Big Data And The FDA: To Mine The Value, First Mind The Gaps
Executive Summary
INFORMED is a new initiative at the FDA to incubate new ideas in applying big data to boost the scientific, economic and social returns from the regulation of drugs and medical devices, with a particular focus on cancer.
- INFORMED is a collaborative exercise with no fixed budget. It pursues linked-up projects with other FDA departments, the wider federal government, academia and industry.
- INFORMED’s work is focused on improving the scale and relevance of clinical trial content and broader regulatory science in RWE, AI/machine learning, EHR, and patient-friendly biosensors and mobile apps.
- So what? Although its potential has been demonstrated on an ad hoc basis, big data’s prospects as a game-changer in the development, delivery and financing of medicines depends on a supportive institutional structure buttressed by coherent methodologies, consistently applied.This is not the case at present.It’s why the emerging active engagement of the FDA on digital and data science is so critical to the future growth of biopharma, which in many ways is now an information business.
Innovation is the spark of creation that waits for no one. Least of all for the regulator, whose reaction to something new is often to regulate it – usually after the need to regulate has passed.
That’s why it’s a surprise to see the FDA take the lead in tackling what’s arguably the key challenge to progress in drug development: how to transform the explosive growth in “big” data into “smart” data, delivering actionable insights to advance efficacy, safety and quality – in other words, more innovation – at every stage of the drug life-cycle.
The objective is to make data analytics an integral part of regulatory decision-making, supported by novel public-private collaborations to engage industry, academia, patients and other FDA stakeholders in a common assault against the silos that limit big data’s potential in fighting disease, particularly for biologically complex conditions like cancer.
Although the FDA has a formal strategy to broaden access to data-based digital health products – 51 such devices and drugs were authorized by the agency last year – its most important actions in the big data space have been taking place behind the scene. Specifically, it’s investing time and treasure – in the form of inter-agency and outsider expertise, not money – to create the first FDA data science “incubator” – the Information Exchange and Data Transformation Initiative (INFORMED). Announcing the formal launch of INFORMED at the AcademyHealth Datapalooza conference on April 26, Commissioner Scott Gottlieb noted how advances in information technologies are forcing “a rethink of our mandate on how we enable safe, effective innovation in this novel area, and to support its benefits to patients.”
Based in the FDA’s Oncology Center of Excellence, INFORMED has actually been at work since May 2016, when the FDA applied to the Department of Health and Human Services (HHS) Innovation, Design, Entrepreneurship and Action (IDEA) Lab as part of a competitive program to help the agency recruit a few “entrepreneur-in-residence” positions under a special HHS hiring authority covering projects of transformative potential in government. The IDEA Lab selected INFORMED for hiring support, allowing the FDA to move forward on a collaborative incubator applying data science to oncology drug development and regulatory decision-making.
Thus, from a purely organizational standpoint, INFORMED is an innovation in itself, seeking to emulate the entrepreneurial mind-set of a Silicon Valley start-up. It has no allocated budget. Staffing relies on these secondments or fellowships from other FDA departments, the federal government, industry and various academic and non-profit research partners.
Busting Those Data Silos
Sean Khozin, a physician oncologist and acting associate director of the FDA Oncology Center, serves as founding director of INFORMED, where he manages pilot collaborations with other oncologists, data scientists, programmers, statisticians and several entrepreneurs-in-residence, whose pay is funded by the FDA. “There is a gap in expertise and understanding between the tech world, the data science profession and the life sciences community, and it happens to be very wide, even today,” Khozin said in an interview with In Vivo. “INFORMED is committed to closing that gap using agile, diverse and consolidated data technologies to address real-world challenges that slow progress for cancer patients.” Khozin is strong on creating a durable infrastructure to process and make accessible large volumes of integrated data as a decision support in drug development. That requires eliminating data silos; supporting novel, evidence-based methodologies to extend the horizon of the traditional RCT and related postmarketing research; and promoting a more data-rich systems approach to the field of regulatory science, a priority of FDA commissioners for the past decade.
In other words, big data sets now rule the day at FDA – and will set the agenda for tomorrow. It’s a commitment reaffirmed by Commissioner Gottlieb’s Datapalooza pledge to “modernize our framework for advancing the most promising digital health tools. This will help us better understand variations in individual patient experience using diverse data from clinical trials, electronic health records (EHR) and biometric monitoring devices.” This modernization is focused on two areas: (1) organizing FDA’s existing clinical trial data sets into a common standard to better facilitate their use in research and regulatory decision-making; and (2) building out the FDA’s data storage and assessment capabilities to incorporate these new and diverse data streams, which, in addition to the above, includes artificial intelligence/machine learning, the emerging fields of DNA- and RNA-based multiomics as well as less structured data in the form of patient-reported outcomes and social media.
