Comparison of pharmacokinetic procedures for dosing lithium based on analysis of prediction error.

Five pharmacokinetic methods for estimating maintenance dosage requirements of lithium carbonate were compared retrospectively in 20 inpatients with acute bipolar illness. Specific pharmacokinetic methods tested included the method of Cooper, the multiple-point method of Perry, the single-point method of Perry, the method of Zetin, and the method of Pepin. Data analysis was based on evaluation of prediction error or the difference between the predicted steady-state lithium concentration and the measured steady-state lithium concentration at equivalent daily doses. Each dosing method was assessed in regard to accuracy and bias of predicted steady-state serum lithium concentrations. Bias was assessed by comparison of the median value of the prediction error with zero. The dosing recommendation based on the Cooper nomogram resulted in a significant positive bias (p less than or equal to 0.05). Intermethod accuracy was assessed by comparison of the absolute prediction errors of each dosing method. Significant differences in accuracy were observed between the method of Pepin when compared with the single-point method of Perry (p less than or equal to 0.05, k-sample sign test). All other comparisons were nonsignificant.