A data-driven model of disability progression in progressive multiple sclerosis

Abstract

Bayesian learning; Multimodal data; Primary progressive multiple sclerosis

Aprendizaje bayesiano; Datos multimodales; Esclerosis múltiple progresiva primaria

Aprenentatge bayesià; Dades multimodals; Esclerosi múltiple progressiva primària

This study applies the Gaussian process progression model, a Bayesian data-driven disease progression model, to analyse the evolution of primary progressive multiple sclerosis. Utilizing data from 1521 primary progressive multiple sclerosis participants collected within the International Progressive Multiple Sclerosis Alliance Project, the analysis includes 18 581 longitudinal time-points (average follow-up time: 28.2 months) of disability assessments including the expanded disability status scale, symbol digit modalities, timed 25-foot-walk, 9-hole-peg test and of MRI metrics such as T1 and T2 lesion volume and normalized brain volume. From these data, Gaussian process progression model infers a data-driven description of the progression common to all individuals, alongside scores measuring the individual progression rates relative to the population, spanning ∼50 years of disease duration. Along this timeline, Gaussian process progression model identifies an initial steep worsening of the expanded disability status scale that stabilizes after ∼30 years of disease duration, suggesting its diminished utility in monitoring disease progression beyond this time. Conversely, it underscores the slower evolution of normalized brain volume across the disease duration. The individual progression rates estimated by Gaussian process progression model can be used to identify three distinct sub-groups within the primary progressive multiple sclerosis population: a normative group (76% of the population) and two ‘outlier’ sub-groups displaying either accelerated (13% of the population) or decelerated (11%) progression compared to the normative one. Notably, fast progressors exhibit older age at symptom onset (38.5 versus 35.0, P < 0.0001), a higher prevalence of males (61.1% versus 48.5%, P = 0.013) and a higher lesion volumes both in T1 (4.1 versus 0.6, P < 0.0001) and T2 (16.5 versus 7.9, P < 0.0001) compared to slow progressors. Prognostically, fast progressors demonstrate a significantly worse prognosis, with double the risk of experiencing a 3-month confirmed disease progression on expanded disability status scale compared to the normative population according to Cox proportional hazard modelling (HR = 2.09, 95% CI: 1.66–2.62, P < 0.0001) and a shorter median time from the onset of disease symptoms to reaching a confirmed expanded disability status scale 6 (95% CI: 5.83–7.68 years, P < 0.0001). External validation on a test set comprising 227 primary progressive multiple sclerosis participants from the SPI2 trial produced consistent results, with slow progressors exhibiting a reduced risk of experiencing 3-month confirmed disease progression determined through expanded disability status scale (HR = 0.21), while fast progressors facing an increased risk (HR = 1.45). This study contributes to our understanding of disability accrual in primary progressive multiple sclerosis, integrating diverse disability assessments and MRI measurements. Moreover, the identification of distinct sub-groups underscores the heterogeneity in progression rates among patients, offering invaluable insights for patient stratification and monitoring in clinical trials, potentially facilitating more targeted and personalized interventions.

This investigation was supported by an award from the International Progressive Multiple Sclerosis Alliance (award reference number PA-1412-02420). This work was supported by #NEXTGENERATIONEU (NGEU) and funded by the Ministero dell’Università e delle Ricerca, Italy (MUR), National Recovery and Resilience Plan (NRRP), project MNESYS (PE0000006)—A Multiscale integrated approach to the study of the nervous system in health and disease (DN. 1553 11.10.2022). The authors acknowledge financial contribution from the Italian Ministry of Health, with the project NeuroArtP3 (NET-2018-12366666). S.G. acknowledges financial contributions from the Istituto Nazionale di Alta Matematica—Gruppo Nazionale di Calcolo Scientifico (INdAM–GNCS) Project (CUP_E53C22001930001). C.T. is currently being funded by a Miguel Servet contract, awarded by the Instituto de Salud Carlos III (ISCIII), Ministerio de Ciencia e Innovación de España (CP23/00117). She has also received a 2023 FORTALECE grant, awarded by the ISCIII (FORT23/00034) and a 2020 Junior Leader La Caixa Fellowship (fellowship code: LCF/BQ/PI20/11760008), awarded by ‘la Caixa’ Foundation (ID 100010434), a 2021 Merck’s Award for the Investigation in multiple sclerosis, awarded by Fundación Merck Salud (Spain), and a Research Grant awarded by the ISCIII, Ministerio de Ciencia e Innovación de España (PI21/01860). M.L. acknowledges funding from the Michael J. Fox Foundation for Parkinson’s Research (grant ID MJFF-021683). F.B. received the 2022 Biostatistic/Informatics Junior Faculty Award (grant code BI-2107-38160) awarded by the National Multiple Sclerosis Society.