Association of Brain Atrophy With Disease Progression Independent of Relapse Activity in Patients With Relapsing Multiple Sclerosis

IMPORTANCE The mechanisms driving neurodegeneration and brain atrophy in relapsing multiple sclerosis (RMS) are not completely understood


M
ultiple sclerosis (MS) is a chronic disease of the central nervous system characterized by inflammatory, demyelinating, and neurodegenerative processes. 1Despite significant progress in the clinical management of patients with MS, the mechanisms driving disability accumulation are not fully understood.
While it is widely accepted that disability accrual may result from the neuroinflammatory events occurring in clinical relapses (relapse-associated worsening), 2 it is much less clear why some patients experience disability progression independent of relapse activity (PIRA).
5][6] Indeed, PIRA has been shown to occur in typical RMS populations, challenging the traditional distinction between an early exclusively relapsing phase and a late secondary progressive MS (SPMS).The pathophysiological determinants of PIRA remain elusive, although it is plausible that PIRA is associated with increased diffuse neuroaxonal loss.
Assessment of brain atrophy by magnetic resonance imaging (MRI) enables the in vivo quantification of ongoing neurodegenerative processes.Accelerated brain tissue loss is a critical phenomenon in MS, presenting close association with clinical disability, and involving multiple central nervous system compartments with as yet poorly understood differences in regional magnitude and temporal evolution. 7Brain tissue loss may be the consequence of acute focal neuroinflammatory events as well as more diffuse primary or secondary neurodegenerative processes that occur independent from lesion activity. 8These include chronic focal "smoldering" activity, 9,10 progressive loss of chronically demyelinated axons outside MS lesions, astrocyte damage and microglia activation in the normal-appearing white matter (WM) tissue, 11 and meningeal inflammation leading to subpial gray matter (GM) pathology. 12n this work, we aimed at investigating whether PIRA is associated with brain volume loss and whether the pace and pattern of brain volume loss in patients with PIRA are distinct from those observed in patients with clinical relapseassociated worsening.In a large cohort of patients with RMS, we therefore assessed the association between global and regional rates of brain atrophy and (1) focal inflammatory activity (evaluated both clinically and radiologically) and (2) PIRA.

Participants
This longitudinal retrospective investigation included patients prospectively followed up in the Swiss Multiple Sclerosis Cohort, an observational multicentric study with standardized collection of demographics, clinical, and MRI data. 13The study was approved by the local ethics committee.Written informed consent was obtained from all participants before study enrollment.The study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline for reporting observational studies. 14tients participating in the Swiss Multiple Sclerosis Cohort study were enrolled in the present study according to the following inclusion criteria: (1) availability of at least 2 brain MRIs, including 1-mm isotropic magnetization-prepared rapid gradient-echo (MPRAGE) and

Clinical Data
Demographic and clinical data included sex, age, disease duration (defined as time since first symptom), and current and previous disease-modifying therapies (DMTs).Standardized clinical assessments with Expanded Disability Status Scale (EDSS) calculation were performed by certified raters every 6 or 12 months. 13,16The occurrence of relapses (defined according to the McDonald criteria 15 ) was recorded at each visit.
Confirmed disability progression was defined as an EDSS score increase, confirmed at least after 6 months, of (1) 1.5 or more points if baseline EDSS was 0, (2) 1.0 or more points if baseline EDSS was 1.0 to 5.5, or (3) 0.5 or more points if baseline EDSS was greater than 5.5.PIRA was defined as an episode of confirmed disability progression with no relapse during the 90 days before the EDSS increase and during the 6-month period between the EDSS increase and the confirmation of disability progression.
According to the longitudinal clinical evolution, we distinguished: 1. Patients with relapse activity and without PIRA: presenting at least 1 relapse (irrespective of whether it was associated with confirmed disability progression) and without PIRA events during the entire observation; The study design is outlined in Figure 1.

MRI Acquisition
Brain MRI scans were performed at each center using protocols optimized for homogeneous signal-to-noise ratio (eTable 1 in the Supplement).All protocols included a 3-dimensional (3-D), T1-weighted, 1-mm isotropic MPRAGE and a 3-D, 1-mm isotropic FLAIR.

