The Pulse of Modeling and Simulation in Clinical Trials

Michelangelo, who was born in 1475, built models for his sculptures from wax and clay. The Wright Brothers, circa 1906, reportedly built a few models of the first airplane in their bicycle shop. And the Model T, so the story goes, was so named because of the 19 models (A-S) that preceded its launch in 1908.

One could argue that these geniuses created models and prototypes because they saw value in investing that time: Where were the flaws in their design? The sweet spots? How could they make their creation more beautiful, efficient, useful? The use of a model, logic would dictate, could help eliminate or minimize the risk of failure in the end result. In the above examples, the end results speak for themselves.

Of course, designing a Ferrari isn’t the same as designing a monoclonal antibody; the latter will behave differently because of human variables.

Variables aren’t anything to fear, say experts in the field of simulation; certainly they aren’t reason enough for the $1.48 trillion pharma industry¹ to forgo simulation of clinical trials, or to patient-test the protocol.

Mark Lovern

Until the pivotal trials begin, lots of time exists to learn about the drug and how patients will respond to protocol demands, experts say.

Once the trials start, a sponsor is trusting fate. If one of the usually two FDA-required Phase III trials fail, says Mark Lovern, senior vice president, Certara, “you’re basically up a creek without a paddle.”

Of the differing types of simulation companies portrayed in this article—predicting drug performance in people, designing optimal protocols, or inviting patients to play-act roles in clinical trials—all say their businesses are growing and repeat customers are the norm. The case studies are impressive, highlighting money saved and FDA approvals achieved.

Alicia Staley

At Medidata, Alicia Staley, vice president, patient engagement, reports that its role-playing studio for patients—who often leave trials because of the onerous demands placed on them—has become so popular with sponsors that within two years of operation it is now a revenue producer.

Exploristics, CorEvitas, and Certara all report growth; in some cases, significant growth.

However, while significant growth is reported here, industry-wide adoption may be lacking, specifically because of time and money.

Reasons vary from a potential fear of adding more time to pre-trial investigation; queasiness over leaning too heavily on AI results; unwillingness to change long-existing exploratory and analytical procedures; and problems with communication between internal company silos. These are the reasons, those interviewed say, for why the use of simulations is still the exception, and not the norm.

Mike Rea, founder of IDEA Pharma, says the few R&D professionals he has spoken with about relying on AI-based simulation results are very wary. “I get the strong sense from R&D that they are skeptical about how far technology can drive things at the moment with AI, including simulations,” he says. AI, adds Rea, helps create ideas and run various scenarios, “but may not reliable enough to make a decision.”

While AI may not be an iron-clad decision, simulation can help point a sponsor in the right direction. Controversies abound over the efficiency and productivity of pharma R&D. In a recent analysis of 16 industry members between 2001 and 2020, the group collectively launched 251 new drugs, representing 46% of all Center for Drug Evaluation and Research (CDER)-related FDA approvals within the time period. The average total R&D expenditure per new drug was $6.16 billion.²

Time and money

Despite any gains that pharma might make in terms of extending lives or making those lives more livable, it is the remarkable length of time that trials take and the stunning financial outlay involved in taking a molecule from lab to regulator define pharma’s measure of success, or not. Simulators think in silico trial models can help decrease time and money spent with more accurate patient selection, a more honed indication, and subsequent endpoint. “Anything you can do in the planning phase will help,” says Lovern. “You can use a model as a framework.”

Sam Miller

“What unites people who are interested in simulations is the feeling that too much of the decision-making in drug development” is made by the medically trained who rely on their gut feelings and past experiences to make major decisions on drug development and subsequent trials, says Sam Miller, head of strategic consulting, Exploristics, a provider of biosimulation software and biostatistics services.

In short, more quantification, less qualification in the decision-making process.

Simulation under review

PubMed doesn’t provide extensive help in assessing whether simulation of any part of the clinical trial process is, or isn’t, a keeper. There were no studies that assessed any type of simulation’s overall value. Wrote Krishnaswami et al. in 2020: “The discipline needs to continue to evolve from one‐off case studies to a paradigm of systematic, best practices‐driven approaches to improving decision-making across the [drug discovery, development, regulatory approval, and clinical utilization] continuum.”³

But there are glimpses of success.

