Purpose: In Canada, regulatory warnings on cerebrovascular effects of risperidone and olanzapine in the elderly population with dementia were issued, respectively, in 2002 and 2004, and those on mortality associated with all atypical antipsychotics (APs) in 2005. These warnings led to a decrease in the prescription of APs, but effects on patterns of usage remain poorly examined. We conducted a study to assess the association between warnings and patterns of AP use in a population of community-dwelling elderly with dementia. Methods: A retrospective cohort of 10,969 community-dwelling elderly (age 66+) with dementia who were new users of APs between 1st January 2000 and 31st December 2009 was assembled through the Quebec drug claims database (RAMQ). Association between regulatory warnings and rate of initiation of AP treatment was evaluated through interrupted time series analysis. Effects of the 2005 warning on cerebrovascular history in treated patients, and AP usage patterns (dosage, duration) were assessed, respectively through multivariate logistic regression and multiple linear regression analysis. Results: The proportion of AP treatments initiated with risperidone decreased over time while that of quetiapine increased and of olanzapine remained stable. Controlling for covariates, the cerebrovascular risk profile of treated patients did not change after the 2005 warning (OR=1.05, 95%CI: 0.90-1.22). A small decrease in mean prescribed daily dose for risperidone was observed after the 2005 warning (-0.05 mg, p<0.001) while an increase was observed for olanzapine (+0.34 mg, p=0.009) and quetiapine (+1.27 mg, p=0.40). No change in treatment duration was observed (p=0.19). Conclusion: Although regulatory warnings led to a decrease in the use of atypical APs, these products are still widely prescribed off-label in the elderly population with dementia. Channelling of APs toward patients with lower cerebrovascular risks and changes in prescription practices were not apparent after the warnings.