Objective This study aimed to define the minimal evidence of disease activity (MEDA) during treatment that can be tolerated without exposing patients with relapsing-remitting multiple sclerosis at risk of long-term disability.
Methods We retrospectively collected data of patients followed up to 10 years after starting interferon beta or glatiramer acetate. Survival analyses explored the association between the long-term risk of reaching an Expanded Disability Status Scale≥6.0 and early clinical and MRI activity assessed after the first and second year of treatment. Early disease activity was classified by the so-called ‘MAGNIMS score’ (low: no relapses and <3 new T2 lesions; medium: no relapses and ≥3 new T2 lesions or 1 relapse and 0–2 new T2 lesions; high: 1 relapse and ≥3 new T2 lesions or ≥2 relapses) and the absence or presence of contrast-enhancing lesions (CELs).
Results At follow-up, 148/1036 (14.3%) patients reached the outcome: 61/685 (8.9%) with low score (reference category), 57/241 (23.7%) with medium score (HR=1.94, p=0.002) and 30/110 (27.3%) with high score (HR=2.47, p<0.001) after the first year of treatment. In the low score subgroup, the risk was further reduced in the absence (49/607, 8.1%) than in the presence of CELs (12/78, 15.4%; HR=2.11, p=0.01). No evident disease activity and low score in the absence of CELs shared the same risk (p=0.54). Similar findings were obtained even after the second year of treatment.
Conclusions Early marginal MRI activity of one to two new T2 lesions, in the absence of both relapses and CELs, is associated with a minor risk of future disability, thus representing a simple and valuable definition for MEDA.
- multiple sclerosis
- treatment response
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
The rapidly expanding range of treatments available for relapsing-remitting multiple sclerosis (RRMS) has led to increasing interest in the issue of treatment sequencing and algorithm.1 In the vast majority of patients, the initial treatment choice relies on an escalation strategy, that is, starting with safer platform therapies and moving on to higher-efficacy treatments in case of persistent clinical and/or MRI activity.2 This approach requires careful evaluation of treatment response, an issue lacking of consensus on definition and monitoring actions.3
To date, there is no predictive and easily reproducible biomarkers beyond clinical assessment and conventional MRI measures to forecast treatment response at individual level.4 In a large multicentre study, the MAGNIMS (MRI in MS) network has classified the risk of short-term clinical worsening into three risk categories (low, medium and high) by combining the numbers of relapses and new T2 lesions after the first year of treatment.5 According to the so-called ‘MAGNIMS score’, patients at low risk (ie, no relapses and minimal focal MRI activity of one or two new T2 lesions) have a negligible probability of disability worsening in the following 2–3 years,5 whereas those at medium risk (ie, no relapses and ≥3 new T2 lesions or 1 relapse and 0–2 new T2 lesions) require an additional clinical and MRI follow-up (ie, ‘re-baseline’) even after the first year of treatment to overcome uncertainty about treatment response.6 The threshold of <3 new focal T2 lesions activity, in the absence of relapses,5 has been advocated as a more realistic goal to achieve in clinical practice than the ‘no evidence of disease activity’ (NEDA), which is based on the ‘zero tolerance concept’ entailing for the absence of relapses, confirmed disability worsening and MRI focal activity.7 However, even if reliable and easily accessible in clinical practice, the MAGNIMS score is not conclusive in terms of long-term prediction1 and does not encompass contrast-enhancing lesions (CELs), whose early occurrence while on treatment was strongly correlated with severe disability 15 years later.8
Here, we analysed data from a large cohort of patients followed up to 10 years after starting a platform treatment with interferon beta (IFNB) or glatiramer acetate (GA) to explore if a low MAGNIMS score can be actually considered the ‘minimal evidence of disease activity’ (MEDA) to tolerate without exposing patients to risk of future disability.3 To conform with our purpose, we aimed at exploring: (1) if the MAGNIMS score is valid and reliable on the long-term risk of disability; (2) if the acquisition of postcontrast MRI is actually dispensable or should be encompassed in the MAGNIMS score; (3) if the predictive value of the MAGNIMS score is consistent even after re-baseline, that is, assuming the possibility of residual disease activity due to the gap between the treatment start and its actual effect.3
Study design and participants
This was an independent, multicentre, retrospective cohort study based on data collected in the real-world setting. We analysed data of patients with RRMS9 regularly attending three tertiary MS Centres in Italy. All data were gathered after approval by institutional ethical committees and an informed consent was obtained from each participant. In no way this study did interfere in the care received by patients. Anonymised data presented in this article will be made available at the request of a qualified investigator; requests should be sent to Dr Prosperini (firstname.lastname@example.org).
