A systematic literature review of the association between global brain atrophy and the Expanded Disability Status Scale score in people with multiple sclerosis

Abstract

Background:

Brain atrophy (BA) is a useful predictor of clinical outcomes in people with multiple sclerosis (PwMS). For this reason, MAGNIMS (Magnetic Resonance Imaging in Multiple Sclerosis), an expert consensus group, recommended that global brain volume loss (BVL) is included as a secondary outcome in therapeutic clinical trials. However, there has not been a recent review of the evidence of the association, or strength of association, between global BA and disability in PwMS.

Objectives:

Our aim is to review articles from 2013 onward measuring the associations between percentage of brain volume loss (PBVL), normalized brain volumes (NBV) or normalized brain parenchymal volume (NBPV), and the Expanded Disability Status Scale (EDSS), or disability progression (DP) measured by EDSS in PwMS.

Design:

Systematic review.

Methods:

We searched Medline, Embase, Cochrane Library, Cochrane Clinical Register of Controlled Trials, Cochrane Database of Systematic Reviews and Cumulative Index to Nursing and Allied Health Literature for observational studies, clinical trials and modelling studies measuring the association between global BVL, PBVL, NBV or NBPV, and EDSS score or DP in PwMS. We included people with clinically isolated syndrome and excluded studies with a population greater than 20% primary progressive multiple sclerosis patients.

Results:

We found 58 studies were eligible for the review. Most longitudinal studies (19/23) observed a significant association between global BVL and change in EDSS score or DP. Similarly the majority of cross-sectional studies (26/29) observed an association between baseline BV measures and EDSS. Most studies investigating the association between baseline brain volume (BV) measures and follow-up EDSS, that is, asking if baseline BV is a predictor of DP, or future EDSS score, did not find an association (4/15 observed an association).

Conclusion:

Around a 1% (range 0.4%–1.3%) decrease in global BV per year was associated with DP, but caution in comparing studies is recommended due to variations in the definition of DP.

Keywords

brain atrophy brain volume loss disability multiple sclerosis

Introduction

Multiple sclerosis (MS) is a widespread neuroinflammatory disease associated with progressive neurodegeneration. Typically onset is at a young age.¹ Symptoms include altered sensation such as numbness, tingling, itching or pins and needles, impaired coordination and balance, weakness, muscle spasms or cramps, fatigue, pain, and visual disturbances.^2,3

Brain atrophy (BA) in people with multiple sclerosis (PwMS) was first observed in the early 1960s.⁴ By the mid-1990s magnetic resonance imaging (MRI) was being used to quantify BA, and its association with physical disability.⁵ These early studies used two-dimensional scans to measure BA.^5,6 More recent studies using three-dimensional scans with analysis by Structural Image Evaluation Using Normalization of Atrophy (SIENA) software suggest that PwMS will lose around 0.5%–1.35% of brain volume (BV)/year, whereas for an average person without MS around 0.1%–0.3% of BV/year is lost due to normal ageing.^7
–9 Measurements of BA are thought to represent the net effect of all the degenerative processes occurring in MS^10
–12 and seem to be better predictors of clinical outcomes then other MRI measures.^6,13
–22 For example, BA seems to be a better marker of clinical disability than conventional lesion measures,^23,24 and people with higher levels of BA are more likely to progress from clinically isolated syndrome (CIS) to MS.^15,25,26 BA has been associated with increased disability and worsening cognition,^{11,17,20,21,27
–41} poorer quality of life,^42,43 increased fatigue⁴⁴ and poorer economic outcomes.^45,46 Several reviews (2009–2016) have described associations between BA and cognition but just three considered physical disability.^23,47
–50 The MAGNIMS (Magnetic Resonance Imaging in Multiple Sclerosis) group recently recommended that global brain volume loss (BVL) is used to define and predict MS severity and is included as a secondary outcome in therapeutic clinical trials.⁵¹

Over the last four decades, the most widely used instrument to assess disability progression (DP) in people with MS (PwMS) has been the Expanded Disability Status Scale (EDSS).⁵² It is an ordinal rating system ranging from 0 to 10. A frequent criticism of the EDSS is the unequal interval distances between points, for example, the difference between the values 1 and 2 has a different relevance to 6 and 7. This is because the lower end of the scale (0–4) measures neurological impairments, 4–6 indicates impacts on walking ability, 6–7 indicates loss of walking other than a few metres and by 7–7.5 wheelchair dependency.⁵³ Furthermore, EDSS is only weakly correlated with neuropsychological impairment or patient-reported outcomes.⁵³

There are several approaches to, and technical challenges in, quantifying global BA using MRI.²⁴ A longitudinal approach analyses two or more MRI scans on the same individual at different time-points that are spatially matched and subtracted. The brain surface is automatically modelled and registered to the outer skull to compare and normalize serial images. Percentage of brain volume change (PBVC) is based on changes in the brain surface relative to the skull and normalized to the skull size. SIENA is the most frequently used image-processing software.⁵⁴ BA can also be estimated or inferred from a single MRI scan. The image is registered to a standard space and normalized to the skull surface and tissue is segmented into regions of brain and non-brain tissue.⁵⁵ The most frequently used software is Structural Image Evaluation, using Normalization of Atrophy-Cross-Sectional (SIENAX).⁵⁴ The normalized brain volume (NBV) measurements can then be categorized as increased or decreased in volume relative to controls or a standard brain atlas and compared with EDSS scores, or compared between groups such as PwMS with or without DP defined by EDSS. Alternatively, the image may be processed using a proportion-based method to estimate the proportion of intracranial volume occupied by brain tissue, that is, the brain parenchymal fraction or normalized brain parenchymal volume (NBPV).⁵⁶

