Objectives The accuracy of ‘no evidence of disease activity’ (NEDA) in predicting long-term clinical outcome in patients with relapsing remitting multiple sclerosis (RRMS) is unproven, and there is growing evidence that NEDA does not rule out disease worsening. We used diffusion tensor imaging (DTI) to investigate whether ongoing brain microstructural injury occurs in patients with RRMS meeting NEDA criteria.
Methods We performed a retrospective study to identify patients with RRMS visiting our centre over a 3-month period who had undergone prior longitudinal DTI evaluation at our facility spanning ≥2 years. Patients meeting NEDA criteria throughout the evaluation period were included in the NEDA group, and those not meeting NEDA criteria were included in an ‘evidence of disease activity’ (EDA) group. Fractional anisotropy (FA) and mean diffusivity (MD) maps were created, and annual rates of change were calculated.
Results We enrolled 85 patients, 39 meeting NEDA criteria. Both NEDA and EDA groups showed longitudinal DTI worsening. Yearly FA decrease was lower in the NEDA group (0.5%, p<0.0001) than in the EDA group (1.2%, p=0.003), while yearly MD increase was similar in both groups (0.8% for NEDA and EDA, both p<0.01). There was no statistical difference in deterioration within and outside of T2 lesions. DTI parameters correlated with disability scores and fatigue complaints.
Conclusions White matter microstructural deterioration occurs in patients with RRMS over short-term follow-up in patients with NEDA, providing further evidence of the limitations of conventional measures and arguing for DTI in monitoring of the disease process.
Statistics from Altmetric.com
Improved therapies for relapsing-remitting multiple sclerosis (RRMS) have led to the concept of ‘no evidence of disease activity’ (NEDA), a measure of treatment success defined by the absence of clinical worsening (relapses or Expanded Disability Status Scale (EDSS) increase) and radiological activity (new/enlarging T2 lesions or gadolinium-enhancing lesions).1
While NEDA is difficult to maintain over long-term follow-up,2 3 it is considered a goal of treatment in clinical practice. Despite new drug availability, patients maintaining NEDA on a stable disease modifying therapy (DMT) regimen would not routinely be considered for treatment escalation. However, patients meeting NEDA criteria can exhibit worsening cognition and walking speed,4 demonstrating that NEDA does not preclude ongoing subtle changes not measured by the standard NEDA definition. Furthermore, routine MRI measures are limited in monitoring the disease. When brain atrophy rate was incorporated as an additional criterion for NEDA (NEDA-4), a substantial proportion of patients meeting standard NEDA criteria nevertheless exhibited abnormal brain atrophy rates,5 demonstrating the need for novel tools to detect subtle worsening during clinical monitoring.
Diffusion tensor imaging (DTI) is an advanced MRI technique that estimates white matter (WM) integrity by measuring mean diffusivity (MD), reflecting overall diffusion and fractional anisotropy (FA), reflecting directionality of diffusion.6 Demyelination/axonal degeneration results in decreased FA and increased MD. DTI abnormalities within multiple sclerosis (MS) lesions correlate with tissue damage.7 Additionally, abnormalities occur even in WM that appears normal on conventional imaging, and these abnormalities correlate only modestly with T2 lesion burden.7 DTI abnormalities are detected very early in MS8 and correlate highly with disability.9–11 Furthermore, DTI is superior to volumetric analysis in differentiating between patients with MS and healthy controls.12 For the above reasons, DTI may enhance detection of subtle disease pathology. While DTI changes occur longitudinally in the general MS population,13 it is unknown whether WM deterioration also develops in patients maintaining NEDA. We used DTI to determine whether microstructural brain WM injury occurs in patients with RRMS meeting NEDA criteria. We hypothesised that patients maintaining NEDA would exhibit deterioration on DTI, although at a lower rate than those who do not meet NEDA criteria.
