Multiple sclerosis (MS) is a chronic inflammatory and degenerative disease of the central nervous system (CNS). Relapses (or “attacks”) affect 85% of patients and are a defining feature of MS,1 differentiating the MS phenotypes: relapsing-remitting (RR), secondary-progressive (SP) and primary-progressive (PP) MS.2 Relapses are associated with acute focal inflammation and demyelination.3 From the patient’s perspective, relapses are disruptive and are a source of anxiety.4

Currently, the primary effect of the most widely used immunomodulatory drugs (IMDs) for MS has been to reduce the relapse rate; the much-hoped-for beneficial effect on disability progression has not been conclusively shown.5 6 Consequently, when treating MS, an understanding of the natural history of the targeted disease process—ie, relapses—is fundamental. This population-defining information is required to facilitate the: prognosis of disease activity (expected relapse rate); evaluation of drug effectiveness in clinical practice; selection of the most appropriate patients to treat in clinical practice (ie, which patients will derive the most benefit from IMD treatment). However, to date, many natural history studies in MS have centered around the (important) outcome of disability progression, with relapses typically being a risk factor rather than an outcome.7 Often, time to a fixed disability milestone or to onset of SP MS is measured;7 however, neither of these outcomes may ultimately be altered by use of an IMD.

Those requiring relapse data might derive this information from the placebo arms of clinical trials.8 9 The relapse information derived from these well-controlled prospective studies can be of considerable value—for instance, in the design or interpretation of other similar studies. However, these cohorts are rarely representative of the RR MS population that are treated in clinical practice. In addition, relapse data are typically only collated for patients with (recent) disease activity who are followed for relatively short periods of time, which is of limited use in a slowly evolving disease such as MS, which spans decades.

Natural history studies can provide relapse information of a more encompassing nature, being representative of the MS population over longer periods of time. However, the nature and duration of these studies means that regular examinations are impractical, such that these studies will probably underestimate the true relapse rates.10 11

Valuable information has been derived from previous natural history studies explicitly describing relapses as an outcome (rather than just a risk factor). However, these studies are now dated, typically being published more than two decades ago.10 12^–14 Divergence in other measures of disease activity (disability progression) between older and more recent studies is now evident,15 16 further highlighting the need for updated relapse information.

We set out to provide a contemporary description of relapses in our British Columbian MS population. Given the well-known limitation of retrospective studies of MS relapses,10 11 we primarily examined the relative pattern of relapse rates over time (rather than absolute values) and also investigated the onset of extended relapse-free periods.

METHODS

This study refers to a cohort of patients previously described.15 17^–20 Briefly, patients with laboratory-supported or clinically definite MS (Poser criteria21) with an initial RR course were selected from the British Columbia (BC)-wide MS database (established in 1980), which linked the four MS clinics in BC, capturing 80% of the MS population in this Canadian province.22 23 Patients were followed prospectively from their initial MS clinic visit. Detailed historical information, including the onset date and subsequent relapses, were recorded on the first visit by an MS specialist neurologist. This information was elicited from both the patient and the accompanying physician(s) referral letters. Thereafter, relapse and other clinical information was recorded by the same MS specialist neurologist, typically on a yearly basis (this frequency of follow-up for a neurological specialty follow-up is typical in BC). The same five core neurologists examined more than 95% of patients over the course of the study. Immunomodulatory drug use (beta-interferon or glatiramer acetate; natalizumab was not licensed during the study) either through clinical practice or clinical trials was captured by the database and examined. The study end date was 1 July 2003. The inclusion criteria are shown in figure 1. To maximise the possibility of a substantial and meaningful follow-up time, patients with first-onset symptoms prior to July 1988 were selected and, to enable establishment of the disease course, a first clinic visit prior to July 1998 was required. No minimum active follow-up time was required; therefore, those who died or moved away could still be eligible. Ten patients were excluded because of the absence of clinical data (no disability scores).

