery staff than the more effective programs. Also, fidelity to the implementation was higher in the more effective programs. Thus, it is not enough to have an effective structured intervention with good content; it is also necessary to develop an effective training and quality control system for the program’s dissemination.

Thomas and colleagues’ (2003) analysis looked at other questions in comparing the research, including whether family interventions were as effective as school interventions. From their analysis, family interventions seem to be as effective as school interventions: five of the RCTs that tested both a family intervention and a school intervention showed significant positive effects for both. However, these authors found that none of the six studies that compared a family program plus a school program with a school program alone showed significant positive effects from adding the family program to the school program. In the one trial that tested a more general risk-reduction intervention (not specific to tobacco) but measured tobacco outcomes, the combined parent-and-adolescent intervention resulted in less smoking than the youth-only intervention.

Limitations of the Studies Reviewed

Most of the RCT studies reviewed here relied on self-reported measures by the students or their parents. A second limitation of those programs that included family components is that for many of the school-based or community-based interventions, the family programming was merely a minimal dose added to a more substantial program for youth. The one exception was the totally family-focused ISFP (Spoth et al. 2001).

A major weakness of the studies reviewed is that few of them tested the family or parenting intervention separately from the youth intervention to determine the unique contribution of the family or parenting intervention. Another limitation is that of all the family studies, only ISFP was replicated by someone other than the program developer. This was, however, a semi-independent replication; Karol L. Kumpfer, the original developer, was a coprincipal investigator on the grant that conducted the replication.

In addition, of all the interventions, only one was either culturally adapted or gender specific, the BE smoke-FREE program designed specifically for Norwegian youth and parents (Jøsendal et al. 1998, 2005). Although cultural adaptations of evidence-based interventions are likely to improve outcomes to get to “deep” structure (Resnicow et al. 1999), even those with “surface” structure cultural adaptations have shown few improvements in outcomes over the generic versions (Botvin et al. 1995b). One exception is that the involvement of parents and families in terms of attendance and retention was found to improve results by about 40% if the evidence-based intervention was culturally adapted (based on a comparison across five studies of the 14-session ISFP in culturally adapted form with the multiethnic version) (Kumpfer et al. 2002).

It is not known whether the programs reviewed in this section are equally effective for girls and boys because analyses for subgroups are rarely reported (Kumpfer et al. 2008). One exception was the D.A.R.E. Plus program, which resulted in significant reductions in smoking, alcohol use, and violence for boys but not for girls or the total population (Perry et al. 2003).

Discussion and Recommendations

Several different programs that involve parenting or families may be effective in reducing tobacco use. Most of the tested programs were interventions added to school-based programs in which the parents were sent materials or homework assignments to complete with their children. The most effective programs in terms of ES or percentage of reduction in smokers had one or more of the following characteristics:

1. Targeted high-risk adolescents with selective interventions;
2. Combined skills training among youth with homework assignments for parents on parenting;
3. Focused specifically on the family, with skills training for the family that included more sessions or included time with the families to learn together;
4. Provided longer periods to train the staff in the intervention methods;
5. Conducted checks on the fidelity of implementation or on quality;
6. Used interventions for skills training among families that were based on behavior change theory; and
7. Stressed active parental involvement and parenting skills and developed social competencies and self-regulation among youth.

Thus, it appears that some well-executed family interventions with sufficient dosage may help to prevent smoking among adolescents, but the reports in the literature on RCTs of family interventions that were less well executed have had mostly limited results. There may, therefore, be a need for more well-designed and properly executed RCTs in the area of family-focused tobacco prevention, particularly those testing the family component separately from the youth component. Studies of disseminating effective family interventions are also needed.

Rates of Delivery of Tobacco Prevention Services to Youth in Health Care Settings

In 2006, 84.2% of adolescents (aged 10–17 years) and 72.0% of young adults (aged 18–24 years) had visited a doctor’s office in the past year, not including hospitalizations, emergency room visits, or surgeries (Mulye et al. 2009). Female young adults were much more likely to have visited a doctor than male young adults (84.7% vs. 59.3%), but the difference by gender was less pronounced among adolescents (85.5% for females vs. 83.0% for males). White adolescents and young adults were more likely to have seen a doctor than were their Black counterparts, with Hispanics having lower rates than Blacks. The rates varied greatly by insurance status, with 87.2% of insured adolescents and 80.1% of insured young adults visiting a doctor compared with 54.9% of uninsured adolescents and 46.2% of uninsured young adults.

National guidelines support providing tobacco prevention services to youth and promote brief tobacco-screening questionnaires under the presumption that they are effective (Benuck et al. 2001). Still, delivery rates of these services have been insufficient among physicians in private practices, community-based practices, and managed care settings. Studies have shown that less than 60% of adolescents were provided guidance about smoking (Marks at al. 1990), and only 1% of office visits by adolescents included advice about smoking cessation (Igra and Millstein 1993). In a large survey of family practitioners, pediatricians, internists, and obstetrician-gynecologists, Ewing and colleagues (1999) found that less than one-half of these physicians routinely inquired about smoking. In a survey of pediatricians and family physicians, Klein and colleagues (2001b) found that these physicians asked more than 90% of their adolescent patients about smoking but discussed tobacco-related health risks with only about 75%. Inquiries about parental smoking, peer smoking, and use of smokeless tobacco were less common, ranging from 32% to 54% of patients. Although more than 80% of the physicians promoted abstinence from smoking among their nonsmoking patients and assessed motivation to quit among those who smoked, less than one-half followed up with cessation materials or referrals.

Halpern-Felsher and colleagues (2000) reported that 77% of adolescents in a managed care setting were screened for tobacco use. Among those patients who reported tobacco use, however, only three-quarters were screened further about the amount they smoked and only 84% were counseled on the risks of smoking. This same study also found that no more than 43% of the patients’ parents were told about the need to monitor youths for risk behaviors, including substance use. Galuska and colleagues (2002) found that less than one-half of pediatricians counseled adolescents about tobacco use by others in the home.

In general, the provision of tobacco prevention services remain low, even for particularly vulnerable adolescent populations, such as low-income, asthmatic, or chronically ill youth (Fairbrother et al. 2005; Rand et al. 2005; Tercyak et al. 2007). In addition, physicians are more likely to ask older adolescents about their smoking status than to deliver preventive advice to preadolescents who might benefit more from prevention messages because they are less likely to have started smoking (Makni et al. 2002). A study of almost 1,000 pediatricians randomly selected from a national sample in 1998–1999 found that only 29% always counseled younger children (6–12 years of age) about tobacco use, but 69% always counseled their 13- to 18-year-old patients about using tobacco (Galuska et al. 2002).

The rates at which adolescents are screened for tobacco use and other risk behaviors vary by physicians’ characteristics, including age, gender, year of graduation, practice setting, and subspecialty. Two studies found that rates of counseling for tobacco use and other preventive services were greater among female providers and among pediatricians who were able to spend relatively more time with their patients (Klein et al. 2001a,b; Galuska et al. 2002). More recently, Perry and Kenney (2007) found that pediatricians were more likely than physician subspecialists, as well as nonphysician providers, to advise patients that smoking in the home is harmful. Earlier, Ewing and colleagues (1999) found that physicians under the age of 50 years were more likely than older physicians to provide tobacco-related clinical preventive services. Still earlier, Blum and colleagues (1996) found that the provision of clinical services was lowest in nonadolescent-focused practice settings, independent of patient age or gender. Halpern-Felsher and colleagues (2000) found greater provision of clinical preventive services among female physicians, recent graduates from medical school, and physicians with a greater number of older adolescent patients. Similarly, Klein and colleagues (2001b) found that rates of delivery for tobacco-related preventive services varied by provider characteristics, with women being more likely than men to ask about smoking behaviors and smoking by parents and peers and men more likely to ask about the use of smokeless tobacco.

Relatively little is known about how to improve the rates at which services to prevent tobacco use are delivered to children and adolescents. Ozer and colleagues (2005) showed that training physicians could increase the rates at which health care providers screen and counsel youth about risk behaviors, including tobacco use. Providers’ self-efficacy to provide preventive services was found to be linked to the actual delivery of services (Ozer et al. 2005) and can be enhanced through trainings (Buckelew et al. 2008). Studies from the literature on adult patients indicate that the use of paper-based, computer-generated, or computerized reminders in patient charts is particularly effective at increasing the delivery rates of smoking cessation services (Dexheimer et al. 2008). In addition, the literature on adolescents suggests that providing charting tools can improve the rates at which services are delivered to younger patients (Ozer et al. 2001, 2005). Electronic health care record systems that require documentation of service delivery may also increase the rates at which preventive services are delivered. More research is needed to determine the extent to which the implementation of provider training, electronic systems, or other charting tools increases the delivery of tobacco-related preventive services to children and youth.

Hymowitz and colleagues (2004, 2007, 2008) focused on increasing the provision of tobacco-related preventive services (and self-efficacy to deliver them) by adding a training program in tobacco to residencies in pediatrics. As Hymowitz and associates (2004) pointed out, few pediatric residents receive any training in addressing the use of tobacco by patients or their parents, and many pediatricians question the efficacy of counseling. In a randomized study, pediatric residents were assigned to either standard training or a new training, Solutions for Smoking (SOS), which used a combination of CD-ROM and Web site programming to provide information on interviewing skills, the use of the “5 A’s,” and behavioral and pharmacologic methods for reducing tobacco use. The researchers found that from baseline to 4 years after the program, residents in the SOS training were more likely to inquire about secondhand smoke in the home and to provide specific advice and materials to help parents stop smoking; those in the SOS program also reported feeling more efficacious for addressing tobacco issues. These studies do not, however, directly address the effects of the intervention on whether the pediatricians trained in the program had any effect on tobacco use by children or adolescents.

