Drug clearance during Continuous Veno-Venous Hemofiltration (CVVH). A mathematical model based on ex-vivo experiments.

摘要

Introduction: Continuous veno-venous hemofiltration (CVVH) with an AN69 membrane is used in hemodynamical unstable patients with acute renal failure admitted to the Intensive Care Unit (ICU) of the Catharina Hospital in Eindhoven (CHE), the Netherlands. Literature search revealed that only the clearance and kinetics of some antibiotics during CVVH were studied. Data obtained with different renal replacement techniques and membranes can't easily be extrapolated to CVVH with an AN69 membrane. (Reetze-Bonorden et al. 1993; Mueller et al., 2003) Due to the lack of clinical research data, dosages in patients treated with CVVH at the ICU of the CHE or are not adjusted or are adjusted on the base of the empirical determined creatinin clearance of 30 ml/min. For aminoglycosides and vancomycin dose adjustment is performed applying therapeutic drug monitoring. Because of this a lot of drugs are possibly suboptimal dosed during CVVH. Aim: The aim of this study was to develop a predictive mathematical model based on ex-vivo experiments for the drug clearance by CVVH with an AN69 high-flux membrane. Methods: The experimental set-up consisted of a PRISMA CFM hemofiltration machine with a AN69 HF 0,9 m(2), 500 ml heparinised bovine blood as blood compartment, blood flow rate 150 ml/min and ultrafiltration rate 1500 ml/h. The following drugs were added to the blood compartment in therapeutic concentrations, covering a broad range in molecular weight (Mw), % protein binding (%PB), logP and charge (C): amiodarone, amitriptyline, carbamazepine, cyclosporine, clozapine, digoxin, phenobarbital, phenytoin, gentamycin, midazolam, acetaminophen, theophylline, tobramycin, valproic acid and vancomycin. The sieving coefficient (SC) was calculated by SC=(2x concentration ultrafiltrate) / (concentration post membrane + concentration pre membrane). Results: All four parameters; %PB, logP, C and Mw contributed statistically significant (p < 0,05) to the SC. The regression line, explaining 75% of the variance (R(2)=0,751) is: SC=0,826 - 0,06%PB - 0,055 logP - 0,085C + 0,000095MW Conclusion: Based on the ex-vivo experiments a mathematical model has been extracted to predict a-priori the sieving coefficient of drugs during CVVH with an AN69 high-flux membrane on the base of protein binding, charge, logP and molecular weight. This model has to be validated in vivo and can then be used, if valid, for drugs which physico-chemical properties in the ranges studied: p%PB 0 - 98%, logP -7,32 - 7, Mw 151 - 1500, C -1 - 3.