Vancomycin population pharmacokinetic analysis in patients with hematological malignancies

PDF of poster

Introduction: The population approach and pharmacodynamic criteria have become available as tools in individualized antimicrobial therapy, leading to increased efficacy and reduced selection of resistance. However, little is known about the pharmacokinetics of vancomycin (VAN) in patients with hematological malignancies.

Objective: This study aimed  to develop  a VAN population pharmacokinetics model in a broad and representative sample of adult patients with hematological malignancies.

Methods: Sparse serum concentration data (n = 1128) from therapeutic drug monitoring of VAN were collected retrospectively from 274 patients. The dataset was divided into two groups according to a sequential time criterion. An index group of 286 courses administered to 215 patients was used for population PK modeling. The remaining 62 courses administered to 59 patients were used as a validation data set.

A one-compartment PK model was selected and  population parameters were generated using the NONMEM program. A graphical approach and stepwise generalized additive modeling (GAM) were used to elucidate the preliminary relationships between PK parameters and the covariates: age, gender, weight, body surface area, serum creatinine, hemoglobin, albumin, time post-chemotherapy, hematology diagnosis, ECOG status, stage of antineoplastic treatment, autologous bone marrow transplantation, neutropenia and concurrent amikacin or amphotericin therapy.

Results: Covariate selection revealed that total body weight (TBW) affected V whereas renal function, estimated by creatinine clearance, and a diagnosis of acute myeloblastic leukemia (AML) influenced VAN clearance. We propose one general and  two specific models customized for AML patients. The former was defined by: CL (L/h) = 1.08 * CL_cr,_{Cockroft and Gault} (L/h); CVCL = 28.16 % and V (L) = 0.98*TBW; CVV = 37.15 %. The AML models confirmed this structure but with a higher clearance coefficient (1.17).

The a priori performance of the models was evaluated in another 59 patients, and their clinical suitability was confirmed. The corresponding standardized prediction errors included zero and an SD close to unity.

Conclusions: The proposed models could be used to estimate appropriate VAN dosage guidelines, which are not clearly defined for this high-risk population. Their simple structure should allow easy implementation in clinical software and application in dosage individualization using the Bayesian approach.