Cannabis as a risk factor for psychosis: systematic review (2005)

Various lines of evidence suggest an association between cannabis and psychosis. Five years ago, the only significant case-control study addressing this question was the Swedish Conscript Cohort. Within the last few years, other studies have emerged, allowing the evidence for cannabis as a risk factor to be more systematically reviewed and assessed. Using specific search criteria on Embase, PsychINFO and Medline, all studies examining cannabis as an independent risk factor for schizophrenia, psychosis or psychotic symptoms, published between January 1966 and January 2004, were examined. Additional studies were also reviewed from references found in retrieved articles, reviews, and a cited reference search (ISI-Web of Science). Studies selected for meta-analysis included: (i) case-control studies where exposure to cannabis preceded the onset of schizophrenia or schizophrenia-like psychosis and (ii) cohort studies of healthy individuals recruited before the median age of illness onset, with cannabis exposure determined prospectively and blind to eventual diagnosis. Studies of psychotic symptoms were also tabulated for further discussion. Eleven studies were identified examining the relationship between cannabis use and psychosis. Seven were included in the meta-analysis, with a derived odds ratio (fixed effects) of 2-9 (95 % confidence interval = 2.4-3.6). No evidence of publication bias or heterogeneity was found. Early use of cannabis did appear to increase the risk of psychosis. For psychotic symptoms, a dose-related effect of cannabis use was seen, with vulnerable groups including individuals who used cannabis during adolescence, those who had previously experienced psychotic symptoms, and those at high genetic risk of developing schizophrenia. In conclusion, the available evidence supports the hypothesis that cannabis is an independent risk factor, both for psychosis and the development of psychotic symptoms. Addressing cannabis use, particularly in vulnerable populations, is likely to have beneficial effects on psychiatric morbidity. (Author' s abstract)