A method for predicting the maintenance dose of parent drug required to give a desired steady-state concentration of metabolite, using a single determination of metabolite concentration in serum following the first dose of parent drug, is described. Clinical evidence that such a method is feasible for the drug-metabolite pair imipramine-desipramine has been reported. The error inherent in an estimation of maintenance dose based on a single determination of metabolite concentration is a function of sampling time and the first-order elimination rate constants for parent drug and metabolite (K and km, respectively). The method is applicable to drug-metabolite pairs in general by selecting the sampling time (t*) to give minimum error: t* equal 1/km + 1.3/K, when km less than or equal to K, and t* equal 1/K + 1.3/km, when km greater than K (bars denote population mean value). The error expected to be encountered in the application of the method to specific drug-metabolite pairs can be analyzed by the graphical methods described.