We collected data of patients with RRMS according to the following criteria:
Subject received any formulation of IFNB or GA as first treatment.
An Expanded Disability Status Scale (EDSS)10 <4.0 at treatment start (henceforth defined as ‘baseline’).
Regular at least biannual clinical visits from treatment start, including EDSS scoring performed by certified neurologists (neurostatus.net).
Complete data on brain MRI scans at baseline (within 1 month before starting treatment) and after 1 and 2 years (±1 month) since treatment start, acquired with the same scanner in each patient according to local guidelines.11
No current participation in experimental trials.
At least 10 years of follow-up after treatment start.
Classification of early disease activity
Level of disease activity was defined on the basis of clinical and focal MRI activity assessed after the first year and the second year of treatment.
Clinical activity included relapses, defined as any new neurological symptom not associated with fever or infection lasting for at least 24 hours and accompanied by new neurological signs.9
Focal MRI activity included the presence of new T2 lesions (when compared with the previous scan) and CELs. New focal T2 lesions were detected on both fluid-attenuated inversion recovery and T2-weighted axial images of the brain when compared with the baseline scan; enlarging T2-hyperintense lesions were not counted due to the poor between-rater agreement for this metric under routine clinical setting.12 Presence of CELs was detected on T1-weighted spin echo axial images of the brain after gadolinium-diethylenetriamine penta-acetic acid administration sequences.11
On the basis of the aforementioned MAGNIMS paper,5 levels of disease activity were scored as follows: low, that is, no relapses and <3 new T2 lesions; medium, that is, no relapses and ≥3 new T2 lesions or 1 relapse and 0–2 new T2 lesions; high, that is, 1 relapse and ≥3 new T2 lesions or ≥2 relapses.
We set as main study outcome the time to reach the disability milestones of EDSS score≥6.0, corresponding to the ability to walk only with unilateral support and <100 m without resting, confirmed in at least two consecutive visits and sustained (stable or higher) over the entire follow-up. We adopted such combined outcome instead of the classical 0.5-point or 1.0-point EDSS worsening13 to set a robust outcome based on a clinically significant milestone for patients with MS.
Continuous data were presented as mean±SD or median (interval); categorical data were presented as count (proportion).
Baseline variables included sex (female or male), age (years), time since first symptom (years), EDSS score, number of relapses in the previous year, number of CELs on brain MRI scan, type of treatment started (subcutaneous IFNB-1b 250 mcg every other day (eod); intramuscular IFNB-1a 30 mcg once weekly (ow); subcutaneous IFNB-1a 22 or 44 mcg three times per week (tpw); subcutaneous GA 20 mg once daily (od)) and calendar year at treatment start.
Disease activity after the first and second year of treatment was modelled as multilevel variable according to the MAGNIMS score as afore defined (low, medium or high). Absence or presence of CELs after the first and second year of treatment was also entered in models as nested term. We ran two hierarchical Cox proportional hazards regression models to ascertain the effects of the MAGNIMS score (step 1) and the presence of CELs (step 2) after the first year (model A) and the second year (model B) of treatment year on the risk of reaching EDSS≥6.0.
All models were stratified by site and adjusted by calendar year at treatment start and baseline variables (covariates of no interest). Exposure to highly intensive treatments was also inserted in the models as time-varying covariate, under the assumption that escalation to intravenous immunosuppressants (mitoxantrone, cyclophosphamide) and monoclonal antibodies (natalizumab, alemtuzumab, ocrelizumab, rituximab) may have affected the patients’ disability status during the follow-up.14–16 To assess robustness of the results, a post-estimation sensitivity analysis was done by re-running all models after removing patients who were escalated to highly intensive treatments.
Adjusted HR and their corresponding 95% CIs were provided for significant predictors; p values<0.05 in either direction were considered significant.
We examined records of 1384 patients starting GA or IFNB who met the inclusion criteria. Of them, 348 were excluded mainly due to incomplete MRI data at year 1 or year 2. Therefore, we analysed 1036 patients (722 female, 314 male) who started subcutaneous GA 20 mg od (n=132), intramuscular IFNB-1a 30 mcg ow (n=296), subcutaneous IFNB-1b 250 mcg eod (n=184), subcutaneous IFNB-1a 22 or 44 mcg tpw (n=424) from 1996 to 2009. Their characteristics at baseline are shown in table 1.
