Takayuki Katsube, Toshihiro Wajima
Shionogi & Co., Ltd., Japan
Objectives: Understanding a contribution of each covariate effect to an interested index (e.g., maximum concentration [Cmax], area under the concentration-time curve) is important to assess an impact of each covariate effect, e.g., for dose adjustment based on the covariates. A variance of inter-individual variability (IIV) for the index is considered to be variances from individual covariates and from individual random variables which are defined as ETA in NONMEM [1]. Variance ratios using global sensitivity analysis approach [2, 3] can be useful to efficiently assess the contribution of each covariate effect and each random effect on the index. In this study, the contribution of covariate effect and random effect on the index was evaluated based on the variance ratios using global sensitivity analysis approach.
Methods: Monte-Carlo simulations were conducted to simulate covariates and ETAs, calculate outputs for an index based on the simulated covariates and ETAs, and calculate ratios of the variance explained by each covariate or of the variance explained by each random variable of ETAs over the overall variance of the index using soboljansen in R library ”sensitivity” [4]. As a simple example model, an oral 1-compartment pharmacokinetic (PK) model with mono-exponential covariate model on each parameter and no IIV for any PK parameters was used to assess the variance ratios of each covariate or ETA to Cmax or time above a specified concentration (T>C) at steady state. As an assessment based on real data, the variance ratio of each covariate or ETA to peak platelet count was assessed using the pharmacokinetic/pharmacodynamic (PK/PD) covariate model (oral 3-compartment PK and 4-compartment pharmacodynamic [PD] linked model) of lusutrombopag, a thrombopoietin receptor agonist to induce thrombopoiesis [5]. The PK/PD model included the effects of body weight on apparent total clearance (CL/F) and apparent volume of distribution (V/F) with power models. Body weight (covariate) was resampled from the original dataset, and ETAs for PK/PD parameters were simulated according to the parameter estimates.
Results: The variance ratios using the oral 1-compartment model demonstrated the variability for Cmax was highly dependent on the covariate on V/F, and the variability for T>C was highly dependent on the covariate on CL/F. These results were consistent with visual inspections based on the simulated concentration profiles. The assessments based on the PK/PD model of lusutrombopag suggested that the variability for peak platelet count was mainly dependent on ETA of 50% effective concentration (EC50) and ETA of baseline platelet count, accounting for 35% with ETA of EC50 and 55% with ETA of baseline platelet count. Body weight did not affect the variability for peak platelet counts, suggesting no clinically relevant effect of body weight on the platelet counts, although body weight was an influential factor for the exposure of lusutrombopag.
Conclusions: The contribution of covariate effects to the index was quantified based on the variance ratios. The variance-based global sensitivity analysis would be proposed as an efficient approach to assess the impact of covariate effect.
References:
[1] Beal SL et al. NONMEM Users Guides. Icon Development Solutions, Ellicott City, Maryland, USA. 1989-2011.
[2] Saltelli A et al. Comput Phys Commun. 2010. 18:259-270.
[3] Zhang XY et al. CPT Pharmacometrics Syst Pharmacol. 2015. 4:69-79.
[4] https://CRAN.R-project.org/package=sensitivity
[5] Katsube T et al. Clin Pharmacokinet. 2016. 55:1423-1433.
Reference: PAGE 28 (2019) Abstr 8819 [www.page-meeting.org/?abstract=8819]
Poster: Methodology - Covariate/Variability Models