We performed a retrospective study utilising chart review to identify patients with RRMS who visited the Corinne Goldsmith Dickinson Center for Multiple Sclerosis between 1 November 2015 and 31 January 2016. Patients were included in the study if they had undergone longitudinal brain MRI scans containing diffusion imaging performed at least 2 years apart at Mount Sinai Hospital (New York, USA) and if they had corresponding electronic medical record clinical documentation encompassing that time period. A total of 1292 patients were screened, of which 777 had a confirmed diagnosis of RRMS based on revised McDonald criteria.14 While DTI is routinely performed at our centre, 676 patients with RRMS were excluded due to insufficient MRI/clinical data (mostly related to MRIs done at other institutions or lack of sufficient longitudinal data), and 16 were excluded due to confounding neurological diagnoses. The remaining 85 patients were enrolled in the study. Patients were then separated into two groups, a NEDA group and an ‘evidence of disease activity’ (EDA) group, based on clinical and radiographic evaluation (figure 1). Patients were included in the NEDA group if baseline and follow-up MRIs bookended a period of NEDA; otherwise, they were included in the EDA group. Median follow-up was 2.7 years (range 2–5) in the NEDA group and 2.9 years (range 2–4.6) in the EDA group.
Standard protocol approvals
The Mount Sinai Hospital Institutional Review Board approved the study protocol.
Demographic and clinical information was obtained through chart review and included baseline age, sex, disease onset, time of last disease activity prior to baseline MRI, medication usage, timed-25-foot-walk (T25FW) and physical examination. Patients at our centre are routinely asked if they have experienced recent changes in fatigue. Charts were analysed for mention of subjective fatigue worsening, and complaints were recorded as binary values (0 for no worsening, 1 for worsening during study duration). A single rater (AH) abstracted an EDSS score from chart documentation at each clinic visit spanning the study duration.
MRI evaluation and analysis
Subjects underwent brain MRIs (GE Medical Systems, GE Electric, Milwaukee, Wisconsin, USA) using one of the two protocols (Protocol A at 1.5T and Protocol B at 3T), with each patient exclusively evaluated using the same protocol at baseline and follow-up. ReadyBrain software was used for standardised orientation of images.15 Imaging included two dimensional (2D) T2, fluid attenuation inversion recovery (FLAIR) and T1 sequences. For DTI, a single-shot diffusion-weighted spin-echo echo-planar-imaging sequence was obtained (Protocol A: 7800/100 repetition time/echo time, 220×220 mm field of view, 256×256 mm matrix, 5 mm slice thickness, 0 mm gap; Protocol B: 4000/86 repetition time/echo time, 240×240 mm field of view, 256×256 mm matrix, 5 mm slice thickness, 0 mm gap). Following acquisition without diffusion sensitisation (b=0 s/mm2), images were acquired with diffusion gradients applied (b=1000 s/mm2 in Protocol A and b=1200 s/mm2 in Protocol B) in six directions.
Diffusion images were processed using the Functional MRI of the Brain Software Library16 17 (FSL) brain extraction and eddy current corrector tools. FA and MD maps were calculated by fitting the diffusion tensor at each voxel using DTIFit. Baseline and follow-up images were transformed to standard space using tract-based spatial statistics18 (TBSS) and a mean FA image was created to derive a WM skeleton representing the centres of all tracts common to the population. The skeleton was subsequently thresholded at 0.35 in order to reduce the effect of cross-subject variability on estimation of DTI parameters, and each subject’s aligned FA and MD images were projected onto the skeleton. Individual mean values were derived. A standardised brain structural atlas (JHU ICBM labels 1 mm, http://neurovault.org/images/1401) was overlaid to obtain mean FA/MD values over the skeleton within selected anatomical structures (corpus callosum (CC), cingulum, superior corona radiata (SCR), internal capsules (IC), external capsules (EC), brainstem corticospinal tracts (BCS) and middle cerebellar peduncles (MCP)). T2 lesions were segmented using a semi-automated tool (Jim V.7.0, Xinapse Systems, England), and lesion maps19 were overlaid with the WM skeleton to determine mean lesional and non-lesional FA/MD.
Determination of NEDA status
NEDA status was defined by the absence of clinical relapses, the absence of EDSS worsening and the absence of radiographic activity (new/enlarged T2 lesions or gadolinium-enhancing lesions). A stringent approach was taken in determination of NEDA in order to ensure exclusion of any worsening on the above metrics, as described below.
