A simple mathematical model for predicting lithium dose requirement.

Available methods of predicting lithium dose have been based on kinetics of a single test dose. A statistically based mathematical model was developed in which lithium dose is derived by stepwise multiple linear regression based on desired level, form of lithium, concomitant tricyclic use, age, sex, and weight. Predictions of the model were correct to within 300 mg in 66% of the 100 initial cases and within 600 mg in 94% of cases. A validation study of 112 additional cases revealed similar percentage. The simple mathematical expression derived from record review allows the clinician to calculate the relationship of steady-state lithium dosage to serum level prior to initiating treatment.