Rather than focus directly on health care providers, Christakis and colleagues (2006) demonstrated the effectiveness of using a Web-based intervention to encourage the parents of younger children (0–11 years of age) to discuss health topics, including tobacco use, with their pediatricians. The authors found that parents were more likely to discuss topics with their pediatrician during a well-child visit if the parents participated in an interactive Web site, thus, in turn, changing the physician’s behavior in a way that produced greater levels of preventive services. Future studies are needed to test the effectiveness of this intervention with parents of older children and adolescents.

Research Support for Tobacco Prevention Strategies Involving Health Care Providers

Unfortunately, there has been little research on whether increased rates of preventive screening, counseling, and education by health care providers actually lower the rates of tobacco use among youth. Nor have studies determined the mechanisms by which these interventions might be most effective (Christakis et al. 2003; Krowchuk 2005). In fact, there is no research at all demonstrating the effectiveness of the “5 A’s” in preventing tobacco use among children and adolescents, although the Prokhorov and colleagues (2010) study did involve pediatricians and demonstrated some success. It also remains unclear how many providers adhere to antitobacco policies in their offices or how effective such policies are in changing smoking norms or preventing smoking initiation among youth.

The first RCT to test the effectiveness of a program for preventing tobacco use among youth involved training orthodontists to deliver eight “prescriptions” to their patients over time (Hovell et al. 1996, 2001). These prescriptions included providing advice on eight tobacco-related topics, such as “tobacco and sports” and “nicotine and tobacco addiction.” Rates of smoking initiation did not differ between the prevention and control groups over a 2-year follow-up, but higher rates of delivering the prescriptions by orthodontists predicted lower rates of smoking initiation. Later, Hollis and colleagues (2005) conducted an RCT to examine the long-term effects of brief counseling, physician advice, and computer-based intervention on prevention of smoking and on cessation among adolescents 14–17 years of age. Compared with controls participating in a dietary intervention, adolescents in the tobacco intervention were significantly less likely to report smoking 1 and 2 years after that intervention than those in the control group. These effects were even stronger for those reporting smoking at baseline. Among that group, 24% indicated at the 2-year follow-up that they had quit. However, this brief intervention had less of an effect on preventing the onset of smoking.

Three other studies that used RCTs of interventions to prevent smoking in medical settings found that preventive services had little or no effect on smoking among youth. In one, screening for smoking behavior and providing pictures of tooth discoloration at annual dental visits did not reduce the prevalence of smoking (Kentala et al. 1999). In a second, mailing age-appropriate information about the advantages of remaining a nonsmoker to primary care patients at 3-month intervals produced a significant but still small difference in smoking rates between youth in the intervention and control groups (Fidler and Lambert 2001). In the third study, Curry and colleagues (2003) implemented and evaluated an RCT of a family-based smoking prevention program in a managed care setting. The intervention included a smoking prevention kit mailed to parents, newsletters for the parents, follow-up telephone calls by health educators, materials for the children, and information placed in medical records and charts as reminders to the physician to deliver prevention messages. Although the intervention had small but significant effects on increasing parent-child communication about tobacco, no differences between the intervention and control groups were found in rates of susceptibility to smoking, experimentation with smoking, or monthly smoking rates.

Another study investigated whether implementing an office-systems approach would prevent or delay adolescents’ drinking and smoking behaviors (Stevens et al. 2002). The idea of the approach in question, as expressed by Klein and Camenga (2004), is that the primary care physician provides anticipatory guidance and screening, the entire office staff endorses the prevention messages, and prevention materials are provided in the office. Stevens and associates (2002) found that despite evidence that their intervention was implemented successfully, it did not significantly affect adolescents’ tobacco use. The authors suggested that their program might have been ineffective in part because it focused on increasing parent-child communication rather than on targeting adolescents’ behaviors.

Ozer and colleagues (2004) presented preliminary results of a study indicating that adolescents who received clinical preventive services in managed care settings were less likely to increase the regular use of tobacco over a 1-year period, but they did not report the effects on initiation of tobacco use. More recently, Brown and associates (2007) examined the impact of a single-lesson course in tobacco cessation given to fourth and fifth graders at a health education center. The lesson focused on improving knowledge of tobacco, the identification of refusal techniques, and lowering intent to smoke. General knowledge about tobacco and refusal techniques significantly increased, but rates of intent to smoke did not decrease, perhaps because the rate was low before the intervention.

In summary, the few studies that have examined the efficacy of provider-based interventions suggest that the strategies they have employed may not be effective. However, the results must be interpreted with caution. Only a limited number of strategies have been assessed, and none of the studies on a specific prevention strategy have been replicated. This problem is complicated by the fact that most youth and many young adults are low-volume, intermittent smokers who often do not think of themselves as smokers. Furthermore, efforts directed at youth have been investigated in just a few health care settings, such as physicians’ and orthodontists’ offices and specialty clinics. Additionally, little is known about the impact of youth-focused efforts to prevent tobacco use that are conducted in specialty services such as asthma clinics, urgent care facilities, or emergency rooms.

Barriers to the Provision of Clinical Preventive Services to Youth

Physicians cite numerous barriers to providing clinical preventive services, including (1) having a large number of patients, which limits their time per patient; (2) competing health care demands during preventive visits; (3) insufficient education and training; (4) lack of information about how to access referral and treatment resources; (5) lack of dissemination to physicians of research that supports positive treatment outcomes and the negative effects of failure to intervene; (6) fear of alienating patients and families; and (7) inadequate reimbursement relative to the time and effort required to provide such services (Cheng et al. 1999; Kulig and American Academy of Pediatrics Committee on Substance Abuse 2005; Oscós-Sánchez et al. 2008; Sanders and Colson 2008). Research also suggests that physicians’ confidence in their ability to screen and advise adolescents about tobacco use is related to how frequently they deliver preventive services (Cheng et al. 1999; Ozer et al. 2004). Education and training of health care professionals can reduce the impact of several of these barriers; indeed, studies have shown that even a few hours of training on tobacco use can significantly improve medical students’ and physicians’ knowledge about this behavior as well as their confidence in delivering preventive services and the likelihood that they do so (Pederson et al. 2006; Fiore et al. 2008).

Summary Regarding Clinical Interventions with Young People

As primary sources of health information and potential role models, health care providers are well suited to address the prevention of tobacco use among youth. National guidelines for the provision of preventive services recognize the pivotal role that health care providers can play in preventing tobacco use and stipulate that prevention be addressed at least once per year throughout adolescence (AMA 1997; USDHHS 1998; USPSTF 2003; Bonnie et al. 2007; Hagan et al. 2008). The available literature indicates that adherence to recommended screening and prevention activities for patients who are children or adolescents, such as implementing the “5 A’s,” is low (Galuska et al. 2002). Studies suggest that tobacco-training programs and paper- and computer-based reminders for health care professionals to deliver services may be viable options for increasing the rates at which services to prevent tobacco use are delivered to children and adolescents (Ozer et al. 2005; Pederson et al. 2006; Dexheimer et al. 2008). Finally, little is known about the effectiveness of tobacco prevention services delivered to children and adolescents in health care settings, although a recent meta-analysis suggests that counseling may be effective in reducing adolescent smoking (Fiore et al. 2008). As a result, there is currently no clear evidence to suggest that any prevention strategies delivered in health care settings are effective in preventing the initiation of smoking in this population, but clinicians may be important in encouraging young smokers to quit.

Review of Meta-Analyses and a Cochrane Review

Flay (2009b) provided a review of meta-analyses and of the Cochrane review by Thomas and Perera (2006) in an effort to determine from past reviews whether school-based smoking prevention can be effective. Among the multiple meta-analyses of school-based programs was one that included 74 studies of smoking prevention among 207 studies on the prevention of substance abuse (Tobler et al. 2000), another that evaluated 65 separate programs (Hwang et al. 2004), a review of 94 randomized trials that reviewed only 23 in detail because of methodologic limitations with the remaining studies (Thomas and Perera 2006), and a review focusing on the quality of 11 evaluations but not their outcomes (Tingle et al. 2003). Reviews of the long-term effects of these programs have varied in scope: one review included 25 studies with at least 2-years of follow-up (Skara and Sussman 2003), and another found only 8 studies with outcome data for grade 12 (or 18 years of age) (Wiehe et al. 2005). The findings range from precise and substantial ESs for some types of programs (Tobler et al. 2000; Hwang et al. 2004) to conclusions that most school-based prevention programs are effective (Dobbins et al. 2008) or do not work (Glantz and Mandel 2005; Wiehe et al. 2005).

Tobler and colleagues (2000), after summarizing a series of meta-analyses, suggested that programs that used interactive learning strategies and involved same- or similar-aged peers as leaders or facilitators were most effective. In addition, Tobler and colleagues (2000) found that smoking prevention programs produced an average ES of 0.16, with interactive programs producing a signifi-cantly larger ES than did noninteractive programs (0.17 vs. 0.05). These authors also found that programs that addressed multiple substances were less effective at reducing tobacco use than were programs that targeted only tobacco (ES_=_0.10 vs. 0.17), but the multiple-substance programs had the added benefit of reducing alcohol and other substance use. These researchers also found program effects to be larger in schools with predominantly special or high-risk populations (characterized by minority populations, high absenteeism or dropout rates, or poor academic records). Hwang and colleagues (2004), in a review of 65 programs, estimated an average short-term ES of 0.19 for outcomes involving smoking behaviors. These authors reported ESs of 0.22 for attitudes and skills and 0.53 for knowledge and found that all program effects were smaller at those follow-ups that did not take place immediately after the intervention. Outcomes involving behavior, however, decayed very little over 1–3 years (from the original 0.19 to 0.18) but, without further programming, they decayed by one-half (to 0.09) at follow-ups of 3 or more years. Knowledge decayed by over 60% by 1-year follow-up (to 0.19), and attitudes and skills decayed to under one-half their original effects by 1-year follow-up (to 0.10 and 0.09, respectively).

