When a substance with a long biologic half-life is administered over an extended period with a varying dose schedule, many concerns about safety arise. In the case of rubidium chloride, which has been reported to have antidepressive activity, some of these concerns are related to the potential for unanticipated accumulation in the intracellular fluid, where replacement of more than 40% of the potassium ions is associated with toxicity in animals. A pharmacokinetic model of rubidium distribution, based on a three-compartment system, is in accord with the empirical data from previous human trials. By comparison of predicted with observed plasma rubidium levels, a sudden change in the ratio of intra- to extracellular distribution of rubidium can be detected, and availability of this warning signal of potentially toxic intracellular accumulation can be useful during extended administration.