Predicting Conversion to Insulin Sensitivity With Metformin

BackgroundInsulin resistance (IR) changes the trajectory of responsive bipolar disorder to a treatment-resistant course. A clinical trial conducted by our group demonstrated that IR reversal by metformin improved clinical and functional outcomes in treatment-resistant bipolar depression (TRBD). To aid clinicians identify which metformin-treated TRBD patients might reverse IR, and given strong external evidence for their association with IR, we developed a predictive tool using body mass index (BMI) and homeostatic model assessment-insulin resistance (HOMA-IR).MethodsThe predictive performance of baseline BMI and HOMA-IR was tested with a logistic regression model using known metrics: area under the receiver operating curve, sensitivity, and specificity. In view of the high benefit to low risk of metformin in reversing IR, high sensitivity was favored over specificity.ResultsIn this BMI and HOMA-IR model for IR reversal, the area under the receiver operating curve is 0.79. At a cutoff probability of conversion of 0.17, the model's sensitivity is 91% (95% confidence interval [CI], 57%-99%), and the specificity is 56% (95% CI, 36%-73%). For each unit increase in BMI or HOMA-IR, there is a 15% (OR, 0.85; 95% CI, 0.71-0.99) or 43% (OR, 0.57; CI, 0.18-1.36) decrease in the odds of conversion, respectively.ConclusionsIn individuals with TRBD, this tool using BMI and HOMA-IR predicts IR reversal with metformin with high sensitivity. Furthermore, these data suggest early intervention with metformin at lower BMI, and HOMA-IR would likely reverse IR in TRBD.