Skip to main content

Software at the heart of drug development for COVID-19

Elsevier and ExactCure recently announced a collaboration to develop and offer to hospitals – without charge – drug-specific exposure models for 20 already approved medicines that are being tested as potential treatments for Covid-19.

Each drug-specific model, generated using ExactCure’s AI-driven simulation platform, will help to predict a drug molecule’s pharmacokinetic properties in each individual patient, according to their age, sex, whether they have other diseases – comorbidities – and other factors, and so give guidance to clinicians on likely optimum dosing.

Development of the drug exposure models will leverage data held in thousands of drug-related documents within Elsevier’s PharmaPendium drug data resource, which contains decades’ of searchable FDA and EMA regulatory approval and related documents on the drugs.

Olivier Barberan, director of translational medical solutions at Elsevier explains: ‘We will provide ExactCure with information held in PharmaPendium that spans more than 50 drug-specific parameters, including PK and pharmacodynamic data, safety data, adverse events and drug-drug interaction records, together with data on drug efficacy. This may encompass many thousands of reports, for example, there were in excess of 13,000 records just for the antiviral drug ritonavir, which is one of the drugs under consideration for Covid-19 therapy.’

Fabien Astic, ExactCure cofounder. states: ‘A growing number of marketed drugs, such hydroxychloroquine, chloroquine, lopinavir/ritonavir and azithromycin, are being assessed as candidate therapeutics for Covid-19.’

‘Our simulation models will give clinicians in hospitals, including ICU departments, guidance on the appropriate dose of a drug. Clinicians may not have had experience with some of the drugs that are being repurposed. This, coupled with the fact that we have yet to fully understand Covid-19 disease, means that informed guidance on dosing from potentially years of use in diverse patients could help to ensure that each individual gets the most appropriate amount of drug.’

‘We have already released the first couple of models to a French hospital. And as additional models are developed and validated, we may roll them out to general practitioners for use in community medicine. We also expect to expand the collection of drugs as additional therapeutics are repurposed against the SARS-CoV-2 virus,’ added Astic.

Creating a repository of data for drugs and vaccines

PharmaPendium is a repository of fully searchable FDA and EMA drug approval documents and extracted information spanning drug safety, pharmacokinetic/pharmacodynamics, efficacy and metabolism. Containing data stretching back to the 1930s, PharmaPendium is used by regulators, the pharma industry and academic organisations for basic drug discovery, R&D, and clinical trials and drug approval programs.

Elsevier is making existing regulatory data and reports on drugs that are being repurposed as potential Covid-19 treatments available freely to researchers through its recently launched Covid-19 hub. ‘We held a Webinar to launch the Covid-19 hub in April, to which there were 14,000 attendees,’ Barberan noted. ‘We are sure that it will represent a valuable resource for regulators, industry and academic scientists globally who are working to develop drugs and vaccines against the virus.’

Having generated a repository of searchable documents, Elsevier is now working to extract the most critical data, including pharmacokinetic data, information on drug inhibition of cytochromes or key enzymes, together with data related to drug-drug interactions, efficacy data from trials, and both premarket and post-market safety data. A key part of this extraction process is working towards standardising that data so that content isn’t lost in search or analyses. Standardisation is key, Barberan commented. ‘Where available, we use existing standards, such as MedDRA, for adverse events.’

‘And while our day-to-day work is still compiling and curating the compendium, we are also working internally, and with partners, such as ExactCure, to make use of the huge amount of pharmacokinetic and other key data, in combination with AI tools, for predictive modelling, where we can provide a complete solution to our clients for drug discovery and development, not just offer them the data,’ Barberan continued.

Elsevier aims to leverage the database in a number of predictive fields, and has already signed an agreement with FDA centred on the development of predictive tools for drug-induced liver injury, which the agency could use as a basis for industry guidance. ‘A separate collaboration is under discussion with the Evidence-Based Toxicology Collaboration at Johns Hopkins University, on generating predictive tools for drug and other chemicals’ development that use human biology-based information and might reduce the need for animal testing.’