An earlier review looked at the associations between BA and physical disability using studies published prior to 2013.⁴⁸ Our aim is to build on this review by systematically reviewing the associations between global BA and physical disability measured by EDSS in people with MS, particularly CIS/RRMS, from 2013 onwards. Our primary objective is to compare longitudinal studies measuring the association between change in whole BV and change in EDSS score or DP progression measured by EDSS. Our secondary aim is to compare cross-sectional studies measuring the association between normalized whole BV, or fraction, and EDSS score. A further aim is to compare longitudinal studies measuring the association between baseline normalized whole BV, or fraction, and future EDSS score at follow-up. As BA is increasingly used as an outcome in clinical trials,⁵¹ we need to improve our understanding of the relationship between BA and physical disability.

Methods

Protocol

The protocol was designed according to the preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement⁵⁷ and finalized in January 2022 before the final literature search was undertaken.

Inclusion criteria

People aged 18 or over, diagnosed with CIS or MS.

Longitudinal observational studies, clinical trials and modelling studies measuring the association between change in whole BV and EDSS score or DP defined by EDSS score. Confirmed or non-confirmed DP was included.

Cross-sectional studies measuring the association between NBV, or fraction, and EDSS score or DP defined by EDSS score.

Original article written in English and published between 1 January 2013 and 3 February 2022 in a peer reviewed journal.

Exclusion criteria

Study population includes 20% or more people with primary progressive multiple sclerosis (PPMS). A pragmatic decision to keep the review focus on CIS/RRMS.

Studies measuring the association with cord or regional atrophy only.

Studies where EDSS was only reported as a composite of ‘No Evidence of Disease Progression’. (NEDA-3 is defined by no new MRI lesions, no new relapses and no change in EDSS over 1 year. NEDA-4 is defined by no clinical relapse, no confirmed disability progression, no new or enlarging T2 lesion on brain MRI and no brain volume loss ⩾0.4% per year.)

Studies where the association was not reported separately for PwMS and healthy controls.

The review is focused on CIS/RRMS because BVL is thought to proceed at a faster rate in the earlier stages and disease modifying therapy (DMT) are more effective at this time. However, many studies were not restricted cleanly to CIS/RRMS and sometimes patients progressed to secondary progressive multiple sclerosis or PPMS during follow-up. We wanted to include as many papers as possible that focused on people with CIS/RRMS so made a pragmatic, a priori, decision to exclude studies with more than 20% of people with PPMS diagnoses.

Electronic searches

We searched Medline, Embase, Cochrane Library, Cochrane Clinical Register of Controlled Trials (CENTRAL), Cochrane Database of Systematic Reviews (CDSR) and Cumulative Index to Nursing and Allied Health Literature (CINAHL). The full search strategy is listed in Table 1.

Table 1.

Literature search terms for review of the association between global brain atrophy and the EDSS score in people with multiple sclerosis.

Search	Description	Search terms	Hits
S1	Multiple sclerosis	EMB.EXACT (‘multiple sclerosis’) OR MESH.EXACT.EXPLODE (‘Multiple Sclerosis’) OR MESH.EXACT (‘Multiple Sclerosis, Relapsing-Remitting’) OR MESH.EXACT (‘Multiple Sclerosis, Chronic Progressive’) OR TI (‘multiple sclerosis’ OR RMS OR RRMS OR ‘relapsing forms of multiple sclerosis’ OR ‘relapsing multiple sclerosis’ OR ‘relapsing-remitting multiple sclerosis’ OR ‘clinically isolated syndrome’ OR ‘secondary progressive multiple sclerosis’ OR ‘secondary progressive MS’ OR SPMS OR ‘primary progressive MS’ OR ‘primary progressive multiple sclerosis’ OR PPMS)	224921
S2	Brain volume loss	EMB.EXACT (‘brain size’) OR EMB.EXACT (‘brain atrophy’) OR (MESH.EXACT (‘Atrophy’) AND ‘brain’) OR ‘brain volume’ near/3 (loss OR atrophy OR reduction OR change OR low) OR TI, AB (BVL) OR (‘gray matter’ OR ‘cortical thickness’ OR ‘thalamic’ OR volume) near/3 (loss OR atrophy OR reduction OR change) OR TI, AB (‘cerebral atrophy’) OR TI (atrophy)	271928
S3	Clinician-measured outcomes	EMB.EXACT (‘Expanded Disability Status Scale’) OR EMB.EXACT (‘California verbal learning test’) OR EMB.EXACT.EXPLODE (‘walk test’) OR EMB.EXACT (‘symbol digit modalities test’) OR TI, AB (‘Multiple Sclerosis Functional Composite’ OR ‘MS Functional Composite’ OR MSFC OR ‘Timed 25 Foot Walk Test’ OR T25FWT OR ‘Hole Peg Test’ OR 9HPT OR NHPT OR ‘Paced Auditory Serial Addition Test’ OR PASAT OR ‘Brief Visuospatial Memory Test’ OR BVMT OR ‘California Verbal Learning Test’ OR CVLT OR ‘Symbol digit modalities test’ OR SDMT OR ‘Expanded Disability Status Scale’ OR EDSS OR ‘disability status scale’ OR DSS OR ‘motor status scale’ OR MSS OR ‘ambulation index’ OR ‘functional independence measure’ OR FIM OR ‘Kurtzke functional system scores’ OR FSS OR ‘test performance’ OR ‘neuropsychological performance’ OR ‘cognitive domains’) OR EMB.EXACT (‘clinical outcome’) OR MESH.EXACT.EXPLODE (‘Cognition’) OR EMB.EXACT (‘cognition’) OR TI (disability)	918094
S4	Combined search	S1 AND S2 AND LA (English) NOT (EMB.EXACT (editorial OR ‘case report’ OR letter OR note) OR DTYPE (‘Editorial’ OR ‘Comment’ OR ‘Letter’ OR ‘Case Reports’ OR ‘News’ OR ‘Newspaper Article’) OR TI, AB (‘case study’ OR ‘case report’))	7385^a
S5	Clinical results	S4 AND S3	2924^a

