Pieter Colin, Stijn Jonckheere, Hans De Beenhouwer, Nikolaas De Neve, An Vermeulen, Jan Van Bocxlaer
Ghent University
Objectives: Cefepime (CFP), is extensively used for serious infections in the intensive care unit. Several groups described extensive variability in plasma concentrations and a high proportion of ICU patients not achieving suitable PKPD targets following currently used clinical treatment regimens. Nevertheless, due to discrepancies between published models for CFP and uncertainties with respect to the correlation between CFP clearance and different GFR biomarkers, therapeutic drug monitoring is not yet routinely implemented for this compound. We set out to compare the predictive performance of the different published models and propose a new model / sampling strategy to guide CFP dosing in the ICU.
Methods: Plasma and urine samples were obtained from patients who were treated with CFP, per standard of care, in the ICU of the OLVZ hospital in Aalst, Belgium. Patients were treated with a median of 10 doses resulting in a median follow-up of 80 hours. Using NONMEM, median prediction error and root mean square error were chosen as a performance metric to compare the different models. As a step up from the available models we investigated whether adding different biomarkers for GFR as time-varying covariates could significantly improve the model’s predictive performance.
Results: The predictive performance of earlier published models was found unsatisfactory in our patient cohort. Furthermore, the non-renal clearance of CFP, which accounts for approx. 1/3 of CLTOT is frequently ignored, making accurate predictions of CFP clearance impossible. Since, in our institution, 1/5 patients on CFP undergo hemodialysis we developed a PopPK model including in addition to parameters describing the renal and non-renal clearance, a CLDialysis parameter describing the kinetics of the “on-dialysis” moments. Finally, by incorporating CKD-EPI as a time-varying covariate on the renal CFP clearance, the model’s goodness-of-fit was further improved.
Conclusions: A therapeutic drug monitoring-based adaptation of our PopPK model could allow to adequately control the level of variability in this vulnerable patient population. Nevertheless, more accurate, non-serum creatinine related, GFR estimators might be more suitable to follow-up on changes in CFP clearance over time (and are currently under evaluation), reducing the necessity of daily TDM measurements.
Reference: PAGE 24 () Abstr 3600 [www.page-meeting.org/?abstract=3600]
Poster: Drug/Disease modeling - Infection