Some results are already in place, primed and ready to go. One example is a new Premarketing Digital Safety Program for drugs, announced by Gottlieb on April 26. Its key feature is the establishment of a unified data standard for the notification of adverse events occurring during a clinical trial conducted as an investigational new drug (IND) application, a category much larger than trials for new drug (NDA) or biologic license (BLA) applications. The FDA has long been aware that such reporting, conducted on an analogue, paper-based platform, was inadequate to meeting the seven- to 15-day time frame required by law, and often led to lags in signal detection and medical review. INFORMED took the responsibility to test the new digital data standard in a pilot project with four of the biggest biopharma companies (see sidebar), with positive results that will allow the initiative to enter into force this year.
Another collaboration is moving the FDA a step forward to incorporating real-world evidence (RWE) in its regulatory decision-making, particularly in finding new insights on how drug regimens shape the patient experience where it counts most – directly at the point of care. The focus is a data-sharing partnership agreed in June 2017 between INFORMED and the American Society of Clinical Oncology (ASCO), covering ASCO’s CancerLinQ inventory of thousands of de-identified patient treatment records from nearly 100 oncology practices in the US.
Ongoing review of these records will focus on the latest immunotherapy advances in melanoma cancers to plumb insights such as the impact of combination therapies and to use this to identify areas of improvement in design of clinical trials as well as the approval evaluation process at the FDA itself, including the approach to drug labeling. “There is strong pressure on the FDA to render the right decisions on access to patients as the science of cancer pharmacology improves. We see such partnering as a way to ensure the FDA is responsibly in synch with the pace of change, and how it affects patients in the actual clinical setting,” Khozin says.
Khozin notes that INFORMED is supporting similar work involving a unique crowdsourcing model to test the impact of RWE on outcomes for patients with prostate cancer, in conjunction with Project Data Sphere, an independent non-profit sponsored by the CEO Roundtable on Cancer (Also see "Free And Open: The Next Wave In Clinical Trial Data?" - In Vivo, 10 May, 2017.). Likewise, INFORMED has forged an ongoing work relationship with Flatiron Health Inc., a leader in the development of digital data technologies for cancer acquired by Roche in February. (Also see "'Watch This Space' Roche Execs Say, Outlining RWE Rationale For Flatiron Buy " - Scrip, 26 Apr, 2018.) Again, the focus is on application of RWE, using EHR files involving patients in community-based cancer clinics. A joint study presented last June at the 2017 ASCO Annual Meeting found that patients with non-small cell lung cancer (NSCLC) receiving PD-L1 checkpoint inhibitors experienced different outcomes than revealed through the RCT format, a result that the FDA hopes will be analyzed to shape the design of future trials involving checkpoint inhibitors, particularly for patients subject to previous co-morbidities, in addition to NSCLC.
Now that INFORMED has been publicly launched, Khozin expects his group’s work output to increase. “This year we aim to demonstrate how big data is going to extend the boundaries of translational research into new areas focused on oncology. It’s foundational work that needs to be done if the FDA, industry and academia are to fulfill the public’s expectations for the changes in drug development, laid out in the 21st Century Cures Act passed by Congress in December 2016.” The tools of choice will be publication as well as additional new collaborations. “We publish for reasons very different than the norm of seeking a credential embellishment – instead, it’s to get ideas that we think are important out to the community of users, who relate to the actual experience of patients.” It’s a long-term game, says Khozin, but one where risks can be controlled with metrics to guide product development, making the process more predictable to research sponsors.
What’s Ahead: Practical Apps
Four areas of investigation will attract the most attention from INFORMED this year. The intent is very grounded: to build a framework of foundational learnings around big data and then help translate these to the actual practice of drug development and FDA regulatory science.
The first is to continue looking at the use of EHR to test the utility of RWE both inside and outside the trial space, involving drugs under the FDA regulatory authority, and which will hopefully shape the development of future agency guidance to industry in this area.
Second, INFORMED is actively examining the potential of the Blockchain in addressing concerns about privacy, security and operability of sharing data, especially for the large-scale transactions required in the oncology space. “We have underway a very exciting project with IBM Watson Health looking at whether we can use Blockchain to create a worldwide web for health data exchange – one that, like the Internet today, will belong to no one and everyone, at the same time,” Khozin says. “It’s a data ecosystem that will be both secure and decentralized.” He tells In Vivo that INFORMED and Watson now have a technological solution reliant on Blockchain that is in the late alpha, early beta stage of testing. “Later this year, we are going to demo this solution and launch a public discussion on ways to take the model forward, to feed into this big data ecosystem we and our collaborators are seeking as the end game.”