Hwang and colleagues (2004) also estimated the effects of different approaches to school-based smoking prevention: social influences, cognitive-behavioral interventions (programs that included the elements of the social influences approach plus at least two cognitive skills), and life skills. They found that social influences approaches had average ESs of 0.12 at short-term follow-up, 0.15 at 1–3 years, and 0.07 at more than 3 years; cognitive-behavioral approaches had average ESs of 0.21 at both short-term follow-up and 1–3 years; and life skills approaches had average ESs of 0.29 at short-term follow-up and 0.16 at 1–3 years. There were too few studies in their meta-analysis to provide estimates of the longer-term effects (more than 3 years) of cognitive-behavioral or life skills approaches.

Hwang and colleagues (2004) also distinguished between programs based only at a school and those in school-plus-community settings. They found that school-only programs reported average ESs of 0.22, 0.16, and 0.06 in the short term, at 1–3 years, and more than 3 years, respectively, and school-plus-community programs reported average ESs of 0.16 in the short term and 0.21 at 1–3 years. In an earlier systematic review of school and school-plus-community programs in preventing substance abuse, Flay (2000) concluded that school-plus-community programs produced about double the effect of the school-only programs when the type of school program was held constant.

Rooney and Murray’s (1996) meta-analysis of 131 smoking prevention programs adjusted for studies with an error in the unit of analysis (i.e., the group analyzed was not the correct one, a common error in the relevant literature at that time), but this adjustment had little or no effect on the overall ESs. The average ES was around 0.10 at long-term follow-up, which would be about a 5% relative reduction in smoking (Rosenthal 1984). Using a modeling approach, the authors estimated that the impact of programs could be increased if they began around sixth grade as part of a multicomponent health program, gave same-age peer leaders a role in program delivery, and used booster sessions. They estimated that this might achieve a relative reduction in smoking of between 19% and 29% (or ESs in the 0.5–0.8 range).

Thomas and Perera (2006), who completed the most thorough systematic review of school-based smoking prevention studies to date (it is included in the Cochrane Database of Systematic Reviews), required a minimum of 6 months’ follow-up after the completion of the intervention. They restricted their reviews to RCTs and found 94 of them. The authors rated the methodologic biases of each study and classified them as having minimal, medium, or high risk of bias; they analyzed in detail only the 23 studies they judged to be of the highest quality. They determined statistical significance from their own analysis of ORs—the odds of those who were lifetime nonsmokers at baseline starting to smoke by the posttest in the intervention group compared with the control group. When intra-class correlation coefficients (ICCs) were not reported, they assumed an ICC of 0.097, the average found in a limited set of older studies (Siddiqui et al. 1996). Another criterion imposed by Thomas and Perera (2006) was requiring a minimum of one assessment at least 6 months beyond the end of the intervention. As interventions have become more comprehensive and longer in duration, it is becoming more difficult to meet this standard; it is not clear that a study should be excluded from consideration because the last posttest was less than 6 months after the last session, especially if the bulk of the intervention took place several years earlier. The only outcome reported by Thomas and Perera (2006) was the prevalence of smoking among participants who were never smokers at pretest, and thus they did not include such possible outcomes as changes in the proportion or prevalence of ever, weekly, or monthly smokers.

In terms of program types, Thomas and Perera (2006) assigned the 94 studies to five groups: (1) information only; (2) social competence (e.g., the Good Behavior Game, the Seattle Social Development Project); (3) social influences (e.g., Project CLASP, Waterloo Smoking Prevention Project); (4) combined social competence/influences (e.g., LST, Project Towards No Tobacco Use, Child Development Project); and (5) multimodal (i.e., including family or community components). However, as Thomas and Perera (2006) acknowledged, it is extraordinarily difficult for people not intimately involved in the field to determine how to group the different interventions. In addition, over time, the programs have become more alike in principle as they incorporate ideas from each other.

Based on the above inclusion criteria, Thomas and Perera (2006) concluded the following about school-based programs to prevent smoking:

1. There is little evidence that information alone is effective.
2. Nine of 13 studies of social influences that met their criteria for inclusion demonstrated positive effects.
3. The longest-lasting test (65 lessons over 8 years) of a social influences program (Hutchinson Smoking Prevention Project) found that the program was not effective.
4. There was limited evidence for the effectiveness of social competence programs (only two studies met criteria for inclusion).
5. Of only three high-quality studies of the combination of social competence and social influences, just one showed a significant effect overall, and one showed a significant effect only for the condition in which the program was led by a health educator (not significant for the self-instruction condition).
6. Three of the four studies of multimodal approaches that met criteria for inclusion produced positive effects.
7. There is little evidence of the long-term effectiveness of school-based programs to prevent smoking.

Assessing Short- and Long-Term Effects of Prevention Programs

Although there are multiple studies of school-based programs that have demonstrated short-term effects (at the completion of the program), there has been some concern about the maintenance of these outcomes in the long term (end of high school or beyond). Wiehe and colleagues (2005) conducted a meta-analysis of the eight studies they could locate with results reported at 12th grade or 18 years of age. Of the studies reviewed, only the LST program, an interactive program of 30 sessions (15 in 7th grade, 10 in 9th grade, and 5 in 10th grade) that incorporates the social influences approach, as well as the teaching of other general personal and social skills, was effective at long-term follow-up.

Skara and Sussman (2003) reviewed studies of 25 programs to prevent the use of tobacco or other drugs that included follow-up of at least 24 months. Eighteen of the studies reported significant short-term effects, and 15 reported significant long-term effects. Of 17 studies with both pretest and posttest data, 11 (65%) reported significant long-term effects, with an average reduction in the percentage of baseline nonusers who initiated smoking in the program (using the rate of initiation in the control group as the comparison) of 11.4% (range: 9–14.2%, ES = 0.28). Of the studies with significant short-term effects, 72% (13 of 18) had significant long-term effects. Program effects were less likely to decay when there was extended programming or booster sessions were given.

The Task Force on Community Preventive Services (Zaza et al. 2005), on behalf of the CDC, examined the effectiveness of school-based tobacco use interventions that were published from 1980 to 2001. The Task Force examined 117 studies from 154 published papers. Of these, 48 studies were excluded due to limited quality of implementation or poor study design, leaving 69 studies that were seen to provide the “best evidence” concerning the effectiveness of school-based interventions. Fifty-two studies measured changes in tobacco use prevalence among adolescents. A summary of these studies and their outcomes is shown in Table 6.9. The Task Force noted an overall median effect of nearly −1.0% in absolute difference in smoking prevalence between control and intervention groups (with a range from −10% to +4%). The Task Force concluded that school-based tobacco use interventions can be effective in the short term, but that evidence was insufficient from their review to include a recommendation of implementation at the national level, given the lack of long-term outcomes for most studies.

Table 6.9

Studies of the effectiveness of school-based interventions to reduce tobacco use.

In a second review, the Task Force on Community Preventive Services (Zaza et al. 2005) reviewed studies of comprehensive community-wide programs that included a school-based tobacco use prevention intervention. Community-wide programs included mass media campaigns, clean indoor air legislation or ordinances, excise tax increases on tobacco products, community education efforts conducted by local groups, and interventions to restrict minors’ access to tobacco products. The Task Force reviewed studies that had been published from 1980 to 2001 and identified 17 studies that (1) evaluated community or statewide multicomponent interventions that included a school-based intervention and (2) measured differences or changes in student tobacco use. Of these, one study was excluded because it did not provide measurements of differences or changes in student tobacco use behavior. A summary of the studies deemed sufficient quality for inclusion (n = 16) is found in Table 6.10. Of the 16 studies reviewed, 14 found significant reductions in student tobacco use. In particular, the combination of school-based programs, mass media campaigns, and community education demonstrated a consistent and strong reduction in adolescent tobacco use over time, with a median effect of −4.5% in absolute difference in smoking prevalence between control and intervention groups (with a range of −13% to −2%). The Task Force recommended school-based tobacco use prevention programs be implemented in combination with mass media campaigns and additional community-wide educational activities (The Community Guide 2011).

Table 6.10

Studies of the effectiveness of multicomponent interventions that include school-based programs to reduce tobacco use.

Dobbins and colleagues (2008) conducted a comprehensive review of the effectiveness of school-based tobacco use prevention programs, examining all systematic reviews and meta-analyses from 1985 to 2007. From an initial analysis of 10,163 abstracts and titles, 92 papers were potentially relevant, and 12 reviews were considered relevant with moderate or strong methodologies. Smoking behavior was reported in 11 of the 12 reviews, with 6 reviews showing a positive effect of school-based programs, 2 showing promising effects, and 3 reporting no impact on smoking outcomes. The reviewers concluded that school-based tobacco use prevention programs are effective in reducing smoking prevalence, onset, and intentions to smoke in the short term. Flay (2007) provided a review of the long-term effectiveness of school-based smoking prevention for the 2007 Institute of Medicine report on tobacco control (Bonnie et al. 2007, Appendix D). From an examination of the previous reviews and meta-analyses reviewed above, Flay concluded that school-based programs to prevent smoking can have significant long-term effects if they have the following attributes: (1) They are interactive programs based on social influences or social skills; (2) 15 or more sessions are involved, including some up to at least ninth grade; and (3) substantial short-term effects are produced.

