Results

GLOBAL RATING OF MESSAGES

Subjects generally found all three messages believable (rated at 4.2 (SE 0.02) on a five point scale ranging from “not believable at all” to “very believable”). However, smokers found the Feel (4.1 (0.04)) and Vents (4.1 (0.05)) messages significantly less believable than the Health message (4.5 (0.04); p < 0.001). This was equally true for all types of smokers. There was no interaction between type of smoker and message group.

BELIEFS ABOUT LIGHT AND ULTRA LIGHT CIGARETTES

We examined change in smokers' beliefs about L/ULs after exposure to the messages, by message group. (It should be noted that Light cigarette smokers assessed beliefs about Lights, and Ultra Light smokers assessed beliefs about Ultra Lights, so beliefs were assessed about different cigarettes depending on the type of smoker.) On average, exposure to all the messages reduced beliefs that L/ULs yield lower tar and nicotine delivery, as evidenced by an overall mean (SE) decrease of −0.49 (0.02) in stated beliefs (p < 0.0001 for the hypothesis of no change). More importantly, beliefs were affected differentially by each message (fig 1). Those hearing the messages that addressed L/ULs—the Feel and Vents messages—demonstrated significantly more movement toward skepticism than those hearing the generic Health message (table 2). Further, the Feel message was significantly more effective than the Vents message in promoting movement away from the belief that L/ULs have diminished delivery. Regular, Light, and Ultra Light smokers reacted similarly to the different messages (interaction p = 0.56).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Changes in beliefs about Light and Ultra Light cigarettes from pre- to post-message.

The messages also significantly and differentially affected beliefs about the safety of L/UL cigarettes. Smokers who heard the Feel message showed significantly greater movement away from believing in the safety of L/ULs compared to those who heard the Health message (fig 1, table2). Additionally, smokers exposed to the Feel message showed significantly more movement toward scepticism than those exposed to the Vents message. Not surprisingly, collapsing across all three messages, a greater change in safety beliefs was seen for Light and Ultra Light smokers than for Regular smokers (contrasts among smoker types, p = 0.006, respectively). However, there were no differential effects of the three messages by type of smoker (p = 0.15).

Participants also assessed the perceived risks of Lights and Ultra Lights, respectively, rating their risk compared to Regular cigarettes, on the one hand, and to non-smoking, on the other. Following exposure to the messages, the greatest change in the perceived risk of Light cigarettes was produced by the Feel message; its impact was significantly greater than either the vents or Health message (table2). The impact of the Vents message on perceived risk was not significantly different than that of the Health message. There was no difference in perceived risk of Light cigarettes (p = 0.92) by type of smoker. As shown in table 2, the results for Ultra Light cigarettes were exactly parallel. The Feel message produced greater impact than either the Vents or Health messages. The impact of the Vents message was equal to that of the Health message. There was no effect by type of smoker (p = 0.59).

INTEREST IN QUITTING

The effect of message exposure on interest in quitting was assessed by analysing pre-post changes in quitting interest and intention, as captured by the “quit index”. Exposure to the messages resulted in a modest but significant overall increase in quitting interest (0.04 (0.02), p = 0.04). As shown in table 2, the Feel message produced the greatest increase in average quit index scores, significantly higher than either the Vents or Health messages, each of which actually resulted in lower quit index scores (fig 2). The impact of the Vents message on interest in quitting was not significantly different than that of the Health message.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Changes in “quit index” from pre- to post-message.

We also tested each message to assess whether the observed shifts following the message exposures were reliable (that is, we tested against the hypothesis of “no change”, or the hypothesis that the change score was = 0). Only the Feel message was found to produce significant movement towards quitting (testv no change, p < 0.0001); there was no significant change in the groups that heard the Vents and Health messages. Analysis by individual change rather than by group mean change revealed similar findings. Smokers exposed to the Feel message showed significant net positive change: 15.0% shifted towards greater quit interest, while 8.2% regressed; 2.6% increased their “quit index” by 3 or more. In contrast, neither the Vents nor the Health message produced significant positive gain. Of those who heard the Vents message, 9.6% increased their quit interest (0.4% by 3 points or more), but 11.0% regressed. Among those who heard the Health message, 8.3% increased their quit interest (0.7% by 3 points or more), and 6.9% regressed. The net positive movement (distribution by positive, versus negative or no change) was significantly greater in the Feel group than in the other two groups (χ² = 17.4, p < 0.002).