In a recent study, researchers simulated the effects of a central nervous inhibitor called HSK3486.⁴ Designed as an anesthetic, HSK3486 was developed in 2019 to guide model-informed drug development (MIDD) on certain populations like infants and the elderly. HSK3486 was designed to have better anesthetic properties and less systemic exposure than the widely used propofol, which tends to stay in the liver. Using MIDD to test for exposure, the researchers found that predicted systemic exposure went up somewhat in those with hepatic impairment and the elderly, which was similar to the data gathered in a subsequent clinical trial.

FDA supports MIDD; it has done so since 2018. The agency runs what it calls its MIDD-paired meeting program, so sponsors and agency experts can discuss different modeling approaches. A successful modeling application, writes the agency, can raise the odds of a drug’s approval and help determine the right dosages for various populations “in the absence of dedicated trials.”⁵

Certara, says Lovern, has enabled 100 novel drugs, found using physiological based pharmacokinetic modeling and simulation, to obtain label claims in lieu of clinical trials. Most are for drug-drug interactions, he says.

An Exploristics’ case study also shows simulation success. A biotech working on a molecule for a rare disorder came to Exploristics to see the probability of this molecule in a particular indication. It was showing possible efficacy in a few different ones. The biotech was looking for a Phase II trial. It had been told it needed 1,000 patients.

Miller says they ran many assumptions regarding various endpoints and different patient groups to recruit. All those scenarios showed less than 50% of probable success. So it pared down the indication and honed the endpoint to mortality. The result: a new enrollment of 180 patients, and at least 80% success rate. The results showed benefit to patients by enabling a trial that otherwise was unfeasible, says Miller.

The technology and the data

Computer power and data retrieval have made simulation possible. Before the 20th century, the technology didn’t exist for researchers to look at bodily processes, AKA pharmacokinetics, to measure the minute amounts of drug present in blood and tissues, says Lovern. It wasn’t until the 1980s that computing power enabled analysis of the large quantities of data needed to inform predictive pharmacological models.

Jane Myles

However, even with the technology, data at scale, remains elusive. It’s either proprietary or outrageously expensive to get from a provider. Jane Myles, program director, Decentralized Trials & Research Alliance, says that even within a pharma company itself, the silos don’t share. “Our computational vision is limited to that silo,” she notes.

Exploristics and Certara know about data limitations. Both mostly rely on public sources for data and engage teams of scientists to scour PubMed and the like. Certara uses about 30 data scientists. And Exploristics, says Miller, increases its data pool in a singular way. Its clients agree that the company can access the simulated information to add to future simulations. So far no one has said no, he says. According to Miller, Exploristics’ business has grown at least 30% year over year for some time.

Non-tech based simulation

Myles says audience members attending one session at the Drug Information Association (DIA) Annual Meeting in June were asked if they incorporated patient voice into their protocols. Fewer than 40% was the response—a far different percentage than found in the DIA-Tufts Center for the Study of Drug Development patient engagement studies conducted a few years ago. Asked a similar question, 65% of those surveyed said they are “investing in patient-centric initiatives in drug development.”⁶

The following are good examples that our interviewees shared in regard to patient-centric simulations.

Five years ago, Medidata wanted to know how being patient-centered would impact clinical trial design. The company’s so-called design studio was the result. The studio, Staley tells Applied Clinical Trials, is an “opportunity for [the in-house] team to bring problems into the studio.”

The studio is in operation every month; Medidata has a set group of patient advocates and others they can call for specific disease states. Product managers, designers, and engineers bring their problems to the patient team. And these aren’t all new problems—some have been worked and reworked over the years.

One tool that has received such attention is eConsent. The patient advocates—all of whom have some disability—go through a mock scenario and assess every aspect: ease of setting up an account, language comprehension, even if the screen is presented in an “empathetic way.”

