Providence, Microsoft team on AI to speed up clinical trials

It takes years to study potential drugs and treatments, but the Providence health system and Microsoft have been partnering in an effort to use AI to speed up the process.

The health system and tech giant have developed TRIALSCOPE, an AI-enabled framework that uses patient data to reproduce the results of large clinical trials.

TRIALSCOPE scours data from electronic health records and can simulate clinical trials of treatments.

Dr. Carlo Bifulco, the medical director of cancer genomics and precision oncology at the Providence Cancer Institute, tells Chief Healthcare Executive® that TRIALSCOPE could help research institutions overcome the thorny problem of finding enough participants for clinical trials. And he says that offers the potential to speed up research of promising therapies for cancer patients.

“They're the engine for medical care progress,” Bifulco says of clinical trials. “They represent the best option for the patient, but also they're the way we can actually improve future care for patients.”

Bifulco notes that getting patients to enroll in clinical trials has been a chronic problem for years. And he is hopeful that the AI-powered platform can help overcome a hurdle to cancer research.

Providence researchers studied the data of over 1 million cancer patients in their work. Researchers published the results of their testing of TRIALSCOPE in NEJM AI last September.

“TRIALSCOPE was able to extract cancer treatment data from EMRs, overcoming limitations of manual curation,” the researchers wrote. “We were also able to show that TRIALSCOPE could reproduce the results of lung and pancreatic cancer clinical trials from the extracted real-world data.”

With TRIALSCOPE, Bifulco says researchers were able to demonstrate that the majority of the trials in the immuno-oncology area could be reproduced.

Bifulco says researchers were able to show that the drugs studied were achieving the results that they did in a clinical trial, “which seems trivial, but actually it's not.”

Using TRIALSCOPE, researchers can simulate the selection criteria for a clinical trial.

By doing so, Bifulco says, “It enables you to do better design for clinical trials in the future.”

The AI-powered platform can also simulate the control arm of clinical trials, he says. In clinical trials, researchers will examine the drug or treatment they are studying against a competitor’s medication or a placebo. The TRIALSCOPE platform offers a virtual way of simulating that step, Bifulco says.

Researchers also note that simulated trials using real patient data can reduce the risk of exposing more patients to medications that are being studied.

Hoifung Poon, general manager of real-world evidence at Microsoft Research, told Chief Healthcare Executive® in a recent interview that the work with Providence aims to tackle the fundamental tension in research trials.

“On one hand, we need to develop more and more of this kind of personalized, individualized medicine … but at the same time, each trial takes so much to run,” Poon says.

Poon said his “moonshot dream” would be to use the technology to eventually help design a digital twin for patients that could project how patients would respond to different treatments.

In his vision, Poon said, “To some extent, it's almost kind of like a weather forecast system for the patient.”

Providence officials note that TRIALSCOPE can review data much faster than an entire team of researchers.

Bifulco also noted that by using existing patient data, researchers can test theories and also shorten the process of designing the trial at the beginning and running it through to the end. He says researchers could eventually shorten the timeline to find out if a new treatment has potential.

Just as critically, if a therapy isn’t effective, researchers could learn that sooner and move on to a different treatment for the patient.

AI advances that have taken place since researchers have begun work on TRIALSCOPE offer more potential for alternative approaches to clinical trials, he says.

“We think that it's going to be a very powerful tool in the future for facilitating the design and implementation of trials,” Bifulco says.