Abstract
The use of in silico trials is expected to play an increasingly important role in the development and regulatory evaluation of new medical products. Among the advantages that in silico approaches offer, is that they permit testing of drug candidates and new medical devices using virtual patients or computational emulations of preclinical experiments, allowing to refine, reduce or even replace time-consuming and costly benchtop/in vitro/ex vivo experiments as well as the involvement of animals and humans in in vivo studies. To facilitate and widen the adoption of in silico trials, InSilicoTrials Technologies has developed a cloud-based platform, hosting healthcare simulation tools for different bench, preclinical and clinical evaluations, and for diverse disease areas. This paper discusses four use cases of in silico trials performed using the InSilicoTrials.com platform. The first application illustrates how in silico approaches can improve the early preclinical assessment of drug-induced cardiotoxicity risks. The second use case is a virtual reproduction of a bench test for the safety assessment of transcatheter heart valve substitutes. The third and fourth use cases are examples of virtual patients generation to evaluate treatment effects in multiple sclerosis and prostate cancer patients, respectively.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
Introduction
Modeling and simulation (M&S) approaches are playing an increasingly important role in the design and development of new medical products for assessment of their safety and/or efficacy, and furthermore in supporting regulatory submission and decisions. In silico techniques can provide predictions about toxicity, efficacy, optimal dosing strategies and study design, and are currently employed by pharmaceutical companies to guide the different stages of the drug development process, from discovery through the preclinical and clinical phases, drug registration as well as to leverage post-market real-world data.20 Similarly, in silico tools can be applied to different phases of medical device design, from concept through verification and certification, to make predictions about mechanical performance, durability and resistance, deployment effects and interaction with human anatomy.8
One of the powerful advantages of simulation techniques is that drug candidates or new devices can be tested on virtual patients within in silico clinical trials. This allows to investigate several “what if” scenarios while minimizing interventions on real individuals, by far ethically sounder than traditional “trial and error” approaches. A virtual patient is a computational model representing the anatomy and/or (patho-)physiology of interest, as well as the pharmacokinetic (PK) and pharmacodynamic (PD) properties of pharmaceutical compounds or the interaction with the medical device. Virtual patients are generally created by directly measuring or estimating the model parameters from experimental or clinical data, or based on literature knowledge, so that simulation outputs are physiologically plausible, i.e., they can reproduce processes occurring in realistic human subjects.
A virtual patient cohort consists of multiple virtual patients with a common model structure but distinct parameter values, which allow to account for inter-patient variabilities in anatomy, physiology, and/or PK/PD properties, depending on the specific context of use for the model.
Different strategies can be applied to generate virtual cohorts. A first method consists of sampling the model parameters from inferred parameter distributions. This approach is commonly applied in semi-mechanistic PK/PD models, where population distributions of the model parameters are usually inferred under the mixed effects modeling framework.16 More computationally intensive quantitative system pharmacology (QSP) models require different sampling techniques, such as random generation of the model parameters in conjunction with verification of acceptance criteria, where parameter sets are accepted only if the resulting model output falls within observed physiological ranges.1
For medical device applications, human anatomy is usually reconstructed by medical image segmentation, and then combined with other numerical parameters (e.g., tissue mechanical, thermal and electrical properties) and boundary conditions (e.g., external forces during daily living activities, blood flows and pressures at geometry inlets and outlets) which represent the input for the mechanistic computational model.24 Additional methods such as principal component analysis, statistical shape modeling and population analysis are then used to generate virtual populations starting from real clinical datasets.26
In silico trials methodologies are not restricted to clinical contexts, being also applied to conduct virtual preclinical studies, allowing to refine, reduce and potentially replace benchtop, in vitro and ex vivo experiments as well as in vivo studies in animals.36 For example, the model developed by Paci et al.,22 permits the virtual emulation of the cardiac safety in vitro test on human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs). Similarly, in silico reproduction of existing standard testing allows to refine and reduce the number of physical experiments to be performed.15 Recently, standard practices for finite element analyses have been codified to virtually test medical devices.2
In silico trials methodologies are nowadays encouraged by regulatory authorities such as the US Food and Drug Administration10 and the European Medicines Agency7, and included in the new EU Medical Device Regulation.28 Nevertheless, the requirements for technical computational skills and the need of extensive verification, validation, and credibility assessment, along with substantial investment in software licenses and IT infrastructure, have considerably limited their adoption to M&S consulting companies and large pharmaceutical and MedTech companies, while remaining an insurmountable challenge for small and medium-sized enterprises. To disrupt the M&S adoption barrier, we have developed an easy-to-use cloud-based commercial platform, where advanced healthcare simulation solutions, developed by internationally recognized universities and research centers, are made available, often for the first time, to medical device and pharmaceutical companies to accelerate time and to reduce costs of their research and development processes (Fig. 1).