Duplicates between Embase and MEDLINE removed.

EDSS, Expanded Disability Status Scale; MS, multiple sclerosis; PPMS, primary progressive multiple sclerosis; RRMS, relapsing-remitting multiple sclerosis; SPMS, secondary progressive multiple sclerosis.

Screening process

PROQUEST was used to remove duplicate articles from Medline and Embase which was then combined with articles from Cochrane and CINAHL. Additional duplicates were removed by searching within excel. The initial title and abstract screening consisted of two researchers independently completing a standard data collection form. Any discrepancies between the researchers were resolved by discussion. Full-text articles were obtained for all the abstracts selected for the second screen. The second screening of the full-text articles was undertaken independently by two researchers followed by discussion to reconcile any differences. A third reviewer was available in case the two reviewers did not reach agreement but was not required.

Data extraction

A standardized Excel spreadsheet was used to capture the data list in Box 1. The data were extracted by one researcher and validated by a second researcher.

Box 1.

Data extracted from each study.

1. Study design
2. The primary aim/endpoint of the study
3. Volumetric brain measurements reported
4. Description of MRI acquisition parameters
5. Description of image analysis
6. Description of disability measure – EDSS score or disability progression based on EDSS
7. Date of data collection
8. Setting – source of recruitment of PwMS source, country
9. Study size
10. Study population – phenotype of MS, age, gender, baseline EDSS score, disease duration
11. Inclusion and exclusion criteria
12. Length of follow-up and time points at which BA and EDSS measured for longitudinal study
13. Loss to follow-up (if applicable)
14. Time between MRI and EDSS for cross sectional study
15. Statistical methods
16. Description of association between BA and EDSS
17. Name of study/funder

BA, brain atrophy; EDSS, Expanded Disability Status Scale; MRI, magnetic resonance imaging; MS, multiple sclerosis; PwMS, people with MS.

Data analysis

Studies were grouped first according to study design, that is, longitudinal or cross-sectional. Longitudinal studies were then grouped according to the way EDSS was used to describe increase in disability – either as change or loss in BV (percentage of brain volume loss, PBVL) or as DP. If EDSS scores were transformed in to DP then further groupings were used to count how many studies had equivalent definitions of DP given the range of baseline EDSS reported for the study. Cross-sectional studies were also further grouped into those comparing baseline BA with baseline EDSS and those comparing baseline BA with follow-up EDSS. Within these four groups, the studies were described and compared qualitatively. Results that were methodologically aligned and used similar outcome measures were reported individually to allow for comparisons.

Results

Screening process

The screening process is shown in Figure 1. Briefly, 3114 abstracts were screened resulting in 715 papers and abstracts for full screening. At this point, the review was limited to full-text articles published from 2013 onwards resulting in 320 articles for full-text screening. In total 164 articles published before 2013 and 231 conference abstracts published from 2016 onwards were not considered further. The full-text screening resulted in 197 articles measuring BA; 131 of which also focused on disability outcomes. Studies only addressing regional BA, or not reporting on the associations with EDSS or DP, or the study population included more than 20% PPMS diagnoses were excluded. The remaining 58 studies were included in this review.

Figure 1.

Flowchart of the selection process.

Study designs and patient characteristics

The study designs and patient characteristics are summarized in Table 2. Twenty-three were longitudinal analyses and 37 were cross-sectional. Two studies included both cross-sectional and longitudinal analyses.^58,59 The longitudinal studies had slightly larger study populations and included slightly younger PwMS with slightly shorter disease duration compared to cross-sectional and follow-up studies (Table 2).

Table 2.

Summary of study population characteristics by study design.

