Effectiveness of a Universal Drug Abuse Prevention Approach for Youth at High Risk for Substance Use Initiation 1

Transcription

1 Preventive Medicine 36, 1 7 (2003) doi: /pmed Effectiveness of a Universal Drug Abuse Prevention Approach for Youth at High Risk for Substance Use Initiation 1 Kenneth W. Griffin, Ph.D., M.P.H., 2 Gilbert J. Botvin, Ph.D., Tracy R. Nichols, Ph.D., and Margaret M. Doyle, M.P.H. Institute for Prevention Research, Weill Medical College, Cornell University, New York, New York Background. Universal school-based prevention programs for alcohol, tobacco, and other drug use are typically designed for all students within a particular school setting. However, it is unclear whether such broad-based programs are effective for youth at high risk for substance use initiation. Method. The effectiveness of a universal drug abuse preventive intervention was examined among youth from 29 inner-city middle schools participating in a randomized, controlled prevention trial. A subsample of youth (21% of full sample) was identified as being at high risk for substance use initiation based on exposure to substance-using peers and poor academic performance in school. The prevention program taught drug refusal skills, antidrug norms, personal selfmanagement skills, and general social skills. Results. Findings indicated that youth at high risk who received the program (n 426) reported less smoking, drinking, inhalant use, and polydrug use at the one-year follow-up assessment compared to youth at high risk in the control condition that did not receive the intervention (n 332). Results indicate that a universal drug abuse prevention program is effective for minority, economically disadvantaged, innercity youth who are at higher than average risk for substance use initiation. Conclusions. Findings suggest that universal prevention programs can be effective for a range of youth along a continuum of risk American Health Foundation and Elsevier Science (USA) Key Words: drug abuse; prevention; adolescents; minority; middle school. 1 This research was supported by funds from the National Institute for Drug Abuse (P50DA7656). 2 To whom reprint requests should be addressed at Institute for Prevention Research, Weill Medical College, Cornell University, 411 East 69th Street, New York, NY Fax: kgriffin@med.cornell.edu. INTRODUCTION Research concerning the onset and developmental progression of adolescent substance use indicates that most youth initiate use by experimenting with alcohol and cigarette smoking during early adolescence [1 3]. Such early experimentation can lead to later heavy use of these substances. A recent longitudinal study found that those who had experimented with cigarette smoking during early adolescence, including having tried only a few puffs, were three times as likely to be pack-a-day smokers as high school seniors compared to adolescents who did not experiment with tobacco during early adolescence [3]. A related concern is that early onset of smoking and drinking may contribute to more serious subsequent involvement in illicit drug use. According to the gateway hypothesis of adolescent drug use [4], youth who use cigarettes and alcohol are more likely to experiment with marijuana, and those who use marijuana are more likely to progress to the use of depressants, stimulants, hallucinogens, narcotics, and other dependency-producing drugs. Furthermore, empirical evidence has accumulated showing that the early initiation of substance use is associated with a variety of additional negative outcomes in later adolescence and early adulthood such as violent and delinquent behavior, poor physical health, and mental health problems [5,6]. Because this general pattern of substance use initiation and escalation is well documented, many prevention programs for adolescent drug abuse aim to prevent early-stage substance use or at least delay the initiation or onset of use among youth. Many of these programs are provided to middle school or junior high school students because this is when many youth begin to experiment with substances. These programs typically target the use of tobacco, alcohol, and marijuana because these are the most widely used substances in our society and because preventing the use of these gateway substances may reduce the risk for later negative outcomes. Such broad-based (universal) preven /03 $ American Health Foundation and Elsevier Science (USA) All rights reserved.