The Covid-19-focused collaboration with ExactCure, to which Elsevier is also providing the PharmaPendium data free of charge, is an evolution of a commercially focused partnership between the two firms.

Using AI in the fight against Covid-19

ExactCure is exploiting the PharmaPendium data and its own AI tools to build a simulation-based digital companion – a Digital Twin – app for smartphones that patients would use to help make sure that they use medicines safely, and at an appropriate dose and frequency, ‘whether that be an OTC painkiller, or an antiviral medicine,’ Astic notes. ‘This can help to prevent underdosing, overdosing, and to prevent drug-drug interactions or adverse events relating to the individual’s health status or even genetic profile.’

The Digital Twin is designed to simulate the efficacy and interactions of medicines in each individual’s body, based on the personal characteristics that have a proven influence on a specific medication. The easy-to-use app can indicate, for example, whether and when it is safe for the patient to take the next dose of their drug – and how much they can take – dependent on how much and when they dosed previously, and when they last took another medication that they require.

‘The AI-based software derives the personalised guidance according to key patient-specific characteristics such as weight, age, gender, renal and liver function, smoking status, and genetic background. Importantly, it could also feed back information to the prescribing physician, so that they will know how well the patient is sticking to their drug schedule,’ Astic suggested.

‘Our first model, for paracetamol, could help dramatically reduce overuse of the drug and even prevent overdose-related deaths.’ ‘We signed a partnership with Vidal, key player of medical information in France, to integrate our technology into their Vidal Sentinel platform designed for hospitals. They call our API [Application Programming Interface] and the pharmacist or doctor can run personalised simulations until they reach what they estimate to be the best posology for a given patient,’ added Astic.

Digital companion apps could also be utilised by the pharma industry to support clinical drug trials and potentially speed time to market and reduce attrition at the trials stage.

The app would effectively enable remote, on time reporting, which would give investigators real-time insight into the effects of the drug within the cohort, as well as correlate drug effects to concentration, in the context of the individual patient. ‘Using this technology could allow trial sponsors to spot outliers before the whole study is jeopardised, and potentially reduce attrition rate from failed studies,’ Astic suggested.

Dosing optimisation for Covid-19

During early April, Certara launched the Covid-19 Pharmacology Resource Center, an online resource giving scientists around the world access to simulation and modelling tools to aid the design of clinical trials and optimise dosing regimes for candidate drugs such as hydroxychloroquine, and lopinavir/ritonavir, against SARS-CoV-2. Funded by the Bill & Melinda Gates Foundation and supporting global collaboration in the drive to develop new treatments for Covid-19, the Center offers researchers a workbench of in silico modelling tools, integrated with existing and emerging data.

The new Center offers an accessible outreach of the expertise that Certara provides within the global Covid-19 Therapeutics Accelerator, which has been set up with an initial $125 million in funding from the Gates Foundation, Wellcome and Mastercard, through which WHO, governments, healthcare providers and industry are collaborating to speed the development of therapeutics to treat Covid-19 or prevent SARS-CoV-2 infection.

Craig Rayner, president, integrated drug development at Certara explains: ‘Certara is providing expertise in translational and clinical pharmacology, quantitative science and regulatory strategy to support critical stage decisions, clinical trials design and dosing optimisation for Covid-19.’

For a virus like SARS-CoV-2, for example, Certara researchers invest significant efforts in bringing the biology and math together to help improve decision making for new therapeutics. One can now evaluate how the virus enters cells, how it interacts and replicates, what the immune system is doing in response, via sophisticated quantitative pharmacology frameworks and predictive tools and then simulate new situations, he suggested.

‘We can take huge amounts of data from preclinical models, in vitro testing and clinical experience, as a fundamental foundation on which to use math engines to model what will happen in different trial scenarios, start to simulate clinical trials accurately, and then add data derived from new trials back into the model, and validate in silico learning,’ said Rayner.

Certara modelling and simulation software and services are harnessed by the pharmaceutical industry globally, together with regulators and academic institutions, Rayner continued. ‘Our in silico tools have evolved through a combination of learning through historical experimentation, and the ability to apply ‘math’ and modelling across the entire disease process.’ The ultimate aim is to develop more effective treatments, while also reducing the attrition rate, development timeline and associated costs.

