Population pharmacokinetic study of colistin for dose optimization in patients with acute on chronic liver failure (ACLF) patients

• Published in: International journal of infectious diseases Vol. 152; p. 107691
• Main Authors: Bhandari, Dr Ritika, Shafiq, Prof. Nusrat, Murali, Dr. Naveen, Vij, Dr. Soumya, Verma, Dr. Nipun, Pandey, Dr. Avaneesh Kumar, Gota, Prof. Vikram Prakash, Rajan, Dr. M Surulivel, Duseja, Prof. Ajay Kumar, Malhotra, Prof. Samir
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.03.2025; Elsevier
• ISSN: 1201-9712
• DOI: 10.1016/j.ijid.2024.107691

Acute on chronic liver failure (ACLF) is characterized by severe systemic inflammation and multiple organ failures, posing high mortality risks, particularly in cirrhotic patients. The complex alterations in drug distribution and elimination kinetics observed in ACLF and sepsis patients often result in considerable variability in antibiotic concentrations. Conducting a population pharmacokinetic (PK) study among individuals with hepatic dysfunction is crucial for guiding initial antibiotic dosing in this patient population. Here, we aimed to conduct a population pharmacokinetic study of colistin in patients with hepatic failure to elucidate drug kinetics and optimize dosing. We enrolled patients with ACLF receiving colistin for suspected or proven hospital-acquired infections caused by multi-drug resistant Gram-negative bacteria. A sparse sampling approach was employed for pharmacokinetic analysis. Clinical data, including body weight, serum creatinine, calculated creatinine clearance (using the Cockcroft formula), and various liver disease severity scores (MELD, AARC score, CLIF-C ACLF score), were collected. Colistin plasma concentrations were measured using LCMS/MS. We developed a population PK model using nonlinear mixed-effect modelling with the software program PUMAS AI. Fourteen patients (mean age: 45.64±11.90, Male/Female) were included in the study. A one-compartment model best described the data. Random variability was accounted for in both clearance and volume of distribution. The combined error model (additive and proportional) provided the best fit for the data. The final model-fitted parameter estimates were: Clearance (Cl) = 1.84 ml/min; Volume of distribution (V) = 41.15 ml with a proportional error of 50% and an additive error of 0.59 mcg/ml. Between-subject variability was observed to be 124% for clearance and 44% for volume of distribution. Covariate model building, utilizing forward addition and backward elimination, was initiated. Body weight and serum creatinine were found to be non-significant covariates, while creatinine clearance was found to be significant in decreasing the objective function value and improving goodness of fit plots of the model. Understanding the pharmacokinetic properties of a drug is crucial for optimizing dosing regimens and ensuring therapeutic efficacy. Preliminary analysis of the covariate model indicated creatinine clearance as a significant covariate, which will be further explored. The findings suggest significant between-subject variability in drug clearance and volume of distribution, highlighting the importance of individualized dosing regimens. The preliminary findings of this population PK study underscore the necessity for dose optimization of colistin in ACLF patients, given the fluctuating pharmacokinetic parameters attributed to complex bodily changes. Further population pharmacokinetic modelling, incorporating significant covariates, holds promise in devising an optimized dosing regimen tailored to this patient cohort.