A translational cardiovascular systems model to quantify drug effects on contractility and other hemodynamic variables
Yu Fu1, Tingjie Guo1, Hadi Taghvafard1, Medhat M. Said1, Cleo Demeester1, Victor Dudal1, Piet H. van der Graaf 1,2, Nelleke Snelder3, J. G. Coen van Hasselt 1
1 Leiden Academic Centre for Drug Research, Leiden University, the Netherlands; 2 Certara QSP, Canterbury, United Kingdom; 3 LAP&P Consultants BV, Leiden, the Netherlands
Objectives:
Cardiovascular safety issues are among the major causes of safety-related attrition in drug development.1 Early prediction and quantification of drug effects in the cardiovascular systems (CVS) are crucial to support decision making during drug development. The interpretation of preclinical data from cardiovascular safety studies, designed to determine the drug mode-of-action (MoA) for specific hemodynamic variables, can be challenging due to underlying complex homeostasis and feedback interrelationships.
A hemodynamic systems model was established for eight compounds with diverse modes-of-action in rat characterizing drug effects on the interrelationship between heart rate (HR), stroke volume (SV), total peripheral resistance (TPR), cardiac output (CO) and mean arterial pressure (MAP).2,3 In this model Invasive and challenging CO measurements were required, which limits the integration of this model into a translational modelling platform. Measures for contractility (CTR), such as dP/dtmax, could replace CO in the hemodynamic modelling, as it can be measured more easily using telemetry.
However, variability in hemodynamic readouts can influence the parameter estimation of baselines and circadian rhythms, which need to be quantified using placebo data. In addition, pharmacokinetic (PK) information is not always available in preclinical studies, which may not be necessary in the development of the hemodynamic model.
Therefore, we aimed to 1) develop a CVS contractility model using atenolol as proof-of-concept, and investigate the identifiability of this model to evaluate if the MoA of new drugs can be identified; 2) quantify the baselines and circadian rhythm effects of each hemodynamic variable using placebo data, as a basis for initial estimates in systems modelling; 3) apply the established CVS contractility model to multiple compounds to further calibrate the system-specific parameters, and evaluate the contribution of PK information in the prediction of hemodynamics; 4) develop a user-friendly web application to predict drug effect based the established model.
Methods:
Development of the CVS contractility model:
Previously collected experimental data was used to develop the model, and included measurements for HR, CO, MAP and dP/dtmax after administration of atenolol (0.3-30 mg/kg) from three in vivo telemetry studies in conscious Beagle dogs. Interrelationships between HR, SV, TPR, CO and MAP are expressed as: 1) CO = HR * SV; 2) MAP = TPR * CO.2,3 Principles of pressure-Volume (PV) loop theory was used as a basis for integration of CTR in the model.4 The model consists of four differential equations for HR, end-diastolic volume (EDV), CTR and TPR with negative feedback from MAP to the dynamics of HR, CTR and TPR.
Subsequently, we evaluated the structural identifiability of the CVS model using the MATLAB (version R2020a) with toolbox GenSSI 2.0, with two different sets of observations, (HR, dP/dtmax, CO and MAP) and (HR, dP/dtmax and MAP).5 Stochastic simulation and estimation (SSE) using Perl-speaks-NONMEM (PsN, version 4.8.1, Uppsala University, Sweden) were conducted to determine if the CVS-contractility model can be used to identify the MoA of new drugs.4
Quantification of baselines and circadian rhythm (CR) effects:
We obtained data of placebo groups retrospectively from 37 in vivo telemetry studies in conscious Beagle dog, including individual measurements of MAP, HR, and dP/dtmax. Each experimental study was conducted using 4-8 animals. Cosine functions were used to describe CR in hemodynamics.
