In search of moderators and mediators of hyperglycemia with atypical antipsychotic treatment

Signal detection methods were used to identify values of metabolic variables that predict development of prediabetes or diabetes before (moderators) or associated with treatment (mediators), utilizing data from two multi-center clinical trials of patients with schizophrenia, treated for 6 months with olanzapine (OLZ) or ziprasidone (ZIP). At baseline, participants were often overweight/obese (63% with a body mass index ⩾25.0 kg/m2), dyslipidemic [more than one-third had elevated triglyceride (TG) and low high-density lipoprotein cholesterol (HDL-C) concentrations], and prediabetic (20%). Weight gain was significantly greater in OLZ-treated patients, as was accentuation of dyslipidemia. However, there were no significant correlations between weight gain and lipid changes from baseline to weeks 2, 4, 8 or to last observation. Type 2 diabetes developed in 4% and prediabetes in 18% of the population. Significant baseline predictors of diabetes were a HDL-C concentration <28 mg/dL, or being ⩾58-years-old if HDL-C concentration was ⩾28 mg/dL. Baseline plasma glucose concentration ⩾92 mg/dL was the only significant predictor of developing prediabetes, accounting for 60% of cases. Post-treatment increments in plasma TG concentrations ⩾145 mg/dL or ⩾59 mg/dL were significant predictors of diabetes (23%) or prediabetes (27%), respectively. If the increase in TG was <145 mg/dL, rapid weight gain ⩾6.1 kg in 2 weeks predicted development of diabetes (18%). These findings provide a quantitative approach to identify those at greatest treatment-associated risk to develop glucose intolerance, and emphasize the need to address co-morbid medical disorders in these patients.