Working from these three conclusions, Flay (2009a) reviewed evaluations of programs that had included 15 or more sessions (preferably some in high school), had demonstrated effects at both short- and medium-term follow-up, and followed students to the end of high school and beyond. Only three school-based programs and three school-plus programs (i.e., plus small media, plus mass media, or plus family or community components) fulfilled these criteria. This review was not limited to randomized trials, but most of the studies reviewed by Flay (2009a) were of this type. The two groups of studies (involving school-based and school-plus programs) are labeled as Category 1. All six programs evaluated in the Category 1 studies had been included in the 25 studies with at least 2 years of follow-up reviewed by Skara and Sussman (2003), as well as in the Task Force review (2003). For Category 1, only studies that included follow-up into high school were considered. Few studies included follow-up beyond high school, but for those that did, the reported effects are of interest.

The percentage of RI was used as the indicator of ES since it was readily available for all programs, while the detailed statistics needed to calculate ES were incompletely reported. Also, RI is widely used in calculations of cost and benefit and readily understood. For randomized trials, pretest levels of smoking should be the same in both the program and control groups, and RI would be the difference between posttest control (C) and program (P) groups divided by the level in the control group: (%C − %P)/%C. However, pretest levels in the programs were not always the same (because randomization does not always result in equal pretest levels), and adjustments should be made for these differences. In cases in which pretest data were reported, RI is the posttest difference between groups minus the pretest difference between groups, divided by the control group posttest level—that is, (%ΔC − %ΔP)/%C_post—expressed as a percentage. One may compare the ESs reported in meta-analyses and RIs by translating the ES into an RI on the basis of the area under the curve in the Z distribution (Rosenthal 1984). (For a convenient conversion tool, see Wilderdom [2012]). This approach translates an ES of 0.17 into a 7% relative reduction in smoking (an ES of 0.96_= an RI of 33%).

Category 1 school-based programs included the Tobacco and Alcohol Prevention Project (TAPP) (Hansen et al. 1988b), the LST program (Botvin et al. 1995a), and Project SHOUT (Students Helping Others Understand Tobacco) (Elder et al. 1993; Eckhardt et al. 1997) (see Table 6.9). On average, these three social influences/social competence programs, counting only those instances of 15 or more sessions during 2–4 years, preferably with some content in high school, had significant short-term effects (i.e., at grades seven or eight) of 21.8% (a range of 9–30%) and significant long-term effects (i.e., at grades 10–12) of 27.6% (a range of 19–44%) in terms of relative reduction in smoking. TAPP was the only one without any high school content and for which short-term effects decreased over time. Project SHOUT (Elder et al. 1993; Eckhardt et al. 1997) produced effects that may have been due to added content on activities of the tobacco industry, the teaching and encouragement of advocacy skills, and personal attention during high school. The long-term effects for the three programs suggest that a minimal personal-contact intervention of this kind in high school could increase the effects of any other program delivered in middle school. From these studies, Flay (2007) concluded that programs oriented to social influences/social competence that are of proven effectiveness and well implemented can produce long-term RIs of between 25% and 30% or ESs between 0.7 and 0.8.

The Category 1 school-plus studies included the North Karelia Project (Vartiainen et al. 1983, 1986, 1990, 1998), the Minnesota Class of 89 project (Perry et al. 1989, 1992, 1994), and MPP (Pentz et al. 1989b–e; Johnson et al. 1990). These programs produced mean short-term RIs of 40.7%, almost twice as high as the school-only programs, a finding consistent with a previous review by Flay (2000). These effects decayed over time an average of 21% to reach 32% RI. The long-term effects of school-plus-community or mass media programs were 12% better than school-only programs. It should be noted, however, that program effects were maintained at a higher level (at almost 40%, or 31% better than school-only programs) for those programs that included a high school component (North Karelia and Minnesota Class of 89), suggesting that programming in high school may reduce the decay of effects.

That the use of multiple delivery modalities increases a program’s effectiveness over that obtained from school-only programs (Flay 2000) is consistent with theories about the influences on behavior that exist across multiple domains of life (Bronfenbrenner 1977, 1979, 1986; Flay and Petraitis 1994; Petraitis et al. 1995; Flay et al. 2009). Thus, it has been argued that prevention programs will be more effective if students receive consistent messages across community contexts and over time. On the basis of the Category 1 school-plus studies, Flay concluded that ongoing programs of proven short-term and intermediate-term effectiveness that combine school intervention with mass media or a community program can produce a long-term RI of between 35% and 40% or an ES between 1.0 and 1.3.

Additional School-Based Smoking Prevention Programs, 2008–2011

The systematic reviews discussed above cover the peer-reviewed literature up to 2008. Since that time, six studies have been published that provide further support for the effectiveness of school-based smoking prevention programs and comprehensive community-wide interventions that include a school-based program. The studies also point to the potential for dissemination and adaptation of programs in other countries, peer involvement, new technologies, and community-wide strategies.

Ariza and colleagues (2008) examined the effects of a school-based program (16 sessions over 3 years), smokefree policies, smoking cessation for teachers, parent education, and community-based activities, using a quasi-experimental design in schools in Barcelona, Spain. At 36 months, when the cohort was 15 and 16 years of age, 18.6% of the boys and 31.2% of the girls were regular smokers in the intervention group, compared with 21.6% of the boys and 38.3% of the girls in the control group (p <.001).

Campbell and associates (2008) evaluated the ASSIST intervention in a randomized trial of 59 schools in England and Wales. The intervention consisted of training influential students to act as peer supporters outside the classroom in informal interactions with their peers and to encourage their peers not to smoke. Using data from all three follow-ups, the odds of being a smoker in an intervention school compared to a control school was 0.78 (0.64–0.96), although annual data were not as compelling.

Prokhorov and colleagues (2008) examined the long-term efficacy of the computer-based ASPIRE program for culturally diverse high school students in Houston, Texas. ASPIRE is a computer-based theoretically driven program on smoking prevention and cessation for high school students. Students randomized into the ASPIRE program had significantly lower smoking initiation rates than did students in the control group (1.9% vs. 5.8%, p <0.05) at the 18-month follow-up.

Perry and colleagues (2009) assessed the effectiveness of Project MYTRI, a 2 year multicomponent school-based tobacco intervention, in Delhi and Chennai, India. MYTRI, based on social cognitive theory, included peer-led activities, posters hung in the school, parental postcards, and peer activism outside of the classroom. Students in 32 schools, in sixth and eighth grades, were recruited and schools were randomized into either intervention or control groups; baseline, intermediate, and outcome data were collected for the two cohorts. At the end of the 2-year period, all students in the intervention group were significantly less likely to have smoked either bidis (p <0.01) or cigarettes (p = 0.05) than students in the control group.

Lotrean and associates (2010) examined the effectiveness of a video and peer-led school-based smoking prevention program among students 13 and 14 years of age in Romania. Pretest and posttest data were collected 9 months apart from 1,071 students. The program was focused on increasing both self-efficacy and cigarette refusal skills. At follow-up, 4.5% of students receiving the intervention reported weekly smoking compared with 9.5% in the control group; multivariate logistic regression demonstrated that nonsmokers in the control groups were twice as likely to become smokers (OR = 2.23, p <0.01) compared to nonsmokers in the intervention group.

Hawkins and colleagues (in press) evaluated the effectiveness of the Communities that Care (CTC) prevention program on levels of risk and adolescent problem behaviors, including cigarette use. Twenty-four communities were matched and randomly assigned to either the intervention or the control group; 4,407 5th-grade students, in 2004, were recruited and surveyed annually through 10th grade, in 2009. Students in the CTC communities were 21% less likely to report smoking cigarettes in the past 30 days compared to students in the control communities (adjusted OR = 0.79, p <0.05).

Summary of Review of Reviews

Ultimately, the purpose of reviews of smoking prevention programs is to provide guidance to schools and communities as to what approaches might be effective. In a field such as school-based smoking prevention, which compares disparate programs with differing formats, theoretical orientations, targeted behaviors, and targeted populations and age groups, the application of meta-analysis methods can be difficult. Despite the challenges, the meta-analyses by Tobler and colleagues (2000) and Hwang and associates (2004) both provide clear directions on what types of programs are most effective. From a systematic review of reviews and individual studies of mediators, boosters, peer-directed versus adult-led programs, and community components of drug prevention programs, Cuijpers (2002) developed a useful summary of the important ingredients of effective prevention programs that can be set forth as follows:

1. They use interactive delivery methods.
2. They employ the social influences model (defined more broadly than by Hwang and colleagues [2004]).
3. They include components on norms and commitments not to use tobacco and intentions not to use this product.
4. They add community components.
5. They include the use of peer leaders rather than relying totally on adult providers.
6. They include training and practice in the use of refusal and other life skills.

In addition, meta-analyses have established that programs that have relatively more sessions and continue for multiple years are more effective. From a systematic review of the long-term effects of school-based prevention, Flay (2007) concluded that programs with demonstrated short- and intermediate-term effectiveness could have large long-term effects in the range of 35%–40% reductions in the proportion of youth who smoke.