The third focus is on artificial intelligence (AI) and machine learning. In a collaboration with the Harvard Medical School Program in Therapeutic Science (HiTS), INFORMED is pursuing work that leverages the significant knowledge about AI (new technology around neural networks is a prime example) accumulated since the 1970s. INFORMED is supporting a post-doctoral two-year fellowship to build contacts with academic institutions (Massachusetts Institute of Technology, Harvard University and Tufts University) and research hospitals (Dana-Farber Cancer Institute, Massachusetts General Hospital and Brigham and Women’s Hospital) to design, test and implement AI and machine learning models to support improvements in regulatory review, drug development and patient outcomes.
In addition, the Harvard group was expanded this month to include a partnership with Aetion Inc., a New York-based, privately held software company specializing in RWE. In the project, which is being funded and led by the FDA Center for Drug Evaluation and Research (CDER), Office of Medical Policy, Aetion will seek to replicate, with its own proprietary analytically based RWE platform, the results of 30 FDA published RCTs, covering CVD, and endocrine, musculoskeletal and pulmonary conditions, this time in actual clinic settings. The goal for this three-year effort – results of which will roll out sequentially, with some early milestones due later this year – is to determine whether RWE would have produced the same regulatory decisions as the FDA took originally, in relying on the RCT standard. If so, it will help to validate greater use of RWE to accelerate drug approval and access decisions, and thereby help fulfill a key legislative mandate of the 21st Century Cures Act.
“The FDA is the standard bearer on evidence in medicine for the entire nation, so I believe the methodology we hope to develop through this project will give more certainty and structure to use of RWE in access decisions taken by the major commercial payers and health plans,” Sebastian Schneeweiss, MD, ScD, Aetion’s co-founder and science lead, who also serves as a professor at Harvard Medical School and vice chief of the Division of Pharmacoepidemiology and Economics at Brigham and Women’s Hospital, tells In Vivo. Adds Aetion CEO Carolyn Magill, “If we find that RWE can indeed play a role to augment the traditional RCT, or, in some cases, allow the FDA to support a trial entirely using real-world data, the result will be lower costs in bringing drugs to market – and faster access for patients.”
Specific to machine learning, INFORMED will disclose results from another collaboration with Project Data Sphere, where a study has been conducted on using sophisticated algorithms to detect tumor dynamics based on computer vision and AI as well as to map changes in tumor composition over time. These algorithms will expand on the current human visual inspection of radiological screens and, through a combination of mathematics and predictive analytics, may accurately automate the review of the changes in tumor growth, in a three-dimensional format. “Obviously, the results using these technologies have to be validated,” Khozin observes, “but one can imagine how measuring tumor progression using precision mathematics and pattern recognition will minimize human error and thus be very useful in making FDA guidance more relevant and effective in the clinic. To FDA, that’s the bottom line.”
The fourth priority is research on the use of biometric sensors, in close collaboration with providers and patients. With the National Institutes of Health's National Cancer Institute (NCI), INFORMED is conducting a small observational study to apply sensors to track the progress of cancer patients on basic indicators of performance (physical and cognitive functionality) that will then be analyzed against a cohort of patients without sensors, to determine criteria for clinical trial participation and administration of drug therapy. The idea is to also create incentives for industry device-makers to develop new biomarkers guided by the specific technology needs identified in the study.
The active sponsorship of Commissioner Gottlieb and high-profile staff such as Director of the Office of Hematology and Oncology Drug Products Richard Pazdur, MD, will be crucial to the momentum of this highly leveraged initiative going forward. Broad support from a cross-section of interests is necessary to avoid INFORMED projects from becoming “data dumps” that fail to drive changes in regulatory practice carrying a measurable impact on drug developers, providers and patients. So far, however, reaction from one key constituency – biopharma – has been positive. Bill Louv, a former corporate venture capitalist for GSK and new president of the CEO-backed cancer trial open-access advocacy group, Project Data Sphere, tells In Vivo, “In the competitive world of pharma today, it takes a set of truly visionary leaders from industry, government and academia to work together. The FDA is a big part of this collaborative vision, so we look forward to continue working with INFORMED and the agency to advance transparency in regulatory science and improve the lives of cancer patients.” Sounds like an endorsement – and an invitation to FDA to deliver.