2 2 GRIFFIN ET AL. tion programs are typically designed for all youth in a particular school setting regardless of level of risk. However, a criticism of school-based drug abuse prevention is that any single program may not meet the needs of all youth in a particular school setting, particularly if there is a large spectrum of risk among students [7]. Accordingly, it is argued, universal prevention programs may be ineffective for youth at high risk for engaging in substance use. Instead, high-risk youth may require selective interventions that target at-risk youth who show signs of potential drug involvement, or indicated interventions that target youth already engaging in drug use. An issue that has not been adequately addressed in the prevention literature is whether effective universal prevention programs can be effective for those further along the continuum of risk, such as young people with one or more risk factors for substance use initiation. Research on the etiology of adolescent substance use has shown that several factors are associated with increased risk for early onset or escalation of use during adolescence [8,9]. Peer social influences are among the most important risk factors, with a large literature demonstrating that teens that have peers that engage in substance use are more likely to engage in use themselves [10 12]. The effects of peer influences on substance use can take a variety of forms including increased availability and direct modeling of substance use behavior, peer pressure, encouragement, and social reinforcement of substance use, enhanced perceptions regarding the social acceptability of use, and increased normative expectations regarding the prevalence of adolescent substance use [13,14]. These social influence processes are likely to be even more powerful when a young person s direct group of friends are involved rather than one s peers in general. A second important risk factor for youth involvement in substance use is poor school performance. Youth who perform poorly or fail in school are more likely to engage in a number of problem behaviors compared to academically successful youth, including alcohol, tobacco, and other substance use [15,16]. Although the mechanisms by which poor school performance is associated with early onset of substance use are unclear, one possibility is that youth who cannot succeed in school engage in alternative behaviors in an attempt to establish an identity, enhance the way others view them, or enhance the way they view themselves. While some of these youth may turn to activities that are sanctioned by adults (e.g., sports or music), others may engage in behaviors that are considered problematic by adults such as substance use. This proposition is consistent with problem behavior theory [17], which states that adolescents engage in negative behaviors in order to achieve developmental goals that they believe they cannot achieve in more adaptive ways. The goal of the present study was to examine the effectiveness of a universal drug abuse prevention approach for youth at high risk for early onset of substance use based on their friends use of alcohol and tobacco and poor grades in school. METHODS Sample The data for the present study were collected as part of a larger school-based randomized drug abuse prevention trial for inner-city adolescents [18]. Students from the longitudinal sample were identified as high social risk if they had friends that smoke cigarettes and drink alcohol, and at high academic risk if they reported poor academic achievement in school (i.e., grades of C or less). About 21% (n 758) of the original sample was classified as high risk for substance use initiation based on social and academic risk, and this subsample is the focus of the present analyses. This final sample of high-risk students was 49% male and 51% female, and was predominantly African- American (58%) and Hispanic (29%). The sample included a large proportion of economically disadvantaged youth as shown by the fact that 61% of students received free lunch at school. More than one-third (35%) of students lived in mother-only households. Procedure Prior to randomization, schools were surveyed and divided into high, medium, or low smoking prevalence, and were randomized into the experimental or control conditions from within these groups. Students in the experimental condition received a drug abuse prevention program consisting of a primary year of intervention in the seventh grade and a booster intervention in the eighth grade. Regular classroom teachers administered all intervention sessions. Students in the control group received the substance use curriculum normally in place in New York City schools. Additional information on the research design and methods can be found elsewhere [18]. Prevention Program The preventive intervention, called Life Skills Training (LST), teaches drug resistance skills, norms against substance use, and material designed to facilitate the development of important personal and social skills. The LST program aims to provide young people with the knowledge and skills needed to resist social influences to engage in substance use. The program also aims to increase general personal and social competence skills in order to reduce potential motivations to use alcohol, tobacco, and other drugs. The intervention teaches students a variety of cognitive behavioral skills for building self-esteem, resisting advertising

