Limited Sampling Strategies Fail to Accurately Predict Mycophenolic Acid Area Under the Curve in Kidney Transplant Recipients and the Impact of Enterohepatic Recirculation

Abstract

Background: Therapeutic drug monitoring for mycophenolic acid (MPA) is challenging due to difficulties in measuring the area under the curve (AUC). Limited sampling strategies (LSSs) have been developed for MPA therapeutic drug monitoring but come with risk of unacceptable performance. The authors hypothesized that the poor predictive performance of LSSs were due to the variability in MPA enterohepatic recirculation (EHR). This study is the first to evaluate LSSs models performance in the context of EHR. Methods: Adult kidney transplant recipients (n = 84) receiving oral mycophenolate mofetil underwent intensive MPA pharmacokinetic sampling. MPA AUC0–12hr and EHR were determined. Published MPA LSSs in kidney transplant recipients receiving tacrolimus were evaluated for their predictive performance in estimating AUC0–12hr in our full cohort and separately in individuals with high and low EHR. Results: None of the evaluated LSS models (n = 12) showed good precision or accuracy in predicting MPA AUC0–12hr in the full cohort. In the high EHR group, models with late timepoints had better accuracy but low precision, except for 1 model with late timepoints at 6 and 10 hours postdose, which had marginally acceptable precision. For all models, the good guess of predicted AUC0–12hr (±15% of observed AUC0–12hr) was highly variable (range, full cohort = 19%–61.9%; high EHR = 4.5%–65.9%; low EHR = 27.5%–62.5%). Conclusions: The predictive performance of the LSS models varied according to EHR status. Timepoints ≥5 hours postdose in LSS models are essential to capture EHR. Models and strategies that incorporate EHR during development are required to accurately ascertain MPA exposure.