Contributions
Abstract: P674
Type: Poster Sessions
Abstract Category: Clinical aspects of MS - Natural course
Future rate of disease progression in multiple sclerosis patients is difficult to predict, and clinical factors which indicate more severe disease course are limited. Identifying subsets of patients with more rapid progression with common features could lead to future biomarkers of progression and mechanistic discovery.
To determine if topological data analysis can recover subgroups of more rapidly progressing patients, with common characteristics.
Data was extracted for 14318 multiple sclerosis patients from the Swedish MS Registry with complete information regarding the patient´s age at disease onset, sex, and first EDSS score as static variables. For each neurologist visit, the age at visit and EDSS was also included, as time-varying. Each patients´ outcome was determined using the rate of progression of the EDSS score over age as determined using a linear mixed effect model. To form a binary outcome, only the most and least severely progressing patients were retained, leaving 1689 patients with an average of 11.0 EDSS scores. A topological based network graph was constructed using this information, onto which a cross-validated classifier was built. The cross-validated accuracy was used to select the best metric over a 1000 run bootstrap. In the final model, nodes with high concentrations of more severe patients were identified.
The classifier had an accuracy of 83.4% when classifying out of sample data. The model required 102 total nodes for classification, making a large number of smaller subgroups. Sixteen nodes had greater than 90% concentration of the most severe patients, with four nodes being comprised only of the severe group. Notably, one node contained 1053 patients, with 99.0% being severe patients. These groups are promising to further investigate for genetic associations that may serve as both biomarkers of more severe progression or to elucidate mechanisms of disease drivers.
Subgroups of more severe multiple sclerosis progression can be identified using topological data analysis, and these groups may be useful for further analyses.
Disclosure: Ali Manouchehrinia has received speaker honoraria from Biogen. Wojciech Chachólski and Ryan Ramanujam have nothing to disclose. Jan Hillert has received honoraria for serving on advisory boards for Biogen, Sanofi-Genzyme and Novartis and speaker´s fees from Biogen, Novartis, Merck-Serono, Bayer-Schering, Teva and Sanofi-Genzyme. He has served as P.I. for projects, or received unrestricted research support from, Biogen, Merck-Serono, TEVA, Sanofi-Genzyme and Bayer-Schering. This MS research was funded by the Swedish Research Council and the Swedish Brain foundation.
Abstract: P674
Type: Poster Sessions
Abstract Category: Clinical aspects of MS - Natural course
Future rate of disease progression in multiple sclerosis patients is difficult to predict, and clinical factors which indicate more severe disease course are limited. Identifying subsets of patients with more rapid progression with common features could lead to future biomarkers of progression and mechanistic discovery.
To determine if topological data analysis can recover subgroups of more rapidly progressing patients, with common characteristics.
Data was extracted for 14318 multiple sclerosis patients from the Swedish MS Registry with complete information regarding the patient´s age at disease onset, sex, and first EDSS score as static variables. For each neurologist visit, the age at visit and EDSS was also included, as time-varying. Each patients´ outcome was determined using the rate of progression of the EDSS score over age as determined using a linear mixed effect model. To form a binary outcome, only the most and least severely progressing patients were retained, leaving 1689 patients with an average of 11.0 EDSS scores. A topological based network graph was constructed using this information, onto which a cross-validated classifier was built. The cross-validated accuracy was used to select the best metric over a 1000 run bootstrap. In the final model, nodes with high concentrations of more severe patients were identified.
The classifier had an accuracy of 83.4% when classifying out of sample data. The model required 102 total nodes for classification, making a large number of smaller subgroups. Sixteen nodes had greater than 90% concentration of the most severe patients, with four nodes being comprised only of the severe group. Notably, one node contained 1053 patients, with 99.0% being severe patients. These groups are promising to further investigate for genetic associations that may serve as both biomarkers of more severe progression or to elucidate mechanisms of disease drivers.
Subgroups of more severe multiple sclerosis progression can be identified using topological data analysis, and these groups may be useful for further analyses.
Disclosure: Ali Manouchehrinia has received speaker honoraria from Biogen. Wojciech Chachólski and Ryan Ramanujam have nothing to disclose. Jan Hillert has received honoraria for serving on advisory boards for Biogen, Sanofi-Genzyme and Novartis and speaker´s fees from Biogen, Novartis, Merck-Serono, Bayer-Schering, Teva and Sanofi-Genzyme. He has served as P.I. for projects, or received unrestricted research support from, Biogen, Merck-Serono, TEVA, Sanofi-Genzyme and Bayer-Schering. This MS research was funded by the Swedish Research Council and the Swedish Brain foundation.