Study population characteristic	Longitudinal studies n = 23		Cross-sectional studies n = 29		Follow-up studies n = 15
Study population characteristic	Median value	Range	Median value	Range	Median value	Range
Study size (PwMS)	180	16–3635	80	20–3635	102	23–1214
Follow-up time (years)	4	1–12	n/a	n/a	5	1–30
Age at baseline (years)	37	29–48	41	31–61	40	31–61
Proportion female at baseline	70%	54%–80%	68%	57%–80%	68%	59%–80%
Proportion with PPMS at baseline	0%	0–17%	0%	0–17%	0%	0–14%
Mean EDSS score at baseline	2	1.3–6	2.5	1.2–6.5	2.5	1.2–5
Mean disease duration at baseline (years)	7.5	1–14	9.5	0.5–27	8.5	4–30

EDSS, Expanded Disability Status Scale; PPMS, primary progressive multiple sclerosis; PwMS, people with multiple sclerosis.

Methods and results of longitudinal studies

The methods and results for the longitudinal studies are described in Table 3. The patient characteristics are summarized in Table 4. All studies measured BA as PBVL but one study categorized PBVL for analysis.⁶⁰ Eight studies investigated the association between PBVL and change in EDSS; 6 of which observed a significant association (p < 0.05). Two small studies were inconclusive, one observed a non-significant association (n = 38)⁶¹ and another observed no change in EDSS (n = 16).⁶² Fifteen studies investigated the association between PBVL and DP defined by changes in EDSS score; 13 of which reported a significant association between PBVL and DP. One study observed a non-significant association (n = 62),⁶³ and one study observed no difference in NBV (n = 82).⁶⁴ There were six different definitions of DP (see 1–6, Table 5). Twenty longitudinal studies were retrospective analyses of existing cohorts and clinical studies. Three longitudinal studies were prospective, two of which observed a significant association between PBVL and DP^46,65 and one observed a significant association between PBVL and EDSS with Icobrain software but not SIENA.⁵⁹ All except one of the longitudinal studies used SIENA software (one used FreeSurfer⁶⁶) but two used other software in addition to SIENA (Icobrain⁵⁹ and SPM12⁶²). Eighteen studies used regression or survival models adjusted by various combinations of sex, age, education, disease phenotype, disease duration, study cohort, lesion volume, field strength and other baseline variables. The majority of the study populations were treated with DMT and just one study included mostly untreated patients.⁴⁶ Many studies also reported measures of regional BA (14/23, 61%) and regional atrophy was the main focus for many studies. A few studies reported additional measures of physical disability, specifically 9HPT and T25FW (4/23, 17%).^61,62,67,68 Most studies didn’t report an annualized PBVL tending to report the PBVL over the observation period or by categories but for the 5 that did report annualized PBVL, the range was −0.48% ± 0.93 to −0.9% ± 1.0 with observation periods ranging from 2 to 7.5 years.

Table 3.

Description of analysis and outcomes of longitudinal studies reporting associations between PBVC and change in EDSS scores (raw EDSS score or categorized as disability progression).

