Pharmacodynamics must inform statistics: an example from a cocaine dependence pharmacotherapy trial

Background. There is no FDA-approved medication for cocaine dependence or consensus on the statistical approach for analyzing data from cocaine dependence pharmacotherapy trials. The goal of this paper is to illustrate the importance of understanding medication's pharmacodynamics when specifying the statistical model to test its efficacy. Method. Data from a double-blind placebo controlled trial of reserpine for cocaine dependence are analyzed. Since the antihypertensive properties of reserpine are well established, blood pressure data are utilized to evaluate the ability of two statistical models, one that does not take the pharmacodynamics of reserpine into account and one that does, to detect reserpine's antihypertensive effect. Results. The statistical model specified without regard to reserpine's pharmacodynamics failed to find a significant medication effect for either systolic (P = 0.49) or diastolic (P = 0.59) blood pressure. Contrariwise, the model based on the pharmacodynamics of reserpine found a significant effect for both systolic (P = 0.002) and diastolic (P = 0.004) blood pressure. Conclusions. If the pharmacodynamics of a study medication are not considered when specifying statistical models, then erroneous conclusions may be reached. This trial is registered with NCT00033033.