Collapsing across all messages, Ultra Light smokers showed the largest increase (0.19 (0.06)) in quit interest, significantly greater than both the Light smokers (0.02 (0.02)) and Regular smokers (0.01 (0.03); p = 0.01 for each comparison); Regular and Light smokers showed similar overall change (p = 0.76). However, the differential effects of specific messages were not related to smoker type—that is, there was no smoker type by message interaction (p = 0.28).

MODERATORS OF THE MESSAGE EFFECTS

We examined relevant demographic and smoking history variables to explore whether the messages might have differential effects on quitting interest among different subgroups of smokers. Means and standard errors for the change in “quit index” scores by various moderators are shown in table 3. We compared the effects on young adults (ages 18–25 years) versus others: age moderated the effect of the messages (p < 0.01). The superiority of the Feel message was most pronounced among young adults (fig 3). Number of years smoking demonstrated effects similar to those of age (p < 0.025; not shown). There was a trend for sex to also moderate the message effect (p = 0.08); women who heard the Feel message had the greatest increase in quit interest. Analysis of ethnic differences was complicated because the sample was predominately Caucasian, precluding a detailed breakout of non-Caucasian subgroups. We assessed ethnic differences by comparing Caucasians to non-Caucasians. The largest increase in quit interest was noted among non-Caucasians who heard the Feel message, but the interaction between ethnicity and message was not significant (p = 0.54).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Changes in “quit index” from pre- to post-message by message and age.

We also assessed the effects of the messages at different levels of baseline quitting interest (after removing those initially committed to quitting within 30 days (“quit index” = 6), as they could not increase their “quit index”). Because of ceiling effects and regression to the mean, we expected that the most positive change would occur among smokers who were initially less interested in quitting at baseline, and this was confirmed: more change was evident among those initially less interested in quitting (“quit index” < 4; 0.22v −0.08; p < 0.0001). More importantly, there was a baseline level by message interaction (p < 0.005). As table 3 shows, the superiority of the Feel message was most evident among those initially less interested in quitting (“quit index” < 4).

Theoretically relevant baseline characteristics were also examined as potential moderators. In particular, we examined how the messages interacted with the beliefs they were meant to address. This addresses the specificity of the messages and the fidelity of the message strategy. The effects of the Feel message (which addressed the effects of the sensory experience of smoking L/ULs) were expected to be strongest among those who, at baseline, believed that L/ULs produced less harsh sensations. Indeed, this was the case. There were differential effects of the message by baseline beliefs (p < 0.03). Only among those who initially believed that L/ULs felt lighter did the Feel message have significant impact on interest in quitting (p < 0.0001); the effect was not significant among those who did not endorse these beliefs (p = 0.64). We also assessed the differential impact of the messages according to whether respondents who smoked L/UL cigarettes cited lighter sensations as a reason for doing so (based on an open ended question). The largest increase in quitting interest was noted among those who heard the Feel message and cited sensory factors as a reason for smoking L/UL brands, however the effect was not quite significant (p = 0.11).

Since the Vents message was designed primarily to address beliefs about the delivery characteristics of L/UL cigarettes, we assessed whether baseline delivery beliefs moderated the effect of the messages. There was no interaction between baseline beliefs about the delivery of L/UL brands and the message effects (p = 0.58).