The purpose of this paper is to discuss a selection of in silico trial applications for drug and device development through the illustration of four use cases built by the authors using simulation solutions available on the InSilicoTrials.com platform.
Materials and Methods
InSilicoTrials.com is a unique web-based cloud-based platform offering advanced M&S tools to perform in silico trials analyses. The platform is proposed as a pay-per-use Software as a Service solution. It integrates various original computational solutions (e.g., ordinary differential equations-based semi-mechanistic population PK/PD models, mechanistic QSP agent-based models, partial differential equations-based finite element methods and computational fluid dynamics, AI-based models, as well as hybrid methodologies) provided by the research community and integrated into user-friendly online applications. The different computational solutions are organized by therapeutic areas (including cardiology, oncology, neurology, nephrology, orthopedics) and collected into suites of complementary applications (e.g., “Drug Safety Suite” for torsadogenic risk assessment). The integrated M&S workflows on the cloud platform allow users to upload their own input data, automatically set up trial scenarios, run simulations and process outcome. Finally, the user has the possibility to export the outcome of the simulations in an automatically formatted report.
The proposed platform embeds a variety of programming languages (e.g., C++, Python, R, Matlab) and simulation engines (e.g., NONMEM, ANSYS, Abaqus, CodeASTER, OpenFOAM) with no direct access to the solvers by the user. It is built on the Microsoft Azure cloud environment, in compliance with the highest standards of security and privacy (amongst others HIPAA Privacy and Security Rules; ISO/IEC 9001, 20000, 22301, 27017, 27018 and 27001; FDA 21 CFR Part 11 (GxP); Protection Directive 95/46/EC), protecting the IP of model providers against the downloading, copying, and changing of their models, while providing a safe environment for users to manage their own data.
Results
This section presents four use cases of in silico trials built by the authors using simulation tools available on the InSilicoTrials.com platform. The first application illustrates how in silico approaches based on virtual human myocardial cells can complement in vitro tests to improve the early assessment of drug-induced cardiotoxicity risks. The second use case is a virtual emulation of a bench standard test to characterize the mechanical properties of transcatheter heart valve replacements. The third and fourth use cases are examples of in silico approaches to simulate clinical trials and evaluate treatment effects in relapsing remitting multiple sclerosis (RRMS) and prostate cancer (PCa) patients, respectively.
Early Assessment of Drug-Induced Proarrhythmic Risk Using In Silico Reconstructions of Human Myocardial Cells
Torsade de pointes (TdP) is a life-threatening cardiac arrhythmia that has led to the withdrawal of many drugs from the market. Although the adoption of the preclinical and clinical safety guidelines (ICH S7B, ICH E14) has been effective in preventing torsadogenic compounds from reaching the market, their lack of specificity might have stopped a number of safe compounds from further development.11
In 2013 the Comprehensive in vitro Proarrhythmia Assay (CiPA) initiative was established to improve the accuracy of torsadogenic risk predictions by using—in an evaluation phase—in vitro experiments of drug effects on multiple ion channels incorporated into in silico reconstructions of the action potential of human cardiomyocytes, and—in a subsequent validation phase—in vitro assessment of drug effects on multiple ion channels in hiPSC-CMs to detect any missed torsadogenic effects.4 In 2020 the ICH S7B/E14 Q&As was released to update the previous safety guidelines in light of the CiPA initiative.
In line with the CiPA paradigm, we have developed two online applications to complement the in vitro early screening of a compound’s proarrhythmic risk: QT/TdP Risk Screen and STrhiPS.
QT/TdP Risk Screen
QT/TdP Risk Screen (qttdp.insilicocardio.com) is a web-based implementation of an in silico risk classifier for the assessment of drug proarrhythmicity.18 QT/TdP Risk Screen was built using a machine learning tool, based on cellular simulations of the O’Hara et al. human ventricular action potential model21 and existing pharmacological data (IC50 and effective free therapeutic plasma concentration) for 109 drugs of known torsadogenic risk from the CredibleMeds database.37 Cellular simulations were used to derive four in silico arrhythmogenic biomarkers, which were then combined into ensemble decision tree classifiers. The final classifier combines biomarkers calculated taking into account effects on the four most prominent ion currents—IKr (fast component delayed rectifier current), IKs (slow component delayed rectified current), ICaL (L-type calcium current) and INaL (late sodium current)—and achieved accuracy, sensitivity and specificity values of 94.5%, 94% and 95%, respectively,18 in cross-validation on the CredibleMeds database.
Prediction of a compound’s proarrhythmic potential with the QT/TdP Risk Screen tool requires the in vitro IC50 and (optionally) Hill coefficient values for up to the four ion currents, and the compound’s test concentration values for which the proarrhythmic risk needs to be assessed. The tool’s outcome is the compound classification in one of the possible clinical risk categories: unsafe, probably unsafe, probably safe, or safe.
