OBJECTIVE: To review the baseline similarities and differences between olanzapine- and risperidone-treated patients, as described in epidemiologic studies, and recommend a set of potential confounders that should be addressed in such studies. METHOD: All 24 non-randomized olanzapine-risperidone comparisons cited in Medline or presented as posters at major psychiatric conferences (APA, ECNP, WPA) through 1999 were included. The direction, magnitude, and statistical significance of all baseline comparisons of the two treatment groups were recorded, if available. RESULTS: Across all use, diagnosis distribution and mean age differed without consistent direction. Comorbidities, number of different prior antipsychotics, and prior use of second line treatments were greater for olanzapine users, while percent with any prior antipsychotic use was not significantly different. Within schizophrenia, olanzapine users were more likely to be male, have earlier age of onset, and lower mean age. Most schizophrenia-specific studies included no baseline service utilization or clinical measures; however, one study showed higher previous hospital costs for olanzapine-treated patients. CONCLUSIONS: Mean age, gender, diagnosis, age at onset, history of antipsychotic use and prior costs differed for olanzapine- and risperidone-treated patients, with olanzapine-treated patients having a history indicative of higher cost and treatment-failure. Non-randomized studies must address these factors to avoid misleading interpretations of their results.