Combining an in silico proarrhythmic risk assay with a tPKPD model to predict QTc interval prolongation in the anesthetized guinea pig assay

Abstract

Human-based in silico models are emerging as important tools to study the effects of integrating inward and outward ion channel currents to predict clinical proarrhythmic risk. The aims of this study were 2-fold: 1) Evaluate the capacity of an in silico model to predict QTc interval prolongation in the in vivo anesthetized cardiovascular guinea pig (CVGP) assay for new chemical entities (NCEs) and; 2) Determine if a translational pharmacokinetic/pharmacodynamic (tPKPD) model can improve the predictive capacity. In silico simulations for NCEs were performed using a population of human ventricular action potential (AP) models. PatchXpress® (PX) or high throughput screening (HTS) ion channel data from respectively n = 73 and n = 51 NCEs were used as inputs for the in silico population. These NCEs were also tested in the CVGP (n = 73). An M5 pruned decision tree-based regression tPKPD model was used to evaluate the concentration at which an NCE is liable to prolong the QTc interval in the CVGP. In silico results successfully predicted the QTc interval prolongation outcome observed in the CVGP with an accuracy/specificity of 85%/73% and 75%/77%, when using PX and HTS ion channel data, respectively. Considering the tPKPD predicted concentration resulting in QTc prolongation (EC_5%) increased accuracy/specificity to 97%/95% using PX and 88%/97% when using HTS. Our results support that human-based in silico simulations in combination with tPKPD modeling can provide correlative results with a commonly used early in vivo safety assay, suggesting a path toward more rapid NCE assessment with reduced resources, cycle time, and animal use.

Keywords: Anesthetized Cardiovascular Guinea Pig; In Silico modeling; QT corrected interval; Safety pharmacology; Torsade de Pointes; Translational PKPD modeling.

Fig. 1

In silico APD₉₀ and EMw. The EMw is defined here as the difference between the Ca²⁺ Transient Duration at 90% repolarization (CTD₉₀) and the action potential duration at 90% repolarization (APD₉₀).

Fig. 2

Matrix results (with 95% confidence intervals indicated) of hERG and MK-499 IC₂₀ values relative to the anesthetized CVGP QTc EC₅ using ion channel data generated using A) the PX platform (n = 73), and B) the HTS platform (n = 51). PPV: Positive predictive value, NPV: Negative predictive value.

Fig. 3

Matrix results (with 95% confidence intervals indicated) using the in silico model endpoints (APD₉₀ and EMw) relative to the CVGP QTc EC₅ using in vitro ion channel data obtained from PX (A) or HTS (B) and matrix results when adjusted with tPKPD model projected QTc EC₅ using PX (C) or HTS (D) ion channel data. PPV: Positive predictive value, NPV: Negative predictive value.

Fig. 4

Translational Pharmacokinetic/Pharmacodynamic (tPKPD) results: A) tPKPD projected QTc EC₅ as compared to actual CVGP QTc EC₅ when using PX ion channel data. B) tPKPD projected QTc EC₅ as compared to the actual CVGP QTc EC₅ when using high throughput (HTS) ion channel data. Data points are labeled with their NCE number found in Table 3, Table 4, Table 5.