All relapses were confirmed by an MS specialist neurologist, and were defined by new or worsening symptoms lasting more than 24 hours in the absence of fever or infection. Episodes occurring within 30 days of each other were considered to be part of the same relapse. Relapse rate patterns were described over time, from both the onset of MS and according to the current age of the patient as well as age at onset and by sex. Sex differences might be biased by the frequency of visits/follow-up for men and women, affecting the frequency of reported relapses.10 11 We therefore examined the potential for frequency bias by comparing the visit rate (number of visits/follow-up) between men and women using the Mann–Whitney U-test. The first (onset) attack was also defined as a “relapse” but was excluded from the analyses unless otherwise specified (inclusion of this first attack could artificially inflate the magnitude of the decline in relapse rates overtime).

Statistical analyses

Longitudinal analysis

A longitudinal analysis of the relapse count in each 5-year incremental interval from onset (>0–5 years (the onset relapse was excluded); >5–10; >10–15; >15–20; >20–25; >25–30; >30) was carried out using a (quasi) Poisson regression model with a free scale parameter (to account for over-dispersion).24 The generalised estimating equation (GEE)25 method with an independent “working correlation matrix” was used to fit the model. The logarithm of exposure time was included as an offset to account for the differences in follow-up time for each patient. This model was used to examine the relationship between the relapse rate after onset of MS and baseline characteristics (sex, age at onset (>20 years, >20–30, >30–40 and >40) and onset symptoms (motor, sensory, optic neuropathy and cerebellar, ataxia or brainstem)), as well as disease duration. Interaction between age at onset and disease duration was allowed in a further analysis as the relapse rate appeared to vary over time among the different age at onset groups. Results were expressed as a percentage change in relapse rate with 95% confidence intervals (adjusted for the other factors included in the model).

The main longitudinal analysis included all patients (regardless of whether SP MS had been reached or not). This, in part, mimics clinical practice whereby IMD treatment that was initiated in the RR MS phase may well continue regardless of whether SP MS is subsequently reached. However, this analysis was repeated only considering relapses occurring in the RR phase.

Relapse-free periods

A relapse-free period of 5 years occurring during the RR phase was considered to be an outcome of clinical interest. Time to the first 5 years relapse free was estimated from Kaplan–Meier curves as the time from onset to the first relapse-free year, which heralded the start of five consecutive relapse-free years. Those patients who did not reach the outcome of interest were censored at the time of their last recorded relapse.

Although IMD use was low in the population,20 the main longitudinal analysis was repeated, excluding data collected subsequent to the first use of an IMD.

Some patients exhibited a longer lag time between onset of MS and first clinic visit than others. Although those referred to clinic earlier probably exhibit a more active disease, recall bias might also be the lowest in this group. The main longitudinal analysis was therefore repeated in those who came to clinic within the first few years from onset.

SPSS version 15 was used for all statistical analyses. Ethical approval was obtained from the University of British Columbia.

RESULTS

Aspects of this population have been described previously.15 17^–20 Of the 2837 patients eligible for this study (see fig 1), 2485 (87.6%) patients exhibited a relapsing course from onset, the remainder being primary-progressive.19 Eight patients were excluded from this current analysis as their relapse history was unclear (typically, the patient had no referral notes when being transferred from a clinic outside of BC). The characteristics of the 2477 considered in the current analysis are shown in table 1. The mean follow-up time from onset of MS was 20.6 years. There were 51,120 person-years of follow-up, during which time 11,722 post-onset relapses were recorded.

Relapse rate from onset and by age

The annualised relapse rates for the entire cohort steadily decreased over time from disease onset (see fig 2A), averaging a 17% reduction every 5 years (time period considered: >5–30 years from onset). However, the relapse rate did not decrease over time for all patients, with those aged <20 years exhibiting a smaller increase in relapses, peaking at 20–<30 years of age (see fig 2B and 3). When expressed according to the patients’ age, there was an observed increase in relapse rates, peaking in the third decade of life (see fig 2B).

Predictors of relapse rate—longitudinal analysis

The estimated percentage change in relapse rates are shown in figures 4, 5A and 5B. Women had a 14.3% (95%CI: 5.8% to 23.3%) higher relapse rate than men over the course of the study (p = 0.001, adjusted for onset symptoms, age at onset and disease duration; see fig 4). This gender difference is also depicted in figures 2A and 2B. The actual clinic visit rate (number of visits/follow-up) was not different between men and women (p = 0.450, Mann–Whitney U-test, data not shown), indicating that the observed gender differences were unlikely to be due to frequency bias.