STrhiPS
STrhiPS (Safety Trials on Human Induced Pluripotent Stem Cells, strhips.insilicocardio.com) enables to conduct in silico safety trials on a virtual population of hiPSC-CMs, accounting for drug effects on the seven main ion currents indicated by the CiPA initiative.11 The tool is based on a state-of-the-art hiPSC-CM action potential model integrating multiple current-drug interactions.22 The cellular model was validated against current blocking experiments and was further extended to a population of 1774 in silico cell models, calibrated by using a multi-objective genetic algorithm together with optically recorded action potential and Ca2+ transient data.22
Trial simulations with the STrhiPS tool require the number of hiPSC-CM cells to be simulated (up to 1774 cells), the in vitro IC50 and (optional) Hill coefficient values for up to seven ion currents (IKr, IKs, ICaL, INa, INaL, Ito, IK1) and the concentration values for which the compound needs to be tested. Detailed simulation results are provided to the user both in terms of arrhythmogenic biomarkers distributions for the virtual cell population and in terms of plots of membrane potential traces in each virtual cell.
In silico Predictions of the Torsadogenic Potential for Eight Example Compounds
We applied QT/TdP Risk Screen and STrhiPS to eight different drugs with known cardiac risk at the therapeutic drug concentration, Cmax (Fig. 2a). For each compound, cardiac channel data for hERG, hCav1.2, and peak/late hNav1.5 were manually obtained at the Drug Safety Testing Center Co., Ltd. We provided these data as input to run each product according to their requisites. QT/TdP Risk Screen was used to predict the torsadogenic clinical risk of each compound. STrhiPS was applied to simulate experiments on a population of 110 hiPSC-CMs. Action potential duration at 90% of repolarization (APD90) values were estimated in absence and presence of the drug, and drug-induced repolarization abnormalities were automatically detected.
According to the CredibleMeds classification, all drugs are class 1 (unsafe) except for ranolazine (class 3—probably safe) and verapamil (class 4—safe). All drugs were correctly classified by QT/TdP Risk Screen except for ranolazine which was misclassified as class 1—unsafe (Fig. 2a). Misclassification of ranolazine by QT/TdP Risk Screen is not unique to this case study employing newly derived experimental data, but was also observed with literature-derived IC50 and Hill coefficient values and has been thoroughly discussed elsewhere.18
STrhiPS predictions of drug effects on the safety biomarker APD90 were in line with the CiPA and CredibleMeds TdP classifications: APD90 was prolonged in all unsafe drugs—particularly in the high-risk drug bepridil and flecainide—and it was shortened in the low-risk drug verapamil (Fig. 2b). For ranolazine some prolongation was estimated, likely owing to some difference between the Hill parameters obtained in this work and the ones found in literature.18 In our simulations, cells abnormalities were rarely observed at therapeutic concentrations (< 6% of cells in the population).
The two tools performed well individually. When used in combination they provided complementing information and strengthen each other’s outcomes in line with CiPA recommendations.
Virtual Representation of a Self-expandable Heart Valve
Heart valve disease is prevalent throughout the world, and the number of heart valve replacements is expected to increase rapidly in the coming years. The implantation of transcatheter heart valves provides safe and minimally invasive means for heart valve replacement.13 Self-expandable valves (generally composed by a bioprosthetic valve sutured on a Nitinol stent) have the capacity of self-expanding when released from the catheter, enabling them to be fully repositionable and recapturable to optimize valve positioning before full release.
Computational M&S can help medical device companies to lower costs of multiple prototypes tests before manufacturing and collect scientific evidence for regulatory submissions.
SEV Radial Force Test
SEV (Self-Expandable Valve) Radial Force Test (sevradial.insilicocardio.com) is a web-based implementation of a new computational non-linear implicit finite element model that predicts the Radial Resistive Force and Chronic Outward Force in self-expandable heart valves, reproducing the ISO 5840-3 standard testing.14 The input requirements are the device mesh, nitinol super elastic mechanical properties and the minimum crimping diameter value to be tested. The tool can be used to test different prototype designs before manufacturing, or to test the final device design for regulatory submission, as a report following FDA guidelines9 is automatically generated.
In silico Testing of a Commercial Transcatheter Aortic Valve
We applied SEV Radial Force Test to a 3D model of a SEV resembling the Corevalve 29 mm (Medtronic, Minneapolis, MN, USA).5 Nitinol material parameters were adopted from literature12 and a minimum crimping diameter of 10 mm was considered. Radial Resistive Force and Chronic Outward Force in function of the valve diameter were predicted (Fig. 3). The curves show the typical hysteretic behavior of Nitinol throughout a loading–unloading cycle, with higher force exerted by the device in correspondence with the imposed crimping diameter. Results show similarities to those found in literature for the same commercial product,12 with discrepancies in terms of maximum force related to the adoption of different modeling and meshing techniques, finite element solvers, and slight differences in the simulated SEV geometry and material. In addition to International Organization for Standardization (ISO) standard requirements, the tool allows to inspect strain values within the valve geometry through color maps of maximum principal strains.
