The studies all used a biological criterion of validity, such as urinalysis or hair analysis. The participants included addicts committed to a civil addict treatment program for criminal offenders under court order, parolees, probationers, clients of an inpatient drug treatment program, juvenile detainees, methadone maintenance clients, persons being admitted to a psychiatric hospital, arrestees, and infants' mothers referred by health providers. Coefficients of chance-corrected agreement between self-reports and validity criteria were calculated from published data to aid comparisons across studies. Results revealed that the median conditional kappa value was 0.42, which was considerably below the kappa value of 0.80 that represents acceptable reporting accuracy. Findings indicated that the magnitude of underreporting of drug use documents in this review could seriously bias prevalence estimates and treatment outcome studies. Table and 60 references (Author abstract modified)