The presence of sensory symptoms at onset were associated with a 19.1% (95%CI: 9.1% to 30.1%) higher relapse rate (compared to an absence of sensory symptoms at onset, p<0.0005, GEE; see fig 4). No other onset symptom exhibited a substantial effect (p>0.05; see fig 4).

The relapse rate tended to decline as the onset age increased (see fig 4), with patients 40+ years at onset having a reduced relapse rate during follow-up compared to those who were younger at onset (<20 years, p<0.0005, GEE; see fig 4).

The relapse rate also varied with increasing disease duration (p<0.0005; see fig 5A). For every 5-year increase in disease duration, there was a significant decrease in the relapse rate (compared to the first 5 years’ disease duration; see fig 5A). The largest decrease occurred in those with >5–10 years disease duration where a 22.7% (95%CI: 17.6% to 27.6%) relapse rate reduction occurred (onset relapse was excluded). However, this pattern was influenced by the age at onset of MS (see fig 5B). Each onset age group started with a different relapse rate and declined with a different rate over time, as follows: 30.5% (95%CI: 24.5% to 35.9%) decrease for every 5 years’ disease duration in those oldest at onset (40+ years); a 22.9% (95%CI: 19.4% to 26.2%) decline in those aged 30–<40 years at onset; a 16.9% (95%CI: 14.3% to 19.4%) decline in those aged 20–<30 years and a 6.9% (95%CI: 2.1% to 11.5%) decline in those subjects aged <20 years (estimations were derived from the longitudinal analysis adjusted for gender and onset symptoms, with disease duration included as a continuous predictor as the decline in relapses followed an approximate linear trend (on the log scale); only data gathered prior to 30 years from onset were considered).

Relapse-free periods (during the RR phase only)

In those patients with more than 5 years of follow-up in the RR phase (n = 2189; 30 were excluded because they reached SP MS at an unknown date20), 1692 (77.3%) experienced a 5-year relapse-free period. The median time to the start of the first 5 years’ relapse-free period was 4.0 years (95%CI: 3.6 to 4.4).

Events occurring before and after the first 5 years relapse free

The relapse rate prior to the first 5 years’ relapse-free period varied considerably, averaging 1.04 relapses per year (SD: 0.561, range: 0.27–6.0). Relapses did occur after a relapse-free period, albeit at a seemingly lower rate (see fig 6; again, only relapses in the RR phase are shown).

We examined whether the first 5-year relapse-free period (experienced during the RR phase) merely heralded the onset of SP MS. This did not appear to be the case, as those patients who experienced a 5-year relapse-free period took longer to reach SP MS, taking an estimated median 23.0 years (95%CI: 22.2 to 23.9) compared to 10.9 years (95%CI: 10.0 to 11.9) in those without such a relapse-free period (p<0.0005 (log-rank test)).

Use of IMDs

Four hundred and sixteen of 2477 (16.8%) patients started an IMD at sometime during the study; 78.2% (326/417) were women, being a mean 45.1 years (SD = 7.92) at the start of therapy, which was initiated a mean of 17.8 years post-onset of MS (range: 2–49 years). The fraction of time spent on an IMD was relatively small—the mean follow-up time decreased slightly from 20.6 years (SD = 9.79) for the whole study to 19.9 years (SD = 9.83) when “IMD contaminated” data was removed. Consequently, findings differed little when the main longitudinal analysis was repeated after removing any follow-up that occurred subsequent to starting an IMD (data not shown). The direction of findings also did not differ when the main longitudinal analysis was repeated, excluding relapses occurring in the SP MS phase.

Seen within 5 years from onset

The mean time to the first clinic visit from onset was 12.1 years (median = 10.4 years). In all, one-quarter of patients (626/2477) were seen within 5 years from onset. The main longitudinal analysis was repeated in this group. Although relapse rates were initially higher in this subgroup—averaging 0.9 per year for the first 5 years of disease duration, as expected—these rates were not sustained and declined steeply, ranging from a 57.5% (95%CI: 44.3% to 67.6%) decrease for every 5 years’ disease duration in those aged 40+ years at onset to 40.8% (23.6% to 54.1%) in subjects aged <20 years at onset. As in the whole cohort, those oldest at onset (40+ years) exhibited a reduced relapse rate during follow-up than those youngest at onset (p = 0.012), and women had an 18.4% (95%CI: 2.6% to 36.7%) higher relapse rate than men (p = 0.021, adjusted for onset symptoms, age at onset and disease duration). Although there was a trend for the presence of sensory symptoms to be associated with a higher relapse rate (by 10.6% (95%CI: −6.0–30.0), this did not reach significance in this smaller subgroup (p = 0.223).

