The study examined the validity of 1848 self-reported uses of drugs determined within an Addiction Severity Index interview in comparison with urinalysis results among drug-dependent subjects undergoing treatment in outpatient clinics (Aquitaine area, southwest France, 1994-2005). Agreement and kappa statistics were calculated for each substance. Factors associated with agreement were defined using a multivariate analysis. The conditional kappa coefficients were excellent for all substances assessed. The accuracy between self-reports and urinalysis results was influenced by factors that only slightly affected conditional kappa coefficients. Clients did not underreport their substance use in naturalistic clinical assessment conditions.