To compare the risk of treatment-emergent diabetes(TED) in schizophrenic patients treated with atypical(AAP) versus typical(TAP) antipsychotic medications.

We conducted a retrospective database analysis on episodes of care initiated after 1/1/2000 using data from the California Medicaid program. We included episodes for patients 18 years or older with schizophrenia who switched medications with a minimum “wash out” period of 15 days and no evidence of diabetes in the previous 6 months. If selection bias was present we used a simultaneous bivariate probit model to estimate the risk of TED in patients treated with AAP in comparison to TAP, otherwise we used a univariate probit model. Sensitivity analyses estimated the effect of olanzapine, risperidone and quetiapine independently versus TAP.

A Wald test of the correlation coefficient of the disturbances suggests that treatment selection is exogenous in our model(rho=0.005(p=0.95)) using a Huber-White sandwich estimator of the variance. The univariate probit model results suggest that AAPs were not associated with an increased risk of TED relative to TAPs(p=0.324). Sensitivity analysis showed quetiapine to be associated with a statistically significant decreased risk of TED relative to TAPs. No statistically significant association was shown with olanzapine or risperidone. A bivariate probit model omitting numerous variables demonstrates selection bias(rho=-0.650(p=0.0029)).

The results of this study show that AAPs are not associated with an increased risk of TED relative to TAPs. Explanatory variables that may explain treatment selection that were included in our model were sufficient to control for choice of therapy.