Additional Comments on Reviews and Meta-Analyses

Evidence-based programs. In recent years, evidence-based practice and related terms have become part of the language for clinicians and health care researchers in the United States and other countries. Multiple agencies have reviewed evaluations of programs to prevent substance abuse and produced lists of scientifically proven or evidence-based programs (CDC 2009), and the University of Colorado at Boulder (2010) has provided a comparative matrix.

The stated purpose of such lists and guides is to help decision makers at both the federal and local levels choose programs supported by the best available evidence (Petrosino 2003). After the U.S. Department of Education compiled one such list (of 9 “exemplary” and 33 “promising” programs) with the help of a panel of eminent researchers in prevention, school districts using federal funds were strongly encouraged to select a program from that list (Weiss et al. 2005). These lists of programs are very useful as guides; of course, content and fit for a given community need to be considered.

Cultural sensitivity. Cultural sensitivity is believed to be important for effective prevention (Schinke et al. 1987, 1988, 1990; LaFromboise et al. 1993; Lynagh et al. 1997; Klonoff and Landrine 1999; Litrownik et al. 2000; Vélez 2001; Sussman et al. 2003; Chen 2004; Flay et al. 2004; Shelley et al. 2004; Miranda et al. 2005; Hecht and Krieger 2006; Ferketich et al. 2007). Many studies have evaluated the effectiveness of untargeted or targeted prevention curricula in White, minority, or diverse samples, but few studies have directly compared culturally relevant curricula for smoking prevention with curricula that do not address cultural issues (Johnson et al. 2005). In one study, Botvin and colleagues (1995a) found that culturally targeted and nontargeted versions of their life-skills program were more effective than a control condition in preventing smoking among African American and Hispanic adolescents. Later, another group of researchers (Gosin et al. 2003; Hecht et al. 2003, 2006; Hecht and Krieger 2006; Warren et al. 2006) compared prevention curricula targeted to the values of several cultural groups: Mexican Americans, Blacks/Whites (the study was conducted in a region with a very low prevalence of Blacks), and a multicultural group. All three curricula were more effective than a control curriculum and the Mexican American and multicultural curricula affected more outcome variables (regardless of the students’ ethnic characteristics) than did the Black/White curriculum.

In a study in ethnically diverse schools (Hispanic, Asian American, and White) in Southern California, Johnson and colleagues (2005, 2007) compared two eight-session curricula based on the social influences approach. One, Project CHIPS (Choosing Healthy Influences for a Positive Self), a version of Project SMART (Self –Management and Resistance Training) (Hansen et al. 1988a), had content that emphasized “looking after yourself.” The other, Project FLAVOR (Fun Learning about Vitality, Origin, and Respect), included cultural values from Hispanic and Asian cultures that emphasized group objectives, interdependence of family members, respect for ancestors, and harmonious interpersonal relations. The authors found that the multicultural curriculum (Project FLAVOR) was effective for Hispanic students in mostly Hispanic schools. In contrast, the curriculum framed for individuals (Project CHIPS) was effective only for Asian students in Asian/multicultural schools.

The results reported above suggest that caution is needed when implementing programs with different ethnic groups or in different cultures. Some programs seem to be equally effective with many different groups, but studies suggest that making programs culturally relevant might be very important. Clearly, more research is needed on this issue. In the meantime, any community or country adopting a program will need to evaluate it rigorously to determine its effectiveness in the new setting or culture.

The role of school policies. Before the 1994 Surgeon General’s report (USDHHS 1994), several researchers and educators had suggested that school smoking policies could reduce smoking among youth. School policies generally include rules about tobacco use on campus by students, teachers, staff, and visitors and rules about possession of tobacco products. For example, Pentz and associates (1989a) examined the effects of school policies on adolescents in California and concluded that they were associated with reduced smoking in that group. Overall, the literature on the effectiveness of school smoking policies is surprisingly small, perhaps because such policies are now universally and widely applied to students and schools.

By the late 1980s, most school districts had some type of policy or regulation on tobacco smoking (CDC 1989), and the federal Pro-Children Act of 1994 prompted the majority of schools to create additional tobacco-related policies. However, although research exists relative to facilitating the adoption of tobacco-free school policies (Goldstein et al. 2003), once such policies are implemented, their enforcement and application to students and staff vary considerably. Kumar and colleagues (2005) examined the association between certain variables related to school policies and smoking among middle school (8th grade) and high school (10th and 12th grades) students using the1999–2000 MTF survey to obtain smoking prevalence and relying on data about school policies provided by administrators. The authors found that permissive smoking policies for school staff were positively but not significantly associated with student smoking in middle schools and that this was the only school policy variable associated with the prevalence of smoking in high school. The level of monitoring of smoking in the school was inversely related to the prevalence of smoking among middle school but not among high school students. The severity of consequences was not related to the prevalence of smoking in either group of students, a finding consistent with previous research (Pentz et al. 1989a). This research suggests that to be successful, schools need to take a proactive approach to implementing school no-smoking policies. Similarly, in a study of nearly 5,000 Australian students, Hamilton and colleagues (2003) found that rates of smoking among students were lower in schools that provided education or counseling rather than a discipline-only approach.

Wakefield and associates (2000) examined 1996 survey data for high school students across the United States as part of the Study of Smoking and Tobacco Use Among Young People; the authors examined the effects of both the existence of a smoking ban (as reported by students) and whether the ban was enforced. They found no effect on youth smoking from the existence of a ban but found that an enforced ban was associated with a lower likelihood of progressing from a lower to a higher intensity of smoking. Later, Powell and colleagues (2005) examined the same data on students but used information from administrators on the existence of a smoking ban; they found that the effect of bans on the prevalence of smoking was attenuated by including levels of peer smoking in the statistical model, although the effect of the smoking ban remained significant.

Students’ perceived enforcement of their school’s smoking policy may also be an important factor in reducing the risk of smoking. Murnaghan and colleagues (2008), in a study of 10th-grade Canadian students, found that students who believed that tobacco policies were enforced were less likely to smoke. Similarly, using a random sample of schools from five Canadian provinces, Lovato and colleagues (2007) reported that students’ perception of enforcement was a significant predictor of the prevalence of smoking.

The research reviewed above highlights the importance of implementing and enforcing school tobacco policies and ensuring that students perceive that the policies are enforced. Thus, to provide accurate conclusions when evaluating a policy, studies should evaluate its enforcement (Murnaghan et al. 2007).

Students’ attitudes toward school policies may also have an impact on their decisions to smoke. Using data from a representative sample of 10th-grade California students, Unger and associates (1999) explored adolescents’ attitudes toward no-smoking policies, including school-based policies. Attitudes toward no-smoking policies varied widely and were associated with smoking status, other psychosocial variables, and smoking-related advocacy efforts by the students. The researchers suggested that attitudes toward no-smoking policies may be either a determinant or a consequence of smoking behavior.

In summary, school policies on tobacco use have been recommended as an important component of comprehensive, multicomponent efforts to prevent use (CDC 1989; Barnett et al. 2007). Overall, research has shown that, to be effective, tobacco-related policy needs to be enforced and should foster a proactive approach by schools to prevention.

Ineffective programs. Many programs and prevention activities that have received a lot of attention have been shown to be ineffective, especially in the long term, when they were evaluated fully. Examples include onetime visiting speakers, other 1-day special events, poster competitions, lotteries, and other similar efforts. Other programs that are more similar to the multiple-session school-based prevention programs reviewed above have also been shown to be ineffective.

The D.A.R.E. program was developed by the Los Angeles Police Department and the Los Angeles Unified School District in the early 1980s. These groups essentially took the two variants of Project SMART being tested with seventh-grade students in Los Angeles schools at the time (Graham et al. 1990), combined them, and added a great deal of information about drugs (including, in some variants of the program, what they looked like, where to get them, and how they were used). Police officers were to deliver the program to students in fifth and sixth grades. The results of a randomized trial of the two Project SMART variants found that the program in resistance skills was effective but that the self-management component led to increased drug use relative to that of control group students (Hansen et al. 1988a; Graham et al. 1990). These results, combined with evidence that providing only information does not generally influence behavior change (Goodstadt 1978, 1980), and the use of police officers who are not trained to be highly skilled teachers, indicate that D.A.R.E. is most likely an ineffective program.

Although early nonrandomized studies suggested that D.A.R.E. sometimes had small effects for elementary school students, multiple randomized trials (Ennett et al. 1994a; Rosenbaum et al. 1994; Clayton et al. 1996; Dukes et al. 1996; Rosenbaum and Hanson 1998; Lynam et al. 1999) and two meta-analyses (Ennett et al. 1994b; West and O’Neal 2004) have established that D.A.R.E. has little or no impact on drug use in the short term and no impact in the long term, indicating its ineffectiveness. Even so, D.A.R.E. has been disseminated widely (Rogers 1995a; Des Jarlais et al. 2006). In response to the increasing evidence of the program’s ineffectiveness, the D.A.R.E. organization has developed new programs for junior and senior high school students, but the program for junior high also has been shown to be ineffective (Perry et al. 2000, 2003), and evaluations of the high school program are not yet completed (Sloboda et al. 2009).

The Hutchinson Smoking Prevention Project (conducted at the Fred Hutchinson Cancer Research Center, University of Washington) has received much attention because the outcome evaluation was of such high quality and conducted over the long term. The project was designed to be a multiyear (grades 3–10) social influences program. A large randomized trial (20 school groups per condition) of the project produced no significant effects either by the end of grade 12 or 2 years later (Peterson et al. 2000). The findings of the trial are, however, quite difficult to interpret. The investigators did not report what its effects were at any time other than the two times noted above, including before entering high school (when most other programs report short-term and immediate-term results) or at the end of the program (grade 10). The effects of an intervention are generally measured immediately or shortly after the program ends to see the maximum impact, and the long-term measurement should serve to assess how permanent the effect was or how quickly it decayed.

