Tobacco Documents Online

NOT FOR REPRODUCTION, QUOTATION, OR CITATION WITHOUT AUTHOR'S PERMISSION. 13 Moskowitz (1985) has stated that it is premature to conclude that these programs are effective in preventing cigarette smoking. He points out that the studies have yet to demonstrate effects on the more established'habitual smoking that generally does not occur until the high school years. A fourth generation of studies consisted of large scale trials typically involving five or more units per condition. All of these evaluations were based on the social influences approach. Botvin, Renick, and Baker (1983) have asserted that psychosocial smoking prevention strategies are capable of producing initial reductions in smoking behaviors of about 50% and that longer-term~results look promisimg.F1ay (1985b) concludes that the data:from these studies suggest that the social influences approach to smoking prevention can be effective some of the time. He warns, however, that this conclusion is a tentative one because of the inconsistencies in the patterns of results. The importance of viewing some of these results as tentative is emphasized in a review of one of the more successful intervention programs, the Waterloo Study (Best et al., 1984, Flay et al., 1985). Flay (1985b) concludes that the results of the Waterloo study represent one of the most rigorous tests of the social influences approach to smoking prevention. Flay interprets these results as suggesting reasonably good'.maintenance of long term effects and notes the importance of the findimg that the program was most effective for students at high risk. The results from the Waterloo study do indeed appear encouraging. Among students in the experimental program who were not smoking at the beginning of the study, 60% were still non-smokers at the end of the eighth grade, whereas the comparable figure for the control group was 47%. Among students who said they had tried smoking at the beginning of the study, 43% of the students in

NOT FOR REPRODUCTION, QUOTATION, OR CITATION WITHOUT AUTHOR'S PERMISSION 14 the experimental program quit, as opposed to 25% of the comparison group--a statistically significant difference. Even more impressive was the finding that among students classified as being at high social risk, 77.8% of the students in the experimental group remained never smokers, while the figure was only 44.48 for those in the control group. These results appear to suggest that focusing on social influence processes in order to reduce the onset of smoking can be efficacious. However, from a public health perspective, it is important to consider not only the relative differences, but also the total number of students affected and the robustness of the effects. For example, only 44% of the students in the Waterloo study were non-smokers at the beginning of the study and 33% were regular smokers. Partly as a result of this stratification, only one of the 17 contrasts between the experimental and control groups at the end of grade eight was statistically significant.. Similarly, there were only 36 students in the high social risk group who were non-smokers at the beginning of the study. Thus, the difference of 78% versus 44% reported by Best et al. (1984) reflects the fact that 14 smokers in the experimental versus 8 in the control group remained non-smokers--a difference of 6 students. As Flay et al. (1985) assert, these results are "fragile." This is especially true if one takes into account how unstable these reported patterns at the end of gzade eight are likely to be. Most program evaluations conducted to date have been concerned primarily with the efficacy of specific programs. With apologies to McGuire, one might say that the research has focused on finding the best vaccine for inoculating students. In conducting an efficacy trial, it is appropriate to focus on persons at risk. Before preventive measures are implemented, however, it is necessary to ask what the impact on the entire population wi11 be. In the

NOT FOR REPRODUCTION, QUOTATION„ OR CITATION WITHOUT AUTHOR''S'PERMISSION 15 case of adolescent smoking, this means that instead of asking how many non- smokers are prevented from smoking in a six month period of a specific trial,, we should try and estimate what proportion of all adolescents will be prevented from smoking. This is a complex question that cannot be answered definitively with existing data. However, in order to at least sensitize policy makers to this issue we have calculated two statistics for a number of representative intervention studies. The statistic that may be of most interest to public health officials is what we refer to as Attributable Prevention in the Population. This simply refers to the proportion of all students affected by the program. For example, if two schools with 100 students each are assigned to an experimental and control program respectively, and the proportion of new smokers in the experimental school is 10% and the proportion of new smokers in the other is 5%, we say that the Attributable Prevention in the Population is 5%. Therefore, the upper limit of Attributable Prevention in the Population is the prevalence of the behavior. Experimental and control groups often~exhibit different behavior before the start of a study and both groups usually change behavior over the course of time. The methods available for adjusting for such differences are varied and sophisticated (Cleary, 1983) but for simplicity we simply subtract base rates from rates at the end of the study. For example, if in an experimental program~the rate of smoking increased from 4% to 10% whereas the control group increased from 3$' to 5%, we would say the Attributable Prevention in the Population was 4% (10-4 versus 5-3). This is not a statistically correct way of adjusting for group differences„ but in the absence of the original data, it provides a rough way of comparing studies. Data are sometimes reported only for students "at risk." For example, if 70% of both schools already have experimented with smoking, one might

NOT FOR REPRODUCTION, QUOTATION, OR CITATIWWITHOUT AUTHOR'S PERMISSION 16 analyze data for only non-smokers. In the example given above, if only 30% of both schools were used for analyses, the results would indicate that about 33% of the,non-smokers in the experimental school were prevented from~smoking and about 17% were prevented from smoking in the control group--a difference of 16%. To facilitate comparisons among studies, if data are presented only for a "risk group" we calculate what the Attributable Prevention in the Population would be. These types of calculations are a start towards answering the question of what proportion of students would be prevented from smoking for the period of the study if the intervention was applied to that type of population. These figures, along with some description of the studies are presented in Table 1. These studies were drawn from the list compiled and reviewed by Flay (1985a, Table 1).