In silico Trial with Virtual Multiple Sclerosis Patients
Relapsing–remitting Multiple Sclerosis (RRMS) is characterized by stochastic attacks of the auto-reactive immune system on the body’s own myelin sheaths, damaging the central nervous system and causing the characteristic relapses and remissions. Patients affected by RRMS have dramatically different disease courses, necessitating a wide variety of treatment options with unique efficacy and safety profiles. Clinical trials to develop such treatments are necessarily expensive and long running as a result of RRMS’s generally slowly moving, unpredictable course. Recently, RRMS clinical trial costs have increased further39 as the current preference of active comparator over placebo and the substantial fall of relapse rates over the past few decades have led to a need for longer trials and larger groups to retain statistical power and the ability to observe treatment effects. Virtual patients can help reduce trial design by providing an in silico environment in which the (cost) effectiveness of the trial can be optimized.
MS TreatSim
MS TreatSim (Multiple Sclerosis Treatment Simulator, mstreat.insiliconeuro.com) is a web-based RRMS simulator, where virtual patients are personalized instances of a mechanistic, agent-based model,23 that emulates the immune system and the auto-immune response that characterizes RRMS and includes the option of the commonly prescribed treatment options interferon-β1a, teriflunomide, natalizumab and ocrelizumab. MS TreatSim can be set up by the user to simulate a clinical trial design, requiring input of population characteristics, selection criteria, trial groups and treatments, and trial timelines and milestones.
In silico Recreation of an Historical Trial of Natalizumab
In this use case, MS TreatSim was applied to recreate a historical phase III trial of the RRMS treatment natalizumab. In the AFFIRM clinical trial for natalizumab,27 the treatment group consisted of 627 patients receiving 300 mg of natalizumab every 4 weeks, and the control group was a placebo group (315 patients).
Due to lack of raw trial data availability, the recreation focused on recreating the global characteristics of the trial. The trial population was defined as follows: the age of onset distribution was set to 49.4%:36.2%:14.4% (18–29, 30–39 and 40–49 years), lesion load was set to high and oligoclonal bands status to present for all virtual patients, simulation duration prior to the trial was set to 5 years. Patients inclusion criteria were selected in order that patients were only included if they had experienced at least one relapse in the year preceding inclusion, but not in the final month. Two patients groups were created, mirroring the setup of the historical trial by maintaining a treatment: control ratio of 2:1. The natalizumab group (n = 80) was treated in silico with 300 mg of natalizumab every 4 weeks for 104 weeks, while the control group (n = 40) remained treatment naïve. Simulated time-to-relapse were analyzed using the Kaplan–Meier estimator with 95% Greenwood confidence intervals.
The in silico recreation of the historical trial with MS TreatSim generated full simulations of all virtual patients, which can be characterized at the trial level (Fig. 4a), or examined at the individual level (see e.g., Fig. 4b). In the simulated control group, the percentage of relapse free subjects after 104 weeks was estimated to be 40% (95% CI 25–55%) which very well included the historical trial relapse free subjects outcome of 41%. In the treatment group, natalizumab treatment effect was clearly reproduced with virtual patients showing substantially fewer relapses (over 90% of relapse free subjects at weeks 104) than in the control group. In comparison with the observed outcome of the historical trial (67% relapse free subjects at week 104), however, the simulation overestimated the natalizumab treatment effect. One possible explanation for this result may reside into the framework’s limitations to fully account for complex disease histories, which are prone to induce treatment resistance.25 This is supported by the observation that other metrics in the AFFIRM trial, as well as real world data on the effectiveness of natalizumab all support a larger treatment effect than that expressed in relapse free subjects alone.25
Treatment Effects of a GnRH Agonist on a Virtual Cohort of Prostate Cancer Patients
PCa is one of the most common malignant tumors in men worldwide and the second leading cause of cancer related deaths in US men.34 Androgen deprivation therapy, achieved either by surgical or chemical castration, is the first line treatment against advanced prostate cancer and is also used to treat high-risk disease in combination with local radiation.30 Androgen blockade can be achieved through the administration of gonadotropin-releasing hormone (GnRH) agonists.30
Because treatment of prostate cancer with GnRH agonists requires chronic administration, administration of these drugs is usually performed in the form of sustained-release formulations. In this context, in silico clinical trials can help to optimize and investigate the PK profile of new sustained-release formulations that allow simplified dosing regimens, thus enhancing therapeutic compliance of patients and, consequently, their quality of life.
PCa GnRH Agonists Simulator
PCa GnRH Agonists Simulator (gnrhagonists.insilicoonco.com) is a web-based, cloud-based tool that allows to simulate the effects of GnRH agonists on a virtual population of prostate cancer patients. The tool is based on a PK/PD model that describes the suppression effect of testosterone after the administration of a GnRH agonist,29 simulates the drug and testosterone concentration levels in a population of virtual subjects and estimates the percentage of virtual subjects for whom testosterone levels are reduced below the castration limit during a trial.