Clinical relapses were defined as new neurological abnormality present for ≥24 hours, occurring in the absence of fever/infection. Radiographic confirmation was not required for a relapse and symptoms considered as ‘possible relapses’ were included, even in the absence of radiographic change.
EDSS progression was defined as an increase of ≥1 point sustained for at least 12 weeks, a more stringent definition than has been used in prior studies.4 5 In the majority of cases, clinical documentation at 12 weeks after an EDSS change was present; in the absence of follow-up clinical documentation at that time, sustained progression was assumed.
Screening for radiographic activity was performed using official MRI reports. Although not all patients underwent routine spinal cord imaging (23/85 patients had cervical spine imaging spanning the full study duration, 11/39 NEDA and 12/45 EDA), worsening on such scans was nevertheless counted as disease activity. However, isolated cord radiographic activity (ie, in the absence of relapse or EDSS change) did not occur in our population and therefore did not alter the proportion of patients maintaining NEDA and EDA. After screening, direct comparison of baseline and follow-up MRIs was performed by an MS trained neurologist (AH) for confirmation, and, when unclear, a consensus was reached with an independent neurologist (RB).
Additionally, in order to be included in the NEDA group, patients were required to be clinically stable for ≥3 months prior to baseline scan, and baseline scans could not demonstrate enhancing lesions.
Statistical analysis was performed with SPSS V23.0. Significance was set at a p value of 0.05.
A paired sample t-test was used to compare baseline and follow-up FA/MD values. An independent sample t-test was used to compare FA/MD change between NEDA/EDA groups. Distribution of DMT use in NEDA/EDA groups was compared using Χ2, while multilinear regression was used to assess the effect of DMT use on FA/MD change. To assess the effect of MRI protocol, a binary logistic regression was performed, with MRI protocol as the dependent variable and FA/MD parameters as predictors. Relationships between baseline MRI metrics and clinical variables were evaluated with a partial correlation, taking into account the effect of age, sex and MRI protocol. The relationship between FA/MD parameter change and continuous clinical/demographic variables were determined using a hierarchical multiple linear regression analysis. Models including binary (ie, fatigue worsening) and categorical (ie, NEDA/EDA status) variables were estimated via logistic regression. For all models, the first block included age, sex and MRI protocol as covariates, while the second block included MRI metrics entered in stepwise order, with variable entry at p<0.05 and variable removal at p>0.10.
Thirty-eight patients (45%) had radiographic activity, 26 patients (31%) had clinical relapses and 10 patients (12%) exhibited disability worsening. Thirty-nine patients (46%) met criteria for NEDA group inclusion, while the other 46 were included in the EDA group. Baseline characteristics are highlighted in table 1. Notably, disease duration and baseline EDSS, T2 lesion volume (T2LV) and T25FW did not statistically differ between the groups, suggesting no significant difference in baseline disease severity.
Forty patients (47%) were on a stable DMT regimen throughout the study, 21/39 patients (54%) in the NEDA group and 19/46 (41%) in the EDA group. As expected, of patients on a stable regimen, DMT distribution differed between the two groups (NEDA: 11 interferon beta-1a, four glatiramer acetate, two natalizumab, two dimethyl fumarate and two no therapy; EDA: four interferon beta-1a, 14 glatiramer acetate and one no therapy; p=0.02).
Longitudinal change in DTI parameters over WM skeleton
Annual rates of change in FA and MD were calculated for each group (figure 2). In the NEDA group, there was a statistically significant decrease in FA (0.5%±1.3 %/year, p<0.0001) and increase in MD (0.8%±1.9 %/year, p<0.01), consistent with deterioration in tract integrity even in the absence of overt disease activity. The EDA group exhibited a faster rate of FA decline (1.2%±1.2 %/year, p<0.01), suggesting faster tract deterioration in the active disease group. The rate of MD increase did not differ between the two groups.