There were no significant differences in baseline variables (sex, age, time since first symptom, EDSS score, number of relapses in the previous year and number of CELs) between the 1036 patients who were analysed and the 348 who were not, except for the calendar year at treatment start (patients who started treatment in early years were more likely to miss MRI scan; p<0.05).
Early disease activity
Disease activity after the first year of treatment was scored as low in 685/1036 (66.1%) patients, medium in 241/1036 (23.3%) patients and high in 110/1036 (10.6%) patients. One or more CELs were detected in 244/1036 (23.6%) patients; notably, 20/1036 (1.9%) patients presented 1 (n=14) or 2 (n=6) CELs in the absence of focal new T2 lesions, suggesting a re-enhancement of pre-existing lesions.
Disease activity after the second year of treatment was scored as low in 788/1036 (76.1%) patients, medium in 178/1036 (17.2%) patients and high in 70/1036 (6.7%) patients. One or more CELs were detected in 178/1036 (17.2%) patients; notably, 8/1036 (0.8%) patients presented 1 (n=7) or 2 (n=1) CELs in the absence of focal new T2 lesions, suggesting a re-enhancement of pre-existing lesions. Description of disease activity after the first and second year of treatment is shown in table 2.
After a median time of 4.5 (interval 1 to 9) years, 312/1036 (30.1%) patients were escalated to highly effective treatments, namely, natalizumab (n=238), mitoxantrone (n=47), alemtuzumab (n=15) and cyclophosphamide (n=12). They were exposed to highly effective treatments for a median time of 5 (interval 1 to 9) years. The remaining 724/1036 (69.9%) patients had been treated with the same self-injectable treatment (n=417), or had been switched from any IFNB formulation to GA or vice versa (n=182), or had been moved to oral treatments, such as dimethyl fumarate (n=48), fingolimod (n=45), teriflunomide (n=22) and azathioprine (n=10).
A total of 148/1 036 ( 1 4.3 %) patients reached an EDSS score ≥ 6.0 after a median time of 7.5 (interval 2.0 to 9.5) years from treatment start. In patients who did not reach the disability milestone, the median EDSS score at last visit was 2.0 (interval 0 to 5.5).
Long-term disability by MAGNIMS score after the first year of treatment
After the first year of treatment, 61/685 (8.9%) patients with low score (reference category), 57/241 (23.7%) patients with medium score (HR=1.94, p=0.002) and 30/110 (27.3%) patients with high score (HR=2.47, p<0.001) reached an EDSS≥6.0 over the follow-up. In the subgroup of patients with a low score (n=685), an EDSS≥6.0 was reached by 49/607 (8.1%) patients without CELs and 12/78 (15.4%) patients in whom ≥1 CELs was detected after the first year of treatment (HR=2.11, p=0.01). In the remaining two subgroups, that is, medium and high scores, looking at CELs did not contribute to fit better the model A (table 3). Excluding patients who were escalated to highly effective treatments before reaching the outcome (n=286) provided consistent results.
Survival curves showing time to EDSS≥6.0 according to level of disease activity after the first year of treatment are shown in figure 1A.
Long-term disability by MAGNIMS score after the second year of treatment
Re-baseline analyses after the second year of treatment confirmed previous findings. Data on 24 (2.3%) patients who were escalated to highly intensive treatments before completing 2 years of platform therapy were removed from the next analysis, including two patients who reached the outcome at year 2. An EDSS≥6.0 was reached by 70/770 (9.1%) of patients with low score (reference category), 50/175 (28.6%) patients with medium score (HR=2.98, p<0.001) and 26/67 (38.8%) patients with high score (HR=5.12, p<0.001) after the second year of treatment. In the subgroup of patients with low score (n=770), an EDSS≥6.0 was reached by 59/711 (8.3%) patients without CELs and 11/59 (18.6%) patients in whom ≥1 CELs were detected after the second year of treatment (HR=2.29, p=0.03). In the remaining two subgroups, that is, medium and high scores, looking at CELs did not contribute to fit better the model B (table 4). Excluding patients who were escalated to highly effective treatments before reaching the outcome (n=286) provided consistent results.
Survival curves showing time to EDSS≥6.0 according to level of disease activity after the second year of treatment are shown in figure 1B.