Cost-Effectiveness

It is difficult to estimate the costs and benefits of successful prevention programs and, therefore, their cost-benefit ratio (Caulkins et al. 1999; Foster et al. 2003). First, the costs and benefits for a particular program are variable; second, the long-term effectiveness of these programs has varied a great deal as well (Tengs 1996). Nevertheless, several scholars have provided estimates of cost-benefit ratios, using different techniques to do so.

One analysis estimated the cost of an effective 30-session prevention program in the United States at US$150 per student for program materials, training, teacher time, and other expenses (Caulkins et al. 1999). The estimated savings from such programs owing to the benefits of preventing significant numbers of students from initiating smoking and delaying the start date for those who later initiate smoking (and therefore the lifetime consumption) were substantial (Caulkins et al. 2004). For example, the estimated social benefits of smoking prevention alone were about US$300 per student, for a cost-benefit ratio of 2.0, and the estimated total benefits were about US$840, a cost-benefit ratio of 5.6.

The cost of an effective school-based smoking prevention program in Canada was estimated at C$67 per student (Stephens et al. 2000). Assuming a modest 4% level of long-term effectiveness, the benefits of smoking prevention were estimated to be lifetime savings for health care of C$3,400 per person and an increase in (lifetime) productivity of almost C$14,000 (Stephens et al. 2000), a cost-benefit ratio of 15.4. In other words, a moderately successful school-based smoking prevention program could produce a savings of C$15.40 for every C$1.00 spent.

Almost two decades ago, Hodgson (1992) asserted that smokers incur about US$9,379 more in lifetime health costs than do nonsmokers. Using this information, Wang and colleagues (2001) estimated the cost-effectiveness of LST to be about US$13,316 per life saved and US$8,482 per quality-adjusted life year (QALY), with the program costing US$13.29 per student. Given the large increases in unit costs for health care since 1992, these figures would have to be updated, but the results are instructive as to the cost-effectiveness of LST.

A group that looked at Project Towards No Tobacco Use (TNT) estimated its costs at US$48 per student and determined that it would cost about US$20,000 per QALY gained (Tengs et al. 2001). Although TNT was not cost saving, the authors concluded that the prevention of smoking offers gains in both survival and health-related quality of life that make it worth the cost. This latter statement is based on citizens’ demonstrated “willingness to pay” for gains on the order of several hundred thousand dollars per QALY saved. In addition, an earlier analysis by Tengs and coworkers (1995) found that the median cost of 587 medical and public health interventions was US$42,000 per year of life saved and concluded that school-based smoking prevention is more efficient than most health/medical interventions.

The social benefits of even broader programs for improving behavior could be considerably greater. For example, Aos and colleagues (2004), at the Washington State Institute for Public Policy, who analyzed the cost-effectiveness of about 70 prevention programs, estimated that LST cost US$29 per student and led to benefits of US$746 (from the prevention of both smoking and drug abuse), a benefit of over US$25.61 per dollar spent, or a cost-benefit ratio of 25.61. In addition, they estimated that TNT cost US$5 per student and produced a benefit of US$279, a cost-benefit ratio of 55.84. Other programs included in both that review and in this chapter include the Good Behavior Game (cost-benefit ratio = 25.92), the MPP (ratio of 5.29), the Minnesota Smoking Prevention Program (ratio of 102.29), and a category of “other social influence/skills building substance prevention programs” (cost-benefit ratio of 70.34).

Although these cost effectiveness studies have focused on school-based prevention programs, their results support all prevention efforts. From a societal perspective, the costs of effective prevention are justified, both to the individual student and to society as a whole. In the study by Aos and colleagues (2004), cost-benefit ratios ranged from 2 to more than 100 for the prevention programs reviewed.

Summary Regarding School-Based Prevention

There are effective school-based smoking prevention programs that can be adopted, adapted, and deployed with at least short-term outcomes among adolescents. Programs can be found at the National Registry of Evidence-based Programs and Policies. Communities and school districts should invest only in the research-proven programs and avoid spending money on programs with little or no evidence of effectiveness. When implementing programs, decision makers must pay attention to maintaining program fidelity to ensure quality control.

Unfortunately, the inconsistent results and conclusions reported in the literature have caused many researchers, educators, and policymakers to conclude that school-based prevention does not work. Prior reviews have suggested that a more appropriate conclusion would be that many existing school-based programs have demonstrated effectiveness in the short term and that selected programs have demonstrated long-term effectiveness (Skara and Sussman 2003; Flay 2007, 2009a). Importantly, school-based programs produce larger and more sustained effects when they are implemented in combination with supplementary or complementary family-, mass media-, or community-based programs (Table 6.3). Similarly, other kinds of interventions produce larger effects when carried out in combination with other interventions (e.g., mass media plus taxation). Theories from sociology and public health (Bronfenbrenner 1977, 1979, 1986; Flay and Petraitis 1994; Flay et al. 2009) reveal that the more risk and protective factors an intervention or set of interventions addresses, the greater will be the effects. All of these data support the conclusion that a comprehensive, multicomponent approach to tobacco use prevention is more efficacious than a single strategy.

Thus, for school-based prevention to be effective, the programs should be comprehensive, interactive, start early, be sustained, incorporate an appropriate number of lessons, and be integrated into a community-wide approach (Flay 2007). Even among studies that have presented different conclusions regarding the effectiveness of school-based prevention, numerous studies (Sutton 2000; Cuijpers 2002; La Torre et al. 2005; Davis et al. 2007b; Warner 2007b) have concluded that school-based prevention works when combined with a comprehensive approach; that is, prevention efforts must address more distal, social, and community influences, too.

Review

Programming for the cessation of cigarette smoking among adolescents is defined as any type of programming in any setting that targets an age range of 12–19 years, that focuses on persons who smoke cigarettes at baseline (generally at least once in the last 30 days), and that encourages them to quit cigarette smoking. There have been nine systematic reviews of the relevant literature. In the first, Sussman and colleagues (1999) evaluated 34 trials, 17 on smoking cessation and 17 on smoking prevention, for their impact on cessation of cigarette smoking. Next, Sussman (2002) provided an enlarged review of 66 cessation trials and 17 studies of self-initiated quitting, and then McDonald and colleagues (2003) provided a review of many of the same studies (Sussman 2002). Garrison and colleagues (2003) reviewed six studies that used relatively rigorous designs, and Backinger and colleagues (2003) performed a qualitative review of prevention and cessation programs.

In the first meta-analysis of smoking cessation programs for adolescents, Sussman and colleagues (2006) included 48 studies with control groups. Shortly thereafter, Grimshaw and Stanton (2006) provided a Cochrane meta-analysis of 15 studies. The main difference in inclusion criteria between the two meta-analyses was that Grimshaw and Stanton required that studies contain follow-ups at least 6 months after the intervention (the standard used for adult cessation programs), while Sussman and colleagues did not. Both meta-analyses included RCTs, cluster RCTs, and non-RCTs. Next, Gervais and colleagues (2007) empirically reviewed 16 RCTs derived from previous reviews and data searches up to November 2006.

Sussman and Sun (2009) provided the most recent review; their literature search covered January 1970 to December 2007. This review included 64 studies, 16 more than the initial meta-analysis by Sussman and associates (2006), and included any English-language article or report with data on the contents of an adolescent smoking cessation effort, rates of quitting, and an age range of 12–19 years. Studies that included fewer than eight cigarette smokers at baseline were excluded because of the extremely small sample (fewer than five smokers per condition). Tobacco-related interventions for pregnant females were not included, and all reviewed studies included both genders. Data available through surveys of practitioners in the field were not reviewed. Finally, only studies that included a control condition were selected, and multiple-baseline, quasi-experimental, or experimental designs were permitted.

The 64 controlled trials that met the criteria for inclusion in the Sussman and Sun (2009) review were selected from an initial 130 studies; 50% of those 130 lacked control conditions (were single-group designs) and were not included in the review. An estimated one-third of the studies completed after 2000 that were in the initial group of 130 were single-group designs, suggesting some improvement in the design of these types of studies in recent years. Also, about one-third of the original 130 studies were published in 2000 or later (n = 42), an indication of increasing interest in adolescent cessation. A summary of the studies included in this review can be found in Table 6.11.

Table 6.11

Studies on smoking cessation among youth.

The variables examined by Sussman and Sun (2009) included program content, modalities of delivery, number of contacts, and expected rates of quitting at follow-ups. In addition, the means of recruiting and retaining smokers in the programs and suggestions on the lead time needed for a measurable effect were discussed. The results of the Sussman and Sun (2009) review were consistent with all previous reviews, except that of Garrison and colleagues (2003), which reviewed only six studies. The 64 studies in Sussman and Sun (2009) had an average reach to the recruited target audience of more than 35% and an average retention rate of approximately 75% for follow-up. The studies reviewed showed little evidence of disruption during implementation, sessions that were omitted, or restart of the intervention. However, specific documentation of the fidelity of implementation was not provided in most studies. Across the 64 studies, direct interpersonal contact of the treatment provider with potential participants and recruitment in contexts (e.g., classrooms) in which most of the members were potential participants led to relatively higher participation in the programs.

