Logo
DeutschClear Cookie - decide language by browser settings
Hellriegel, B. and Daumer, Martin and Neiß, A. (2003): Analysing the course of multiple sclerosis with segmented regression models. Collaborative Research Center 386, Discussion Paper 355
[img]
Preview

PDF

248kB

Abstract

Multiple sclerosis (MS) is a demyelinating disease of the central nervous system whose cause is still unknown. The disease course shows great inter- and intra-individual variability and this results in insecurity of diagnosis and prognosis. A well-founded knowledge of the natural history of MS, however, is an important prerequisite for developing adequate strategies for therapy and research. In order to increase our understanding we developed a segmented regression model which extracts three main characteristics of the time course of this complex disease from natural history data. For each individual patient this model determines baseline disability (as measured by the Expanded Disability Status Scale = EDSS), the time point where the disease starts to progress and the slope of this progression. The model is applied to data of patient registries from all over the world that are pooled in the database of the Sylvia Lawry Centre for Multiple Sclerosis Research (SLCMSR). The analyses used a random subsample of the entire database and were restricted to patients seen from onset of MS with time series of at least three years. Thereby we were able to avoid some of the problems related to missing data. Our results revealed a weak negative correlation between time to progression (change point) and slope of progression for this group of patients, i.e. those patients who do progressed later and remained stable for a longer time developed disability more slowly than those who progressed earlier. For the two parameters and their interaction we did not find an influence of basic covariates like gender, disease course and mono- or poly-symptomatic disease onset. According to the SLCMSR Policy these results will be subjected to a validation using an independent "validation dataset". This remains to be done.