Previous papers show discordant patterns of monthly and seasonal differences in the frequency of multiple sclerosis relapses. Attacks are more often reported in spring and summer, but there are many variations, mainly as to summer peaks. This paper, an MSBase collaboration substudy, reports multiple series of relapses from 1980 to 2010, comparing ultradecennal trends of seasonal frequency of attacks in different countries. The MSBase international database was searched for relapses in series recording patient histories from 1980 up to 2010. The number of relapses by month was stratified by decade (1981-1990, 1991-2000, 2001-2010). Positive spring versus summer peaks were compared by odds ratios; different series were compared by weighted odds ratio (Peto OR). Decade comparison of the 1990s versus 2000s shows inversion of spring-summer peak (2000s = March; 1990s = July), significant in the whole group (Peto odds ratio = 1.31, CI = 1.10-1.56, p = 0.003) and in Salerno series (OR = 1.97, CI = 1.14-1.40). The global significance persisted also excluding Salerno series (Peto odds ratio = 1.25, CI = 1.04-1.50, p = 0.002). Multicentric data confirm a summer peak of relapses in the 1991-2000 decade, significantly different from the spring peak of 2001-2010. Seasonal frequency of relapses shows long-term variations, so that other factors such as viral epidemics might have more relevance than ultraviolet exposure.