Discussion

This randomised experimental study demonstrated that smokers' beliefs about L/ULs and their quitting intentions could be positively influenced by Health messages, and that different message strategies had differential effects. Two different messages addressing beliefs about L/ULs were compared—one emphasising how the sensory experiences of L/ULs can be deceptive (Feel), and one emphasising how cigarette ventilation can undermine nominal reductions in yield (Vents). The Feel message yielded consistently better results in changing both beliefs and quitting intentions. Similarly, the Feel message resulted in greater changes in beliefs and increased interest in quitting compared to the message about the general hazards of smoking (Health).

The differential effect of the Feel message may be due to its strategy of directly and empathetically addressing smokers' sensory experience of the mildness of L/ULs, which might otherwise be a barrier to accepting abstract scientific data that these cigarettes are essentially equivalent to Regulars. Results from a survey of L/UL smokers in the USA1 showed that L/UL smokers endorse the sensory benefits of L/ULs (more so than abstract health benefits). Indeed, the superiority of the Feel message was evident primarily among smokers who initially stated that L/ULs produced less harsh sensations. The efficacy of this message is consistent with research on symptom perception and common sense models of illness, which suggest that people often apply common sense models to interpret bodily experiences and draw conclusions—often erroneous conclusions—about health threats and appropriate health protective behaviour (that is, if it feels milder, it must be safer (see Brownlee et al 7). Thus, smokers regard the less harsh sensations of smoking L/ULs as concrete, personal confirmation of their safety, reinforcing their decision to smoke L/ULs as a way of protecting their health. This cognitive process may make it hard to displace this belief unless the sensory experiences and associated beliefs are addressed. These sensory phenomena and the cognitive processes associated with them appear to be important influences on smoking behaviour among L/UL smokers, and should be addressed in targeted health messages.

The effect of the Feel message was robust and general, affecting smokers of Regular, Light, and Ultra Light cigarettes. The impact of this message on Regular smokers is perhaps a bit surprising, as it does not speak to their current cigarette brand. However, the Feel message may have addressed beliefs held by smokers of Regulars even about their own cigarette brands, or it may have spoken to prior brief experiences of trying a L/UL cigarette. The Feel message was also notably effective among young smokers, women, and minorities, all of whom are important targets for tobacco control. The broad appeal of the Feel message, and its appeal to special target groups, suggests its broad utility as a media intervention.

In contrast to the Feel message, which addressed smokers' sensory experience, the Vents message may have been too complex (relying on a technical explanation of how L/ULs are constructed) or may have been seen as blaming the smoker for blocking vents (despite the attempt to avoid doing so). Although the Vents message actually provided more objective information, the information appeared to have been psychologically less effective in promoting changes in beliefs, attitudes, and intentions.

As expected, the effect of a single exposure to a Health message appeared modest; it did not convert large numbers of smokers to immediate commitment to quit. However, the effects were meaningful. Small changes in stated intention to quit have substantial impact on subsequent behaviour. Among subjects who heard the Feel message, 5.5% of those who were initially thinking of quitting in the next six months, subsequently indicated that they would plan to quit within a month. Among those not initially interested in quitting in the next six months 5.8% advanced to that stage. In prior research,18advances in quit intention of this magnitude were associated with a doubling or tripling in the likelihood of making a quit attempt in the following month. Over a large population of smokers, effects of this magnitude could be significant.

However, we would not claim that the message exposures tested in the study themselves had such enduring effects. The experimental delivery of the messages lacked almost all of the essential elements of a good information campaign19: it consisted of a single, very brief exposure to a text-only message, read over the phone by an unvalidated source (that is, an unknown interviewer). However, the fact that exposure under even these untoward conditions had an effect suggests that the effects could be quite substantial under more optimal conditions—repeated exposure to well crafted messages from authoritative sources. It is well established that repeated exposure to advertising messages has cumulative effects and dramatically enhances their impact on behaviour.20 21 Research has also shown that elements of execution (for example, visual elements, music, authority) add to a message's impact on behaviour.19 This suggests that a campaign of well crafted media advertisements carrying these messages and/or a campaign of repeated counselling along these lines by health care providers could substantially increase interest in quitting smoking, which would yield significant public health benefit.

