Why our culture of data ownership must evolve to address this global healthcare crisis
Machine learning and predictive analytics are changing the way we approach recruiting for clinical research – this has particular significance for Alzheimer’s Disease (AD) studies. As we mark World Alzheimer’s Disease Day on September 21, I think it’s important to note how these tools coupled with a more collaborative data sharing environment, are essential to accelerating the development of new treatments.
Recruiting is always one of the most challenging and unpredictable aspects of planning a clinical trial, and AD studies face bigger obstacles than most. One of the biggest hurdles with this disease, is the need to identify AD patients early in the disease progression, prior to irreversible brain loss. The focus on prodromal AD (patients with AD who have early signs of memory impairment but are functionally independent) means ideal participants have likely never spoken to their physicians about their risk of AD or sought medical support for the condition, so traditional referral-based recruiting methods are unlikely to deliver results. Diagnosing these patients, using PET scans or lumbar puncture can also be complicated, expensive, invasive and not terribly specific, adding further complications.
Instead, we believe that by leveraging human data analytics, artificial intelligence and machine learning, and emerging biomarkers (blood, digital, imaging, genomics) health systems and sponsors will be able to more confidently identify pools of potential patients for these trials. This will generate faster and more predictable recruitment efforts, while increasing patient and provider awareness of Alzheimer’s risk at the earliest stages of the disease. In addition, these tools may enhance the ability of sponsors to enrich trial populations and identify patients who may be at risk for more rapid progression or who may see greater benefits from treatments.
ROADMAP for collaboration
In her recent blog, my colleague, Olga Uspenskaya-Cadoz, talked about how sponsors are already beginning to use machine learning and predictive analytics to mine multiple real-world data sets to find at-risk prodromal AD patient populations. These data sets can include electronic medical records, AD registry data, prior clinical trial results, data from digital and connected devices, imaging results, lab results and population based claims data.
The challenge now is figuring out how to access and integrate at scale these growing volumes of disparate sources of data so that we can more effectively drive novel AD research.
All of these data sets are housed in different places, in different formats, and controlled by unique owners. Each of them on its own cannot yield the deep knowledge needed to find and differentiate prodromal patients in an optimal way. Joining these pools of information can provide more complete views. To accelerate development of treatments for AD, we need to create collaborations through which traditional healthcare industry vendors, hospitals, patient advocacy groups, academia, and government agencies can easily access and integrate these data sets to develop new tools and methods and unlock novel insights about this disease. Such collaborations would make it possible to identify trends among prodromal patients and to find indicators that suggest certain populations are at higher risk of progression to symptomatic AD, which would make recruiting more efficient and predictable. Such approaches also can help identify patient populations who are more at risk for rapid progression or who may be more likely to benefit from treatment.
We are already seeing some examples of such collaborations, including The Innovative Medicines Initiative’s (IMI) ROADMAP project, which is attempting to create a collaborative framework for using real-world evidence in AD research. The €8.21 million project includes 26 collaborating partners including pharma, universities, research organizations, regulatory bodies, and non-profit groups, who are jointly committed to developing methods for the scalable, transferable integration of multiple data sets on AD patient outcomes in the real world. The tools are already in development and several are being tested through pilot projects. This research will lay the foundations for a Europe-wide platform on real-world evidence in AD.
These collaborations offer a powerful model for the future of clinical research. But there are many obstacles yet to overcome. While the technology exists to mine even the most colloquial data formats, we need to address the culture of data control that has dominated the pharma industry for decades. With the right structure, sharing data promises to deliver value to all stakeholders, improving our ability to discover new treatments faster and at lower costs.
There are also concerns about information governance and data privacy. As more information rich datasets are merged and co-mingled around AD patients, the risk that patients will unintentionally be identified becomes a risk that must be addressed through more rigorous data controls, and the integration of data privacy tools and experts on these programs. Though always a significant concern in dealing with health care data, this is particularly significant in handling sensitive data for patients who may have prodromal AD – people not yet symptomatic but who may progress to severely debilitating – and costly – disease.
The lack of treatment options means many people don’t want to know if they have this disease, particularly in the prodromal phase when they are asymptomatic. Looking at a healthy population and finding false positives for AD is potentially devastating. And, as with all innovations in clinical research, payers and regulators must be on board. It is promising that the UK’s National Institute for Health and Care Excellence (NICE) and the Dutch Medicines Evaluation Board (MEB) are both involved in the ROADMAP project. The US Food and Drug Administration (FDA) has also voiced support for the adoption of real-world evidence for agency decision-making. Conversations will continue across the pharma industry on the best use real-world data to accelerate drug development, and how to combine existing data sets for the benefits of all participating partners, including payers, regulators, physicians and patients.
Due to the considerable challenge in AD and the monumental impact personalized treatment approaches could have on patients and society, it is clear that new models of collaboration and research are needed. Given the growing availability of data, technologies and tools, and research expertise that now exists in healthcare the opportunity to achieve the scale necessary to make a real impact in AD is before us. However this requires collaboration, having the right governance model across partners, and scalable data curation and integration capabilities. The sooner we combine these assets, the faster we can accelerate efforts to treat and cure AD.