Authors: Neuro Compass
Patients with relapsing–remitting multiple sclerosis (RRMS) treated with disease-modifying therapies (DMTs) in the modern era have a more favourable outlook than historical untreated cohorts. This was the conclusion from two recent publications describing disability outcomes and associated predictive factors in treated RRMS cohorts, using 10 years of ‘real-world’ follow-up data [1,2].
The main caveats to these analyses are the increasingly inclusive criteria for the diagnosis of MS that have been applied in recent years and the lack of published data in the historical cohorts on baseline clinico-radiological characteristics, which preclude any accurate comparisons. As with previous clinical, radiological and pathological studies, there remains conflict as to the link between inflammation and neurodegeneration in MS.
Given the conflicting findings and the varied sample sizes, further work is needed to clarify the benefit of setting short-term treatment targets and using regular magnetic resonance imaging (MRI) to assess treatment effectiveness if the therapeutic aim is long-term disability prevention.
The following article is taken from the Hot Topics section of Neuro-Compass. Neuro-Compass is a free, expert-led practical medical education resource developed for all healthcare professionals caring for people with MS. Sign up to Neuro-Compass for free at www.neuro-compass.education to read more Hot Topic articles and for access to extensive clinically relevant education from experts in MS.
Key points from these studies
- Disability outcomes in treated RRMS populations in the modern era appear significantly better in comparison to natural history studies.
- The link between early ‘inflammatory’ activity in RRMS (relapses/MRI activity) and longer-term disability still remains uncertain.
- The utility of an early ‘treatment to target’ policy (including no evidence of disease activity [NEDA]) to reduce long-term disability is challenged by a well-conducted prospective study.
- These studies do not address the utility of long-term MRI monitoring to assess disease activity or treatment response to guide therapeutic decisions.
- These studies do not include patients treated with more recent disease modifying therapies, including oral disease modifying therapies or monoclonal antibodies
Dr Paul Gallagher (Institute of Neurological Sciences, Glasgow, UK) provides an overview of these two recent papers by Jokubaitis and colleagues  and Cree and colleagues , which both assessed factors predictive of long-term disability outcomes in people with multiple sclerosis (MS).
- Jokubaitis VG, Spelman T, Kalincik T, et al. Predictors of long-term disability accrual in relapse-onset multiple sclerosis. Ann. Neurol. 80, 89–100 (2016).
- Cree BAC, Gourraud PA, Oksenberg JR, et al. Long-term evolution of multiple sclerosis disability in the treatment era. Ann. Neurol. 80, 499–510 (2016).
Differences in the findings from the studies
Disability outcomes and associated predictive factors in treated RRMS cohorts were assessed using 10 years of ‘real-world’ follow-up data.
Both studies report notably more benign disability outcomes compared with natural history studies
The sample sizes, methodology, analyses and outcome measures used are somewhat different between the two studies (Table 1) [1,2], making direct comparisons difficult, but both studies report notably more benign disability outcomes compared with natural history studies (Table 2). Conclusions from the papers diverge, however, in the factors associated with worsening disability over the 10-year follow-up period.
Whilst Jokubaitis et al. reported a number of factors, most prominently annualised relapse rate (ARR), associated with disability worsening at 10 years in their international multicentre cohort, Cree et al. found neither early clinical nor radiological factors correlated with long-term disability in their single-centre cohort. Notably, Jokubaitis et al. included events occurring over the entire 10-year period, while Cree et al. focused on predictors within the early years of disease (first 2 years). From a methodological standpoint, the study by Cree et al. included a prospective contemporary data collection with an a priori hypothesis as part of an ongoing study. In contrast, Jokubaitis et al. (whilst using prospectively collected data) undertook a retrospective, post hoc subgroup analysis of the MSBase dataset, with an awareness of the limitations of such an approach.