3 DRUG PREVENTION AMONG YOUTH AT RISK 3 pressure, managing anxiety, communicating effectively, developing personal relationships, and asserting one s rights. These are taught using specific skillstraining techniques that include group discussion, demonstration, modeling, behavioral rehearsal, feedback and reinforcement, and behavioral homework assignments for out-of-class practice. Additional information on the goals and methods of the prevention program can be found elsewhere [18]. Previous evaluation research has demonstrated that the prevention program is effective among suburban, White youth [19], with prevention effects lasting until the end of high school [20,21]. Research has also shown that this approach is effective with inner-city minority youth [18,22,23]. The present study examines a subset of youth from a recent large randomized prevention trial among inner-city middle schools [18]. Students from this larger study were selected to be included in the present analysis if they were at high social and academic risk as defined above. The present study is the first to specifically examine the effectiveness of this prevention approach on youth at high risk for substance use initiation. Measures Data were collected following a detailed protocol approved by Cornell Medical College s Institutional Review Board. Students completed a survey that included items assessing current alcohol and drug use and a series of scales measuring cognitive, attitudinal, and skills variables believed to be associated with the initiation of alcohol and drug use. Although all measures were self-reported, data collectors emphasized the confidential nature of the data being collected in order to enhance the truthfulness of student responses. Carbon monoxide breath samples were collected simultaneously with the questionnaire data to enhance the validity of the self-report data. A number of additional psychosocial constructs were assessed in the survey; however, we limited the analysis of program effects to substance use outcomes in the present study. Demographics. Several standard survey items assessed gender, age, family structure, race and ethnicity, and socioeconomic status (receive free or reduced school lunch). Risk factors for substance use. Social risk was assessed by asking participants how many of your friends do you think smoke cigarettes? and how many of your friends do you think drink beer, wine, or liquor? Response options for these items ranged from 1 (none) to 5 (all or almost all). Those who responded at baseline that they have friends that smoke and friends that drink were considered to be at high social risk. Academic risk was assessed by asking participants what grades do you generally get in school? with response options of 1 (mostly A s), 2 (mostly B s), 3 (mostly C s), 4 (Mostly D s), and 5 (D s or lower). Substance use. Frequency of smoking was measured by asking About how often (if ever) do you smoke cigarettes with response options on a 9-point scale anchored by 1 (never) and 9 (more than once a day). Similarly worded items were used to assess frequency of drinking, drunkenness, marijuana use, getting high or stoned from marijuana, and inhalant use. Quantity of smoking was measured by asking If you smoke cigarettes, how much do you usually smoke with response options ranging from 1 (none at all) to 8 (more than two packs a day). Quantity of alcohol use was measured by asking If you drink alcohol, how much do you usually drink each time you drink? with response options ranging from 1 (I don t drink) to 6 (more than six drinks). Composite substance use scores were created in the following manner. A smoking composite score consisted of the mean of the frequency and quantity scores for smoking; a drinking composite score consisted of the mean of the frequency of drinking and drunkenness scores and quantity of drinking score; and a marijuana use composite score consisted of the mean of the frequency of marijuana and getting high scores. A polydrug use score was calculated by adding the number of substances used in the past month out of the four listed above with a possible range of 0 to 4. Program fidelity. In order to assess how much of the program was actually provided to students, project staff in randomly selected classrooms monitored program implementation. Trained staff members observed teachers and recorded on forms developed for each session how much of the material allocated for each session was actually covered in the classroom. Most teachers were observed for two or three class sessions. From these classroom observations, implementation scores were calculated by dividing the number of objectives actually covered by the teacher during the session by the total number of curriculum objectives in the curriculum for that particular session. Thus, the implementation fidelity score was determined at the classroom level, not the individual level. The mean number of program points covered was 48% (SD 19.8). Data Analysis Data were analyzed using chi-square tests, generalized linear models (GLM) analysis of variance (ANOVA), and generalized estimated equations independent (PROC GENMOD) in SAS 8.0 [24]. Listwise deletion was used for all analyses. First, a series of chi-square tests were computed to determine pretest comparability of the intervention and control conditions. Second, the longitudinal sample used in this