Consistent with the above, a regression model that included yearly FA and MD per cent change significantly predicted NEDA/EDA status, after controlling for age, sex and MRI protocol (p=0.02), with yearly FA per cent change being an independent predictor of clinical status (Beta −0.410, p=0.03).
Thirty-three per cent of patients meeting NEDA criteria (13/39) and 54% of patients meeting EDA criteria (25/46) exhibited ≥1.0% yearly declines in FA. More than 1.0% yearly increase in MD were exhibited by 31% of patients meeting NEDA criteria (12/39) and 33% of patients meeting EDA criteria (15/46).
Lesional and non-lesional change in DTI parameters
Since DTI abnormalities were previously demonstrated outside T2 lesions within the normal appearing WM7 (NAWM), we hypothesised that some of the tract deterioration found in patients meeting NEDA criteria took place outside of discrete lesions. We overlaid lesion maps with the WM skeleton (online Supplementary eFigure 1) and determined annual change within and outside the lesions in the NEDA group. We included only patients maintaining NEDA whose T2/FLAIR sequences were performed using the same technique at baseline and follow-up (n=23). We found that FA decreased and MD increased both within and outside lesions and that the rate of change was not statistically different in the two locations (table 2). Hence, longitudinal tract deterioration can be detected in NAWM.
Supplementary file 1
Change in DTI parameters within specific anatomical structures
We aimed to determine whether tract deterioration within the NEDA group is a global process or whether specific anatomical structures are preferentially affected. For this, image masks of selected WM structures (CC, cingulum, SCR, IC, EC, BCS and MCP) (online Supplementary eFigure 2) from an anatomical atlas were overlaid with the WM skeleton. We found that, within the NEDA group, FA decreased and MD increased longitudinally within the CC, cingulum, SCR and IC (table 3). However, FA and MD did not change significantly in the EC, BCS and MCP, suggesting relative preservation of these structures in patients with RRMS meeting NEDA criteria.
Association between DTI parameters and clinical metrics
To determine clinical relevance of DTI parameters, we performed exploratory analysis to determine relationships with demographic and clinical variables.
Within the total study population, after controlling for age, sex and MRI protocol, baseline EDSS was negatively correlated with baseline FA along the CC (r=−0.252, p=0.02), EC (r=−0.219, p=0.05), SCR (r=−0.283, p=0.01) and MCP (r=−0.238, p=0.03) and positively correlated with baseline MD over the IC (r=0.283, p=0.01), EC (r=0.255, p=0.02) and SCR (r=0.225, p=0.04). We did not detect any association between FA/MD change and EDSS or T25FW.
Interestingly, a clinical measure that correlated with worsening DTI parameters over time was fatigue. After controlling for age, sex and MRI protocol, subjective complaints of worsening fatigue were predicted by decreasing FA over the BCS (p=0.01) and SCR (p=0.04) and increasing MD over the SCR (p=0.04) and IC (p=0.01). After controlling for NEDA status, relationships remained between worsening fatigue and change in BCS FA (p=0.008), SCR FA (p=0.03) and IC MD (p=0.01). This demonstrates that, even in the NEDA population, complaint of worsening fatigue, inherently not leading to an EDSS change, was associated with worsening DTI metrics.
While the analyses above were controlled for age, sex and MRI protocol, there was no association between these parameters and any DTI parameter. Use of different DMTs had no impact on change in FA or MD over time in the NEDA population, although the study likely did not have enough power to detect a potential effect.
Our study shows evidence of ongoing brain WM microstructural injury in patients with RRMS meeting NEDA criteria as measured by DTI. This suggests that NEDA status does not preclude deterioration of axonal or myelin integrity, providing further evidence that conventional measures are limited in monitoring subtle disease worsening. The accuracy of NEDA in predicting long-term MS disability is unproven,2 20 and there is growing evidence that absence of MRI activity, relapses and EDSS worsening does not necessarily rule out ongoing disease.4 5 Furthermore, though the consequences of this ongoing disease worsening may not be apparent in the short term, cumulative effects may result in long-term accumulation of cognitive and/or physical disability over time.
