Previous papers show different patterns of seasonal distribution of multiple sclerosis attacks. This paper compares long-time modifications. Salerno MS registry (Southern Italy), was reviewed, including 189 patients, age onset 12-51 years (mean = 29.88, SD = 8.4), disease duration mean = 6.94 years (1-29), attacks mean = 4.5 (2-25, SD = 3.41). Data were stratified by decades. Number of events/month was analyzed by odds ratios and forecast modeling (ARIMA); means by ANOVA and post hoc tests, and correlations by multiple regression. We found 869 relapses: J = 72, F = 48, M = 122, A = 75, M = 68, Jn = 59, Jl = 81, A = 74, S = 63, O = 70, N = 72, D = 65. In 2001-2008 there was one significant peak (March); in 1991-2000 many (greatest = July), and in 1984-1990, one positive (June), one negative (April). Differences between 1990s and 2000s are significant. It is the first study addressing ultradecennal trends, and finding that the season distribution of MS attacks is significantly different: the study confirms frequency peaks in early spring and summer, but they are different in different decades. This significant ultra-decade difference might support hypotheses more linked to infections or toxic substances than to sunlight, UV, or similar.