Modeling the effects of bivalirudin in cardiac surgical patients

Bivalirudin is direct thrombin inhibitor used in patients with heparin-induced thrombocytopenia. A pharmacokinetic and--dynamic model that predicts the partial thromboplastin time (PTT) based on the past infusion rates of bivalirudin following dose adjustment would be useful to guide optimal therapy. In this retrospective study we randomized 132 patients to a derivation and a validation cohort, and tested two models. The first model is a single-state linear model; the other incorporates a non-linear element to account for renal elimination of bivalirudin. Both models predicted PTT changes equally well with root-mean squared errors of 15 to 16 seconds (Pearson correlation coefficients for both were 0.67). Intra- and inter-individual variability of response to bivalirudin was significant. Although a high percentage of patients had moderate to severe renal dysfunction at one point during the bivalirudin infusion, the non-linear model that incorporates variable renal clearance of drug did not perform better than the linear model. This finding persisted even in the subgroup analysis of patients with moderate and low estimated glomerular filtration rates.