Cannabis and Neuropsychiatric Harm: Evidence on cannabis use and its effects on mental health and development

Cannabis exposure can affect psychiatric symptoms, cognition, development and everyday functioning. Clinical consequences vary with THC dose, frequency, age, prior vulnerability and the outcome measured. This critical review integrates experimental and observational evidence, published meta-analyses, selected secondary syntheses and outcome-specific probability models, with attention to recovery and evidence that constrains claims of persistent harm. The evidence base combines reference tracing with six public MEDLINE-index searches through 18 October 2026. The date-restricted searches yielded 1,043 unique records, all assessed by title; 221 also received abstract assessment. Prioritized primary-source appraisal distinguished access depth, cohort overlap, exposure timing and outcome definition. Six selected frequentist and Bayesian meta-analyses examined compatible psychiatric contrasts. Twenty-five probability-model specifications were fitted using exact transitions, logistic sufficient statistics and exact or rounded grouped outcome data. Survey-weighted adult risks supplied an additional pathway. Evidence is strongest for acute THC effects, cannabis use disorder and withdrawal. Converging findings support a contributory role in psychosis and an adverse clinical course with continued use after psychotic illness. Birth-growth outcomes show more consistent associations than many later developmental diagnoses. Cognitive and biological findings include both persistent differences and recovery; imaging or molecular associations alone do not establish clinical injury. Social and functional outcomes remain consequential despite substantial confounding and variation by design. Selected secondary pools gave relative-effect estimates of 1.03 for depression, 1.25 for anxiety and 2.67 for higher-exposure psychosis, with uncertainty strongly affected by compatibility and small study counts. A forty-five-item questionnaire produces source-compatible outcome reports with frequencies, odds and modeled ranges. Internal CUD discrimination, measured by area under the receiver operating characteristic curve, was 0.938 across the selected sample, 0.781 among regular users and 0.733 among baseline non-cases. Pregnancy comparisons revealed appreciable subgroup miscalibration despite close overall agreement for one outcome. Joint coefficients, baseline risks and reproducible calculations support restricted prognostic applications; broad external calibration of the questionnaire remains unknown. Cannabis-related harms are clinically important, unevenly distributed and specific to exposure and outcome. Quantification is most informative when it preserves these distinctions and reports the assumptions governing individual applicability.