MRI Analysis
Brain MRI analysis was conducted in Basel, Switzerland.Before analysis all native images were visually assessed to ensure sufficient quality.T2 lesion volume (T2LV) was calculated automatically on FLAIR images using the multidimensional gated recurrent units algorithm, 17 and results were manually reviewed.Longitudinal changes of WM lesions were automatically assessed with LeMan-PV, 18 and the outputs, in terms of new and enlarged lesions, were manually reviewed (A.Cagol, R.A.T., N.S.).
Estimations for the regions of interest, including total brain volume (TBV) and volumes of total GM, total WM, cerebral cortex, deep GM, thalamus, cerebellum, and ventricular system, were obtained from FreeSurfer; for symmetric structures, left and right volumes were summed.Mean cortical thickness (CTh) was quantified for the whole cortex, as well as for each lobar and regional cortical area (according to the Desikan-Killiany atlas 25 ), as the average of the thickness obtained in the 2 hemispheres.Total intracranial volume (TIV) was used as a covariate in the analyses to adjust for between-patient differences in head size.SPM12 26 was preferred to FreeSurfer for measuring TIV because it provides a direct quantification instead of an indirect estimation. 27n this study, we chose the longitudinal pipeline of Free-Surfer to explore brain volume changes over time because (1)  it gives the opportunity to consider several brain regions, including structures characteristically affected in MS pathology, and (2) it provides measures of global and regional CTh.In FreeSurfer, the inherent variability associated with the independent analysis of each time point of a patient is addressed with a dedicated longitudinal pipeline by creating an unbiased, robust, within-patient template. 24Such approach proved to increase sensitivity and statistical power in detecting subtle longitudinal changes. 24o support the reliability of brain volumetric measures in our data set, TBV and deep GM volume were quantified in all MRI scans also using SPM12 26 and FIRST, 28 respectively, and brain atrophy rates were obtained also with Structural Image Evaluation, using Normalization, of Atrophy (SIENA). 29The  agreement between software packages is reported in the eMethods in the Supplement.

Statistical Analysis
All statistical analysis was conducted in R version 3.6.3. 30Data were analyzed between January 2020 and March 2021.Linear mixed-effect models 31 were performed using the lme4 package 32 to do 5 types of analyses.First, we investigated the cross-sectional association at baseline between brain volumetric measurements (dependent variable) and ( 1) EDSS score and (2) T2 lesion load (estimated as the logarithmic transformation of T2LV).Models included TIV, sex, age, and disease duration as covariates and MRI protocol (defined by the combination of center and scanner) as random intercept.
Second, a linear mixed-effect model was used to quantify the longitudinal rates of atrophy during follow-up.Brain measurements at each given time point were used as dependent variables.To estimate annual percentage change in brain volume from the slope over time, brain measurements were log-transformed. 33Models included as covariates time (to estimate the rate of volume/thickness change), sex, TIV, age and disease duration at baseline, as well as the interactions between sex and baseline disease duration with time (to adjust the rate of change for sex and disease duration).In addition, both random intercepts (for participants and MRI protocols) and a random slope (on time) were included.
Third, we assessed the association between longitudinal rates of atrophy and (1) brain parenchymal fraction (calculated as the ratio between TBV and TIV) at baseline and (2) T2 lesion load at baseline.The association was investigated by introducing in the model the interaction term between time and the variables of interest.
Fourth, we explored the association between rates of volume/thickness change (dependent variable) and (1) the rate of change in lesion burden, (2) the annualized relapse rate, and (3) the rate of change in EDSS score.The association was investigated by introducing in the model the interaction term between time and the variables of interest.
Fifth, the rates of brain atrophy were compared between (1) patients with exclusive PIRA activity and stable patients; (2) patients with relapse activity but no PIRA and stable patients; and (3) patients with exclusive PIRA activity and patients with relapse activity but no PIRA.The mean difference in annual percentage volume/thickness change (MD-APC) was assessed as the interaction term between patient group and time.The 3 comparisons were performed after a 1:1 nearestneighbor propensity score matching, including duration of follow-up, age, sex, disease duration, number of scans available, and treatment with DMTs as criteria.The balance between groups was assessed with Pearson χ 2 and Mann-Whitney U tests.Comparisons of demographic, clinical, and MRI measures between groups were investigated with Welch t test, Pearson χ 2 test, and Mann-Whitney U test as appropriate.
Results were corrected for multiple comparisons using the false discovery rate approach; reported P values are adjusted for false discovery rate.The threshold of statistical significance was set at P < .05.Graphical results for CTh analysis were displayed with the fsbrain package. 34To assess the reliability of brain volumetric measures obtained with different software packages, an intraclass correlation coefficient was calculated. 35,36As a sensitivity analysis, the effect of DMTs on brain atrophy rates was investigated.