Author, year	Is there an association between EDSS or DP and PBVL (p ⩽ 0.05)	Description of statistical method and summary of association between change in EDSS (or DP) and PBVL	Mean PBVL	Study design	Sample size at follow up	MRI analysis software	Mean follow-up time, years	Disease progression definition (Table 5)
Uher,⁶⁹ 2021	Yes	Significant association between annualized PBVL and log annualized EDSS change using mixed models adjusted for sex, study cohort, T2 lesion volume, NBV at baseline, proportion of time spent on low-efficacy DMTs, and proportion of time spent on high/moderate-efficacy DMTs. (β = 0.15; 95% CI = 0.09–0.21; Cohen f2 = 0.006; p < 0.0001).	−0.48 ± 0.93 annualized	Retrospective cohort of 2 clinical studies combined	1903	SIENA	3.4	n/a EDSS score used
Dekker,⁶⁸ 2019	Yes, at 6 years follow-up, but not 12 years	Association between EDSS score at 6 year follow-up and PBVL (baseline to year 2). Univariate β = 0.158, p = 0.09; multivariate linear regression β = 0.143, p = 0.03 adjusted for baseline age, education, disease phenotype, T1 and T2-lesion volumes and sex. No significant association at 12 year follow-up.	−0.54 (±0.55) over 2 years	Retrospectively selected from observational prospective cohort	115/97 at 6/12 years	SIENA	6 or 12	n/a EDSS score used
Beadnall,⁵⁹ 2019	Inconclusive	Non-annualized PBVL by Icobrain was rank correlated with EDSS change by Kendall Tau (τ = 0.148, p = 0.04). PBVL by SIENA was not correlated with EDSS change	SIENA −0.64 ± 0.73 Icobrain −0.59 ± 0.65 over 1 year	Prospective cohort	102	SIENA Icobrain	1.05	n/a EDSS score used
Chu,⁶² 2018	Inconclusive	Proportional odds model with random effects found no association between PBVL and EDSS change. EDSS at follow-up was unchanged from baseline (1.3 ± 1.0, p = 0.47).	−0.79 ± 1.1 over 5 years	Not stated, clinical cohort	16	SIENA SPM 12	5	n/a EDSS score used
Ghione,¹¹ 2018	Yes	Annualized PBVL was the dependent variable with EDSS change as a predictor in multivariate linear mixed-effects models with interaction terms across time adjusted for age, sex, field strength. An increase of 1 in EDSS score was associated with annualized PBVL of −0.134 p = 0.0001	−0.9 ± 1.0 annualized	Retrospective analysis of routine clinical data	1651	SIENA	4.7	n/a EDSS score used
De Stefano,⁷⁰ 2015	Yes	Significant univariate correlation between annualized PBVL and EDSS change (Pearson’s r = −0.21, p = 0.003). Multivariate analysis used a linear model adjusted for age, disease duration, EDSS and disease type showed a significant partial r = −0.23, p = 0.001	−0.51 ± 0.27 annualized	Retrospective analysis of phase III RCT	206	SIENA	7.5	n/a EDSS score used
Radue,⁵⁸ 2015	Yes	Significant correlation between 1 year PBVL from baseline and 1 year change in EDSS score Pearson’s r ≈ −0.075, p < 0.001); At 4 years Pearson’s r ≈ −0.23, p < 0.001	n/s but 2 year PBVL > −1 = 64% of cohort; PBVL = −1 to >2.5% = 28% of cohort; PBVL −2.5 to −5 = 9% of cohort	Retrospective analysis of 3 phase III trials combined	3635	SIENA	2 or 3	n/a EDSS score used
Maghzi,⁶¹ 2014	Borderline	Mixed model regression with PBVL as the predictor and EDSS as the dependent variable. Random intercepts and slopes allowed for repeated measures and time trends. Each 10% decrease (−24% to 1%) in brain volume was associated with increase in 1 EDSS point, p = 0.08	n/s	Retrospective analysis of RCT	38	SIENA	2 or 3	n/a EDSS score used
Rocca,⁷¹ 2021	Yes	Multivariate logistic regression models adjusted for follow-up duration, treatment and centre were used to assess the association between PBVL and DP. PBVL with DP = −1.00% (±0.93) and without DP = −0.28% (±1.07); p = 0.01	−0.38 (1.07) over 1 year	Retrospective analysis of existing study cohort	144	SIENA	1.08	<6 DP = +1 ⩾6 DP = +0.5
Koch,⁷² 2021	Yes at 96 weeks No difference at 48 weeks	Compared PBVL at 48 and 96 weeks with DP using Student’s t-test. Used multivariate logistic regression models with DP as outcome and categorized PBVL as predictor adjusted for age, sex, treatment arm. At 48 weeks PBVL with DP = −0.54 (±0.62) and without DP = −0.53 (±0.56), p = 0.82; at 96 weeks PBVL with DP = −1.12 (±0.81) and without DP = −0.90 (±0.74), p = 0.02	−0.32 (0.5) over 24 weeks −0.53 (0.57) over 48 weeks −0.75 (0.68) over 72 weeks −0.95 (0.76) over 96 weeks	Retrospective analysis of phase III placebo controlled RCT	889	SIENA	1.8	<5.5 DP = +1 ⩾5.5 DP = +0.5
Sprenger,⁶⁰ 2020	Yes	Kaplan–Meier estimated risk of disability worsening and reaching EDSS ⩾4 or ⩾6 compared between PBVL categories using a Cox proportional hazards model adjusted for baseline EDSS and region. Hazard ratio (HR) disability worsening for PBVL group 1 vs group 3 = 0.58, p = 0.001; HR group 2 vs group 3 = 0.68, p = 0.009). HR reaching EDSS ⩾4 for PBVL group 1 vs group 3 = 0.39, p = 0.0003; HR group 2 vs group 3 = 0.59, p = 0.0128. HR reaching ⩾6: for PBVL group 1 vs group 3 = 0.32, p = 0.0013; HR group 2 vs group 3 = 0.31, p = 0.0001.	PBVL categories were grp 1 (least BVL) ⩾ −0.52; group 2 < −0.52 to −2.18; group 3 (most BVL) < −2.18 over 2 years	Retrospective analysis clinical study extension	709	SIENA	7	<6 DP = +1 ⩾6 DP = +0.5