Application of CVS contractility systems model:
Previously collected experimental data for verapamil, flecainide, sumatriptan, pimobendan, and atenolol were used to develop the systems model, including measurements for HR, MAP, CO and dP/dtmax from 6 preclinical in vivo telemetry studies in conscious Beagle dog. Linear and Emax models were used to describe the drug effect on HR, CTR and/or TPR. Predictive performance of both kinetic(K)-PD and PKPD models were evaluated using goodness-of-fit and visual predictive check plots.
Shiny application:
A web application was developed using the R package Shiny and other widgets packages (shinydashboard, shinyalert and shinyWidgets). The RxODE R package was used to perform the simulations.6 Simulation of PK, HR, CO and MAP for reference drugs and investigational drugs can be plotted together with user’s uploaded dataset for comparison of different MoA.
Results:
Development of the CVS contractility model:
The developed CVS contractility model adequately described the effect of atenolol on HR, CO, dP/dtmax and MAP dynamics and allowed identification of both system- and drug-specific parameters with good precision. Model parameters were structurally identifiable, and the true mode of action can be identified properly.4
Quantification of baselines and CR effects:
The established model well described the circadian rhythm in MAP, HR, and dP/dtmax profiles and quantified the inter-individual and inter-site variability on each parameter. In the model-based power analysis, the minimal drug effect needed for the model without CR were 2-7 folds higher than that of the model with CR, indicating that model with CR can better identify the drug effect compared to model without CR.
Application of CVS contractility systems model:
The developed CVS contractility systems model adequately described the drug effect on HR, CO, dP/dtmax and MAP dynamics for all five compounds. The estimated system-specific parameters (Kout and feedback) were consistent with the ones in atenolol model. The MoA of each compound was found as: 1) Verapamil: inhibition on both CTR and TPR; 2) Flecainide: inhibition on CTR; 3) Sumatriptan: inhibition on CTR; 4) Pimobendan: inhibition on TPR; 5) Atenolol: inhibition on HR and CTR. There was no significant difference between the prediction of KPD and PKPD models.
Shiny application:
The shiny app was developed and the source code is available on Github. Model-based prediction can be obtained following these steps:
- select species and strain for simulation;
- select reference drug and define dosage for simulation;
- simulate investigational drug by define drug-specific parameters and dose regimens;
- input data file for plotting following pre-set dataset template;
- generate a PDF report with the plots and all the information for simulation.
Conclusions:
The developed model characterizing the interrelationships between contractility and other hemodynamic biomarkers, can be used to quantify the drug effects on hemodynamics for multiple compounds with diverse MoA. PK information was not required to identify the drug effects. The shiny application with user-friendly interface can help safety evaluation and decision making in drug development. Ultimately, a set of system-specific parameters was identified for dog, so that the developed model has the potential to be of relevance to support translational cardiovascular safety studies.
References:
[1] Munos, B. Lessons from 60 years of pharmaceutical innovation. Nat. Rev. Drug Discov. 8, 959–968 (2009).
[2] Snelder, N. et al. PKPD modelling of the interrelationship between mean arterial BP, cardiac output and total peripheral resistance in conscious rats. Br. J. Pharmacol. 169, 1510–1524 (2013).
[3] Snelder, N. et al. Drug effects on the CVS in conscious rats: Separating cardiac output into heart rate and stroke volume using PKPD modelling. Br. J. Pharmacol. 171, 5076–5092 (2014).
[4] Fu, Y. et al. A novel cardiovascular systems model to quantify drugs effects on the interrelationship between contractility and other hemodynamic variables. CPT Pharmacometrics Syst. Pharmacol. 1–13 (2022).doi:10.1002/psp4.12774
[5] Ligon, T. S. et al. GenSSI 2.0: Multi-experiment structural identifiability analysis of SBML models. Bioinformatics 34, 1421–1423 (2018).
[6] Wang, W., Hallow, K. M. & James, D. A. A tutorial on RxODE: Simulating differential equation pharmacometric models in R. CPT Pharmacometrics Syst. Pharmacol. 5, 3–10 (2016).