4 4 GRIFFIN ET AL. TABLE 1 Adjusted Substance Use Means at One-Year Follow-up for High-Risk Youth, by Experimental Condition Intervention group Control group Mean SE Mean SE 2 df GEE P value Smoking Drinking Marijuana Inhalants Polydrug use Note. N 802. Covariates for all analyses were gender, race, percentage program completed, free lunch, and baseline substance use; 2 represents difference between adjusted means at follow-up assessment; GEE P values represent one-tailed significance levels after adjusting for ICCs at the school level. study was compared to the full pretest sample to determine the impact of attrition using a series of chisquare analyses and GLM ANOVAs. Third, the effectiveness of the prevention program on the substance use outcomes was examined after controlling for pretest scores and several additional covariates (gender, race, free lunch, and percentage program completed). Because the intervention was randomized and administered at the school level, analyses of program effectiveness were conducted controlling for intraclass correlations (ICCs) among students within schools using the GEE-independent option as recommended by Norton et al. [25]. The GEE method adjusts the estimated standard error to account for the within-cluster correlation and generally provides for a more conservative test of the hypothesis when a positive ICC is present. As in previous research [18,20,22], one-tailed significance tests were used for the analyses of intervention effects as warranted by the unidirectional nature of hypothesized effects and the results of previous research using similar prevention approaches. RESULTS Base rates of substance use at the seventh-grade baseline assessment were low for this high-risk sample, with 10% reporting having smoked cigarettes, 10% having used alcohol, and 4% having used marijuana within the past month. A series of t tests was conducted and showed that there were no significant differences between the experimental and the control groups in terms of substance use at baseline, nor were there differences between conditions in terms of gender, race, or grades in school. Further analyses were conducted to examine attrition from the pretest assessment to the one-year follow-up assessment. These analyses showed that rates of attrition were similar in the intervention and control conditions, with 38% of controls dropping out of the study compared to 36% of intervention youth, 2 (1) 0.5, P Based on substance use at the pretest assessment, further analyses showed that those who reported lifetime smoking were more likely to drop out of the study (42%) than nonsmokers (34%), 2 (1) 8.9, P 0.003; lifetime marijuana users were more likely to drop out of the study (49%) than nonusers at the pretest (35%), 2 (1) 13.6, P 0.001; and lifetime polydrug users were more likely to drop out of the study (42%) than nonusers at the pretest (35%), 2 (1) 6.0, P Lifetime drinkers and inhalant users were not more likely to drop out than nondrinkers and those who never used inhalants. However, an analysis examining differential attrition of substance users across experimental conditions revealed that rates of attrition of substance users were equivalent in the intervention and control groups. Several analyses were conducted to examine the effect of the intervention on the substance use composite scores at the one-year follow-up, comparing the follow-up substance use scores across the two conditions after adjusting for relevant covariates. In each analysis, experimental condition was the independent variable, and covariates were gender, race, percentage of program completed, free lunch, and the substance use outcome of interest as measured at baseline. For each substance use outcome, an effect size for the intervention was calculated using Cohen s d statistic [26]. As shown in Table 1, the preventive intervention had a significant effect on several of the substance use outcomes. The composite smoking score for the intervention group was lower than that of the control group ( 2 (1) 6.4, p 0.006, d 0.22); the composite alcohol use score for the intervention group was lower than the control group ( 2 (1) 5.8, p 0.008, d 0.22). Significant intervention effects were also observed for inhalant use ( 2 (1) 2.9, p 0.043, d 0.14) and for polydrug use ( 2 (1) 7.0, p 0.004, d 0.21), with intervention students scoring lower than control students at the one-year follow-up assessment. In summary, in this sample of youth at high risk for substance use initiation there were significant program effects at the one-year follow-up assessment for all substance use outcomes except marijuana use, with the intervention group reporting less substance use than the control group.

