Assessment of the glomerular filtration rate (GFR) in kidney transplant recipients using Bayesian estimation of the iohexol clearance

Camille Riff,Joevin Besombes,Philippe Gatault,Christelle Barbet,Matthias Büchler,Hélène Blasco,Jean-Michel Halimi,Chantal Barin-Le Guellec,Isabelle Benz-de Bretagne

DOI: https://doi.org/10.1515/cclm-2019-0904

2020-01-11

Clinical Chemistry and Laboratory Medicine (CCLM)

Abstract:Abstract Background Plasma iohexol clearance (CL iohexol ) is a reference technique for glomerular filtration rate (GFR) determination. In routine practice, CL iohexol is calculated using one of several formulas, which have never been evaluated in kidney transplant recipients. We aimed to model iohexol pharmacokinetics in this population, evaluate the predictive performance of three simplified formulas and evaluate whether a Bayesian algorithm improves CL iohexol estimation. Methods After administration of iohexol, six blood samples were drawn from 151 patients at various time points. The dataset was split into two groups, one to develop the population pharmacokinetic (POPPK) model (n = 103) and the other (n = 48) to estimate the predictive performances of the various GFR estimation methods. GFR reference values (GFR ref ) in the validation dataset were obtained by non-compartmental pharmacokinetic (PK) analysis. Predictive performances of each method were evaluated in terms of bias (ME), imprecision (root mean square error [RMSE]) and number of predictions out of the ±10% or 15% error interval around the GFR ref . Results A two-compartment model best fitted the data. The Bayesian estimator with samples drawn at 30, 120 and 270 min allowed accurate prediction of GFR ref (ME = 0.47%, RMSE = 3.42%), as did the Brøchner-Mortensen (BM) formula (ME = − 0.0425%, RMSE = 3.40%). With both methods, none of the CL estimates were outside the ±15% interval and only 2.4% were outside the ±10% for the BM formula (and none for the Bayesian estimator). In patients with GFR ≤30 mL/min/1.73 m 2 , the BM formula performed very well, while the Bayesian method could not be evaluated in depth due to too small a number of patients with adequate sampling times. Conclusions GFR can be estimated with acceptable accuracy in kidney transplant patients using the BM formula, but also using a Bayesian algorithm.