This analysis illustrates the importance of compliance in understanding intrapatient variation in plasma drug concentrations during 2 weeks of repeated (4 times daily) administration. Plasma concentration data are presented from 4 illustrative patients enrolled in a dose-ranging randomised clinical trial comparing diltiazem with placebo for the prevention of painful vaso-occlusive crises in sickle cell disease. Nonlinear regression was used to fit a 1-compartment model (using 1 elimination constant for the first dose and another for subsequent doses) to the observed diltiazem concentration curves for individual patients, taking into account the time of administration of each dose. Actual dosage intervals were obtained from an electronic device (the Medication Event Monitoring System). The parameters estimated from fitting the actual compliance curves were then used to predict the curves obtained if compliance had been as prescribed (perfect compliance curves). Comparison of the actual and perfect compliance curves shows that within-patient variation in plasma diltiazem concentrations over time can only be understood when the timing of drug administration is included in the analysis. We conclude that dynamic compliance data are important in those situations where close monitoring of treatment is required and may be important in population pharmacokinetic modelling when limited data are available from each patient.