Central to this suite of in silico predictive and intelligence tools is the firm’s Simcyp Simulator, which is applied at multiple points in the drug development and trials timeline to help determine optimum drug dose. This is critical to ensure that enough of a given drug is given to achieve efficacy, but that overdosing is avoided, both to reduce wasting the drug, and also to reduce the likelihood of drug-related side effects.

The Simcyp Simulator has been developed as a suite of modules that simulate drug pharmacokinetics (PK) and so can predict and describe how the body affects the drug-drug absorption, distribution, metabolism and excretion (ADME), and how PK may be altered by formulation, patient variables like age, gender or genotypic information, or concomitantly administered medications.

The Simcyp Simulator links laboratory data to in vivo ADME data and when integrated with and extended to pharmacodynamic (PD) information (how the drug effects the body) such as biomarkers or clinical efficacy and safety, is a powerful tool to support dosing decision making in new trials.

‘Designing and running clinical trials for any drug or vaccine is hugely expensive and time consuming, so there is a great need to boost efficiency, and improve the likelihood of success,’ commented Keith Nieforth, senior director, software division.

‘The Certara tools can also model drug activity at particular sites of action, and look at the physicochemical properties of that molecule in the context of other molecules with similar structure and activity, to make predictions on whether the drug will reach target tissues, such as the lung, if we consider SARS-CoV-2,’ Nieforth said.

‘In the case of Covid-19 drug development, the Certara models integrate simulations of drug pharmacokinetics and pharmacodynamics, alongside virus interaction with the host and symptoms. ‘You can then link those models together and that enables you to simulate what you think might happen in clinical trials. Ultimately, modelling and simulation can reduce the number of patients, or trial arms required, as well as evaluate the influence of other design factors on trial outcomes, and so improve the probability of success,’ added Nieforth.

In the race to develop therapeutics against COVID-19, the Simcyp Simulator has been instrumental in predicting whether it will be possible to achieve a therapeutic dose of existing drugs that were reportedly effective against the virus in the lab. ‘Hydroxychloroquine was shown to inhibit SARS-CoV-2 in lab-grown cells, but there were concerns that the plasma concentrations wouldn’t be high enough to achieve therapeutic efficacy in infected people,’ Rayner commented. ‘However, the virus accumulates in the lungs, and our simulation demonstrated that the concentration of drug that would reach the lungs would be sufficient to inhibit the virus.’

‘That data has been instrumental in opening the way to hydroxychloroquine trials that are now ongoing worldwide,’ he added. ‘The drug worked in vitro and there’s pharmacological plausibility and there’s proof of hope that it would work because it would achieve the concentrations that we think are relevant to the site of primary viral action. Within the Accelerator, we worked with the University of Washington and New York University primary investigators, to achieve hydroxychloroquine dose selection, clinical study design and institutional review board approval, in just eight days, a process that would normally take months.’

Another existing drug that had shown potential against SARS-CoV-2 in vitro was the antiparasitic drug ivermectin, Rayner continued. ‘This is the flipside to our experience with hydroxychloroquine, as in this case Simcyp simulations indicated that the drug would not achieve sufficient lung concentration to be effective against the virus in vivo, and we argued against the start of clinical trials that would likely fail.’

‘We are learning from data emerging when is the most effective treatment point during the infection timeline. Give an antiviral drug too late, for example, and the virus has effectively finished its infectivity, and the patient is now dealing primarily with the immunological effects of the infection, which is almost a completely different ‘disease.’ What we are trying to do is to put models around viral kinetic data that is coming in from around the world, to understand the infection and viral load time course, and so simulate when to most appropriately effect therapy.’

‘Our aim has been to develop these tools to help explore, quantify, predict, and confirm,’ Nieforth said. ‘The endpoint is that they can be used together to provide recommendations that have impact. Appreciating that we have a very diverse user base, we have evolved our tools to cover the full scope of drug development and user experience, both for individual pieces of the workflow and also to tie those individual pieces together very nicely.’



Topics

Read more about:

Laboratory informatics

Media Partners