Input requirements are the PK and PD properties of the drug, and the trial design. A library of sustained release formulations of known GnRH agonists is available including their PK and PD properties. Additionally, the tool can be customized for a new compound or drug of interest. In this case, the user is asked to setup a compartmental model to describe the PK of the drug. For the PD model, only the in vitro ligand–receptor equilibrium dissociation constant (Kd) of the GnRH agonist is required.
In silico Recreation of the Virtual Cohorts Treated with Different Sustained-Release Formulations of Leuprolide
Leuprolide acetate is a GnRH agonist and it is commonly used in the treatment of prostate cancer.30 In this use case, PCa GnRH Agonists Simulator was applied to three cohorts from different leuprolide clinical trials. Description of the cohorts is provided in Table 1.
To replicate in silico the three cohorts, the three trial schedules (3M, 4M, and 6M) were simulated on a virtual population of 1000 prostate cancer patients. PK and PD properties were taken from the PCa GnRH Agonists Simulator library.17 Testosterone concentration castration limit was set to 0.5 ng/mL, as done in the real clinical trials. The study duration of each cohort is defined in Table 1.
The primary efficacy endpoint of the in silico clinical trial was the portion of patients achieving and maintaining the serum testosterone concentrations under the castration limit (0.5 ng/mL) from day 28 to the end of the study. The percentage and 95% confidence intervals of subjects achieving chemical castration was computed using the Kaplan–Meier method for right-censored observations.
In Fig. 5, observed and simulated data for the three cohorts are displayed. Observed data were extracted from publications.31,33,35 Results of the in silico clinical trials were in good agreement with the outcomes of the actual trials. Visual predictive checks of the testosterone concentrations in plasma well contained the mean testosterone levels measured in actual patients. The percentages of virtual and real patients31,33,35 achieving and maintaining the testosterone levels below 0.5 ng/mL are summarized in Table 2.
In silico clinical trials using the cloud-based tool PCa GnRH Agonists Simulator can reliably reproduce testosterone suppression in prostate cancer patients. The results of this use case obtained in few minutes and with a mean prediction error over the three schedules equal to 4.8% further corroborate this conclusion.
Discussion
Recent analyses estimated research and development investment to bring a new compound to the market to be $1.3 billion on average, with costs depending on therapeutic areas, ranging from $765.9 million for nervous system agents to $2771.6 million for antineoplastic and immunomodulating agents.38 Similarly, in 2010 it was estimated that the average total cost to bring a new medical device from concept to approval in US varies from approximately $31 million for a low-to-moderate-risk 510(k) product, up to $94 million for a higher-risk PMA product.19
The use of in silico trials in the development or regulatory evaluation of new treatments is meeting expectation of beneficial results in terms of cost and time to market reduction.6 Indeed, in silico trials provide unique data that may be used to: predict performance in the short and mid-long term and avoid failures in development; quickly evaluate the potential applicability of a medical product in various clinical scenarios; evaluate adverse events without risk to animals or patients.36 This paper presented two preclinical use cases and two clinical use cases of in silico trials performed via simulation tools available on the InSilicoTrials.com platform.
Traditionally, the preclinical phase of a drug development program evaluates the toxicity and efficacy of a new compound by means of in vitro and in vivo experiments. The first use case illustrates how in silico trial methodologies based on virtual human adult and fetal cardiomyocyte cells models can support the preclinical evaluation of drug-induced cardiotoxicity by leveraging in vitro experimental data to i) anticipate torsadogenic clinical risk-outcomes (QT/TdP Risk Screen) and ii) reduce the number of time-consuming and expensive in vitro tests on hiPSC-CMs (STrhiPS), while at the same time improving the predictive potential of in vitro results.
For medical devices, standard tests performed during preclinical phases are used to evaluate device safety. The second use case shows how in silico methodologies can support the characterization of the mechanical behavior of a cardiovascular device, providing prediction comparable to literature results. The SEV Radial Force Test application allows for in silico reproduction of bench tests to virtually assess a new medical device design, and to identify the worst-case within a series of different prototypes to reduce and refine the physical test burden in terms of time and costs.
The clinical phase of a drug development program evaluates the toxicity and efficacy of a new compound by means of clinical studies. The MS TreatSim application presented in the third use case simulates realistic virtual patients and clinical trials, broadening the original scope of this model from individual applications23 to population-level simulations. Such population-level simulations can be leveraged throughout the clinical development of a new drug and beyond, with applications including: the design and de-risking of clinical trials, the selection of sub-populations that will respond best to a new therapy, the augmentation or even replacement of control arms in phase II and III studies, the evaluation of efficacy and safety in long-term scenarios to support label extension requests or post-marketing trials. In addition, the mechanistic nature of the model integrating a detailed and comprehensive description of the immune system is a key feature of the tool and provides opportunities for several applications. For example, patient specific simulations could be applied for individual prognosis and treatment planning, whereas incorporation of mechanism of action for new or existing drugs could be used to explore novel therapies and combination strategies.
