To accelerate drug development for a rare disease, AL amyloidosis, we needed to evaluate a biomarker using data from isolated randomized trials. Pooling patient-level trial data siloed across academic research institutions and pharmaceutical manufacturers in the US, Europe, and China is not possible in the foreseeable future. The federated analytics approach we used allowed learning from the disparate data sources without requiring the data to leave the institutions holding them. We developed a common data model and pre-packaged analytics, then supported teams at each trial location in running harmonized analyses locally. When pooled centrally, the outputs of the local analyses produced a sufficiently robust, broad-based analysis.
To help develop a research plan, a team from the Amyloidosis Research Consortium (ARC), a patient-led, research-focused nonprofit organization, brought together leading academic experts in AL amyloidosis, drug developers, representatives from the US and European regulatory authorities, and experts in data science and statistical methodologies, through a Public-Private Partnership between ARC and the US FDA, known as the Amyloidosis Forum.
ARC also collaborated with data scientists from Analysis Group to design and implement a federated analytics system. ARC conducted outreach to the collaborators contributing data from academic and industry sponsored clinical trials. Each collaborator invested time and personnel to organize data into the common data model, run the pre-packaged analytics, and conduct quality control of results and applicability of methods and study designs to the local data.
The research team brought together a global group of individuals of varying career stages from a number of different stakeholder groups, including patients, nonprofit organizations, a multidisciplinary and cross institutional team of academic experts, drug developers, and commercial institutions.
The study was initiated in Q1 2021 and is ongoing. The ultimate goal is to ensure that evidence for (or against) the surrogacy of biomarkers studied in clinical trials of AL amyloidosis can be evaluated and published by the research community without delay. Such data are necessary for evaluating surrogacy and accelerating development of effective therapies. A scientific abstract reporting preliminary results was accepted at the International Symposium on Amyloidosis 2022.
Central to this initiative is the fact that patient-level data do not need to leave the institutions where they currently reside. We employed a federated analytics approach to develop an open and collaborative platform that could be used locally by each data center.
We strongly support the sharing of clinical trial data with centralized repositories that can make the data more widely available and accessible for research. If such sharing is not possible for cost, timeline, or privacy reasons, however, our federated analytics approach can be employed to overcome such barriers.
For a rare disease, each clinical trial contains valuable data that we cannot afford to exclude. The data we are studying are held by drug developers and academic sites in the US, Europe, and China and cannot be pooled for analyses. This would normally result in delayed and/or suboptimal evidence for the research and drug development community. By engaging all data holders in sharing harmonized analysis results via federated analytics, we removed an important barrier to scientific evidence generation, allowing the advancement of analytical research for drug development.
Our long-term vision is that all randomized trials in AL amyloidosis will be included collaboratively in this system.
We leveraged open-source analytical tools from the R-project for statistical computing, and developed a common data model that is loosely based on CDISC data formats but is designed to be easy for multiple collaborators to use even when their data is not already in CDISC format.
All code and documentation will be made available to others wishing to adapt it to other therapeutic areas. In addition, the system may be used to address additional research goals raised by collaborators in AL amyloidosis, and may also be adapted for other disease areas.
Federated analytics can be more broadly accessible and manageable to advance collaborative learning from data, and to avoid delays associated with data transfers. This is especially important as barriers to some data transfers can be prohibitively high in terms of costs, international legal requirements, and timelines. We support more open access to data and data sharing, but advocate for greater use of federated approaches to accelerate and broaden learning from data that cannot be readily transferred.