Conclusions about the impact of contemporary treatments on long-term disability cannot be derived from these studies
The studies incorporated cohorts diagnosed and commenced on treatment over a decade apart, with the MSBase cohort  (n=2466) including patients diagnosed in the late 1980s/early 1990s and the Californian group recruited between 2004 and 2005 (data available for 471 patients followed for 10 years). Neither study evaluated the oral DMTs (teriflunomide, dimethyl fumarate [DMF] or fingolimod) or an early induction approach with alemtuzumab, for example. The therapeutic landscape has changed since these cohorts were initiated onto treatment and, hence, conclusions about the impact of contemporary treatments on long-term disability cannot be derived from these studies. The lack of correlation between early relapses and MRI activity with either long-term disability or detailed multi-modal functional outcomes (timed 25-foot walk [T25FW], 9-hole peg test [9HPT], paced auditory serial addition test [PASAT]) from Cree et al. does, however, challenge the utility of treating to such targets in the early stages of disease. Further work is required to confirm these findings, given the relatively small sample size and the conflicting findings reported previously.
Table 1. Design of the studies and comparison with natural history cohorts
Table 2. Key results and comparison with natural history cohorts
a58.4% discontinued treatment after a median of 4.3 years (proportion spent on IFN/GA over 10 years=0.79 [0.01 for oral Rx/0.03 for IV Rx/high efficacy]).
bMedian 2.8 years from diagnosis to first DMT.
cData for total population, including the progressive MS population.
9HPT, 9-hole peg test; CIS, clinically isolated MS; DMT, disease-modifying therapy; EDSS, Expanded Disability Status Scale; GA, glatiramer acetate; IFN, interferon; IQR, interquartile range; IV, intravenous; NEDA, no evidence of disease activity; RRMS, relapsing–remitting MS; PASAT 3, 3 seconds version of the paced auditory serial addition test; PPMS, primary progressive MS; SD, standard deviation; SPMS, secondary progressive MS; T25FW, timed 25-foot walk.
Summary of studies and comparison with natural history cohorts
Both studies highlight the disease-modifying capabilities of immunomodulatory treatments in RRMS
Overall, whilst the value of early disease characteristics and the relationship between ‘inflammatory’ disease activity and long-term outcomes in MS is challenged, both studies highlight the disease-modifying capabilities of immunomodulatory treatments in RRMS when outcomes are compared with the pre-treatment era. The most obvious confounding factor in this conclusion, however, is the incorporation of milder disease phenotypes with each iteration of diagnostic guidelines over time. Indeed, it is clear that more recent diagnostic guidelines have widened diagnosis to include patients with less clinical disease activity in comparison to the natural history cohorts studied in the 1970s and 1980s, or the cohorts from early DMT trials. Unfortunately the lack of detailed published baseline cohort characteristics from pivotal natural history studies limits comparisons with more contemporary studies such as these.
Detailed commentary on Jokubaitis et al.
Jokubaitis et al. utilised the MSBase dataset to compare EDSS at baseline and at 10 years in patients initiated onto either interferon (IFN) or glatiramer acetate (GA). Imaging data were not included. Overall, 2466 patients were assessed with 93.8% having an RRMS phenotype at baseline (Table 1). The excluded patients were, however, largely comparable to the included cohort, except for a lower inclusion of clinically isolated MS (CIS) (6.2% vs 13.2%), higher proportion of RRMS (93.8% vs 86.8%) and that 1.5% (n=39) of the included cohort died during follow-up in comparison to 0.5% of those excluded. Overall, this would suggest that the included cohort had a bias towards more active disease. It is notable that the authors chose to include only patients initiated onto first-line therapies and not higher efficacy DMTs, but perhaps this reflects the limited choice available at the time.
A total of 1183 treatment switches for 932 (37.8%) patients occurred over the 10-year period but these were horizontal and intra-class for the majority. Overall, 19% of the 10-year period was spent untreated for the cohort.