Some patient advocates have left their mark, says Staley. A lung cancer patient advocate that she worked with had been in seven lung cancer trials and had received 60 brain scans. The patient had contrast dye (gadolinium) in her brain, of which there are no clear safety guidelines to determine a maximum exposure to gadolinium, notes Staley. The clinical trial teams had not paid attention to this aspect of the patient’s care plan. A team will not act, says Staley, if information isn’t in front of them. “No one is overseeing all the data points,” she points out.

What patient simulations provide is control, notes Staley. “What we are delivering to the market is a team of patients who have looked at it, taken it apart, and put it back together. That is a cost savings right there, that gets back to industry.”

Stephanie Terrey

At CorEvitas, Stephanie Terrey, senior director, patient experience, has a team that gathers recruited patients, study coordinators, the principal investigators, and other physicians. An outreach manager is responsible for patient recruitment. Terrey says the protocols are at the almost finished stage. Once the patients and site staff role play key aspects of the protocol, debrief interviews are conducted with all participants and the feedback is “analyzed and then [CorEvitas] reports to the client,” she explains.

Terrey says there have been times when the patient response has influenced protocol change, such as reducing the number of visits, adding a car service, or easier samples drop-off; even the trial’s design from traditional to decentralized.

In another recent simulation between two principal investigators, Terrey relays that one spoke with empathy, the other went straight down the list of requirements. “We are doing the same protocol, but how patients responded was very different, which is valuable to inform site selection and training,” she says.

A decade ago, at Roche, Myles and colleagues built an “imaginarium” for Roche colleagues who were designing trials. Once a team had assessment schedules and the primary endpoint was established, imaginarium staff looked for real patients. It also brought in important team members—key opinion leaders, clinical scientists, and biostatisticians.

The team also included a graphic artist, for visual depictions done in real time. Myles says, “It was so useful, he would be listening for different things, for example, the emotion in the room.”

But those who were most surprised by what they learned from patients in the two-year patient imaginarium were the biostatisticians and clinicians. “They didn’t know how much they were asking of patients,” says Myles, for example, such as staying in an infusion suite for 12 hours.

Even communication among simulators can be difficult. Lovern says oftentimes, bioengineers and statisticians are speaking a language the decision-makers do not understand. The “modeling nerds” have the information, but they can’t communicate that meaningfully.

“It is like they use different vocabularies,” he says. “The engineers have to explain things in non-tech terminology, or even better, in a therapeutic context. This way, the physician can say, ‘I understand what you’re trying to say and why it’s important to the decision that I am trying to make.’”

Reality wins out

With the successes and challenges of simulations, it still comes down to time and money. For industry, time is a business driver. Staley, a three-time cancer survivor, says, for the patient, time is life. Industry must move faster for patients.

Christine Bahls is a freelance writer for medical, clinical trials, and pharma information.

References

1. Global Pharmaceutical Industry - Statistics & Facts. Statista.com. July 14, 2023, https://www.statista.com/topics/1764/global-pharmaceutical-industry/#topicOverview
2. Schuhmacher, A.; Hinder, M., von Stegmann Und Stein, A.; Hart, D.; Gassmann, O. Analysis of Pharma R&D Productivity - a New Perspective Needed? Drug Discov Today. 2023. 28 (10). https://pubmed.ncbi.nlm.nih.gov/37506762/
3. Krishnaswami, S.; Austin, D.; Della Pasqua, O.; et al. MID3: Mission Impossible or Model-Informed Drug Discovery and Development? Point-Counterpoint Discussions on Key Challenges. Clin Pharmacol Ther. 2020. 107 (4), 762-772. https://pubmed.ncbi.nlm.nih.gov/31955417/
4. Zhang, M., Yu, Z.; Liu, H.; et al. Model-informed Drug Development: The Mechanistic HSK3486 Physiologically Based Pharmacokinetic Model Informing Dose Decisions in Clinical Trials of Specific Populations. Biopharm Drug Dispos. 2023. 44 (3), 259-273. https://pubmed.ncbi.nlm.nih.gov/37313580/
5. FDA, Model-Informed Drug Development Paired Meeting Program (July 31, 2023). https://www.fda.gov/drugs/development-resources/model-informed-drug-development-paired-meeting-program
6. Patient Engagement Study Findings. DIA. 2018, https://engage.diaglobal.org/DIAPatientEngagement.html