Defining ‘minimal evidence of disease activity’ (MEDA)
On the basis of the afore reported Cox models, MEDA corresponds to having no relapse and ≤2 new focal T2 lesions in the absence of CELs. To further confirm that the proposed MEDA definition actually identifies patients at very low risk of disability accumulation, we ran an additional analysis comparing the NEDA and MEDA status on the risk of reaching the disability milestone. We found no statistically significant difference between NEDA and MEDA: after the first year of treatment, an EDSS≥6.0 was reached by 41/514 (8.0%) of patients with NEDA and 8/93 (8.6%) patients with MEDA (HR=1.29, 95% CIs 0.57 to 2.96, p=0.54); after the second year of treatment, an EDSS≥6.0 was reached by 50/616 (8.1%) of patients with NEDA and 9/95 (9.5%) patients with MEDA (HR=1.20, 95% CIs 0.54 to 2.37, p=0.61).
Figure 2 shows boxplots representing the EDSS score at 10-year follow-up by level of disease activity after the first and second year of treatment. Lastly, table 5 shows the diagnostic accuracy of MEDA definition in identifying patients who reached (or not) an EDSS≥6.0 at 10-year follow-up.
In this multicentre study, we performed a retrospective analysis in the attempt to provide a MEDA definition that holds a favourable prognostic value in the long-term (10 years) risk of reaching the disability milestone of EDSS≥6.0 (main outcome). Our analysis was based on a definition of level of disease activity which was borrowed from a seminal paper of the MAGNIMS network, showing that the one or two new T2 lesions without clinical relapses after the first year of treatment produces negligible short-term (3 years) risk of disability worsening.5 We also repeated the analysis after re-baseline (ie, after the second year of treatment), assuming that residual disease activity in the initial weeks or months of treatment (when the therapeutic effect is not completely established yet) could have influenced the risk estimation in the first year of treatment.3
Our findings clearly indicate that the MEDA definition fits with the low MAGNIMS score, but only on condition of absence of CELs. In other words, while we confirmed the robustness of the MAGNIMS score in predicting future disability even in the long-term period, we also found that, among patients in the low score subgroup, the presence of CELs implied a twofold increased risk of reaching the disability milestone. This is not surprising, given that the presence of CELs early on IFNB therapy was strongly related to severe disability 15 years later.8 The major advantages of looking at CELs than new T2 lesions are based on the greater reliability, less influence by technical issues and by residual disease activity (ie, the gap between the treatment start and its actual effect).3 12 17 Again, new isolated focal lesion in confluent areas of white matter abnormality could be missed using only serial non-contrast MRI images.18 Moreover, CELs are typically short-lived,3 and their size and duration can be reduced by treatment, especially IFNB19 20; consequently, their detection under treatment would be closely related to non-response to treatment. However, the pros of counting CELs to predict treatment response should be weighed against the cons of brain deposition of gadolinium-based contrast agents,21 the reason why researchers have proposed alternative solutions such as novel macrocyclic agents,22 surrogate for gadolinium-enhancement,23 or even administration of gadolinium-based contrast agents only after screening precontrast sequence.24
Our definition of MEDA differs from that provided by Rio et al 25 who argued that a minimal clinical activity (ie, a relapse without an impact on disability) can be tolerated in the first 2 years of treatment; however, their definition has low specificity (36%) and overall accuracy (44%) in predicting severe long-term disability.25 Our definition has greater overall accuracy (63% to 73%) and an impressive negative predictive value (∼92%), with well-balanced values for sensitivity (60% to 67%) and specificity (63% to 73%).
In our study, patients experiencing one or more relapses in the first and second year of treatment had a greater risk of future disability, as also found in the seminal MAGNIMS study.5 This is quite expected because the occurrence of relapses does not represent only the manifestation of new lesions in clinically eloquent site26 but also the failure of mechanisms of brain plasticity and tissue repair preventing long-lasting disability outcome.27 28 Notably, a recently published long-term post hoc analysis of the trial on subcutaneous IFNB-1b suggested that a clinical definition of NEDA (ie, no relapses and no disability worsening), even without encompassing MRI activity, is sufficient to predict long-term disability outcomes; however, gadolinium-based contrast agents were not administered as part of that study.29
Here, we defined the MEDA as a composite outcome measure that is easily accessible in clinical practice setting and has good predictive value in discriminating patients with long-term favourable outcome. This comes from the need to provide a more realistic goal than the NEDA to achieve in clinical practice. Notably, the predictive value of our MEDA definition did not differ when applied after both the first and the second year of treatment. This latter finding should be considered as merely confirmative of the robustness of the proposed MEDA definition; therefore, we recommend that the assessment of treatment response should be done soon after the first year of treatment.
