B. Martin (1), E. Courcelles (1), A. L’Hostis (1), L. Etheve (1), E. Bechet (1), N. Ceres (1), E. Jacob (1), E. Peyronnet (1), B. M.W. Illigens (1), M. Hommel (1), J.P. Boissel (1)
1. Novadiscovery, Lyon, France
Introduction: Simulated clinical trials were run using a disease model of cardiac ischemia-reperfusion injury (IRI). The simulation suggested that:
- Myocardial Infarction (MI) was reduced by reactive oxygen species (ROS) inhibition: 5% infarct size (IS) reduction
- Best responders characterized by: final thrombolysis in myocardial infarction (TIMI) flow grade of 3, and left anterior descending artery (LAD) lesion location of proximal or mid: 10% IS reduction
In addition, trial simulations using result 2 showed that:
3. Upcoming trial size could be reduced by 75%
Objectives: Heart disease is the leading cause of death, with 805,000 cases of MI a year in the US [1]. According to ACC/AHA guidelines, standard of care is reperfusion of the heart (thrombolysis and/or percutaneous coronary intervention (PCI)). However, restoring blood flow can lead to cardiac IRI, in part caused by an acute increase of ROS [2].
Pre-clinical studies have shown cardioprotective effects of ROS inhibition at the mitochondrial C1 level [3], but no clinical trial to date has successfully shown a benefit in patients and no specific drug is available.
We hypothesized that we can simulate a clinical trial to study the clinical benefits of ROS inhibition at the C1 level in myocardial IRI.
We therefore ran an in silico clinical trial in ST-elevation MI (STEMI) patients treated with PCI to test C1 inhibition, to study the degree and duration of C1 inhibition and to measure the effect on tissue injury (change in left ventricular (LV) IS (% volume)), LV ejection fraction (LVEF), creatine phosphokinase (CPK) and troponin I (TnI).
Methods:
Model development
The model of myocardial IRI pathophysiology in STEMI patients linked IS to LVEF with 4 submodels: (1) mitochondria, (2) cardiomyocyte, (3) myocardium and (4) ventricular function.
Model development with NOVA’s 3-step in silico approach:
- Model Building using a Knowledge and a Computational Model: Relevant biological entities and their functional relationships were analyzed and translated into ordinary differential equations (ODE). The final model had 496 parameters and 173 ODE states.
- Model Calibration with available pre-clinical and clinical data.
- Validation: The model and its Virtual Population representative of real patients were validated [4] with an independent data set for 4 outcomes: IS, LVEF, CPK and TnI. Validation was evaluated with: (a) Spearman rank correlation (through permutation testing) to test the model’s capacity to rank patients by their outcome severity; (b) AUCROC to test the model’s accuracy in separating patients with severe outcomes from others (a threshold of 0.7 was previously set to define acceptability).
Effect model
The Effect Model (EM) [5] describes for each patient the rate of disease-related events with and without C1 blockade after PCI initiation. The sum of the differences between these rates provides the absolute benefit.
Results: 1000 virtual patients with STEMI (1 to 12 hours ischemia) treated with PCI were simulated.
Three days post-PCI the simulated mean IS was 31.6% (SD=14.8) and the mean LVEF was 41.7% (SD=10.1). The model’s physiological representativeness was confirmed by CPK and TnI showing comparable post-MI dynamics.
In a subset of those virtual patients the effect of increasing degree and duration of C1 inhibition was tested. The maximal effect was found after at least 10hrs of maximal inhibition and led to a mean IS of 26.2% (SD=15.5, p<0.0001 vs control, t-test) and a mean LVEF of 45% (SD=10.1, p<0.0001 vs control, t-test).
Using the EM, an optimal responder group characterized by final TIMI flow grade 3 and LAD occlusion location Mid or Proximal was found with an IS reduction over 10%.
Based on our data, with IS as primary endpoint, we calculated that the necessary sample size for a real trial could be reduced by 75% when optimal responders are selected, i.e. from n=236 with the general population to n=60 with the selection.
Conclusion: This in silico trial demonstrates via simulation that ROS inhibition at the C1 level ameliorates the ROS burst and reduces IRI leading to clinically important improvements.
Moreover, this approach can be used to study new targets and optimize treatment strategies. Characterization of optimal responders can reduce sample size of clinical trials and identify patients benefiting most from a given treatment plan.
In silico clinical trial simulation is a promising approach that can support go/no-go decisions made by clinical researchers, biopharma and regulatory agencies.
References:
[1] CDC – https://www.cdc.gov/heartdisease/facts.htm, Accessed May 6, 2021
[2] Carden et al. – Pathophysiology of ischaemia-reperfusion injury. – The Journal of Pathology (2000) 190 (3) 255-266
[3] Chen, Q., Hoppel, C. L., & Lesnefsky, E. J. (2006). Blockade of electron transport before cardiac ischemia with the reversible inhibitor amobarbital protects rat heart mitochondria. Journal of Pharmacology and Experimental Therapeutics, 316(1), 200-207
[4] The American Society of Mechanical Engineers – Assessing Credibility of Computational Modeling through Verification and Validation: Application to Medical Devices VV40(2018)
[5] Boissel et al. – Bridging Systems Medicine and Patient Needs – CPT Pharmacometrics Syst Pharmacol (2015) 4 (3)
Reference: PAGE 29 (2021) Abstr 9646 [www.page-meeting.org/?abstract=9646]
Poster: Drug/Disease Modelling - Other Topics