Predictors of change in EDSS at 10 years
Higher cumulative exposure to a disease modifying therapy was associated with a lower EDSS score at 10 years
The predictors of median 10-year EDSS change are summarised in Table 3. ARR was the strongest independent clinical predictor of long-term outcome. Other predictors included older age at onset and longer disease duration at baseline. Higher cumulative exposure to a disease modifying therapy (IFN-β/GA) was associated with a lower EDSS score at 10 years (beta coefficient –0.86 [95% CI –1.13 to –0.58]), meaning a 0.86 reduction in EDSS for proportionally more time spent on treatment, which the authors translate into a prevention of increased EDSS of 1 point at 10 years for every 11.6 years of IFN-β/GA exposure. This is later described as ‘a strong protective effect of DMT against long-term disability’, a judgement that is debatable.
Table 3. Predictors of median 10-year EDSS change
The goal was to define clinical or demographic predictors associated with increased EDSS 10 years after baseline. The change in EDSS did not have a normal distribution, hence quantile median regression was used instead of the more familiar linear regression. This analysis is more robust against outliers within the response variables and provides the median change for each unit increase of the predictor variable. That is, the 10-year EDSS score is the variable of interest for all analyses and changes in this, associated with the predictor variable, are reported as the beta coefficient: this represents the number of EDSS points increased at 10 years for the given predictor variable. A beta coefficient of 0.9 translates to a 0.9 increase in EDSS at 10 years for each unit increase of the predictor variable, for example for ARR=1, the EDSS at 10 years increases by 0.9 for every increase in ARR by 1 (1.8 for ARR=2, etc.). If the beta coefficient is negative, this means a reduction in EDSS at 10 years.
The median EDSS point increase for the entire cohort over 10 years was 1 (IQR 0–2), but this varied depending on baseline EDSS (Table 4). Notably, 21% of patients with EDSS ≥6.0 at baseline improved, but only one patient reduced from EDSS ≥6.0 at baseline to <3 at 10 years.
Table 4. Summary of EDSS change after 10 years by baseline EDSS
Interestingly, completed pregnancy was associated with a 3.17 reduction in EDSS score at 10 years, in comparison to 0.71 for DMT treatment for the same proportion of time. Notably, time spent pregnant, rather than whether the pregnancy went to completion or not, was equally associated with reduced EDSS at 10 years (–0.32 EDSS points at 10 years for every 10% of pregnancy completed). It does not appear that these figures were corrected for age or EDSS, however. It seems likely that the younger age of those falling pregnant and their reduced disability in comparison to those not falling pregnant would contribute to the lower long-term disability, confounding this finding. The authors cite the fact that approximately half of those falling pregnant were on a DMT at the time as evidence of active disease.
Relapses on treatment were associated with a worse EDSS at 10 years in comparison to relapses off treatment
Relapses on treatment were associated with a worse EDSS at 10 years in comparison to relapses off treatment. An ARR of 1 on treatment was associated with a statistically significant 0.86 EDSS point increase at 10 years in comparison to a non-significant 0.05 increase for relapses off treatment. It is not stated why patients were off treatment but it may suggest they were non-responders or had developed secondary progressive MS (SPMS), in which relapses do not appear to influence longer-term disability, as shown previously by Tremlett et al.
The development of SPMS was not assessed in this study, but 17.8% of the entire cohort had reached EDSS ≥6 at 10 years. This is comparable to the 19% from the entre Ontario natural history cohort after a mean of 11.9 years, but is notably higher than the 10% in the group seen from onset in that study. This is the only metric which suggests lack of benefit in the use of DMTs, in comparison to an untreated cohort.
Detailed commentary on Cree et al.
Cree et al. report data collected prospectively from a single tertiary referral centre in California, USA. This is part of an ongoing study (EPIC) seeking to use ‘real-world’ data to phenotype and genotype a large cohort of patients with MS, to identify prognostic factors or possibly biomarkers.
Data collection started in 2004, with the 2001 McDonald criteria used for diagnosis. A total of 471 patients had 10-year data available for analysis. The cohort was predominantly RRMS (n=337) but patients with CIS (including imaging diagnostic of MS, n=70) were included to create a ‘relapsing’ cohort (RMS) which is the main focus here. Analysis excluding the patients with CIS did not change the findings significantly.
Median disease duration at baseline was 7 years (range 2–13.5) for the entire cohort. Median disease duration, age at onset and baseline EDSS are shown in Table 1.