DISCUSSION

We report relapse patterns over time in a sizable and largely IMD-free MS population. The ubiquitous use of IMDs reduces the likelihood of future ethically designed population-based natural history studies. Our findings must be taken in context of the retrospective study design, with higher absolute relapse rates being expected in a prospective study.10 11 However, in recognition of this, we focused findings on the relative relapse frequencies rather than absolute rates and, in addition, replicated findings in patients examined relatively soon after onset. Our study has additional practical relevance, being more representative of the evaluation frequency and patient characteristics seen in clinical practice, with relapse histories taken, on average, every 1.1 years from the first clinic visit. Additionally, the consistency of neurological care in our study (more than 95% of patients were examined by the same five core neurologists over the course of the whole study) minimises problems associated with inter-rater variability.

From the population perspective, the impact of any therapeutic agent targeting the inflammatory processes in MS, and hence ability to modify recurrence of relapses, has the greatest potential during periods of high relapse activity. Relapse rates were found to be age-related, peaking for those in their third to fourth decades of life. Similar age-related patterns have been found with gadolinium-enhanced lesions,26 a biomarker for inflammatory white matter lesions.27 Relapse rates were also time-dependent, declining from onset for the whole population, averaging a 17% reduction for every 5 years post-onset; an exception being in those <20 years at onset in which a small initial increase in relapses was observed. When both onset age and disease duration were considered, the decline in relapse rates was greater for those who were older at onset than for those who were younger. These observations have several clinical implications. First, any drug that is able to modify relapse rates has the greatest potential for a population impact in those aged <40 years and within the first demi-decades of disease when the risk of a future relapse is at or approaching its peak (see fig 3 and 5B). Second, continuation of a relapse-modifying drug much beyond these periods may result in the risk of adverse effects from drug treatment, outweighing any possible benefits. Third, caution is required when interpreting long-term follow-up studies or those without a placebo arm—not only must “regression to the mean”28 be taken into account, but also the magnitude of a natural reduction of relapses with increasing age and disease duration, coupled with periods of time with little or no clinical disease activity.

There has been a lack of consensus with respect to the effect of gender on relapse rates; some studies found relapse rates to be similar for men and women,10 11 others found rates higher for women.13 Our findings support the latter, with women exhibiting a higher relapse rate than men. The sex difference appeared to be unrelated to the onset of SP MS or to a possible visit frequency bias. We were, however, unable to examine any possible reporting or physician recording biases that might occur based on the gender of the patient. MRI studies of SPMS also found women to exhibit a more inflammatory disease course than men.29 Why sensory symptoms at onset should be associated with a higher relapse rate is unclear and warrants further investigation. Our study was not designed to examine the existence of temporal changes. Neither were we able to examine MRI data—this information was not collated in our database, which was created before MRI and other imaging techniques were introduced.

Exposure to any IMD carries an inherent risk of an unwanted adverse event. Unnecessary exposure to IMDs might be avoided if a “low-risk relapse period” could be identified and predicted. These low-risk periods existed in a high proportion of patients, with more than three-quarters experiencing a demi-decade relapse free (in those with more than 5 years of follow-up in the RR phase). Interestingly, the onset of a 5-year relapse-free period (during the RR phase) did not merely herald the imminent onset of SP MS. First, the median time to the first 5 years relapse free was much shorter than the median time to SP MS, which was previously shown to be 18.9 years in the entire RR MS cohort.20 Second, patients actually took longer to reach SPMS if they had experienced a 5-year relapse-free period. Future research efforts to elucidate an accurate method of predicting the onset of extended relapse-free periods, in which initiation or continuation of a therapeutic agent aimed at reducing relapses is unlikely to be beneficial, could prove useful.