OBJECTIVE: To identify risk factors for suicide attempts (SA) in individuals commencing treatment for a manic or mixed episode. METHOD: A total of 3390 manic or mixed cases with bipolar disorder (BD) type I recruited from 14 European countries were included in a prospective, 2-year observational study. Poisson regression models were used to identify individual and treatment factors associated with new SA events. Two multivariate models were built, stratified for the presence or absence of prior SA. RESULTS: A total of 302 SA were recorded prospectively; the peak incidence was 0-12 weeks after commencing treatment. In cases with a prior history of SA, risk of SA repetition was associated with younger age of first manic episode (P = 0.03), rapid cycling (P < 0.001), history of alcohol and/or substance use disorder (P < 0.001), number of psychotropic drugs prescribed (P < 0.001) and initiation of an anticonvulsant at study entry (P < 0.001). In cases with no previous SA, the first SA event was associated with rapid cycling (P = 0.02), lifetime history of alcohol use disorder (P = 0.02) and initiation of an anticonvulsant at study entry (P = 0.002). CONCLUSION: The introduction of anticonvulsants for a recent-onset manic or mixed episode may be associated with an increased risk of SA. Further BD studies must determine whether this link is causal.