Pontillo,⁷³ 2020	Yes at 10 years No at 2 years	Binary logistic regression models adjusted for age, gender, disease duration, baseline EDSS score found that DP at 10-year follow-up was significantly predicted by PBVL, β = −0.95 (−1.97 to −0.32), p = 0.04. PBVL at 2 years was not significantly associated with DP.	−0.49 ± 0.58 annualized baseline to 2 years	Retrospective analysis of clinical data	87	SIENA	10	All +1
Ghione,⁶⁷ 2020	Yes	ANCOVA compared DP groups and annualized PBVL adjusted for age, race, T2LV and baseline NBV. DP group had significantly higher PBVL than DI or SD groups (p = 0.001)	DP group = −1.1 ± 1.1 DI group = −0.8 ± 0.7 SD group = −0.8 ± 0.7 annualized	Retrospective analysis of clinical data	980	SIENA	4.8	<5.5 DP = +1 ⩾5.5 DP = +0.5
Zivadinov,⁷⁴ 2019	Yes	ANCOVA adjusted for age, sex and treatment change was used to compare PBVL and DP at 10 years. PBVL with DP = −7.52 ± 3.84; PBVL without DP = −5.23 ± 3.00, p < 0.001	−6.52 over 10 years	Retrospective analysis of open label extension of clinical trial	176	SIENA	10	0 DP = +1.5 >0 DP = +1
De Stefano,⁷⁵ 2018	Yes	Annualized PBVL correlated with cumulative probability of DP. Cox proportional HR adjusted for treatment = 0.67 (95% CI, 0.571–0.787); p < 0.001	n/s but IQR for annualized −0.41 to −1.08	Retrospective analysis of phase III placebo controlled RCT	1025	SIENA	2	0 DP = +1.5 >0 DP = +1 ⩾5 DP = +0.5
Ciampi,⁶⁶ 2017	Yes	PBVL over 3 years with DP = −1.26% and without DP = −0.19%; p (t-test) = 0.02	−1.13 in year 1 No significant change in years 2/3	Retrospective analysis of clinical data	24	Free surfer	3	<5.5 DP = +1 ⩾5.5 DP = +0.5
von Gumberz,⁶⁴ 2016	No	Annualized PBVL was not a significant predictor of DP using linear and logit models with, or without, adjustment for treatment status: gender, age, number of T1-, T2- and Gd-lesions and the absolute change of lesion numbers.	−0.14 ± 1.13 annualized	Post hoc analysis of prospective cohort	82	SIENA	2.9	0 DP = +1.5 >0 DP = +1 ⩾5 DP = +0.5
Jeffery,⁷⁶ 2016	Yes	PBVL at 2 years was categorized into quartiles. ANCOVA showed the highest PBVL quartile had significantly higher risk for DP relative to the lowest PBVL at 2 years (OR, 95 % CI 2.13, 1.33–3.40; p = 0.002) and at 4 years (1.98, 1.22–3.23; p = 0.006). Time-to-reach the EDSS ⩾ 4 and ⩾6 was determined by Kaplan–Meier analysis and a logistic regression model adjusted for baseline EDSS was used to estimate the proportion of patients reaching the EDSS milestone. Highest PBVL quartile had 2.8 (1.5–5.1), p = 0.01 times the risk of reaching EDSS ⩾ 4 over 4 years relative to the lowest PBVL quartile. For EDSS > 6 it was 5.7 (2.1–15.5), p = 0.0005.	n/s overall Q1 −1.45 Q2 −0.60 Q3 −0.25 Q4 −0.23	Retrospective analysis of phase III placebo controlled RCT and its extension	1029	SIENA	4	<5 DP = +1 ⩾5 DP = +0.5
Sastre-Garriga,⁶³ 2015	n/s trend	Binomial models with PBVL as a predictor found DP was (non-significantly) associated with PBVL adjusted for baseline EDSS, age, disease duration and Gd-enhancing lesions at baseline. 1 year OR = 1.7, 95% CI: 0.91–3.47, p = 0.09); 3 year OR = 1.55, 95% CI: 0.94–2.54, p = 0.084)	Overall n/s Year 1 −1.1 ± 1.4 Year 2 −0.47 ± 0.82 Year 3 −0.46 ± 0.71	Retrospective analysis of clinical data	62	SIENA	3	<5.5 DP = +1 ⩾5.5 DP = +0.5
Sormani,⁷⁷ 2015	Yes	DP at 2 years was modelled by unconditional logistic regression adjusted for treatment group, number of relapses in year 1, number of active T2 lesions in year 1 with annualized PBVL as the predictor. A reduced risk of DP was associated with a lower PBVL (unadjusted OR 0.62, 95% CI = 0.49–0.78, p < 0.001, adjusted OR 0.74, 95% CI = 0.57–0.97, p = 0.026)	Overall n/s At 1 year PBVL untreated −0.63 ± 1.08; treated −0.46 ± 1.02 At 2 years PBVL untreated −1.30 ± 1.51; treated −0.86 ± 1.36	Retrospective analysis of phase III placebo controlled RCT	992	SIENA	2	All +1
Jacobsen,⁷⁸ 2014	Yes	At 5 years PBVL with DP = −3.8% ± 2.2 and without DP = −2.0% ± 1.3, p (t-test) <0.01, Cohen’s d = 0.98. At 10 years PBVL with DP = −5.5% ± 2.7 and without DP = −3.7 ± 1.9%, p (t-test)<0.01, Cohen’s d = 0.78. DP at 5 and 10 years was also modelled by binary logistic regression adjusted for age, sex, scanner type, baseline disease duration, T1-LVs and T2-LVs, MS subtype and months and type of DMT with PBVL as the main predictor. Adjusted 5 year β = −0.57 p < 0.01 Adjusted 10 year β = −0.39 p < 0.01	Overall n/s At 5 years −3.8 with DP and −2.2 without DP. At 10 years −5.5 with DP and −3.7 without DP	Prospective cohort study	81	SIENA	10	<6 DP = +1 ⩾6 DP = +0.5
Uher,⁶⁵ 2014	Yes	A mixed-effect model adjusted for age, gender, time from the first event to baseline assessment, and treatment status showed significantly increased NBVL (p = 0.014) in patients with DP compared SD or DI. PBVL with DP = −3.0% ± 2.4; PBVL with SD = −2.5 ± 1.9; PBVL with DI = −2.1 ± 2.3; −1.5, p (t-test)<0.05 for comparison between DP and SD/DI combined	Overall n/s At 4 years −3.0 with DP −2.5 with SD −2.1 with DI	Prospective observational clinical study	210	SIENA	4	0 DP = +1.5 >0 DP = +1
Zivadinov,⁷⁹ 2013	Yes	Mixed-effect models with a random intercept adjusted for age, disease duration and treatment status over 5-year follow-up with PBVL as a predictor of DP p < 0.001. Also a significant interaction between PBVL and DP (p = 0.001) meaning patients with DP had a steeper decline.	Overall n/s At 5 years −4.8 with DP −2.6 without DP	Retrospective analysis of 2 year placebo controlled RCT and its 3 year extension	180	SIENA	5	0 DP = +1.5 >0 DP = +1