Sussman and colleagues (2006) and Sussman and Sun (2009) examined the mean estimated effects for four main predictors of outcomes (Tables 6.12–6.15) from their reviews. The five types of focus were social influences, cognitive-behavioral, motivational, medical, and other (e.g., reduction of supply and clarification of affect). The nine modalities of delivery were classified as classroom, school clinics, medical clinics, family, systemwide, computer, sensory deprivation, court diversion, and interventions in other public settings (e.g., worksite, shopping mall, and dormitory). The number of sessions varied from one to four, five to eight, to nine or more (three categories). Length of follow-up ranged from 0 to 3 months, 4 to 12 months, and more than 12 months past immediate post-test (three categories).

Table 6.12

Youth cessation treatment means, 2006 and 2007 analyses, stratified by duration of follow-up.

Table 6.13

Youth cessation treatment means, 2006 and 2007 analyses, stratified by theory.

Table 6.14

Youth cessation treatment means, 2006 and 2007 analyses, stratified by modality.

Table 6.15

Youth cessation treatment means, 2006 and 2007 analyses, stratified by number of sessions.

Most studies on adolescent cessation were underpowered statistically; in this case, the samples tended to be too small to detect significant differences between the program and control means with reasonable certainty (Cohen 1988). Also, most studies failed to use analyses that were appropriate for clustered data; in this instance when one unit, such as a cessation group, is nested within another, such as a school, the study should account for the confounding of the association between the cessation group and the school to permit a more accurate interpretation of rates of quitting. In addition, randomization generally is most effective with large sample sizes, so differences in treatment groups at baseline needed to be considered.

On average, almost twice as many in the treatment groups quit as in the control groups: 13.4% versus 7.4% (RI = 6.4%; p <0.001). The most effective studies used programs based on social influences, cognitive-behavioral theory, or programming to enhance motivation as the theory behind their intervention design. Results also appeared promising for medical-/recovery-based programming, but the number of studies here was too small (n = 3) to infer consistent effects. The modalities in which programming achieved the strongest effects were classroom-based educational programs, school-based clinics, and computer-based programming.

One limitation in trying to differentiate the effects of theory from modality is that these were not independent categorizations. In the current sample, 7 of 11 classroom-based studies involved manipulations of social influences; 20 of 29 school clinic studies were cognitive-behavioral; 8 of 9 medical clinic studies were based on enhancement of motivation; 4 of 6 systemwide studies were in the “other” theory category; and 2 of 3 computer-based studies were based on enhancing motivation.

Relatively higher rates of quitting were found for programs having five or more sessions (none with fewer than five sessions produced significant findings, but those with five or more sessions showed a 5% increase in quitting compared with controls). In addition, effects for programs with five or more sessions were also maintained at short-term (1 year or less) and long-term (greater than 1 year) follow-ups. Eight studies examined follow-ups longer than 12 months, and in these studies, the short-term effects were maintained. More studies with long-term follow-ups are needed; even so, these data are promising, suggesting that adolescent cessation rates tend not to decrease much over time.

Summary Regarding Smoking Cessation Programs for Youth

Overall, several smoking cessation programs for adolescents have been found to be efficacious. Many of the findings for youth programs are consistent with those found in the literature on adults, particularly regarding the importance of using cognitive-behavioral strategies and achieving a sufficient dosage of programming (Fiore et al. 2000). For example, the N-O-T Program targeting 14–19-year-old daily smokers is based on social cognitive theory and includes 10 hour-long sessions (plus 4 boosters) covering such topics as self-management, social influences, relapse prevention, and managing nicotine withdrawal (Horn et al. 2005a). One difference is that, at present, there is little evidence of the efficacy of pharmacologic adjuncts for youth, in contrast with the strong efficacy for adults. Future work on the metabolism of pharmacologic adjuncts, patterns of tobacco use among youth, and self-reported withdrawal symptoms might help researchers and policymakers improve their understanding of the potential effectiveness of pharmacologic adjuncts among youth.

There is a strong need for more research on youth cessation that makes use of appropriate controls, uses more standard measures of cessation, and conducts longer follow-ups (12 months and perhaps longer). Research on how to effectively recruit young smokers is needed. Also, metrics such as the cost-effectiveness of treatment per QALYs gained and years of disability avoided should be examined in future studies on youth smoking cessation. The use of such metrics could demonstrate even greater cost-effectiveness for early interventions than would be found for smoking cessation programs among adults. There is also a need for evaluating whether different cessation programs are needed for different levels of use or for different kinds of tobacco products, such as smokeless tobacco.

Research on Smokeless Tobacco Cessation with Youth

Of the few publications describing programs to quit smokeless tobacco for youth (Table 6.16; Sussman et al. [2006]), most have focused on high school or college athletes who are known to have higher rates of use (Boyd and Glover 1989; Colborn et al. 1989). Some reviews of more broadly targeted programs, designed to reduce the adoption of overall tobacco use by middle school and high school youth, have examined the impact of these programs on cessation among those students who were using tobacco products at baseline. A handful of these studies have included smokeless tobacco as a part of their comprehensive focus on the tobacco problem, but none has teased out the results for smokeless tobacco in a manner that provides guidance as to which components of the intervention are most effective for quitting the use of this product, nor do they provide long-term cessation results for smokeless tobacco that serve as useful benchmarks (Mermelstein 2003; Skara and Sussman 2003; Sussman et al. 2003). One highly relevant report described results from a focus group of 27 adolescents on the acceptability and appeal of a Web-based smoking prevention program (Parlove et al. 2004) and formative data suggest that this could be a promising avenue to providing assistance with cessation for smokeless tobacco, but no outcome data were reported.

Table 6.16

Studies on smokeless tobacco cessation for youth.

Eakin and colleagues (1989) tested a three-session, multicomponent, cognitive-behavioral intervention that included self-monitoring of smokeless tobacco use, a component designed to increase the user’s awareness of health risks, behavioral coping strategies, frequent telephone contact, and training in the prevention of relapse. Biochemical (carbon monoxide and cotinine) verification of self-reports was obtained. Twenty-one of the 25 adolescents in the original study (14–18 years of age, averaging five to eight dips per day) completed treatment, 9 (36%) were abstinent at the conclusion of the program, and 4 (16%) maintained abstinence at the 3-month follow-up. Participants who did not achieve complete abstinence reported substantial reductions in use of smokeless tobacco. Of those who also were cigarette smokers, none reported an increase in cigarette consumption as a result of reducing the use of or quitting smokeless tobacco. Predictors of cessation for smokeless tobacco included lower baseline consumption levels and involvement in school athletics.

In the study of high school baseball players in rural California (Eakin et al. 1989; Walsh et al. 2000, 2003) described above, 44 high schools were randomly assigned to a treatment condition (516 participants) or a no-treatment control (568 participants). The intervention included discussion of the harmful effects of using smokeless tobacco, refusal skills, a strong peer opinion leader who encouraged cessation of smokeless tobacco, a meeting with parents and coaches to obtain their support and a self-help guide for quitting, a dental exam (with advice on cessation from a dentist and behavioral counseling from a dental hygienist), and booster sessions to prevent relapse. Cessation was observed in 27% of the athletes attending the intervention schools and 14% of athletes in the control schools (RI = 17.8%). The results were based on self-reports, but the authors did take saliva samples from participants who were told that the samples could be used to confirm the veracity of the self-reports (Evans et al. 1977; Murray and Perry 1987), even though there was no intention to test all of them (this is the “bogus-pipeline” procedure). The multiple intervention components, including the use of oral health screening exams, brief counseling, and peer-led educational sessions were successful in doubling the rate of quitting over that obtained by participants in control schools. Previous studies on cessation with adults have reported that oral exams can be a significant motivator for users of smokeless tobacco to quit (Severson and Hatsukami 1999; Ebbert et al. 2007).

A study similar to the one in rural California was designed to determine the efficacy of a college-based intervention that targeted athletes at 16 of the public colleges in California (Walsh et al. 1999). Permission was sought from participating schools to assess all varsity athletes at a team meeting early in the season to seek their participation. Players completed a questionnaire assessing their tobacco use, and the 16 colleges were matched on the prevalence of smokeless tobacco use within these institutions. The intervention schools had 171 participants, and the control schools (no intervention) had 189. The groups did not differ on demographics, characteristics of tobacco use, or motivation to quit.

The intervention was based on cognitive social learning theory (Bandura 1986). A dentist examined the oral soft tissues of each team member in the intervention schools, advised users to quit, pointed out tissue changes related to smokeless tobacco, showed photographs of facial disfigurement caused by oral cancer, provided a self-help cessation guide, and offered the smokeless users a single 15- to 20-minute session of individual counseling. Players who wanted to quit were offered 2 mg of nicotine gum to mitigate their withdrawal symptoms. Dental hygienists met with nonusers in small groups to discuss the quitting process and encourage them to support the efforts of the users to quit. Two follow-up telephone calls were made to users attempting to quit. On average, the observed self-reported rates of quitting were 34.5% for intervention schools and 15.9% for control schools (RI = 28%; p <.008) at 1-year follow-up. In addition to doubling the rate of quitting, the intervention led to significant reductions in reported use of smokeless tobacco for participants who did not quit. The use of the nicotine gum did not appear to be related to success in quitting.

