Comparison of three a priori methods and one empirical method in predicting lithium dosage requirements

The precision and bias of three a priori methods and an empirical method for predicting lithium dosage requirements were studied. Data on serum lithium concentrations were collected from inpatient medical records at a state psychiatric hospital. Predicted daily lithium doses were calculated by using a priori methods proposed by Zetin et al., Jermain et al. and Pepin et al., and an empirical method and compared with the patients' actual dosages. Similar comparisons were made with respect to serum lithium concentrations at steady state. Absolute mean error and mean error were calculated to assess the precision and bias of each a priori method. The records of 47 patients were used in the study. Average mean error for dosage predictions was -130.41, -187.69, 170.80, and -357.23 mg/day for the Jermain, Pepin, Zetin, and empirical methods, respectively. Average mean error for serum lithium concentration predictions was 0.11, -0.09, and 0.37 meq/L for the Jermain, Pepin, and empirical methods, respectively. The Jermain and empirical methods significantly overpredicted concentration and underpredicted dosage. The Zetin method overpredicted dosage. The Pepin method underpredicted dosage but not concentration. The average difference in dosage error among the methods was only 73.3 mg/day. Three a priori dosing methods were similar to an empirical method in their ability to predict lithium dosages. All methods were biased. Although all a priori methods were more precise than the empirical method, the clinical significance is unclear.