CI, confidence interval; DI, disability improvement; DMT, disease modifying therapy; DP, disability progression; EDSS, Expanded Disability Status Scale; HR, hazard ratio; IQR, interquartile range; MRI, magnetic resonance imaging; NBV, normalized brain volume; OR, odds ratio; PBVC, per cent brain volume change; PBVL, percentage of brain volume loss; RCT, randomized controlled trial; SIENA, Structural Image Evaluation, Using Normalization, of Atrophy.

Table 4.

Patient characteristics at baseline for longitudinal studies reporting associations between PBVC and change in EDSS scores (raw EDSS score or categorized as disability progression) and other study data collected including treatment, measures of physical disability and regional brain atrophy.

Author, year	%CIS/RRMS/PMS	%PPMS	% female	Mean age (sd) at baseline	EDSS score (baseline)	Mean disease duration, years (sd or range)	Country	Treatment at baseline	Other reported measures of physical disability (extra to EDSS)	Regional measures of brain atrophy reported
Uher,⁶⁹ 2021	34/66/0	0	70	36 (±9)	2 (1–3)	7.4 (±4.7)	Czech Republic	None 27% GA, DMF and IFNB-1a 59% Fingolimod, Mitoxantrone and Natalizumab 14%	None	None
Dekker,⁶⁸ 2019	31/59/10	<10	66	35 (±9)	2 (1.5–3)	0.92 (0.6–1.6)	NL	DMT 56%/63% at 6/12 years follow-up	T25FW 9HPT	PVVC
Beadnall,⁵⁹ 2019	1/97/2	0	80	38 (±9)	2 (IQR 1.88)	7.4 (±7.4)	Australia	DMT 72%	None	None
Chu,⁶² 2018	0/100/0	0	69	45 (±8)	1.3 (0–3.5)	13.4 (±10.5)	USA	DMT 100%	T25FW	Total DGM, thalamus, caudate, putamen, globus pallidus
Ghione,¹¹ 2018	8/74/18	2	76	45 (±11)	MS 2.5 (±2.5) CIS 1.5 (±1.0)	12.2 (±9.8) 3 (±4.7)	USA	GA 18% Immunotherapy 1% IFNB-1a 40% Natalizumab 7% Other therapy 5% No therapy 16% Missing data 13%	None	LVV
De Stefano,⁷⁰ 2015	0/87/14	6	71	37 (18–63)	2.25 (0–8)	6 (0–32)	Italy	DMT 85%	None	None
Radue,⁵⁸ 2015	0/100/0	0	n/s	n/s	n/s	n/s	Multiple countries^a	Fingolimod 100%	None	None
Maghzi,⁶¹ 2014	CIS/RRMS n/s	0	71	36 (±9)	2 (0–5.5)	7.5 (±4.9)	USA	IFNB-1a 100% Riluzole 50%	T25FW 9HPT	nGMV nNAWMV
Rocca,⁷¹ 2021	9/57/35	17	54	42 (18–72)	3.5 (0–8.5)	14.1 (0.1–46)	Multiple countries^b	DMT 61%	None	nGMV nWMV
Koch,⁷² 2021	0/0/100	0	62	48 (42–53)	6 (5–6.5)	n/s	Multiple countries^b	Natalizumab 50%	T25FW 9HPT	nCGMV nWGMV
Sprenger,⁶⁰ 2020	7/93/0	0	74	37 (18–55)	2.5 (0–5.5)	7.7 (±6.5) 8.9 (±7.0)	Multiple countries^c	Teriflunomide 67%	None	None
Pontillo,⁷³ 2020	0/100/0	0	70	34 (±8.5)	2 (1.5–4)	3.5 (±4.0)	Italy	IFNB-1a 98.8% Natalizumab 1.2%	None	PFHWC PICDC PTVWC P4VWC
Ghione,⁶⁷ 2020	0/81/19	2	76	45(±11) to 47(±10)	2.5(±2.5) to 3.5(±3.0)	13 (±9.5)	USA	DMT 79%	None	LVV
Zivadinov,⁷⁴ 2019	0/100/0	0	78	31 (±7.9)	2 (1–2.5)	4.9 (±5.2)	Czech Republic	Mean time on IFNB-1a 87.3 (±99.5) months over 10 years	None	PCVC PVVC
De Stefano,⁷⁵ 2018	0/100/0	0	68	39 (±10)	2.9 (±1.3) (0–⩽5)	8.7 (±7.4)	Multiple countries^d	Cladribine 67% Placebo 33%	None	None
Ciampi,⁶⁶ 2017	n/s	n/s	n/s	n/s	n/s	n/s	Spain	Natalizumab 100%	None	GMF WMF CTh
von Gumberz,⁶⁴ 2016	0/100/0	0	65	41 (±9.6)	1.6 (±1.4) (0–6)	7.5 (±7.7)	Germany	DMT 51%	None	nGMV, nWMV
Jeffery,⁷⁶ 2016	0/100/0	0	71	37 (±9)	2.1 (±1.2) to 2.7 (±1.3)	7.7 (±6) to 8.5 (±7)	Multiple countries^a	Fingolimod 100%	None	None
Sastre-Garriga,⁶³ 2015	0/95/5	0	63	35 (±8.3)	4 (1.5–6.5)	10.4 (±6.6)	Spain	Natalizumab 100%	None	None
Sormani,⁷⁷ 2015	0/100/0	0	64	37 (±8.7)	2 (0–5.5)	3.1 (0.1–29.1)	Multiple countries^a	Fingolimod 68%	None	None
Jacobsen,⁷⁸ 2014	0/77/24	10	68	42 (±9.5)	3.5 (0–6)	8.3 (±7.3)	Norway	IFNB-1a 11% GA 2% None 87%	None	nGMV, nWMV nCV nLVV nSDGMV
Uher,⁶⁵ 2014	100/0/0	0	66	27(±7.2) 30 (±8.2)	1.5 (0–3.5)	<0.33	Czech Republic	IFNB-1a 100%	None	PGMVC PCVC PSCDGMVCPTVC PHVC
Zivadinov,⁷⁹ 2013	0/100/0	0	78	29(±7.8) to 32(±7.9)	1.8 (±0.8) to 2 (±1.0)	4 (±4.3) to 6 (±5.8)	USA, CZ, Australia	IFNB-1a 33% IFNB-1a/AZA 32% IFNB-1a/AZA 35%	None	nGMV nWMV nLVV nCV nSDGMV