Real-world experience showed NEDA as an extremely stringent treatment goal to achieve and maintain over time,30–34 and growing data suggest that a minimal amount of clinical or MRI activity is not necessarily associated with poor long-term outcomes.25 29 On the other hand, combined scores based only on clinical and conventional MRI activity are not exempt from ambiguity and criticism because they are mainly driven by focal inflammatory activity rather than neurodegenerative processes,3 thus failing to capture all aspects of disease activity and progression and not predicting necessarily long-term stability, especially in case of silent progression.34 35 In this regard, it is remarkable that in our study approximately 8%–9% of patients with either NEDA or MEDA after the first and second year of treatment reached an EDSS≥6.0.
The main limitations of our study are intrinsic in its retrospective and observational design:
Our findings were limited to patients treated with GA and IFNB to obtain 10-year data, since oral drugs became available only in the past few years.3
As per eligibility criteria, we analysed only patients with complete 10-year follow-up, thus not entirely reflecting the real-world population attending our MS centres and potentially representing a selection bias towards the inclusion of more compliant patients.36
Even if our analysis was stratified by site, we cannot exclude a certain degree of between-centre variability, especially because brain MRI scans were acquired at different sites without any central readout of MRI data.11
We did not include data on lesion topography (ie, infratentorial and/or spinal cord lesions) which could be more related to disability accumulation over time than the simple lesion count.37 38
However, we are confident that our study has also some strengths:
We set a ‘hard’ outcome, that is, the disability milestone of EDSS≥6.0, that is not only clinically relevant for neurologists, patients and healthcare providers, but also indicates exhaustion of compensatory mechanisms due to structural damage.39
Findings are based on a relatively large sample size.
Our analyses encompassed treatment strategy and were corrected by calendar year to account for the growing availability of new treatments over time.16
Taken together, our findings suggest that, in the first few years of treatment, we can tolerate a minimal MRI activity of one or two focal new T2 lesions without exposing patients at risk of future disability. Presence of CELs should instead represent a red flag for future disability, even in case of marginal new T2 accrual and in the absence of relapses. However, when interpreting the present findings, we should bear in mind the low positive predictive value (23% to 29%) of our classification. In this regard, there are at least three possible explanations: (1) the low proportion of patients reaching such hard outcome could have influenced the estimation; (2) the proportion of misclassified patients will diminish with longer follow-up; (3) escalation to highly effective treatments could have probably modified the disease course.
Further efforts are now warranted to replicate these findings in other independent cohorts and to verify that they can be translated even in patients treated with oral treatments.
Contributors Conception and design of the study, drafting a significant portion of the manuscript/figures: LP, CG. Acquisition and analysis of data, revision of manuscript content: CM, SH, CC, LDG, NDR, S Rasia and S Ruggieri. Supervision and drafting the final version of the manuscript: SG, CT, CP.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests LP: consulting fees from Biogen, Novartis and Roche; speaker honoraria from Biogen, Genzyme, Merck Serono, Novartis and Teva; travel grants from Biogen, Genzyme, Novartis and Teva; research grants from the Italian MS Society (Associazione Italiana Sclerosi Multipla) and Genzyme. CM: no disclosures. SH: travel funding and/or speaker honoraria from Biogen, Roche, Genzyme, Novartis and CSL Behring. CC: fees as invited speaker and travel grants for attending meeting from Serono, Biogen, Teva and Novartis. LDG: travel grants from Biogen, Novartis and Teva. NDR: speaker honoraria from Biogen Idec, Genzyme, Novartis, Sanofi-Aventis; funding for participation in advisory board to Novartis and Genzyme-Sanofi and for travel to scientific meetings from Biogen Idec, Teva, Sanofi-Genzyme, Roche, Almirall and Novartis. SG: fees as invited speaker or travel expenses for attending meeting from Biogen, Merck-Serono, Teva, Almirall, Sanofi-Aventis, Novartis and Genzyme. S Rasia: fees as invited speaker or travel expenses for attending meeting from Biogen, Merck-Serono, Teva, Sanofi, Novartis and Genzyme. S Ruggieri: speaking honoraria from Merck Serono and Teva. CT: honoraria for speaking and travel grants from Biogen, Sanofi-Aventis, Merck Serono, Bayer-Schering, Teva, Genzyme, Almirall and Novartis. CP: scientific advisory boards for Actelion, Biogen, Genzyme, Hoffmann-La Roche, Merck-Serono, Novartis, Sanofi and Teva; consulting and/or speaking fees, research support and travel grants from Allergan, Almirall, Biogen, Genzyme, Hoffmann-La Roche, Merck-Serono, Novartis, Sanofi and Teva. CG: fees as invited speaker or travel expenses for attending meeting from Biogen, Merck-Serono, Teva, Sanofi, Novartis and Genzyme.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement Data are available on reasonable request.