1. Michael Prendergast UCLA Integrated Substance Abuse Programs Deborah Podus UCLA Integrated Substance Abuse Programs John Finney Veterans Affairs Palo Alto Health Care System Lisa Greenwell UCLA Integrated Substance Abuse Programs John Roll Washington State University, Spokane Contingency Management for the Treatment of Substance Use Disorders: A Meta- Analysis of Comparison-Group Studies

2. Assessed the effectiveness of contingency management (CM) techniques in treating substance use disorders (i.e., illicit drugs, alcohol, and tobacco). Used meta-analytic techniques to examine study characteristics, calculated effect sizes, and examine moderators. 47 comparisons that included treatment-control group contrasts and for which an effect size could be calculated. Introduction

3. CM entails the use of positive or negative reinforcers to promote a desired behavioral outcome. In drug abuse treatment, CM interventions are based on the premise that drug use and addiction are a form of operant conditioning in which behavior is controlled or shaped by its consequences. Most common CM technique is voucher-based reinforcement therapy. Other variations of CM use different reinforcers to promote abstinence or reduced use. CM is also used to promote behaviors associated with improved drug-use outcomes (e.g., treatment attendance). Definition of Contingency Management

4. Study Eligibility Criteria Inclusion criteria –Outcome evaluations of CM treatment for dependence on alcohol, tobacco, or illicit drugs delivered to juveniles or adults –Published in English between 1970 and 2002 –Used a treatment-control group design –At least one drug use outcome –Quantitative data needed to calculate an effect size Exclusion criteria –Studies where different magnitudes of reinforcers or different schedules of reinforcement were compared with one another –Total sample size less than 10

5. Search Strategy Databases searched Current Contents, BIOSIS, Embase, MEDLINE, PsychInfo, Sociological Abstracts, Cork Database, and Cochrane Library. Search terms “contingency management,” “voucher-based reinforcement,” “behavioral contracting,” or “token economy” + “addiction,” “drug abuse,” “ alcoholism,” “cocaine,” “opiates,” or “tobacco” Article reference lists searched for other potentially relevant documents.

6. Coding Codebook: study context, methodology, participant characteristics, treatment characteristics, and dependent variable characteristics and effect size calculations. Each study was coded by one of five masters- or doctoral- level coders. Decisions and clarifications regarding specific questions were recorded in a policy manual that was regularly updated for use by the coders. Coding for each study was reviewed for quality control before data entry.

7. This analysis focused on studies that tested the effects of a CM intervention compared with a no-treatment group or when added to standard treatment (e.g., methadone maintenance, psychosocial treatment). All studies took place in the United States. All documents were journal articles. Nearly all studies were funded by the National Institute on Drug Abuse or other NIH agency. 92% of the studies had experimental designs employing random assignment procedures. Characteristics of Study Comparisons

8. Sample size (median) = 69, Range = 12 - 844 70% of studies conducted during the 1990s Type of reinforcers –Vouchers (51.5%) –Methadone take-home doses or dosage adjustments (29.7%) –Cash (21.3%) –Prizes, recreational activities, housing, and social reinforcement. Characteristics of Study Comparisons (Cont.)

9. Analyses Effect size calculated for drug use variables. Main outcome was an average drug use measure averaged over the course of treatment or a measure of drug use taken at the end of treatment. Within each study, we averaged the effect sizes for multiple drug use measures. The effect size index was the standardized mean difference, corrected for small sample size. In combining effect sizes, we weighted each effect by the inverse of its variance. Summary Statistics are shown in Table 1.

10. *p < 0.05. Table 1. Summary Statistics

11. Some studies reported drug use outcomes measured at various points following the end of the CM intervention. If a study reported data at multiple time points, the time point of the last follow-up was selected. Although the number of studies reporting outcomes within each quarter is small, the general trend in effect size is downward (see Table 2). Effect Sizes by Follow-up Period

12. Table 2. Average Effect Size by Follow-up Period * Total n=48; includes 47 comparisons with last/average in treatment wave plus 15 follow-up waves (one without a last/average in treatment wave). ** Fixed-effects weighted mean effect size *** Measurement occurred while still in treatment

13. Studies conducted during the 1990s had a significantly smaller effect size (d = 0.35) than those conducted in the 1970s and 1980s (d = 0.64). Studies in which the researcher was more involved in the design or delivery of treatment had a larger mean effect (d = 0.46) than studies in which the researcher was less involved (d = 0.14). CM was more effective in treating opiate use (d = 0.65) and cocaine use (d = 0.66), compared with tobacco (d = 0.31) or multiple drugs (d = 0.42). Studies in which the duration of the intervention was from 1 week to 11 weeks had an effect size of d = 0.58. For interventions of 12-25 weeks, the effect size was d = 0.44, and for 26 weeks and over, d = 0.34. Moderators

14. Findings support the effectiveness of CM techniques for treating clients who are dependent on alcohol, tobacco, or illicit drugs. After clients are no longer subject to contingencies, the magnitude of the treatment effect declines. CM is able to establish and maintain abstinence for many clients during treatment, thereby permitting clients to more productively engage in treatment services that promote the broader psychosocial aspects of recovery. Further research on this topic should include examination of the relative effectiveness of different types of CM, further investigation of moderators of the impact of CM, and comparison of the effects of CM and of other treatment approaches. Conclusion

15. Work on this meta-analysis was initiated by the Department of Veterans Affairs Substance Use Disorders Quality Enhancement Research Initiative and was supported by Contract V261P-1447 from the Program Evaluation and Resource Center, VA Palo Alto Health Care System. Funding was also provided by NIDA grants R01-017407 and R01-017084. Thanks are due to Karen Perdue, Jinnie Rhee, Christie Rizzo, and Jennifer Wishner for study coding; to Todd Helmus for editing and quality control; to Sarah Barnett for database searches, document management, and project support; to James Anderson for database searches; to Ron Zuniga for creating the data entry program; to Carolyn Potter, Isa Campbell, and Cora Garcia for data entry and cleaning, and to Stacy Calhoun for manuscript preparation. Acknowledgements