Background

Previous papers show, in patients with multiple sclerosis (MS), discordant patterns of monthly and seasonal differences in the frequency of relapses. Peaks of attacks are more often reported in spring, also supported by a meta-analysis [1–3], and in summer, up to 2010 [4–6], but there are many variations, mainly as to summer peaks. Other papers show higher frequency in winter [7], or multiple peaks (summer and winter [8], summer and spring [9]), a negative peak in late summer [10], and also only small [11] or no monthly differences [12, 13].

Latitude and climate can account only partly for the differences: summer peaks seem to be reported mainly in warmer countries [3, 6, 9], and more spring or winter attacks in northern countries [1, 3, 7], but there are warm countries with no peaks in summer [10, 11], or summer peaks (such as in North Dakota) [5]. In the Southern Hemisphere the main peak is not in summer but in autumn [13]. Papers report a positive correlation with temperature [2, 8, 13], and a negative correlation with precipitation [14]; others do not find any correlation [12].

The viral hypothesis is the most investigated factor possibly explaining seasonal differences; however, vitamin D, which is supposed to be one of the factors for latitude differences [15–23], is also taken into consideration for the spring peaks of relapses.

In a polycentric study, global data from the MSBase registry [24, 25], showed different findings: a significant seasonal variation in both hemispheres, with a main peak in June, on about 22,000 relapses, confirmed on about 5,000 first events [24], and different spring peaks (April in northern, October in southern countries) with secondary peaks in autumn, on 14,000 relapses [25].

A retrospective observational study using high-frequency serial magnetic resonance imaging (MRI), performed in 1991–1993 on 44 cases [26], shows that the peak of T2 MRI activity was in summer (March–August), strongly correlated with higher temperatures.

Recent data published by one of us [27], stratifying in the long term the seasonal frequency, found significant differences between the last two decades, with a spring peak in the 2000s versus a summer peak in the 1990s. A MSBase collaboration substudy, including only series which recorded relapses in different decades, was proposed to confirm these data.

This paper reports the data of this MSBase collaboration substudy, based on multiple series of relapses reported from 1980 to 2010, so as to compare ultradecennal trends of seasonal frequency of attacks in different countries.

Methods

MSBase is an ongoing, longitudinal, strictly observational registry open to all practicing neurologists worldwide. In collaboration with participating physicians, the MSBase registry establishes a unique international database dedicated to sharing, tracking, and evaluating outcomes data in MS [28].

Currently, seven series from six countries are included in the substudy: Argentina (Buenos Aires), Belgium (Brussels), Canada (Montreal), Italy (Chieti), Italy (Salerno), Spain (Madrid), and Turkey (Trabzon).

To minimize selection bias, only patients with complete data were included. Demographic data are presented in Table 1.

Table 1 Demographic data by series: number of patients, age at onset range (mean; standard deviation), disease duration mean (range), total relapse number
Full size table

Only attacks with a definite onset were considered; dates conventionally used to indicate relapses of unknown onset were excluded: January 1 for all series, June 1 for Spain, June 15 for Italy, Chieti.

Criteria for diagnosis of relapses were: new symptoms or signs lasting at least 24 h; exclusion of pseudorelapses or paroxystic episodes; minimal interval between two attacks of at least 30 days [29].

Austral hemisphere series (Argentina) was registered, seasonally, from July (first month) to June (last).

The number of relapses by month was calculated and stratified by decade (1981–1990, 1991–2000, 2001–2010). Positive spring (March) versus summer (July) peaks were compared by odds ratios and chi-square tests; different series were not cumulated, but compared by weighted odds ratio (Peto).

Results

The total number of relapses by month is summarized in Table 2. The 1970s and 1980s decades show of course a lower number of relapses, because MSBase registries started in the late 1990s, and previous data are anamnestic. So, only the last two decades were compared.

Table 2 Total relapses: distribution by month
Full size table

Relapse numbers for 1990 (1991–2000) and 2000 (2001–2010) decades are tabulated in Tables 3 and 4; The same data are shown graphically in Fig. 1. The number of relapses in 1990 has its peak (mode) in July (Fig. 1a); in 2000 the peak is in March (Fig. 1b).

Table 3 Relapse number 1991–2000: distribution by month
Full size table
Table 4 Relapse number 2001-2010: distribution by month (* Argentina: Jan = Jul; Dec = Jun)
Full size table
Fig. 1
figure 1

Number of relapse by series (Y-axis), distribution by month (X-axis): a 1991–2000, b 2000–2010

Full size image

However, the single series show many variations and cannot be simply cumulated, so we calculated and compared spring versus summer peaks, series by series and decade by decade. Statistical comparison between the two last decades is shown in Figs. 2 and 3. Decade comparison of the 1990s versus the 2000s shows inversion of spring–summer peak (March in 2000s, July or August in 1990s), which is 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) (Fig. 2).

Fig. 2
figure 2

Comparison of spring peak of relapses (events) versus summer peak (Total = spring and summer peak), series by series, by Peto odds ratio

Full size image
Fig. 3
figure 3

Comparison of spring peak of relapses (events) versus summer peak (Total = spring and summer peak), series by series, by Peto odds ratio: sensitivity analysis (Salerno series excluded from calculations)

Full size image

The cumulative significance was confirmed (Peto odds ratio = 1.25, CI = 1.04–1.50, p = 0.002) also in a sensitivity analysis excluding Salerno series, which is the only one showing significant results even alone (Fig. 3).

Discussion

There are no previous multicentric studies comparing different decades; the only paper, to our knowledge, addressing this issue is a single-center study by Iuliano [27].

In the present study we widened the population and geographical areas under study.

Our multiple series confirm the high variability shown in the literature (see “Background”) and can reasonably explain this variability.

Previous papers [1–14] do not seem to show definitive trends for spring or summer peaks, also when classified for year of publication, except, possibly, for some papers published after 2004 and showing multiple and hardly statistically assessable peaks, mainly in spring and summer [8, 10, 11].

There are, however, papers confirming a spring peak published in the 2000s [2, 25], and a summer peak published in the 1990s [4, 5]; in addition, a recently published paper based on MRI serial data collected in 1991–1993 also shows a significant summer peak of MRI activity [26].

There are limitations: as the number of patients and of recorded attacks is, in all series, increasing with time, reliable comparison among different decades was possible only for the last two. It is hardly possible that further addition of patients could allow more comparisons, because recalling old events is increasingly difficult with time.

Another limitation is the impossibility of obtaining old data about climate, epidemics, and environmental variables to study possible correlations, which is even more difficult as they would need to be aggregated as mean decade data.

Anyway, variables such as epidemics or local environment factors may be changing over time; other environmental factors, such as UV exposure or climate, are certainly more stable along the course of a year.

Conclusions

Comparison of series from multiple geographical areas confirms a spring peak of relapses in the 2000–2009 decade, contrary to the 1990–1999 one, where a summer peak is evident. Independently from the year period of peaks, seasonal frequency of relapses might vary in the long term, due to other factors such as viral epidemics or other environmental variables, whose effect could be more evident than climate, ultraviolet exposure, and correlated factors.