Natural history studies have identified factors that predict evolution to multiple sclerosis or risk of disability accumulation over time. Although these studies are based on large multicentre cohorts with long follow-ups, they have limitations such as lack of standardized protocols, a retrospective data collection or lack of a systematic magnetic resonance imaging acquisition and analysis protocol, often resulting in failure to take magnetic resonance and oligoclonal bands into account as joint covariates in the prediction models. To overcome some of these limitations, the aim of our study was to identify and stratify baseline demographic, clinical, radiological and biological characteristics that might predict multiple sclerosis development and disability accumulation using a multivariate approach based on a large prospective cohort of patients with clinically isolated syndromes. From 1995 to 2013, 1058 patients with clinically isolated syndromes were included. We evaluated the influence of baseline prognostic factors on the risk for developing clinically definite multiple sclerosis, McDonald multiple sclerosis, and disability accumulation (Expanded Disability Status Scale score of 3.0) based on univariate (hazard ratio with 95% confidence intervals) and multivariate (adjusted hazard ratio with 95% confidence intervals) Cox regression models. We ultimately included 1015 patients followed for a mean of 81 (standard deviation = 57) months. Female/male ratio was 2.1. Females exhibited a similar risk of conversion to multiple sclerosis and of disability accumulation compared to males. Each younger decade at onset was associated with a greater risk of conversion to multiple sclerosis and with a protective effect on disability. Patients with optic neuritis had a lower risk of clinically definite multiple sclerosis [hazard ratio 0.6 (0.5–0.8)] and disability progression [hazard ratio 0.5 (0.3–0.8)]; however, this protective effect remained marginal only for disability [adjusted hazard ratio 0.6 (0.4–1.0)] in adjusted models. The presence of oligoclonal bands increased the risk of clinically definite multiple sclerosis [adjusted hazard ratio 1.3 (1.0–1.8)] and of disability [adjusted hazard ratio 2.0 (1.2–3.6)] independently of other factors. The presence of 10 or more brain lesions on magnetic resonance increased the risk of clinically definite multiple sclerosis [adjusted hazard ratio 11.3 (6.7–19.3)] and disability [adjusted hazard ratio 2.9 (1.4–6.0)]. Disease-modifying treatment before the second attack reduced the risk of McDonald multiple sclerosis [adjusted hazard ratio 0.6 (0.4–0.9)] and disability accumulation [adjusted hazard ratio 0.5 (0.3–0.9)]. We conclude that the demographic and topographic characteristics are low-impact prognostic factors, the presence of oligoclonal bands is a medium-impact prognostic factor, and the number of lesions on brain magnetic resonance is a high-impact prognostic factor.