Using different criteria for classifying patients into various stages of a disease can modify the stage-specific prognosis, even though the overall disease course remains unchanged. This is known as the "Will Rogers phenomenon," precluding the use of historical controls for treatment trials. We assessed whether the Will Rogers phenomenon may affect multiple sclerosis (MS) prognosis when applying different diagnostic criteria. Patients with a clinically isolated syndrome (CIS) suggestive of MS were studied. After 1 year, each patient was classified as CIS or evolved to MS according to two diagnostic criteria (Poser and McDonald). The outcome for prognosis was the time to reach an Expanded Disability Status Scale score > or = 3.0. 309 patients were studied for a median period of 84 months. After 1 year, 16% of patients had MS according to Poser and 44% according to McDonald criteria. The probability to reach Expanded Disability Status Scale score > or = 3.0 at median follow-up was 11% in CIS patients according to Poser and 7% according to McDonald criteria; it was 46% in MS patients according to Poser and 27% acccording to McDonald criteria. The group with a discordant diagnosis had a worse prognosis than that of CIS patients according to both criteria (p = 0.01), but better than that of MS patients according to both criteria (p = 0.01). The use of different diagnostic criteria may generate spurious improvements in the medium-term prognosis of MS. This calls for caution in using historical controls for MS trials.