5 DRUG PREVENTION AMONG YOUTH AT RISK 5 DISCUSSION The objective of the present study was to examine the generalizability of a successful school-based universal drug abuse prevention program for youth at higher than average risk for substance use initiation. Risk status was based on two factors shown to increase risk for the early onset of substance use: exposure to substance-using peers and poor academic achievement in school. Approximately one-fifth of students from a larger school-based sample were selected as high risk for substance use initiation based on their poor grades in school and their friends use of alcohol and tobacco. Findings indicated that the youth at risk for substance use initiation participating in the LST prevention program reported lower levels of smoking, alcohol use, inhalant use, and polydrug use when assessed one year after the intervention program, relative to similar control students who did not receive the intervention. Significant prevention effects on marijuana use were not found, possibly because the base rates of marijuana use were very low (4%), making it difficult to detect behavioral changes. Over the past two decades, a number of evaluation studies have tested the efficacy of the LST program. These studies have demonstrated positive behavioral effects of the prevention program on smoking, alcohol, and marijuana use as well as on the use of multiple substances and illicit drugs. Long-term follow-up data indicate that the prevention effects of the LST program can last for up to 6 years [20,21]. Research has shown that this approach is effective with inner-city minority youth [18,22,23] with minimal modifications (e.g., graphics, language, and role-play scenarios appropriate to the target population). The present study was the first to examine the effectiveness of the LST program for youth at higher than average risk for substance use initiation. Because the sample consisted of young adolescents, the baseline rates for smoking and drinking were quite low and we selected a subsample of students that were at elevated risk based on their having two salient risk factors for substance use onset: peer exposure and poor school performance. Findings from this study and from previous research on this prevention approach indicate that the LST program is effective not only for broad-based samples of youth, but also for youth who are specifically at risk for drug initiation. Additional studies have examined whether other prevention approaches are effective for youth at higher than average risk. Chou et al. [27] examined the effectiveness of a community-based prevention program in reducing cigarette, alcohol, and marijuana use among students that had already initiated use of these substances and found significant prevention effects. Other researchers have reported that primary prevention programs for adolescent drug use are effective among youth across a spectrum of risk [28]. For example, Johnson et al. [29] found that a community-based prevention program effectively prevented the onset of substance abuse equally well among youth at high and low risk. Taken together, these studies suggest that universal prevention programs can be effective for a variety of youth during early adolescence. However, selective and indicated prevention programs are likely to be most appropriate for older adolescents, including youth already at highest risk based on risk factors such as school failure, school dropout, and existing substance use. Programs for these youth typically address the intrapersonal motivational variables (e.g., affect regulation) that play a role in maintaining or escalating drug use/abuse, and place less emphasis on resistance skills training. These programs may target specific motivational factors for using drugs by emphasizing social network development and group support, skills training (decision-making, interpersonal communication), and emotional well-being, and may take the form of a semester-long personal growth class [30 32]. For those high school youth at highest risk, such as students unable to remain in the regular school system and attending continuation high schools, indicated prevention programming that provides drug counseling in addition to addressing motivational factors through support and skills training may be optimal [33]. Furthermore, while abstinence may be the most appropriate goal for universal prevention programs aimed at younger students (those in elementary and middle/junior high school), preventing drug abuse and drug-related problems may be more appropriate goals for selective and indicated prevention programs targeting older students (i.e., high school and college students). The present study found that the implementation fidelity score, or the proportion of