The study's methodological strengths and limitations moderate interpretation of our findings. The study was based on a large representative sample of US smokers, and generalisation to populations in other parts of the world remains to be tested. A strength of the study was its combination of population survey methods (ensuring representativeness) with a randomised experimental design (ensuring clear assessment of the message effects). Further, the messages tested were developed based on qualitative and quantitative research. A limitation of the study is that it did not assess actual quitting behaviour, but focused on self reported intentions. However, the quit intention assessments incorporated into the “quit index” have been shown to predict subsequent behaviour, including participation in smoking interventions, making quit attempts, and actual successful cessation, as long as two years after an initial assessment.13 18 22 Further research should follow smokers to observe whether their changes in beliefs persist and/or lead to actual changes in behaviour following exposure to Health messages.

The study did not include a control group that was not exposed to any messages. This would have helped assess the effect of simply repeating the assessments without any intervening intervention. However, our focus was on contrasting three different interventions; since all groups were subject to repeated pre-post assessment, this does not confound the results or conclusions. Also, it was notable that in many instances the Health message, which reflects standard practice in warnings, produced no change at all, suggesting little reactivity. It is possible that respondents, having given a particular answer on the first administration of a question, may have felt constrained to give a similar answer when asked the same question again only minutes later. In that sense, our estimates of the amount of change caused by the intervention are likely to be conservative.

In interpreting differential response to the messages, we must consider that the messages were not identically matched in other respects. The Health message was far less novel, both in concept (smoking is harmful) and in wording, as it was adapted from warnings that appear in all cigarette advertising. This is realistic, at least in developed countries—the message that smoking is harmful is timeworn, and newer and more focused messages may have more impact. (It is possible that a novel message on health consequences would have done better.) The Health message was also slightly shorter than the others, and the Feel message was longer than the Vents message. The messages also differed in some secondary content; the Vents message addressed the intentions and actions of the tobacco industry, while the Feel message only mentioned the industry in the context of empirical support for full delivery from L/UL brands. The Feel message mentioned gene damage as a consequence of smoking while the others did not. Thus, the thematic message strategies were not completely isolated for study in the messages, and other components may have added variance to the results. It is possible that some combination of message content and novelty influenced the messages' effectiveness. However, the strong specificity of the findings—notably that the superiority of the Feel message was concentrated among those who initially believed L/ULs were less harsh—strongly suggests that the intended thematic elements were responsible for the observed effects.

The messages were also limited to very brief text passages, and almost certainly understate the impact of well developed media messages or physician communications. The impact of any actual implementation will depend substantially on how the messages are crafted and delivered. Ideally, actual advertisements should be tested for effects on attitudes, beliefs, and intentions before being implemented.

The results of the study have implications for several aspects of tobacco control. Clinically, it suggests that health care providers should determine which patients smoke Light or Ultra Light cigarettes, and find opportunities to dispel misconceptions about L/ULs and to move smokers towards quitting smoking. For public health communication, the testing of alternative message approaches is an important foundation for health promotion campaigns, to ensure that the media messages have the desired impact on beliefs and behaviour. Mass media messages that counter the appeal of L/UL brands (and emerging novel cigarette designs claiming reduced risk) may be an important part of a tobacco control campaign. Since the majority of US and European smokers currently smoke L/ULs, this message could reasonably form the nucleus of a public health campaign against smoking. The fact that exposure to messages debunking beliefs about L/ULs yielded increased interest in quitting helps to validate the idea that smokers' adoption of L/UL cigarettes, and their beliefs that they protect them from risk, deters movement towards quitting. It was notable that a message that recounted the health harms from smoking in general (the Health message) did not substantially increase interest in quitting, whereas one that effectively debunked beliefs about L/ULs did. This research showed that a message that addressed smokers' sensory experience with L/ULs—the fact that they feel less harsh or strong and are less likely to acutely elicit coughing—is likely to have the most impact on smokers' beliefs, attitudes, and quitting intentions, and forms a promising basis for a public information campaign.