As we hypothesised, the rate of FA decline was lower in the NEDA group compared with the EDA group. However, the rate of MD increase did not differ statistically between groups. While the reason for this discrepancy is unclear, FA and MD are influenced by different microstructural changes, with MD potentially more sensitive to change in cellularity and water content, and FA more specific for axonal and myelin change.6 Future studies should be aimed at confirming and further evaluating this interesting finding.
Importantly, longitudinal FA/MD change in the NEDA population was not confined to T2 lesions, and significant tract deterioration was demonstrated in NAWM. This is consistent with prior cross-sectional studies showing abnormalities in tract integrity outside T2 lesions. Furthermore, though we did not correct for multiple comparisons, given the exploratory nature of our analysis, we found significant longitudinal changes in the CC, cingulum, SCR and IC, while the EC and posterior fossa structures tested were relatively preserved. This may be due to methodological reasons, but may also reflect the nature of the RRMS clinical phenotype in the absence of disease activity. Future studies evaluating location of DTI changes in progressive MS, in which there is pronounced gradual cerebellar and motor impairment even in the absence of overt inflammatory activity, may demonstrate significant deterioration in these structures.
Consistent with prior studies, we found that DTI parameters correlated with several clinical metrics in our exploratory analysis. Baseline EDSS negatively correlated with baseline FA within the CC, EC, SCR and MCP and MD within the IC, EC and SCR. Furthermore, we found that decreasing FA within the BCS and SCR and increasing MD within the SCR independently predicted fatigue complaints and that these effects were preserved even after controlling for NEDA status. This not only suggests that patient reporting of fatigue may in actuality reflect motor fatigability but it also demonstrates the potential clinical utility of monitoring with DTI. This further implies that incorporation of fatigue measures into the NEDA definition may be useful, as suggested in the multiple sclerosis decision model (MSDM).21 Finally, this is a compelling finding in light of recently presented unpublished work22 demonstrating that subjective fatigue complaints predict conversion to progressive MS, and a recent study showing that fatigue after clinically isolated syndrome predicts conversion to MS.23
Our retrospective study allowed for selection of a larger proportion of patients meeting NEDA than could a prospective trial of similar size, and prevented bias related to patient attrition. Due to this study’s retrospective design, data were obtained from clinical charts and MRIs, and bias could have been introduced by obtaining documentation completed by multiple practitioners. Radiographically, a stringent approach was taken to ensure that all patients in the NEDA group met criteria. However, routine clinical MRIs may not have offered the same surveillance as do more frequent MRIs performed in clinical trials. Additionally, not all patients underwent routine spinal cord surveillance, although this is low yield in monitoring for NEDA.24 While our study utilised medical record documentation of fatigue complaints, a comparison of DTI changes with formal assessments of fatigue and cognition would be beneficial in future studies. Finally, lack of three dimensional (3D) T1 sequences and variability in available 2D T1 sequences prevented comparison of volumetric data with our DTI findings. Follow-up studies should incorporate 3D T1 images for volumetric analysis to assess for correlations with longitudinal DTI changes.
Although our study did not include a negative control and we cannot definitively rule out a contribution of normal age-related tract deterioration, age-related decline is a slower process that mainly affects the last decades of life. Several studies have failed to detect a significant change in DTI parameters over short-term longitudinal follow-up in healthy controls,25 26 and those that have shown short-term changes have focused on elderly populations with mild cognitive impairment,27 where DTI changes are substantially more profound. Our results showed deterioration out of proportion to what has been described in the elderly, and we demonstrated no relationship between age and the likelihood of decline on DTI metrics during study duration. Interestingly, patients in our study under the age of 30 exhibited a decrease in FA over time, at an age during which healthy subjects would exhibit ongoing myelination and increasing FA.28 For the above reasons, we conclude that the changes demonstrated in our study can be attributed to MS pathology rather than the normal ageing process.