Results
A total of 1904 brain MRI scans from 516 patients were included; 19 scans had to be excluded because of insufficient quality.No patients were excluded.Median (IQR) follow-up was 3.2 (2.0-4.9)years, with a median (IQR) number of scans per patient of 4 (2-5).During the observation period, 46 patients experienced only PIRA events and no relapses, 122 patients relapse activity without PIRA, 14 patients both PIRA and relapse activity, and 334 patients remained clinically stable.The cohort's characteristics are summarized in Table 1.

Comparison Between Patients With PIRA and Stable Patients
During observation, 46 patients presented only PIRA activity with a total of 49 events of PIRA resulting in a median (IQR) annualized increase in EDSS score of 0.20 (0.11-0.50) points; this population was propensity score-matched with 46 stable patients (eTable 5 in the Supplement).The matched groups did not differ at baseline in disability and T2LV.Baseline brain parenchymal fraction was lower in patients with PIRA than in stable patients (median [IQR] in PIRA, 0.756 [0.715-0.788];median [IQR] in stable, 0.769 [0.747-0.807];P = .045).There was no difference in T2LV change during follow-up between groups.

Comparison Between Patients With Relapse Activity and Stable Patients
During follow-up, 122 patients had relapse activity without PIRA events; this population was propensity score-matched with 122 stable patients (eTable 6 in the Supplement).There were no differences between groups at baseline in disability, T2LV, and brain volumetric measurements.Mean (SD) annualized relapse rate in the group with relapses was 0.48 (0.43).

Comparison Between Patients With PIRA and Patients With Relapse Activity
After propensity score matching (eTable 7 and eTable 8 in the Supplement), no significant differences in atrophy rates were detected between patients experiencing only PIRA events and patients who had only relapse activity (Table 2).

Discussion
In this large longitudinal cohort study, we show that patients with RMS and PIRA exhibit increased rates of tissue loss in several brain areas compared with patients who are clinically stable.Our data also indicate that patients with RMS and PIRA are subject to global brain tissue loss similar to that of patients experiencing relapse activity.
To selectively investigate PIRA, we identified patients with confirmed disability progression who were free from relapses during the entire observation period.When compared with a propensity score-matched population of clinically stable patients, patients with PIRA showed a remarkable increase in TBV atrophy, mainly driven by GM loss.Accelerated tissue loss was evident in cortical volume and cortical thickness, especially in frontal and parietal areas.Notably, the 2 groups did not differ in longitudinal radiological inflammatory activity.Moreover, the results were confirmed in a subgroup of patients with PIRA who were completely free from radiological inflammatory activity for the entire follow-up.In clinical practice, the occurrence of PIRA in patients with RMS often remains unrecognized because patients with low levels of dis-ability are infrequently considered to potentially present a progressive disease course. 4,6ur results showing an association between PIRA and diffuse neurodegeneration provide strong evidence of the need to promptly recognize PIRA, to prevent the accrual of irreversible central nervous system tissue damage.Approved DMTs may differ in the ability to prevent disability accumulation due to PIRA, 4 and escalation/induction strategies have not yet been assessed in patients with PIRA.Future clinical trials should therefore aim at identifying the best therapeutic strategy for these patients, both those with and those without relapses.
By reflecting ongoing tissue damage and destruction, brain volumetry has emerged as a useful measure to understand disease progression in MS: in fact, neuroaxonal loss exceeding the reserve capacity of the central nervous system is thought to be the ultimate driver of disability accumulation. 37,38ecause of this, assessing brain volume loss is now recommended in trials evaluating drugs with potential neuroprotective effect, even as a primary outcome measure in progressive MS. 39 Nonetheless, various technical and biological factors limit the measurement of longitudinal brain volume changes in the clinic. 39,40On the other hand, the association between baseline brain atrophy and the risk of PIRA suggests that crosssectional measures of brain volume may help identify patients at risk of neurodegeneration.Associations between rates of atrophy (dependent variables) and clinical group were investigated as the interaction term between clinical group and time in linear mixed-effect models, including total intracranial volume as covariate, random intercepts (for participants and magnetic resonance imaging protocols), and a random slope (on time). b The FDR P value for every structure was P = .88.
Several studies have previously failed in detecting significant longitudinal differences in atrophy rates between patients with RMS and those with progressive MS. [41][42][43] Eshaghi et al 36 reported increased atrophy but limited to temporal cortical GM in SPMS with respect to RMS.It might be speculated that the lack of striking differences between RMS and SPMS results from the uncertainty in the clinical definition of the disease course.Indeed, the traditional distinction between an initial relapse only and a secondary progressive course has been called into question by the evidence of PIRA in RMS. 4 Concordantly, our results showed that patients with RMS may have excess neurodegeneration.Moreover, our data showing that patients with RMS and PIRA exhibit more pronounced brain atrophy than stable patients with RMS support the association of brain volume loss and disability progression in MS.Further, the fact that patients with PIRA show significantly increased brain atrophy compared with stable patients with RMS provides additional important evidence that neurodegenerative changes are more pronounced in a subgroup of patients with RMS than in others.Whether the presence of PIRA represents an early phase of SPMS or a milder form of progression should be investigated in future longitudinal studies.
As described previously, 44,45 we found an association also between focal inflammatory activity and diffuse and regional atrophic changes.Lesions in MS may result in brain volume loss through (1) direct inflammatory damage leading to loss of myelin, oligodendrocytes, and axons within lesions and through (2) indirect tissue loss after Wallerian degeneration. 38,46esides, in our cohort, higher T2LV at baseline was related to lower brain volumes and cortical thinning.Longitudinally, MRI lesion activity was linked to a diffuse increase in rates of tissue loss, involving both WM and GM, with the strongest association in the thalamus.
The association between acute inflammatory activity and atrophy was further corroborated by the evidence of accelerated tissue loss in patients exhibiting only relapse activity.Indeed, this population showed increased TBV loss with respect to a propensity score-matched clinically quiescent population.Interestingly, the acceleration in tissue loss was not seen with WM volume and was driven by both deep GM and cortical GM atrophy.As previously reported, 6 we also found that patients with relapse activity with and without associated confirmed disability progression did not differ in atrophy rates.
Remarkably, no significant differences were detected when atrophy rates were compared between patients experiencing PIRA only and patients experiencing relapse activity only.Our findings are in line with and expand the results of Cree et al, 6 who reported increased TBV loss in association to both silent disease progression and overt inflammatory activity, without significant differences between the two.Our study also showed that accelerated brain atrophy in patients with PIRA and in patients with relapse activity is mainly driven by GM atrophy, with involvement of the cerebral cortex in both patient groups and involvement of deep GM in relapsing patients only.

