How can we analyze natural history across multiple sites without moving patient data? – A feasibility study and early pilot using PUL 2.0

Topic:

Other

Poster Number: 158 M

Author(s):

Francesco Muntoni, MD, Dubowitz Neuromuscular Centre, UCL and Great Ormond Street Hospital Trust, London, UK, Erik Niks, MD, PhD, Leiden University Medical Center, Leiden, Netherlands, Michel Michaels, Leiden University Medical Center, Michel Matthew Brooke, PhD, University College London, Yvonne Meijer-Krom, PhD, Leiden University Medical Center, Brenda Wong, MD, Duchenne Muscular Dystrophy Program, UMass Chan Medical School, Laurent Servais, MD PhD, University of Oxford, Wouter Franke, MSc, The Hyve, Julia Kurps, PhD, The Hyve, Jonathan Freimark, MPA, Analysis Group Inc., Emma Billmyer, MS, Analysis Group Inc., Jess Marden, BA, MPH, ScD, Analysis Group, Inc., Boston, Massachusetts, USA , James Signorovitch, PhD, Analysis Group, Inc., Boston, Massachusetts, USA , Susan J Ward, PhD, cTAP

Context
Analysis of Real World Data from patients with DMD has proven pivotal in advancing understanding of natural history, which in turn has led to better clinical trial design and analysis. However, natural history can evolve. With both care standards and the availability of new therapies evolving differently across centers and geographies, it is imperative that comparisons of outcomes are understood in the context in which they were observed.
Method
To address these concerns, cTAP engaged The Hyve, a knowledge engineering consultancy, to investigate collaboration via federated analyses, an approach that enables analyses of data without moving data outside organizations and geographies. A feasibility study was conducted through semi-structured interviews with four geographies assessing interoperability along three axes: data, technology, and governance. Feasibility was ranked as Low, Medium, or High based on a qualitative assessment of the extent of changes required to processes, technologies or data. A pilot study adapted analytical code to the data model used at LUMC which was then run against patient data at that center.
Results
A federated framework is feasible. For data, sites collect similar data and use the same functional assessments. For technology, there is no existing end-to-end solution, but there are components available that can be assembled. For governance, sites are fully compliant and require no change. cTAP will need to formalize current manual processes for securing iterative site approval of analyses to ensure an auditable governance.

Center Data TechnologyGovernance
NorthStar UK M/H H H
ActiLiège Next BEM/H M H
DDD NL H H H
U-Mass USA H M H

Execution of the code in the pilot study generated a prognostic model for PUL 2.0 in a single run with no errors.
Conclusion
Enabling important scientific analyses requiring multi-institutional collaboration to be addressed without moving patient data is feasible. A decade of successful multi-institutional collaboration with cTAP lowers the barrier to execution.