Population PK and PKPD modeling of TBA-7371 in Healthy Adults and Adults with tuberculosis
Shayne Watson (1), Micha Levi (1), David Holtzman (1), David Salinger (2), Kelly Stinson (1), Aparna Anderson (1), Elizabeth Grekas (1), Charles Wells (1)
(1) Bill and Melinda Gates Medical Research Institute, (2) Certara
Objectives:
- Develop a population pharmacokinetics (popPK) model that characterizes the PK of TBA-7371, a drug showing a potent bactericidal anti-Mycobacterium tuberculosis (Mtb) activity in vitro, from Phase 1 and Phase 2 trials.
- To describe the relationship between TBA-7371 PK parameters and colony forming units (CFU) of Mtb on solid culture and time to positivity (TTP) of Mtb in a liquid culture system.
Methods:
PopPK Model:
Data from a Phase 1 and Phase 2 trial was used to fit a popPK model. Phase 1 data were previously observed to have apparent auto-induction of clearance and was described with a 2 compartment popPK model with first-order absorption, and parallel linear and non-linear elimination (via Michaelis-Menten (MM) kinetics) and a hypothetical enzyme compartment to drive auto-induction of the non-linear elimination.
The following covariates were included in the model:
- Dose on absorption rate (KA)
- Fed versus Fasted on KA
- Weight on clearance (CL), volumes of distribution (VC and VP), and inter-compartmental clearance (Q).
This model structure was used as the base model for the pooled Phase 1 and Phase 2 data. Additional covariates (sex, age, dose, and race) were tested based on the correlation of the covariates with the individual random effects (ETA) from the base model. Additional models evaluated estimation of the allometric exponents on CL, VC, Q, and VP.
PKPD Models:
PK exposure metrics for individual participants on Day 14 from the Phase 2 trial were utilized for PK/PD modeling. To assess the influence of exposure on pharmacodynamic (PD) responses (CFU and TTP), both exploratory graphical analysis and regression analysis were used. The variety of regression models investigated comprised of linear, Michaelis-Menten (MM), Hill, shifted Michaelis-Menten, and shifted Hill structures.
Results:
PopPK Model:
Exploratory data analysis suggested Phase 2 data are in line with expectations based on the Phase 1 data. The base model appeared to describe the data reasonably well. ETAs were normally distributed with minimal shrinkage.
The effects of covariates were evaluated, including the impact of sex on VC, race on KA, race on Vmax, race on VC, and the calculation of the allometric exponent. The model that performed best according to the Bayesian Information Criterion (BIC) was the base model, indicating that adding further complexity does not significantly enhance model performance.
PK/PD Analysis:
The cohort receiving 100 mg TID exhibited the most significant decrease in log10CFU and the greatest increase of TTP on Day 14, with the 400 mg QD cohort following. No clear pattern was observed between the change from baseline (CFB) in log10CFU and the Cmax. However, there was a slight trend observed between CFB in TTP and the quartiles of Cmax. The AUC0-24, Cmin, and time above the preclinical MIC90 (TMIC90) all indicated a trend for both CFU and TTP.
TMIC90 was the best correlated exposure measure; however, maximum values of 24 hours were observed for more than 25% of participants making it hard to distinguish differences in the effect of high exposure on PD outcomes. Thus, Cmin was used for PK/PD modelling.
Four Emax models were estimated using Day 14 Cmin as the predictor. The best performing model by BIC was the MM model for both CFB of log10 CFU and TTP. These models had EC90 values of 2496.73 ng/mL for CFB of log10 CFU and 1022.4 ng/mL for CFB of TTP.
Conclusion:
The initial popPK model, created with data from Phase 1, effectively characterized the pooled phase 1 and 2 data. A covariate analysis did not find additional covariates that significantly improved the model's fit. Consequently, the base model was deemed adequate and selected as the final model.
PK/PD modeling identification of Cmin and TMIC90 as the primary PK drivers of efficacy. CFB log10 CFU appeared to decrease and TTP appeared to increase as Cmin, and TMIC90 increased. The relationship was strongest in TMIC90. However, TMIC90 had limited distinction of high exposure participants.
The EC90 for Cmin were 2496.73 ng/mL for CFB of log10 CFU and 1022.4 ng/mL for CFB of TTP. This suggests that for future dosing should have a median Cmin above 2496.73 ng/mL.