In the same vein, the PCa GnRH Agonists Simulator illustrated in the fourth use case can help to optimize the design of a clinical study by identifying the most promising drug formulations and dosing regimens, allowing to suppress testosterone levels below castration levels in most prostate cancer patients.
Every computational approach is suited to fulfill a given objective. Bottom-up, mechanistic models perform very well for hypothesis testing and are adaptable, allowing to incorporate additional elements. When properly validated, these models may provide useful predictions beyond the data employed for model building. For example, mechanistic PBPK simulations are now an industry standard and accepted by regulators in replacement of real clinical studies for the evaluation of drug-drug interactions.32
To deal with always increasing amounts of data that could be integrated in mechanistic models to improve their predictions, there is great potential in combining mechanistic modeling approaches with machine learning algorithms, described by the term “mechanistic learning”.3 An example of such modeling methodology is provided in the first use case presented in this paper. Indeed, the QT/TdP Risk Screen classifier derives in silico biomarkers from simulations of a mechanistic cellular cardiomyocyte model and utilizes them as inputs of a machine learning classifier algorithm.
Computational M&S is thus an invaluable asset in the healthcare industry. However, the need for advanced know-how and computational resources restricts its adoption mainly to large biotechnology and pharmaceutical companies. Making M&S available to a broad spectrum of potential users (medical device and pharmaceutical companies, hospitals, healthcare institutions) would require an easy and controlled access to M&S resources in a secure environment. Furthermore, a joint effort between academia, industry and regulatory bodies is needed to reach a rapid adoption of such a harmonized approach. The InSilicoTrials.com platform aims to disrupt the healthcare landscape, lowering entry barriers for investment and high-level knowledge requirements. It provides healthcare companies and researchers with integrated and easy-to-use simulation workflows to perform computational testing during the development and validation process of new medical devices (e.g., orthopedics, cardiovascular, wearable) and drugs (e.g., PK/PD, disease progression, QSP). Furthermore, the platform is currently being expanded to integrate additional functionalities, including clinical data storage and sharing, medical image analysis and segmentation tools, and access to virtual patients in different disease areas, enabling performing design of experiments, parametric and optimization studies. Based on a web-based collaboration paradigm, the platform represents a cutting-edge solution for the life science and healthcare environment, finally resulting in an increased pace of innovation and reduced time-to-market for new treatments.
References
Allen, R., T. Rieger, and C. Musante. Efficient generation and selection of virtual populations in quantitative systems pharmacology models. CPT Pharmacomet. Syst. Pharmacol. 5:140–146, 2016.
ASTM. F2996-20 Standard Practice for Finite Element Analysis of Non-Modular Metallic Orthopaedic Hip Femoral Stem. West Conshohocken: ASTM International, 2020.
Benzekry, S. Artificial intelligence and mechanistic modeling for clinical decision making in oncology. Clin. Pharmacol. Ther. 108:471–486, 2020.
Blinova, K., Q. Dang, D. Millard, G. Smith, J. Pierson, L. Guo, M. Brock, H. R. Lu, U. Kraushaar, H. Zeng, H. Shi, X. Zhang, K. Sawada, T. Osada, Y. Kanda, Y. Sekino, L. Pang, T. K. Feaster, R. Kettenhofen, N. Stockbridge, D. G. Strauss, and G. Gintant. International multisite study of human-induced pluripotent stem cell-derived cardiomyocytes for drug proarrhythmic potential assessment. Cell Rep. 24:3582–3592, 2018.
Carbonaro, D., D. Gallo, U. Morbiducci, A. Audenino, and C. Chiastra. In silico biomechanical design of the metal frame of transcatheter aortic valves: multi-objective shape and cross-sectional size optimization. Struct. Multidiscip. Optim. 64:1825–1842, 2021.
EFPIA MID3 Workgroup, S. F. Marshall, R. Burghaus, V. Cosson, S. Y. A. Cheung, M. Chenel, O. DellaPasqua, N. Frey, B. Hamrén, L. Harnisch, F. Ivanow, T. Kerbusch, J. Lippert, P. A. Milligan, S. Rohou, A. Staab, J. L. Steimer, C. Tornøe, and S. a. G. Visser. Good Practices in Model-Informed Drug Discovery and Development: Practice, Application, and Documentation. CPT Pharmacomet. Syst. Pharmacol. 5:93–122, 2016.
EMA. Work Plan for the Modelling and Simulation Working Group (MSWG) for 2018. Amsterdam: European Medicines Agency, 2018.
