Discrepancies between observed and predicted continuous venovenous hemofiltration removal of antimicrobial agents in critically ill patients and the effects on dosing

Objective: Drug dosing during continuous venovenous hemofiltration (CVVH) is based partly upon the CVVH clearance (Cl-CVVH) of the drug. Cl-CVVH is the product of the sieving coefficient (SC) and ultrafiltration rate (Q(uf)). Although it has been suggested that the SC can be replaced by the fraction of a drug not bound to protein (F-up), the Fup values as reported in the literature may not reflect the protein binding in critically ill patients with renal failure. We compared the observed Cl-CVVH (SC x Q(uf)) with the estimated Cl-CVVH ( estimated F-UP x Q(uf)) and determined the effect on the maintenance dose multiplication factor (MDMF). Design and setting: Clinical study in a mixed ICU in a university hospital. Patients: 45 oligoanuric patients on CVVH (21/h). Interventions: Timed blood and ultrafiltrate samples. Measurements and results: Amoxicillin, ceftazidime, ciprofloxacin, fluconazole, metronidazole, and vancomycin were easily filtered ( mean SC > 0.7) but not flucloxacillin (mean SC 0.3). Predicted and observed Cl-CVVH corresponded only for fluconazole and metronidazole. The difference between observed and predicted MDMF was small for all drugs, with the exception of ceftazidime (mean 0.25, 95% CI -0.96 to 1.48) and vancomycin (0.05, -1.34 to 1.45). However, this difference was clinically relevant only for vancomycin, because of its narrow therapeutic index. Conclusions: Dosing based on predicted CVVH removal provides an as reliable estimate than that based on observed CVVH removal except for those antibiotics that have both a narrow therapeutic index and a predominantly renal clearance (e.g., vancomycin).