AIMS: To systematically review the evidence on the psychometric performance and accuracy of screening or diagnostic tools for cannabis use disorder. METHOD: Systematic review and meta-analysis which included studies conducted in clinical settings, schools, universities, community settings and population-based surveys in multiple countries and regions. Participants were adolescents, young adults, general adult populations, people who used cannabis, psychiatric and substance use treatment attendees, and specialised groups such as justice involved youth and military personnel. The tools assessed included The Cannabis Abuse Screening Test (CAST), the Cannabis Use Disorders Identification Test-Revised (CUDIT-R), the Severity of Dependence Scale (SDS), the Cannabis Problems Questionnaire (CPQ), the Cannabis Problems Questionnaire for Adolescents (CPQ-A), the Cannabis Use Problems Identification Test (CUPIT), the Tobacco, Alcohol, Prescription medication, and other Substance use tool (TAPS), the Marijuana Screening Inventory-X (MSI-X), the DSM-Guided Cannabis Screen, the Alcohol, Smoking and Substance Involvement Screening Test (ASSIST), the Marijuana Problem Index (MAPI), the Toronto Cannabis Risk Screening Tool (TCRUST), the Nigerian Cannabis Use Disorder Scale, and the Problematic Use of Marijuana measure (PUM). These were assessed for internal consistency, test-retest reliability, as well as diagnostic accuracy using measures of sensitivity, specificity and area under the curve (AUC). RESULTS: Forty studies met inclusion criteria (2000-2025), including data from N = 23 175 participants. Methodological quality was generally moderate to high, though some studies relied on self-reported symptom checklists rather than structured diagnostic interviews. The most frequently studied tools were the CAST (k = 13 studies), CUDIT-R (k = 8), SDS (k = 5), CPQ and CPQ-A (k = 4), and CUPIT (k = 2). Across instruments, internal consistency was generally acceptable to excellent (α = 0.66-0.92), with fair to excellent discriminative validity (AUC = 0.71-0.96) for detecting cannabis use disorder or dependence. Optimal cut-offs varied statistically significantly by population and setting. In clinical samples, the tools generally performed stronger with the use of standard or higher cut-offs to prioritize specificity and avoid misclassifying non-cases. In general population samples, particularly youth, the tools had better performance with lower cut-offs to prioritize sensitivity. CONCLUSIONS: There is a lack of sufficient studies on screening or diagnostic tools for cannabis use disorder for clear evidence. Based on this limited current evidence, several brief screening or diagnostic tools are useful in identifying disorder or problem cannabis use in diverse settings, including the CAST, CUDIT-R, and SDS. Variation in optimal thresholds by age, clinical status and cultural context suggest that the utility of these screening or diagnostic tools depends on the population.
Leung et al. (Tue,) studied this question.