Unlike the study by Jokubaitis et al., the use of DMTs during follow-up, switches and proportion of time spent on treatment is not reported
Only the use of treatment at baseline was included: 61% of the RMS cohort had first-line therapies, 1.2% were receiving “higher efficacy” treatments and 38% were untreated. Unlike the study by Jokubaitis et al., the use of DMTs during follow-up, switches and proportion of time spent on treatment is not reported. Escalation from no treatment to first-line treatment or first- to second-line treatment is commented upon and analysed as a potential predictor of disability, but detailed data on this are not provided. Treatment was started at a median of 2.8 years from diagnosis.
…a larger cohort of ‘partially’ treated patients than in a contemporary cohort where escalation is likely to be undertaken more readily
It is notable that natalizumab would have been available as a first- or second-line treatment for this cohort, yet only a small proportion were on (any) “high efficacy” treatment. The risks of escalation to this from a platform therapy are higher than the current escalation options including DMF and fingolimod and presumably John Cunningham’s Virus (JCV) status would not be known for most patients. For this reason, perhaps, clinicians and patients may have been more inclined to continue a first-line treatment even in the event of disease activity. This may result in a larger cohort of ‘partially’ treated patients than in a contemporary cohort where escalation is likely to be undertaken more readily.
As well as EDSS, other outcome measures included T25FW, 3 seconds version of the paced auditory serial addition test (PASAT 3), 9HPT and Multiple Sclerosis Spasticity Scale (MSSS). NEDA at Year 2 (defined as no relapse or MRI activity or clinically significant EDSS change) was also assessed for an association with disability outcomes. Only EDSS increases relative to baseline EDSS were considered clinically significant: ≥1.5 for baseline EDSS 0–1, ≥1 for baseline EDSS 1–5 and ≥0.5 if baseline EDSS was already greater than 5.0. Additionally, increases had to be sustained to be considered irreversible: for NEDA, the increase must have persisted at Year 2 and for SPMS the increase must have persisted at both Year 5 and the final assessment. This robust approach avoids any concern regarding transient (reversible) worsening, as outlined by Kalincik et al.
…all baseline EDSS scores were associated with a roughly equivalent risk of long-term disability
Almost half of the cohort experienced no clinically significant worsening of EDSS over the 10-year period. However, 59% of patients experienced significant disability despite the majority being on treatment and 82% of patients with RMS experienced some clinical and/or radiological worsening over the study. An analysis comparing treated and untreated patients is not included, however. Notably, baseline EDSS in the RMS cohort was not strongly associated with clinically significant disability worsening at 10 years, with all baseline EDSS scores associated with a roughly equivalent risk of long-term disability.
Predictors of change in EDSS at 10 years
A number of factors between Years 0 and 2 were assessed for their association with disability worsening at 10 years. A logistic regression model was used. Only 17.9% of patients achieved NEDA at 2 years but this was not a significant predictor of long-term disability. An increase in EDSS during the first 2 years was unexpectedly associated with a lower risk of future disability worsening. The authors cite prolonged recovery from relapses or response to treatment escalation may have contributed to this finding. The alternative possibility is that this is a true finding, dissociating early disease activity from long-term disability. Clinical deterioration in Years 0–2 was associated with a worse outcome for patients with RMS who subsequently developed SPMS.
…lack of association between MRI findings in the first 2 years and 10-year disability outcomes
A prominent conclusion of the study was the lack of association between MRI findings (T2/gadolinium positive (Gd+)-enhancing) in the first 2 years and 10-year disability outcomes. This was the case for those with or without clinical disease activity. Lower vitamin D levels were associated with more Gd+-enhancing lesions, but this was not associated with disability at 10 years. Treatment escalation, despite the higher likelihood of recent relapse and new MRI lesions in these patients, did not improve long-term disability outcomes – on average, disability outcomes were the same whether treatment was escalated or not. This may suggest that the escalation was effective given the lack of relative decline in this more active subgroup or that treatment escalation does not improve outcomes. There is no discussion as to the escalation criteria used or when this occurred. It would be expected that earlier escalation is associated with better long-term outcomes but this analysis is not included.
