N.O. Hoem (1), P. Beringer (2)
HPC, Oslo, Norway; (2) School of Pharmacy, University of Southern California, Los Angeles, U.S.A.
Objectives: Population PK of linezolid (LNZ) has been characterized in several relatively recent publications, but with divergent conclusions regarding linear or non-linear elimination. This analysis applies a non-parametric EM method (Non Parametric Adaptive Grid – NPAG) to data densely collected from patients administered LNZ both orally and intravenously. The study design permitted the fraction absorbed (Fa) to be explicitly defined in the model.
Methods: PK-data were from a prospective, randomized, cross-over study in 17 hospitalized patients; 11 receiving continuous enteral feedings and 6 serving as controls. Patients had been were randomized to receive a single dose of either LNZ 600mg IV or 600mg PO suspension, with a 48 hour washout period between dosage forms. Data were analyzed with an iterative-two stage Bayesian (IT2B) method to estimate initial parameter ranges and subsequently with the NPAG method. Model discrimination was determined using log-likelihood values, mean error (ME), root mean squared error (RMSE) and Akaike’s information criteria. Maximum a-posteriori Bayesian analysis was performed to determine the individualized pharmacokinetic parameters. Estimates of Fa and Total clearance were compared with those obtained by non-compartmental analysis.
Results: In this population of elderly hospitalised patients the oral and/or enteral absorption of LNZ was, regardless of PK-model applied, found to be fast (ka >2 ) and complete (Fa = 1). Volume of distribution was linearly correlated with total body weight for all models tested. Creatinine clearance was not found to correlate with any model-parameter. For models incorporating a distributive compartment, elimination was very well described by either a simple linear or by a Michaelis-Menten (MM) saturable model. Administration of the same dose in all experiments restricted the actual analysis of MM parameters to the Vm/Km ratio (Km fixed at 3.6 mM). In general, non-distributive models gave substantially poorer model fits. Interestingly, for such models, linear-elimination gave superior fits as compared to MM-elimination. Although the predictions given by the individualized PK parameters were excellent (R2>0.97), mean-parameters predicted individual observed levels well (R2>0.8) in both MM and linear 2-compartment models.
Conclusion: At doses below 600mg in this population of elderly hospitalised patients, analysis by the NPAG-algorithm demonstrated excellent predictions for 2-compartment distributive models both with MM- or with linear elimination. The MM- had less bias than the linear model. Taken together, this might indicate that the elimination of linezolid is indeed saturable at high plasma levels, as will be present initially before distribution equilibrium (Deq) is achieved, but becomes close to linear at levels usually present after Deq. Individual variability, at least in this group of patients, seems to be small enough to make the population mean parameter estimates useful for individual prediction.
Analysis of Fa could be directly incorporated into the model, and gave results that were essentially identical to a conventional comparison of AUCs. Explicit determination of Fa with compartmental modelling offers the ability to perform bioavailability assessments in patients with intermixed IV/oral dosage regimens (no washout necessary).
Reference: PAGE 13 (2004) Abstr 525 [www.page-meeting.org/?abstract=525]
Poster: poster