Prediction of outcome in multiple sclerosis based on multivariate models

Abstract

An incidence cohort of 308 multiple sclerosis patients was followed up repeatedly during at least 25 years of disease. In the patients with acute onset, multivariate survival analyses were performed and predictive models created. The endpoints DSS 6 and start of progressive disease were used. A number of variables were tested. The most important of these for prediction and therefore included in these models were: age at onset, sex, degree of remission after relapse, mono- or polyregional symptoms, type of affected nerve fibres, number of affected neurological systems. The relapse rate did not correlate with prognosis. In the predictive models, coefficients and risk ratios are provided that can be used for calculating the risk of progression and DSS 6 or to predict the median time for these endpoints in individual patients. It was also found that the risk of progression is not constant, but has a maximum a certain time after disease onset. For a patient with early onset, the risk is low in the beginning, but reaches a maximum level, which is several times higher, after about 15 years. The patient with a late onset has a much higher risk of endpoint immediately after onset, but reaches the maximum in a few years, and after that the risk decreases