Favre, P., G. Maquer, A. Henderson, D. Hertig, D. Ciric, and J. E. Bischoff. In silico clinical trials in the orthopedic device industry: from fantasy to reality? Ann. Biomed. Eng. 49:3213–3226, 2021.
FDA. Reporting of Computational Modeling Studies in Medical Device Submissions—Guidance for Industry and Food and Drug Administration Staff. Food and Drug Administration. FDA-2013-D-1530. Rockville: FDA, 2016.
FDA. 2021 Advancing Regulatory Science at FDA: Focus Areas of Regulatory Science (FARS). FDA, 2021.
Fermini, B., J. C. Hancox, N. Abi-Gerges, M. Bridgland-Taylor, K. W. Chaudhary, T. Colatsky, K. Correll, W. Crumb, B. Damiano, G. Erdemli, G. Gintant, J. Imredy, J. Koerner, J. Kramer, P. Levesque, Z. Li, A. Lindqvist, C. A. Obejero-Paz, D. Rampe, K. Sawada, D. G. Strauss, and J. I. Vandenberg. A new perspective in the field of cardiac safety testing through the comprehensive in vitro proarrhythmia assay paradigm. SLAS Discov. 21:1–11, 2016.
Finotello, A., R. Gorla, N. Brambilla, F. Bedogni, F. Auricchio, and S. Morganti. Finite element analysis of transcatheter aortic valve implantation: Insights on the modelling of self-expandable devices. J. Mech. Behav. Biomed. Mater.123:104772, 2021.
Howard, C., L. Jullian, M. Joshi, A. Noshirwani, M. Bashir, and A. Harky. TAVI and the future of aortic valve replacement. J. Card. Surg. 34:1577–1590, 2019.
ISO. ISO 5840-3:2021. Cardiovascular implants—cardiac valve prostheses—part 3: heart valve substitutes implanted by transcatheter techniques
Kozlov, M., E. Lucano, and L. M. Angelone. Effects of tuning conditions on near field of MRI transmit birdcage coil at 64 MHz. In Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pp. 6242–6245, 2016.
Lavielle, M. Mixed Effects Models for the Population Approach: Models, Tasks, Methods and Tools. New York: Chapman and Hall/CRC, p. 383, 2014.
Lim, C. N., and A. H. Salem. A semi-mechanistic integrated pharmacokinetic/pharmacodynamic model of the testosterone effects of the gonadotropin-releasing hormone agonist leuprolide in prostate cancer patients. Clin. Pharmacokinet. 54:963–973, 2015.
Llopis-Lorente, J., J. Gomis-Tena, J. Cano, L. Romero, J. Saiz, and B. Trenor. In silico classifiers for the assessment of drug proarrhythmicity. J. Chem. Inf. Model. 60:5172–5187, 2020.
Makower, J., A. Meer, and L. Denend. FDA impact of U.S. Medical Technology Innovation—A survey of over 200 medical technology companies, November 2010.
Marshall, S., R. Madabushi, E. Manolis, K. Krudys, A. Staab, K. Dykstra, and S. A. G. Visser. Model-informed drug discovery and development: current industry good practice and regulatory expectations and future perspectives. CPT Pharmacomet. Syst. Pharmacol. 8:87–96, 2019.
O’Hara, T., L. Virág, A. Varró, and Y. Rudy. Simulation of the undiseased human cardiac ventricular action potential: model formulation and experimental validation. PLoS Comput. Biol.7:e1002061, 2011.
Paci, M., E. Passini, A. Klimas, S. Severi, J. Hyttinen, B. Rodriguez, and E. Entcheva. In silico populations optimized on optogenetic recordings predict drug effects in human induced pluripotent stem cell-derived cardiomyocytes. In 2018 Computing in Cardiology Conference (CinC), 2018.
Pappalardo, F., G. Russo, M. Pennisi, G. A. Parasiliti Palumbo, G. Sgroi, S. Motta, and D. Maimone. The potential of computational modeling to predict disease course and treatment response in patients with relapsing multiple sclerosis. Cells. 9:586, 2020.
Peirlinck, M., F. S. Costabal, J. Yao, J. M. Guccione, S. Tripathy, Y. Wang, D. Ozturk, P. Segars, T. M. Morrison, S. Levine, and E. Kuhl. Precision medicine in human heart modeling: perspectives, challenges, and opportunities. Biomech. Model. Mechanobiol. 20:803–831, 2021.
van Pesch, V., C. J. Sindic, and O. Fernández. Effectiveness and safety of natalizumab in real-world clinical practice: review of observational studies. Clin. Neurol. Neurosurg. 149:55–63, 2016.
Pleouras, D., A. Sakellarios, G. Rigas, G. S. Karanasiou, P. Tsompou, G. Karanasiou, V. Kigka, S. Kyriakidis, V. Pezoulas, G. Gois, N. Tachos, A. Ramos, G. Pelosi, S. Rocchiccioli, L. Michalis, and D. I. Fotiadis. A novel approach to generate a virtual population of human coronary arteries for in silico clinical trials of stent design. IEEE Open J. Eng. Med. Biol. 2:201–209, 2021.
