Diminished P-glycoprotein (P-gp)-mediated transport across the blood-brain barrier may play an important role in several neurodegenerative disorders. In previous studies, a racemic mixture of (R)-[11C]verapamil and (S)-[11C]verapamil has been used as tracer for assessing P-gp function using positron emission tomography (PET). Quantification, however, is compromised by potential differences in kinetics between these two isomers. The aim of the present study was to evaluate the kinetics of pure (R)-[ 11C]verapamil in humans and to develop a tracer kinetic model for the analysis of P-gp-mediated transport of (R)-[11C]verapamil, including the putative contribution of its radioactive metabolites. Dynamic (R)-[ 11C]verapamil PET scans of 10 male volunteers were analysed with various single- or two-tissue compartment models, with separate compartments for N-dealkylated and N-demethylated metabolites, assuming that either (R)-[ 11C]verapamil alone or (R)-[11C]verapamil and any combination of metabolites cross the BBB. In addition, six of the subjects underwent two (R)-[11C]verapamil scans to evaluate test-retest reliability. One hour after injection, 50% of total plasma radioactivity consisted of labelled metabolites. Most models fitted the data well and the analysis did not point to a definite 'best' model, with differences in optimal model between subjects. The lowest mean test-retest variability (2.9%) was found for a single-tissue model without any metabolite correction. Models with separate metabolite compartments lead to high test-retest variability. Assuming that differences in kinetics of (R)-[11C]verapamil and N-dealkylated metabolites are small, a one input, one-tissue model with correction for N-demethylated metabolites only leads to a good compromise between fit quality and test-retest variability.