FDA is enhancing its ability to handle real-world evidence by training reviewers in data science via a curriculum on machine learning and artificial intelligence, said FDA Commissioner Scott Gottlieb.
“We’re working to develop new guidance documents to assist sponsors interested in developing and using real-world evidence,” Gottlieb said at a Jan. 28 panel discussion organized by the Bipartisan Policy Center.
“Our ‘Framework for Real-World Evidence Program’ will apply a consistent strategy for harnessing these tools across our drug and biologic review programs,” said Gottlieb, referring to a framework document published last December. The document evaluates the use of RWE to support additional indications for already approved drugs as well as to satisfy drug post-marketing study requirements (The Cancer Letter, Jan. 4).
“The framework is aimed at leveraging information gathered from patients and the medical community to inform and shape the FDA’s decisions across our drug and biologic development efforts,” Gottlieb said. “The goal is to develop a path for ensuring that RWE solutions can play a more integral role in drug development and regulatory life cycle at the FDA.
“Today, I’m announcing four additional activities that’ll help FDA and stakeholders advance these opportunities for the benefit of patients.”
FDA plans to:
- Support the seamless integration of digital technologies in clinical trials by developing a framework on how digital systems can be used to enhance the efficient oversight of clinical trials. These technologies present important opportunities to streamline drug trials and improve data site integrity by remotely monitoring data trends, accrual, and integrity over the course of a trial.
- Use digital technologies to bring clinical trials to the patient, rather than always requiring the patient to travel to the investigator. More accessible clinical trials can facilitate participation by more diverse patient populations within diverse community settings where patient care is delivered, and in the process can generate information that’s more representative of the real world and may help providers and patients make more informed treatment decisions.
- Explore how reviewers can have more insight into how labeling changes inform provider prescribing decisions and patient outcomes. The FDA’s Information Exchange and Data Transformation—or INFORMED—is using RWD to examine the impact of a recent FDA labeling change for two approved products from weight-based dosing to flat-dosing of immune checkpoint inhibitors. This project is focused on how community practices are adopting the flat dose after the labeling change, and factors that may affect adoption.
- Work with the medical product centers to develop an FDA curriculum on machine learning and artificial intelligence in partnership with external academic partners. The aim of this program is to improve the ability of FDA reviewers and managers to evaluate products that incorporate advanced algorithms and facilitate the FDA’s capacity to develop novel regulatory science tools harnessing these approaches.
FDA’s Oncology Center of Excellence is working with Friends of Cancer Research, NCI, and others to harmonize reference standards for assessing tumor mutational burden—as determined by multiple proprietary assays—to help identify cancer patients who are more likely to respond to immunotherapy.
Harmonizing the measurement of tumor mutational burden across commercial assays used in routine oncology care can help reduce treatment variability, and improve the utility of TMB as a potential biomarker for enriching clinical trials that are designed to test immunotherapies.
OCE is also working on a project exploring whether it’s possible to use real world endpoints, such as time to treatment discontinuation (TTD), as a potential real-world endpoint for pragmatic randomized clinical trials, for FDA approved therapies in the postmarket setting.
“Through ‘Project: Switch,’ OCE is investigating whether well-matched contemporaneous synthetic control arms based on prior clinical trials can be used to make inferences regarding the effect of a new drug, or whether a synthetic control could be used to compare data to active control arms in ongoing randomized controlled trials in rare tumor types where the standard of care remained stagnant, and the prognosis is especially poor,” Gottlieb said.
FDA’s framework for RWE, created in response to a mandate in the 21st Century Cures, spells out the agency’s thinking on the types of guidances that need to be developed before RWE can be routinely used in regulatory science (The Cancer Letter, Jan. 4).
“We really need people to weigh in on the guidances, because one thing I did learn at FDA, pretty much if the FDA says something, the industry is going to do it,” former FDA Commissioner Robert Califf said at the meeting Jan. 28. “So, we’d like to get those guidances right.
“I’m very excited that Amy Abernethy is coming to the FDA [as principal deputy commissioner]. She is an expert on this, I have every confidence that she’ll help guide us through this.” (The Cancer Letter, Jan. 4).
There is a need to better understand AI algorithms, and whether they generate results that are replicable, said former FDA Commissioner Mark McClellan, who is also a former commissioner for the Centers for Medicare and Medicaid Services.
“It’s great to see the progress that’s happening at FDA,” McClellan said at the meeting. “I think Rob [Califf]’s vision for what the future ought to look like, which is a lot of data from a wide variety of sources, including many that a lot of people in the health care industry aren’t really thinking about as important sources of health relevant information—that is the right vision. I think we’re still a long way from getting there. So, great vision, great potential.”
Using real-world data effectively is akin to monitoring jet engines to prevent plane crashes, said Andrew von Eschenbach, former FDA commissioner and former NCI director.
“People won’t die, because planes don’t crash. GE has a system in which their jet engines have an incredible number of sensors that are in those engines and they’re sensing and monitoring those engines in real time, and so they know in real time if there’s anything going wrong,” von Eschenbach said at the meeting.
“I think what we have is the opportunity with the kinds of tools that are now becoming available, be they sensors in humans, or the opportunity to access the data that’s coming in both real time and retrospectively, we’re going to be able to prevent problems. We’re going to be able to see ahead, just like they can, and not only retrospectively correct what’s going on, but prospectively be able to create what needs to be created to save lives.”