the LST program points covered by program providers, was 48% across all program sessions. This rate of implementation fidelity is lower than in similar studies we have conducted with suburban White youth, where the proportion of program points covered for the LST program was found to be approximately 68% [19]. These findings suggest that implementing a prevention program with high fidelity appears to be more difficult in urban school settings relative to suburban school settings. This is an important finding because research has clearly shown that implementation quantity and quality determine how effective prevention programs will be. Studies that have included an analysis of implementation fidelity (i.e., process evaluations) have consistently shown superior outcomes when the program has been implemented with high fidelity [34]. In fact, in one of the largest meta-analyses of school-based substance abuse prevention programs, Tobler and Stratton [35] concluded that problems related to pro-

6 6 GRIFFIN ET AL. gram implementation have the largest impact in decreasing the effectiveness of these programs. In a recent study of prevention programs in 104 school districts in 12 states, only 19% of the schools were found to be implementing research-based prevention programs with fidelity [36]. Common barriers to highquality implementation included lack of teacher training, lack of program materials, low levels of funding, decentralized decision-making, and lack of program guidance from school district personnel [36]. Despite these problems, only recently have evaluation studies started to focus on or measure program implementation. One review of 181 school-based prevention studies published between 1980 and 1990 in seven journals known for behaviorally based interventions found that only 15% measured implementation integrity [34]. Thus, it is critical for future prevention research to include process evaluations and to integrate implementation fidelity into the analyses of outcome effects. This will help to systematically examine the implementation process of prevention programs, identify factors that impede and enhance high-quality implementation across diverse settings, and provide information that may help improve implementation fidelity. The present study has several methodological strengths and weaknesses. Strengths include random assignment, the use of well-established survey measures and established data collection protocols, and appropriate statistical analyses that control for ICCs. Together, these methodological strengths serve to increase confidence in the validity of the present findings. Limitations of the study include the fact that follow-up is limited to one year after the initial posttest. Additional, follow-up data are needed to determine the longer-term durability of these prevention effects with this population. Second, we were not able to rule out some alternative explanations for the differential outcomes linked to self-reported school performance and peer substance use. For instance, school performance may be a proxy for factors other than poor academic achievement, such as disruptive changes in the family or home environment. Future research should measure and examine a broader array of variables to rule out alternative explanations. Finally, since this was a school-based study that relied on students self-reports, the significant relationships among variables may partly reflect shared method variance. Future research should attempt to examine the processes by which social and academic risk factors contribute to adolescent substance use and the processes by which poorly competent youth turn to substance use over time. Prevention studies should investigate the extent to which universal prevention programs can address the needs of youth as they pass through adolescence to adulthood. REFERENCES 1. Johnson PB, Boles SM, Kleber HD. The relationship between adolescent smoking and drinking and likelihood estimates of illicit drug use. J Addict Dis 2000;19: Kandel DB, Yamaguchi K, Chen K. Stages of progression in drug involvement from adolescence to adulthood: further evidence for the gateway theory. J Stud Alcohol 1992;53: Griffin KW, Botvin GJ, Doyle MM, Diaz T, Epstein JA. A six-year follow-up study of determinants of heavy cigarette smoking among high school seniors. J Behav Med 1999;22: Kandel D. Examining the gateway hypothesis: stages and pathways of drug involvement. New York: Cambridge Univ. Press, Ellickson PL, Tucker JS, Klein DJ. High risk behaviors associated with early smoking: results from a 5-year follow-up study. J Adol Health 2001;28: Newcomb MD, Bentler PM. Consequences of teenage drug use: impact on the lives of young adults. Beverly Hills: Sage Publications, Institute of Medicine. Prevention. In: