In contrast to previous studies, two projects did not use DSM-IV or ICD-10 diagnoses as the "gold standard" to compare with but some broader concepts of problem use. Thus, the drug abuse section of the POSIT questionnaire was used to cross-validate CAST scores [15] and several single-item indicators of cannabis-related problems were used to validate CUDIT [9]. In this context, CAST proved to be effective when screening for high risks of abuse showing high values of sensitivity, specificity and predictive power (Table 3). CUDIT revealed quite favourable levels of sensitivity and lower, but in most cases still acceptable, specificity. As in the Steiner et al. study [24], positive predictive value of the scale was low due to the low base rate of cannabis use in the overall sample.