…no analysis correlating longitudinal MRI activity and disability outcomes in treated or untreated patients
The authors surmise that these data challenge the practice of routine surveillance MRI scans in the setting of otherwise quiescent MS. Certainly, their data suggest that MRI activity at Year 2 does not correlate with 10-year disability outcomes but data from MRI scans between Years 2 and 10 are not presented. Similarly, there is no analysis correlating longitudinal MRI activity and disability outcomes in treated or untreated patients, or whether MRI lesions were considered in treatment escalation decisions. Whilst it seems entirely fair to conclude that MRI data at Year 2 in this cohort did not predict 10-year disability, data are not presented upon which to determine the utility of longitudinal MRI in either treatment escalation decisions or disability outcomes.
The median time to SPMS from RRMS onset was estimated at 34–35 years
The median time to SPMS from RRMS onset was estimated at 34–35 years for this cohort. In natural history studies, 36–50% of patients with RRMS would be expected to transition to SPMS by 16.8 years but only 11.3% of this contemporary treated cohort did in the study by Cree et al. At 10 years from baseline, 10.1% of the RMS cohort had transitioned to SPMS; this compares to 37% of the Lorraine, France natural history cohort albeit follow-up was from onset (13 years). However, given the median 6 years of disease duration at baseline in the Cree RMS cohort, the transition is still notably low after an average of more than 15 years from onset. Overall, 4.7% of patients had reached EDSS ≥6 at 10 years, compared with 10% at 11.9 years in the group seen from onset in the Ontario natural history cohort.
This article is taken from the Hot Topics section of Neuro-Compass. Neuro-Compass is a free, expert-led practical medical education resource developed for all healthcare professionals caring for people with MS. Sign up to Neuro-Compass for free at www.neuro-compass.education to read more Hot Topic articles and for access to extensive clinically relevant education from experts in MS.
- Jokubaitis VG, Spelman T, Kalincik T, et al. Predictors of long-term disability accrual in relapse-onset multiple sclerosis. Ann. Neurol. 80, 89–100 (2016). Full article
- Cree BAC, Gourraud PA, Oksenberg JR, et al. Long-term evolution of multiple sclerosis disability in the treatment era. Ann. Neurol. 80, 499–510 (2016). Full article
- Bermel RA, You X, Foulds P, et al. Predictors of long-term outcome in multiple sclerosis patients treated with interferon beta. Ann. Neurol. 73:95–103 (213). PubMed abstract
- Weinshenker BG, Bass B, Rice GP, et al. The natural history of multiple sclerosis: a geographically based study. I. Clinical course and disability. Brain 112 (Pt 1), 133–146 (1989). PubMed abstract
- Debouverie M, Pittion-Vouyovitch S, Louis S, et al. Natural history of multiple sclerosis in a population-based cohort. Eur. J. Neurol. 15, 916–921 (2008). PubMed abstract
- Tremlett H, Zhao Y, Rieckmann P, et al. New perspectives in the natural history of multiple sclerosis. Neurology 74, 2004–2015 (2010). PubMed abstract
- McDonald WI, Compston A, Edan G, et al. Recommended diagnostic criteria for multiple sclerosis: guidelines from the International Panel on the diagnosis of multiple sclerosis. Ann. Neurol. 50, 121–7 (2001). PubMed abstract
- Poser CM, Paty DW, Scheinberg L, et al. New diagnostic criteria for multiple sclerosis: guidelines for research protocols. Ann. Neurol. 13, 227–231 (1983). PubMed abstract
- Debouverie M, Louis S, Pitton-Vouyovitch S, et al. Multiple sclerosis with a progressive course from onset in Lorraine, Eastern France. J. Neurol. 254, 1370–1375 (2007). PubMed abstract
- Kalincik T, Cutter G, Spelman T, et al. Defining reliable disability outcomes in multiple sclerosis. Brain 138, 3287–98 (2015). Full article