How advanced big data and machine learning are redefining the Alzheimer’s research landscape.
Throughout September, my colleagues and I will participate in events for World Alzheimer’s Month, a campaign launched six years ago to acknowledge the devastating physical, mental, and economic havoc this disease has wreaked on the global community. An estimated 46 million people worldwide are currently living with dementia, and that number is set to rise to over 131 million by 2050. More than half of these cases are caused by Alzheimer’s Disease (AD). The economic impact of caring for these patients is staggering. By 2030, the estimated worldwide cost of this disease will reach US$2 trillion.
There are still no treatments on the market, but the push to tackle this disease has spurred several ground-breaking collaborations and research initiatives using machine learning and big data across the globe.
A brief history of AD research
Tackling any disease is challenging, but Alzheimer’s Disease presents unique obstacles that have obstructed researchers for years. Earlier studies proved the biologic processes of AD begin decades before dementia sets-in, and that once patient exhibits symptoms reflecting down-stream neurodegenerative processes which result from build-up of pathological amyloid and hyperphosphorylated tau-protein, those symptoms cannot be reversed. That means the best hope for developing treatments requires reaching patients before they are symptomatic, which is incredibly challenging.
Identification and validation of AD biomarkers and advances in tau imaging for live patients have made it possible to recruit at-risk patients prior to the appearance of dementia. However, finding these patients and recruiting them to trials has been complicated. The lack of current treatment lowers the incentive for patients at-risk to pursue AD diagnosis, and even when they are open to participating in research, physicians are seeing around 90 percent screen failure rates in trials seeking preclinical AD patients and 75% in prodromal trials patients. It is a frustrating roadblock that has added incredible time and cost to these trials.
However, advances in the use of machine learning and big data analytics is pushing AD research in new directions and helping reduce the barriers to clinical research. By working collaboratively and leveraging existing healthcare data sources, pharma companies, government agencies, and academic centers are working together to identify new ways to accelerate recruiting and make it easier to share knowledge in an effort to find a cure.
6 projects that are changing AD research
60 million patient records: Europe has been a hub for innovative AD research with many projects featuring collaborative efforts and state-of-the-art data sharing platforms. One of the largest is EMIF-AD, a five-year Alzheimer’s programme to build a platform that shares more than 60 million patient records, and data from 70,000 people directly involved in AD research studies. By connecting relevant cohort studies across Europe and using analytics and data mining tools, EMIF-AD supports large-scale research on biomarkers and risk factors for neurodegenerative disorders, helping researchers increase their baseline knowledge, and reduce time and cost of their own initiatives.
Patient-data for personalized medicine: On a smaller scale, the three-year IASIS project is an EU funded big data project to inform development treatments for AD and lung cancer through better use of patient data. The project team has built a platform that integrates medical records data, imaging databases, and genomics data into a single source, and applies advanced analytics and machine learning to discover useful patterns to guide research efforts. Sponsors hope this innovative use of analytics will drive more personalised diagnosis and treatment approaches and inform related healthcare policy decisions.
Adaptive trial design platform: EPAD is a European initiative to build an environment for testing interventions that will prevent AD. The project plans to recruit thousands of participants from across the EU who are over 50 and have no dementia symptoms and meet other broad criteria. Then they will funnel them into an adaptive trial where multiple biopharma companies can test and compare drugs in order to streamline the discovery process. The longitudinal cohort study began recruitment in May 2016, and now has 38 partnering organizations, 20 study sites, and nearly 900 patients.
Better predictors for AD: In the US, the AD Neuroimaging Initiative (ADNI) out of the University of Southern California is a longitudinal multi-center study to develop clinical, imaging, genetic, and biochemical biomarkers for the early detection and tracking of AD. ADNI researchers are using shared data, including MRI and PET images, genetics, cognitive tests, CSF and blood biomarkers to identify predictors for the disease, and all of the data from North American study participants is freely available at the website. ADNI is part of the World Wide Alzheimer's Disease Neuroimaging Initiative (WW-ADNI) , which has similar programs underway in Europe, Japan, Taiwan, Korea, China, and Argentina.
Patient data speeds recruiting: At IQVIA, a group of researchers in the Central Nervous System Center of Excellenceused patient data and machine learning tools to improve recruiting strategies for asymptomatic AD patients. Through this program, we were able to combine information from claims databases, pharmacy audits, and geographic information about elderly communities to identify actionable insights into where clinical trial sites should be located, and how to expand referral networks to incorporate larger pools of candidates.
INSIGHT-preAD. This ongoing single-centre observational study in France is study 3018 participants, aged 70–85 years, with subjective memory complaints but unimpaired cognition and memory. Roughly one in four (28%) showed amyloid β deposition and the remainder did not. The authors concluded that Brain β-amyloidosis alone does not predict progression to prodromal Alzheimer's disease within 30 months, though longer follow-up is needed to establish whether this finding remains consistent. These types of smaller academic cohorts conducted in well controlled settings with multiple variables will help us better define predictors of conversion to clinical AD in preclinical patients, and provide valuable insights into the disease progression model.
The future of AD
Treating AD continues to be one of our greatest modern-day healthcare challenges, but these next-generation programs suggest that these obstacles can be overcome. The lessons we learn from these and other innovative AD research efforts will not only help us recruit more AD patients and ultimately find a cure, but they will help evolve the way we use big data and analytics across all drug development efforts.