The results of the present study have far-reaching implications with respect to clinical monitoring of patients with MS and to defining the disease course. While advanced MRI metrics like DTI and volumetric analysis may not yet be ready for routine clinical use,29 a recent multicentre study demonstrated feasibility of standardisation of DTI metrics across scanners from different imaging vendors and across different centres.30 Additionally, while brain volume analysis is hampered by non-tissue-related volume changes (pseudoatrophy)31 32 related to hydration status and medication usage, analysis of microstructural change may not be. Therefore, incorporation of DTI may be advantageous beyond what can be accomplished with volumetric analysis.
The standard definition of NEDA appears lacking, both clinically4 and radiographically,5 as demonstrated by prior attempts at improving detection of subtle disease pathology. Improvements to clinical evaluation beyond the EDSS have been proposed, such as in the MSDM, a composite metric that incorporates relapse quantity and quality, as well as measures of walking speed, arm function, cognition, fatigue, depression and quality of life.21 Radiographically, incorporation of annual brain atrophy rate into NEDA (NEDA-4) has been suggested, and DTI may not only provide complimentary information, but may have advantages over volumetric analysis.12 31 32
It has been hypothesised that MS is a neurodegenerative disease from the onset, although with additional recurrent acute focal inflammation.33 Neurodegeneration is evident early in MS,8 even in radiologically isolated syndrome.34 It is possible that, by studying microstructural change in patients with RRMS free of overt inflammatory disease, we have isolated this neurodegenerative aspect and have demonstrated radiographic ‘progression’, even in the relapsing-remitting phase of the disease. This may have important implications later in the disease process per the functional reserve hypothesis,35 and it is of great interest to determine whether patients with high rates of tract deterioration measured by DTI are at higher risk for eventually developing progressive MS. Furthermore, as newer therapies for RRMS become available and NEDA becomes an increasingly attainable goal, it is of paramount importance to determine whether therapeutic strategies differentially affect the DTI worsening we have described. Finally, unless more refined clinical measures and more advanced MRI techniques such as DTI are incorporated into the composite metric, NEDA will remain a misnomer.
A special thanks to Colleen Farrell, Rachel Brandstadter,
Catarina Saiote, James Sumowski, Avi Maayan and Georgette Smith and to the National Multiple Sclerosis Society for fellowship funding via the Institutional Clinical Training Award. This study was in part supported by the NMSS RG 5120-A-3 to MI.
AH and DS contributed equally.
Contributors AH, DS, IK-S and MI contributed to the study design. AH, DS, MP and AN were responsible for the acquisition of data. AH, DS and MI were responsible for the writing of the article. MI supervised the study. AH, DS, MP, SK, FL, ZW, YL and MI participated in the analysis and interpretation of data. All authors contributed to the critical review of the manuscript for important intellectual content.
Funding This study was funded by National Multiple Sclerosis Society (NMSS RG 5120-A-3).
Competing interests AH has received consulting funds from TEVA pharmaceuticals. SK has received compensation for consulting and advisory board work with Acorda Therapeutics, Bayer, Biogen, EMD Serono (Merck & Co.), Genentech, Genzyme, Mallinckrodt, Novartis and Teva Pharmaceutical Industries and has given non-promotional lectures with Biogen. FL has received funding for research from Acorda Therapeutics, Biogen Idec, Novartis Pharmaceuticals Corp, Teva Neuroscience, Genzyme, Sanofi, Celgene, NIH, NMSS and funding for consulting from Bayer HealthCare Pharmaceuticals, Biogen Idec, EMD Serono, Novartis, Teva Neuroscience, Actelion, Sanofi-Aventis, Acorda, Questcor, Roche, Genentech, Celgene, Johnson & Johnson, Revalesio, Coronado Bioscience, Genzyme, MedImmune, Bristol-Myers Squibb, Xenoport, Receptos and Forward Pharma. He serves as Co-Chief Editor for Multiple Sclerosis and Related Diseases and has stock ownership in Cognition Pharmaceuticals. MI has received research grants from NIH, NMSS, Novartis Pharmaceuticals Corp. and she is a consultant for Vaccinex.
Patient consent Not required.
Ethics approval Mount Sinai Hospital Institutional Review Board.
Provenance and peer review Not commissioned; externally peer reviewed.
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.