PII-062 - COMPREHENSIVE PHARMACOKINETIC ANALYSIS OF ISONIAZID IN A LARGE COHORT OF TUBERCULOSIS PATIENTS: ADVANCED MODELING INCORPORATING EXTENSIVE CLINICAL COVARIATES.

Y. Cho^1,2,3, R. Jayanti^1,2, H. Kim^4,5, J. Ghim^1,2,3, D. Kim^1,2, J. Shin^1,2,3; ¹Center for Personalized Precision Medicine of Tuberculosis, Inje University College of Medicine, Busanjin-gu, Busan, Republic of Korea, ²Department of Pharmacology and PharmacoGenomics Research Center, Inje University College of Medicine, Busanjin-gu, Busan, Republic of Korea, ³Department of Clinical Pharmacology, Inje University Busan Paik Hospital, Busanjin-gu, Busan, Republic of Korea, ⁴Center for Personalized Precision Medicine of Tuberculosis, Inje University College of Medicine, Busanjin-gu. Busan, Republic of Korea, ⁵Department of Pharmacology and PharmacoGenomics Research Center, Inje University College of Medicine, Busanjin-gu. Busan, Republic of Korea.

Background: Previous pharmacokinetic (PK) models for isoniazid (INH) have been limited by small sample sizes and insufficient clinical covariates, hindering their applicability to diverse patient conditions. This study aims to develop a more robust and generalizable PK model by incorporating a comprehensive set of covariates to better explain variability in tuberculosis (TB) treatment.

Methods: Data from 1,544 Korean TB patients were divided into training and test sets (4:1 ratio). A population PK model was created using a nonlinear mixed-effects approach, integrating demographic, clinical, and genetic factors influencing INH metabolism. Internal validation was performed with visual predictive checks and nonparametric bootstrap, while external validation assessed prediction errors in the test set and edge populations

Results: A two-compartment model with allometric scaling effectively described INH PK. Key covariates included NAT2 acetylator status, estimated glomerular filtration rate (eGFR), fed state, and age ≥ 70 years. INH clearance (CL/F) was 24.9, 17.8, and 9.02 L/h for rapid, intermediate, and slow acetylators, respectively. In slow acetylators, moderate-to-severe renal impairment reduced clearance by 25.2%. Both internal and external validations confirmed the model’s superior predictive accuracy, especially in edge populations

Conclusion: This study developed a robust and generalizable PK model for INH by incorporating a wide range of clinical covariates. The model demonstrated excellent predictive accuracy, particularly for patients with multiple clinical covariates, providing valuable insights for personalized TB treatment