Identification of prognostic biomarkers in a large cohort of patients with LGMD R2.
Authors
Affiliations (23)
Affiliations (23)
- The John Walton Muscular Dystrophy Research Centre, Translational and Clinical Research Institute, Newcastle University and Newcastle Hospitals NHS Foundation Trust, Central Parkway, Newcastle Upon Tyne, UK. [email protected].
- The John Walton Muscular Dystrophy Research Centre, Translational and Clinical Research Institute, Newcastle University and Newcastle Hospitals NHS Foundation Trust, Central Parkway, Newcastle Upon Tyne, UK.
- The Jain Foundation, Seattle, WA, USA.
- Newcastle Magnetic Resonance Centre, Translational and Clinical Research Institute, Newcastle University, Newcastle Upon Tyne, UK.
- NMR Laboratory, Neuromuscular Investigation Centre, Institute of Myology, Paris, France.
- St Luc University Hospital, Erasme University Hospital, Brussels and University of Liege, Liege, Belgium.
- Center for Health Outcomes Research and Delivery Science, Division of Biostatistics and Study Methodology, Children's National Health System, Washington, DC, USA.
- Pediatrics, School of Medicine and the Health Sciences, George Washington University, Washington, DC, USA.
- Charite Muscle Research Unit, Experimental and Clinical Research Center, A Joint Cooperation of the Charité Medical Faculty and the Max Delbrück Center for Molecular Medicine, Berlin, Germany.
- Department of Neurology and Neurological Sciences, Stanford University School of Medicine, Stanford, CA, USA.
- The Sydney Children's Hospital Network, and The University of Sydney, Child and Adolescent Health, Sydney, Australia.
- Center for Gene Therapy, Nationwide Children's Hospital, Columbus, OH, USA.
- National Institutes of Health (NINDS), Bethesda, MD, USA.
- Department of Neurology, Washington University School of Medicine, St. Louis, MO, USA.
- Friedrich Baur Institute at the Department of Neurology, University Hospital, LMU Munich, Munich, Germany.
- Neuromuscular Unit, Department of Neurology, Hospital U. Virgen del Rocío/Instituto de Biomedicina de Sevilla, Seville, Spain.
- APHP, Reference Center for Neuromuscular Diseases, Pitié-Salpêtrière Hospital, Institute of Myology, Paris, France.
- Department of Neurology, National Center Hospital, National Center of Neurology and Psychiatry Tokyo, Tokyo, Japan.
- Neuroscience Institute, Carolinas Neuromuscular/ALS-MDA Center, Carolinas HealthCare System, Charlotte, NC, USA.
- Department of Neuroscience, University of Padova, Padua, Italy.
- The Abigail Wexner Research Institute at Nationwide Children's Hospital, Columbus, OH, USA.
- Centro de Investigación Biomédica en Red en Enfermedades Raras (CIBERER), Barcelona, Spain.
- Neuromuscular Disorders Unit, Neurology Department, Hospital de la Santa Creu i Sant Pau, Barcelona, Spain.
Abstract
Limb-girdle muscular dystrophy R2-dysferlin related (LGMD-R2) is a progressive muscle condition with marked variability in disease course, making prognosis challenging. Quantitative MRI (qMRI) has emerged as a complementary tool that may detect progression earlier and more precisely. Integrating different data modalities is challenging with conventional approaches, and artificial intelligence (AI) can help overcome this. Our aim is to develop robust models capable of predicting clinical progression in LGMD-R2 by incorporating AI-based techniques into the analysis pipeline. Data from 188 COS 1 participants were analysed. Disease progression was assessed using the North Star Assessment for Limb Girdle type Muscular Dystrophies (NSAD). Ambulatory individuals with a maximum NSAD ≥ 20 were included, and progression trajectories were identified through hierarchical clustering. Feature selection was performed using a machine learning pipeline, and top predictors were entered into stepwise logistic regression to build clinical-only and combined clinical-MRI models. Two stages of progression were identified, a fast one with a mean three-year loss of 14.4 NSAD points, and a moderate one, with a mean loss of 3.8 NSAD points. The combined model achieved better balanced accuracy than the clinical-only one (83.7% vs 78.7%). Key predictors in the combined model were disease duration and fat content measures in the anterior thigh and gracilis muscle, while the clinical model included disease duration, creatine phosphokinase (CK), and 10 m walk/run test velocity. Progression in LGMD-R2 can be grouped into distinct clinical trajectories. Individuals at a faster stage of progression were younger, had shorter disease duration, higher CK, greater weakness, and relatively preserved vastus intermedius and gracilis muscles. AI enabled efficient integration of heterogeneous data, and qMRI biomarkers provided complementary information that improved predictive accuracy.