Strengths and Limitations
Strengths of our investigation include the large sample size and the availability of prospectively performed standardized neurological assessments.In addition, brain volumetric analysis was performed with a pipeline optimized for longitudinal analyses, 24 requiring long computational processes and manual editing but providing sophisticated evaluations and representing the gold standard in CTh measurements. 47ur study also presents some limitations.First, the inclusion of MRI data acquired with different protocols may have represented a confounding factor.However, as part of the Swiss Multiple Sclerosis Cohort study, optimization was performed of scans' signal-to-noise ratio among different centers.Moreover, statistical analysis accounted for heterogeneity in MRI protocols as a confounding factor.
In our study, we did not include measurements of spinal cord atrophy and cortical lesions.Both may help further characterize the mechanisms underlying PIRA, as they have been previously shown to predict physical disability and disease progression, 48,49 silent progression, and conversion to SPMS. 50,51In addition, the criterion used to determine PIRA was based only on EDSS score.Given the absence of measures of upper and lower extremity function in our cohort, subtle neurological worsening that did not result in EDSS score increase may have been overlooked.Furthermore, DMTs may have constituted a bias in our study, which we tried to overcome by performing propensity score matching for treatment status and considering treatment groups in sensitivity analyses.

Conclusions
Our data show that events of insidious PIRA are associated with increased atrophy rates, likely reflecting ongoing diffuse neurodegenerative processes, especially in cortical GM.These results point to the need to promptly identify patients with PIRA in clinical practice, because they may benefit from optimized therapeutic regimens.In this context, clinical trials to assess the potential benefit of treatment escalation/ induction in patients with RMS and PIRA are warranted.

Figure 2 . 5 GM
Figure 2. Annual Percentage Change in Volume and Regional Cortical Thickness

Figure 3 .
Figure 3. Differences in Rates of Regional Cortical Thinning Comparing Stable Patients vs Patients With Only Progression Independent of Relapse Activity (PIRA) or Patients With Only Relapse Activity Patients with PIRA events vs stable patients A T2/fluid-attenuated inversion recovery (FLAIR), separated in time by at least 6 months; (2) diagnosis of RMS fulfilling the 2017 revisions of the McDonald criteria 15 ; (3) availability of at least 1 annual clinical followup, with temporal proximity between MRI acquisition and neurological evaluation (≤2 months); and (4) age between 18 and 80 years.All clinical and MRI data acquired as part of the Swiss Multiple Sclerosis Cohort study between January 2012 and September 2019 from patients fulfilling eligibility criteria were included.MRI scans with insufficient image quality were excluded.

Table 2 .
Differences in Brain Atrophy Rates for the Propensity Score-Matched Group Comparisons a a