Polman, C. H., P. W. O’Connor, E. Havrdova, M. Hutchinson, L. Kappos, D. H. Miller, J. T. Phillips, F. D. Lublin, G. Giovannoni, A. Wajgt, M. Toal, F. Lynn, M. A. Panzara, A. W. Sandrock, and AFFIRM Investigators. A randomized, placebo-controlled trial of natalizumab for relapsing multiple sclerosis. N. Engl. J. Med. 354:899–910, 2006.
Regulation (EU) 2017/745—L117, 5 May 2017 pp. 1–175
Romero, E., N. Vélez de Mendizabal, J.-M. Cendrós, C. Peraire, E. Bascompta, R. Obach, and I. F. Trocóniz. Pharmacokinetic/pharmacodynamic model of the testosterone effects of triptorelin administered in sustained release formulations in patients with prostate cancer. J. Pharmacol. Exp. Ther. 342:788–798, 2012.
Sharifi, N., J. L. Gulley, and W. L. Dahut. Androgen deprivation therapy for prostate cancer. JAMA. 294:238–244, 2005.
Sharifi, R., L. D. Knoll, J. Smith, and E. Kramolowsky. Leuprolide acetate (30-mg depot every four months) in the treatment of advanced prostate cancer. Urology. 51:271–276, 1998.
Shebley, M., et al. Physiologically based pharmacokinetic model qualification and reporting procedures for regulatory submissions: a consortium perspective. Clin. Pharmacol. Ther. 104:88–110, 2018.
Shore, N. D., S. Guerrero, R. M. Sanahuja, G. Gambús, and A. Parente. A new sustained-release, 3-month leuprolide acetate formulation achieves and maintains castrate concentrations of testosterone in patients with prostate cancer. Clin. Ther. 41:412–425, 2019.
Siegel, R. L., K. D. Miller, H. E. Fuchs, and A. Jemal. Cancer statistics, 2022. CA Cancer J. Clin. 72:7–33, 2022.
Spitz, A., J. M. Young, L. Larsen, C. Mattia-Goldberg, J. Donnelly, and K. Chwalisz. Efficacy and safety of leuprolide acetate 6-month depot for suppression of testosterone in patients with prostate cancer. Prostate Cancer Prostatic Dis. 15:93–99, 2012.
Viceconti, M., L. Emili, P. Afshari, E. Courcelles, C. Curreli, N. Famaey, L. Geris, M. Horner, M. C. Jori, A. Kulesza, A. Loewe, M. Neidlin, M. Reiterer, C. F. Rousseau, G. Russo, S. J. Sonntag, E. M. Voisin, and F. Pappalardo. Possible contexts of use for in silico trials methodologies: a consensus-based review. IEEE J. Biomed. Health Inform. 25:3977–3982, 2021.
Woosley, R., C. Heise, T. Gallo, J. Tate, D. Woosley, and K. Romero. www.CredibleMeds.org. QTdrugs List. Tucson: AZCERT, Inc.
Wouters, O. J., M. McKee, and J. Luyten. Estimated research and development investment needed to bring a new medicine to market, 2009–2018. JAMA. 323:844–853, 2020.
Zhang, Y., A. Salter, E. Wallström, G. Cutter, and O. Stüve. Evolution of clinical trials in multiple sclerosis. Ther. Adv. Neurol. Disord. 12:1756286419826547, 2019.
Acknowledgments
The authors acknowledge Prof. Javier Saiz, Prof. Stefano Severi, Prof. Jari Hyttinen, Prof. Michelangelo Paci, Dr. Dario Carbonaro, Prof. Diego Gallo, Prof. Claudio Chiastra, Prof. Umberto Morbiducci, Prof. Alberto Audenino, Prof. Francesco Pappalardo, Dr. Giulia Russo, and Prof. Iñaki Trocóniz for their contribution to the development of the cloud-based simulation tools, and Dr. Hiroshi Matsukawa and the Drug Safety Testing Center for providing cardiac channel data.
Conflict of interest
Luca Emili and Roberta Bursi are founders of InSilicoTrials Technologies. Chiara Nicolò, Fianne Sips, Cristina Vaghi, Alessia Baretta and Vincenzo Carbone are employees of InSilicoTrials Technologies, a company that has developed the InSilicoTrials.com cloud based platform.
Author information
Authors and Affiliations
Corresponding author
Additional information
Associate Editor Joel Stitzel oversaw the review of this article.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Nicolò, C., Sips, F., Vaghi, C. et al. Accelerating Digitalization in Healthcare with the InSilicoTrials Cloud-Based Platform: Four Use Cases. Ann Biomed Eng 51, 125–136 (2023). https://doi.org/10.1007/s10439-022-03052-6
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10439-022-03052-6