BACKGROUND: The development of an economic model to assess antipsychotics for first episode schizophrenia raised several issues: (i) clinical and economic evidence relates to chronic schizophrenia; (ii) inconsistent reporting of outcomes and adverse events; (iii) comparison of costs and outcomes across several interventions is complicated by (a) an absence of head to head comparisons, (b) intolerance or resistance to 3 plus antipsychotics. OBJECTIVES: To design a decision analytic model to evaluate the relative efficiency of antipsychotics. METHODS: Two composite variables were defined to incorporate 23 efficacy and safety events: (i) acceptable treatment (able/willing to continue allocated therapy, with/without side effects or symptom control), (ii) unacceptable treatment (unacceptable side effects or symptom control, non compliance). Utility values were defined for each variable. Clinical data were extracted from systematic reviews separately for 1st episode and chronic schizophrenia. Age specific resource use data were estimated from national statistics to proxy 1st episode patients. Follow on therapy for people failing 1st therapy was estimated by (i) specifying drug sequence; (ii) treating follow on therapy as a random event. Monte carlo simulation analysis was used to reduce complexity and incorporate uncertainty. RESULTS: It was not possible to identify consistent data for 1st episode patients only. Interpretation of the specific treatment sequences was complex with 32 possible combinations identified. There were substantial differences within classes of antipsychotics. The results were sensitive to: (i)specification of follow on therapy (ii) distributional form applied to data. CONCLUSIONS: The analysis highlighted the poor quality and absence of relevant data. This could not be dealt with by modelling. Current classifications of typical and atypical antipsychotics may not be relevant.