A more recent study involved the direction by athletic trainers of a smokeless tobacco cessation program for collegiate baseball players (Gansky et al. 2005), who are known to be high users of snuff (Severson et al. 2005). This study involved 52 California colleges (27 intervention colleges and 25 control schools) in a stratified cluster RCT to prevent initiation of smokeless tobacco use and promote its cessation among baseball players. Schools were stratified by tertiles on the basis of their baseline prevalence of smokeless tobacco use. The intervention included videoconference training, newsletters, a screening exam for each player, a self-help guide for quitting, and a counseling session for interested players. Players who expressed an interest in quitting received follow-up support and referral. Student athletes who were peer leaders conducted a single 60-minute educational team meeting that included video and slides. The overall program reduced the initiation of smokeless tobacco use at 1-year follow-up, but there was no effect on cessation. The authors attributed the lack of effects to a small number of dependent users who were enrolled in the study.

In an earlier study, Chakravorty (1992) assigned 83 male users of smokeless tobacco (14–18 years of age, averaging 1.5 dips per day) to one of three conditions in a school setting: use of a nontobacco product composed of crushed mint leaves (mint snuff), use of nicotine chewing gum, or attendance at a lecture-only control condition. More than 90% of study participants were reached at posttest, and 13% of the participants in both intervention conditions were found to have quit using smokeless tobacco (confirmed by biochemical validation) compared with no quitters in the control group (p <.05). No long-term follow-up figures were reported.

A cessation study on smokeless tobacco among younger users (10–14 years of age) was implemented in 4-H clubs throughout California (D’Onofrio et al. 2002). Seventy-two clubs were matched and then assigned to the intervention (tobacco education delivered by volunteers in five successive monthly club meetings) or a no-treatment control condition. At the 1-year follow-up, results from 1,438 club members (77.6% of eligible participants) in the intervention condition revealed significantly improved knowledge regarding the harmful effects of using smokeless tobacco. Seven of the 24 program effects (including knowledge, attitudes, and intentions) were significant at 1-year follow-up; however, no significant differences were seen in use of smokeless tobacco between intervention and control clubs at the 2-year follow-up (Lynch and Bonnie 1994; D’Onofrio et al. 2002).

In another study, Stotts and associates (2003) examined whether adolescent users of smokeless tobacco (14–19 years of age) were aided in their cessation attempts by using nicotine patches and receiving several follow-up telephone counseling sessions. Three hundred students were assigned to one of three conditions: (1) counseling only (6 weeks of 50-minute age-relevant behavioral intervention classes based on NCI materials); (2) counseling plus an active nicotine patch and telephone support; and (3) counseling plus a placebo patch and telephone calls. Following completion of the class, students who were enrolled in the counseling-only condition were contacted at 2-week and 1-year assessment points, and participants in the two groups that received a patch (active or placebo) plus telephone support received seven 15-minute telephone calls that included “stage-based counseling” and a $5 gift certificate. Analysis of the 1-year follow-up indicated no differences between the two groups receiving a patch and telephone calls, but these conditions combined were more successful in encouraging cessation of smokeless tobacco (32.8%) than was the counseling-only condition (22.9%) (RI = 14.7%). This was a highly intensive intervention, however, and it is not clear whether the telephone calls or the patch (nicotine or placebo) produced the significant effect. The lack of effects for nicotine replacement (vs. placebo) is consistent with studies evaluating the efficacy of nicotine replacement for cessation of smokeless tobacco use among adults.

Burton and colleagues (2009) reported on a school-based study that compared two models of cessation for both smokers and users of smokeless tobacco in high schools. Students were randomly assigned to one of three groups: an addiction group, a psychosocial dependency group (both were treatment groups), or a control group. Sixteen schools in California and Illinois participated, with two treatment groups per school. Each of the 32 groups met for five sessions spaced over 1 month, with follow-up completed 4 months after the end of treatment. The majority of participants were smokers, but 8% of California’s participants and 17.3% of Illinois’ participants used smokeless tobacco only, and an additional 8% and 9% of participants, respectively, reported both smoking and current use of smokeless tobacco. The treatment groups shared some components of the intervention, and the sessions were divided between presentation of information and group discussion. Video clips were used to elicit discussion, and users of smokeless tobacco were encouraged to use oral substitutes. All participants received incentives for participation and attendance. On the basis of an ITT analysis and according to both verbal reports and biochemical verification of these self-reports, the smokeless tobacco users were more likely to be abstinent from tobacco use at the 4-month follow-up than were smokers. The validated rate of quitting at the 4-month follow-up was 14.3% for smokeless tobacco users, while the control group had no one reporting abstinence (RI = 14.3%).

Consistent with the studies discussed above, a Cochrane review of smokeless tobacco cessation by Ebbert and colleagues (2007) concluded that pharmacotherapy has not been shown to affect long-term abstinence in young adults and adults.

Young people are using computers, smartphones, and the Internet with increasing frequency, and these channels might provide a unique opportunity to engage youth in quitting. Fisher and colleagues (2001) reported on an interactive, computer-mediated intervention designed to help adolescents quit smokeless tobacco. This small pilot study was conducted with 50 high school students who used the cessation program Chewer’s Choice; the study used a baseball field as an interface, which appealed to the mostly male users. Participants were given brief instructions before using the program on their own. The authors reported that 85% of the users had made an attempt to quit, and at the 6-week follow-up, 58% reported having quit all tobacco for at least 24 hours. Neither biochemical verification of self-report nor long-term follow-up was included.

Gala and colleagues (2008) reported on a pilot study in which an Internet-based program on cessation of smokeless tobacco use was evaluated using 17 baseball athletes attending California colleges. The interactive Web site appeared to be feasible, was acceptable to users, and resulted in a 26% self-reported reduction in use of smokeless tobacco at 1-month follow-up, but only one subject reported abstinence at this point.

The use of the Internet to deliver a cessation program to young users is being more fully evaluated in a current study supported by NCI; this randomized clinical trial involves the evaluation of a Web-based cessation program (My Last Dip 2010) offered to young users of smokeless tobacco between the ages of 15 and 24 years (Severson and Danaher 2009). The study will evaluate the efficacy of two Web sites designed for this population of young chewers. One Web condition will provide a text-based site designed to offer a proven cessation program as well as information and resources on quitting smokeless tobacco; the other site will offer a tailored and more interactive site that provides video and other engaging activities in addition to the opportunity to post to blogs. One unique feature of this study is that no parental consent is required to participate; previous research has shown that requiring consent from parents can be a significant deterrent to enrolling young people in cessation or prevention studies (Severson and Ary 1983; Severson and Biglan 1989; Gala et al. 2008).

Although no data are yet available on the efficacy of this program, a previous study with adult users of smokeless tobacco demonstrated the efficacy of providing cessation support through the Internet. That study compared an interactive, tailored, Web-based intervention (enhanced condition) with a more linear text-based Web site (basic condition) in a randomized trial with 2,523 adult users (Severson et al. 2008). The point prevalence of all tobacco use (smoking and smokeless use) at 3 months, 6 months, and both 3 and 6 months was 48%, 45%, and 34%, respectively. The researchers found that participants in the enhanced condition quit at significantly higher rates than those in the basic condition. The intent-to-treat analysis indicated quit rates of 12.6% among those in the enhanced condition and 7.9% for those in the basic condition (p <0.001). With the use of complete case analysis, including those with data at all time points, it was found that abstinence was 41% in the enhanced condition and 21% in the basic condition (p <0.001). Program use was significantly related to the outcomes as well as to attrition. The authors concluded that a tailored, interactive, Web-based cessation program may be a promising method of helping to stop the use of smokeless tobacco. It remains to be seen whether these encouraging results can be replicated with a younger population of users, but given the high use of the Internet by young people and the reach of such a program, a program designed specifically for young users could provide a low-cost alternative for promoting cessation.

Discussion Regarding Cessation of Smokeless Tobacco Use

Although many studies have been conducted on smoking cessation for youth, few have focused on smokeless tobacco in this age group. The relative lack of research on smokeless tobacco may be due to the far lower overall prevalence of using this product (vs. cigarette smoking), particularly in females. In addition, the use of chewing tobacco and snuff varies significantly by region and is viewed as a behavior confined mostly to rural and small-town areas in some parts of the country.

Most of the interventions for smokeless tobacco cessation have been based on multicomponent cognitive-behavioral interventions used in smoking cessation (Severson and Hatsukami 1999). Although the basic elements of these interventions apply equally well to smokeless tobacco, cessation of smokeless tobacco use has some unique aspects. The most obvious is the opportunity presented by oral exams to both motivate users to quit and to show them the direct effects of regular use of smokeless tobacco products. Not all users will have observable oral lesions, but it has been reported that 73% of snuff users will have identifiable oral lesions within 3 years of regular use. The lesions’ severity and ratings are directly related to the amount of tobacco used weekly and the number of years of use (Little et al. 1992). The use of oral exams has been a key element of several interventions described above and, for this reason, it has been recommended that a dentist or dental hygienist be part of the intervention team. Other modifications of the interventions focused on smokeless tobacco users involve modified measures used for assessing dependence and use (Hatsukami and Severson 1999).

There is currently a need for innovative, validated, easily delivered, and low-cost interventions to facilitate cessation in smokeless tobacco users, an underserved population. The Internet and interactive computer-based cessation may offer channels of intervention that are particularly attractive to young users, but the data on the efficacy of these interventions are limited.

Although the literature is not extensive, the outcomes of several well-controlled studies suggest that young users can be effectively helped to quit smokeless tobacco. The focus on male athletes who use smokeless tobacco is encouraging, but studies are lacking that target other high-risk or high-use groups, including Alaska Natives, American Indians, and athletes who are involved in rodeo. The prevalence of smokeless tobacco use is very high in these groups, and specialized interventions may be needed to help them to quit.