Pathways of addiction: opportunities in drug abuse research. Washington, DC: National Academy Press, 1996: Hawkins JD, Catalano RF, Miller JY. Risk and protective factors for alcohol and other drug problems in adolescence and early adulthood: implications for substance abuse prevention. Psychol Bull 1992;112: Petraitis J, Flay BR, Miller TQ. Reviewing theories of adolescent substance use: organizing pieces in the puzzle. Psychol Bull 1995;117: Ennett ST, Bauman KE. Peer group structure and adolescent cigarette smoking: a social network analysis. J Health Soc Behav 1993;34: Oetting ER, Beauvais F. Peer cluster theory, socialization characteristics, and adolescent drug use: a path analysis. J Counsel Psychol 1987;34: Wills TA, Cleary, SD. Peer and adolescent substance use among 6th- to 9th-graders: latent growth analysis of influence versus selection mechanisms. Health Psychol 1999;18: Graham JW, Marks GS, Hansen WB. Social influence processes affecting adolescent substance use. J Appl Psychol 1991;76: Huba GJ, Wingard JA, Bentler PM. Beginning adolescent drug use and peer and adult interaction patterns. J Consult Clin Psychol 1979;47: Paulson MJ, Coombs RH, Richardson MA. School performance, academic aspirations, and drug use among children and adolescents. J Drug Educ 1991;20: Stoiber KC, Good B. Risk and resilience factors linked to problem behavior among urban, culturally diverse adolescents. School Psychol Rev 1998;27: Jessor R, Jessor SL. Problem behavior and psychosocial development: a longitudinal study of youth. New York: Academic Press, Botvin GJ, Griffin KW, Diaz T, Ifill-Williams M. Drug abuse prevention among minority adolescents: one-year follow-up of a school-based preventive intervention. Prev Sci 2001;2: Botvin GJ, Baker E, Dusenbury L, Tortu S, Botvin EM. Preventing adolescent drug abuse through a multimodal cognitivebehavioral approach: results of a three-year study. J Consult Clin Psychol 1990;58: Botvin GJ, Baker E, Dusenbury L, Botvin EM, Diaz T. Longterm follow-up results of a randomized drug abuse prevention

7 DRUG PREVENTION AMONG YOUTH AT RISK 7 trial in a White middle-class population. JAMA 1995;273: Botvin GJ, Griffin KW, Diaz T, Scheier LM, Williams C, Epstein JA. Preventing illicit drug use in adolescents: long-term follow-up data from a randomized control trial of a school population. Addict Behav 2000;5: Botvin GJ, Dusenbury L, Baker E, James-Ortiz S, Botvin EM, Kerner J. Smoking prevention among urban minority youth: assessing effects on outcome and mediating variables. Health Psychol 1992;11: Botvin GJ, Schinke SP, Epstein JA., Diaz T, Botvin EM. Effectiveness of culturally focused and generic skills training approaches to alcohol and drug abuse prevention among minority adolescents: two-year follow-up results. Psychol Addict Behav 1995;9: SAS Institute. SAS/STAT user s guide, Version 8. Cary, NC: SAS Publications, Norton EC, Bieler GS, Ennett ST, Zarkin GA. Analysis of prevention program effectiveness with clustered data using generalized estimating equations. J Consult Clin Psychol 1996;64: Cohen J. Statistical power analysis for the behavioral sciences, 2nd edition. Hillsdale, NJ: Erlbaum, Chou CP, Montgomery S, Pentz MA, Rohrbach LA, Johnson CA, Flay BR, et al. Effects of a community-based prevention program on decreasing drug use in high risk adolescents. Am J Public Health 1998;88: MacKinnon DP, Weber MD, Pentz MA. How do school-based drug prevention programs work and for whom? Drugs Society 1988;3: Johnson CA, Pentz MA, Weber MD, Dwyer JH. Relative effectiveness of comprehensive community programming for drug abuse prevention with high-risk and low-risk adolescents. J Consult Clin Psychol 1990;58: Eggert LL, Seyl CD, Nicholas LJ. Effects of a school-based prevention program for potential high school dropouts and drug abusers. Int J Addict 1990;25: Eggert LL, Thompson EA, Herting JR, Nicholas LJ, et al. Preventing adolescent drug abuse and high school dropout through an intensive school-based social network development program. Am J Health Promotion 1994;8: Thompson EA, Horn M, Herting JR, Eggert LL. Enhancing outcomes in an indicated drug prevention program for high-risk youth. J Drug Educ 1997;27: Sussman S. Development of a school-based drug abuse prevention curriculum for high-risk youth. J Psychoactive Drugs 1996; 28: Gresham FM, Gansle KA, Noell GH, Cohen S, Rosenbaum S. Treatment integrity of school-based behavioral intervention studies. School Psychol Rev 1993;22: Tobler NS, Stratton HH. (1997). Effectiveness of school-based drug prevention programs: a meta-analysis of the research. J Primary Prev 1997;18: Hallfors D, Godette, D. Will the Principles of Effectiveness improve prevention practice? Early findings from a diffusion study. Health Educ Res 2002;17: