program is effective for preventing use among 1 percent of the remaining population. Should the results indicate that the program was effective for preventing heroin use among 1 percent of the population, or among 50 percent of those at risk for using heroin?
This illustrates the point that the small effects observed in studies of prevention programs may mask larger effects—both negative and positive—for different user groups. If this is true, it is important to understand the heterogeneity in the population and target prevention activities accordingly. However, only scant evidence is available on differential effects for different user groups (e.g., nonusers, casual or experimental users, users).
Reviews and meta-analyses of prevention effectiveness fail to differentiate among programs that target at-risk and universal populations. Relatively few effectiveness studies conducted on at-risk populations have reported effects on substance use outcomes. In an ongoing meta-analysis of school-based prevention (Gottfredson et al., forthcoming), only 7 of the 88 studies for which effect sizes could be computed targeted populations that were at elevated risk for developing problem use.
A handful of studies have compared the effectiveness of universal prevention activities for groups that differed according to their level of use at baseline. In one of these studies (Hansen et al., 1988), the researchers compared students who had received a 12-session resistance skills program with a control group. Results were reported separately for students who reported no marijuana use at baseline and for the entire population. Statistically significant program effects were found only for baseline nonusers of marijuana. When these students were combined with those who had already initiated marijuana use at baseline, no effects were found.
Studies of the ALERT program provide another example of differential effectiveness for groups differing in level of baseline use (Bell et al., 1993; Ellickson and Bell, 1990, Ellickson et al., 1993). ALERT is a universal social resistance-skill curriculum consisting of eight lessons taught a week apart in the 7th grade, followed by three 8th grade booster lessons. The researchers reported the results from this program separately for baseline nonusers, baseline “experimenters,” and baseline “users,” and also separately for cigarette use, alcohol use, and marijuana use and by follow-up period.
The program’s most consistent effects were found for marijuana use. Statistically it significantly reduced the use of marijuana among students at each risk level, but the strongest effects were for the lowest risk group: those students who had not initiated either cigarette or marijuana use at the time of the baseline measurement. For all groups, small positive pro-