OBJECTIVE: to estimate the economic impact of non-compliance associated with two adverse events (weight change, extrapyramidal symptoms [EPS]) resulting from treatment of schizophrenia with atypical antipsychotics. METHODS: Differential adverse event profiles were expected to cause differences between antipsychotics in terms of compliance, relapse rate and direct costs. Using a state-transition model, non-compliance and relapse over one year following recovery from a schizophrenic episode were simulated. Three transition probabilities governed movement between states: WELL, COMPLIANT to RELAPSE, derived from literature on relapse with atypical antipsychotics; WELL, NON-COMPLIANT to RELAPSE, derived from a quantitative literature summary of relapse following antipsychotic withdrawal; WELL, COMPLIANT to WELL, NON-COMPLIANT, based on non-compliance estimates from a two-round Delphi study. Twenty-five experts in the treatment of schizophrenia, blind to the identity of the products, reviewed published literature on weight change and EPS liability for olanzapine, quetiapine, risperidone and ziprasidone: respondents assumed them to be identical except for adverse event liability. Bootstrap estimates of differential relapse between the atypicals were obtained, accounting for uncertainty in all three transition probabilities. Published data on direct costs of relapse were applied, giving incremental cost estimates for different adverse event liabilities. RESULTS: quetiapine and ziprasidone were similar in estimated non-compliance and relapse rates. Compared to quetiapine, greater weight change with olanzapine led to higher estimated non-compliance and incremental one-year per-patient costs of ~$530 (95% CI ~ $275-800, p < .00001), using USA-based cost data. Greater EPS liability with risperidone led to higher estimated non-compliance and incremental one-year per-patient costs of ~$485 (95% CI ~ $235-800, p < .00001) compared to quetiapine. Similar increased costs of treatment with olanzapine and risperidone were shown (~ £630 [95% CI ~ £330-960] and ~£575 [95% CI ~ £280-960] respectively) when UK-based cost data were used. CONCLUSIONS: The more favourable adverse event profile of quetiapine is associated with a predicted decrease in one-year per-patient direct costs compared to olanzapine and risperidone.