Prediction of unbound mycophenolic acid concentrations in patients after hematopoietic cell transplantation

A model for predicting unbound concentrations of mycophenolic acid (MPA) was developed in 84 subjects and then tested in an independent group of 19 hematopoietic cell transplant recipients. Total and unbound MPA concentrations and total mycophenolic glucuronide concentrations were measured between weeks 1 and 4 posttransplant. The relationships between MPA unbound concentrations and common clinical biochemical markers were investigated. Mean prediction error and root mean square error of the model were calculated to assess prediction bias and precision, respectively The best model for estimation of unbound concentrations included serum creatinine (SCr) and total bilirubin. The model for predicting unbound MPA fraction percentage was [0.928 + (serum creatinine mg/dL) x 0.392] (x 1.48 if total bilirubin >= 3 mg/dL). Unbound MPA concentration (ng/mL) then can be estimated by total MPA concentration (ng/mL) x predicted MPA unbound percentage. Serum albumin was not significant in the regression model. There was no significant correlation between mycophenolic acid glucuronide and unbound MPA. This model demonstrated low bias (mean prediction error, 0.81 ng/mL) but poor precision (root mean square error, 19.49 ng/mL). In the validation population, this model showed poor bias and precision (mean prediction error, -5.42 ng/mL; root mean square error, 35.71 ng/mL). The models overestimated the measured unbound concentrations in the study population but underestimated in the validation population and did not provide satisfactory predictions for clinical assessments. Direct measurements of unbound MPA concentrations are necessary in patients requiring unbound concentration monitoring.