Australia, Belgium, Canada, CZ, Finland, France, Germany, Greece, Israel, Lithuania, NL, Poland, Russia, Slovakia, SA, Switzerland, Sweden, Turkey, UK.

Germany, Italy, NL, Spain, Switzerland, UK (MAGNIMS).

Austria, Canada, Chile, Czech Republic, Denmark, Estonia, Finland, France, Germany, Italy, NL, Norway, Poland, Portugal, Russian Federation, Sweden, Switzerland, Turkey, Ukraine, UK, USA.

Australia, Austria, Belgium, Brazil, Bulgaria, Croatia, Czech Republic, Denmark, Finland, France, Germany, Greece, Italy, Lithuania, Latvia, Morocco, NL, Poland, Portugal, Russia, Saudi Arabia, Serbia and Montenegro, Turkey, Ukraine, USA; n/s not stated.

AZA, azathioprine; CIS, clinically isolated syndrome; CTh, cortical thickness; DGM, deep grey matter; DMF, dimethyl fumarate; DMT, disease modifying therapy; EDSS, Expanded Disability Status Scale; GA, Glatiramer acetate; GMF, grey matter fraction; GMV, grey matter volume; 9HPT, 9 hole peg test; IFNB-1a, interferon beta-1a; IQR, interquartile range; LVV, lateral ventricle volume; MAGNIMS, Magnetic Resonance Imaging in Multiple Sclerosis; MS, multiple sclerosis; NAWMV, normal appearing white matter volume; nCGMV, normalized cortical grey matter volume; nCV, normalized cortical volume; nGMV, normalized GMV; nLVV, normalized LVV; nSDGMV, normalized sub-cortical deep grey matter volume (includes Caudate nucleus, Putamen, Globus pallidus, Thalamus, Hippocampus, Nucleus accumbens, Amygdala); nWGMV, normalized whole grey matter volume; nWMV, normalized white matter volume; PBVC, percentage of brain volume change; PCVC, percentage cortical volume change; PFHWC, percentage frontal horn width change; PGMVC, percentage grey matter volume change; PHVC, percentage hippocampus volume change; PICDC, percentage intercaudate distance change; PPMS, primary progressive multiple sclerosis; PSCDGMVC, percentage sub-cortical deep grey matter volume change; PTVC, percentage thalamic volume change; PTVWC, percentage third ventricle width change; P4VWC, percentage fourth ventricle width change; PVVC, percentage ventricular volume change; RRMS, relapsing-remitting multiple sclerosis; T25FW, timed 25 feet walk; WMF, white matter fraction; WMV, white matter volume.

Table 5.

Definitions of DP.

Label	Definition	Short DP description
1	B/L EDSS <5.5, DP = +1 point or more; B/L EDSS ⩾5.5, DP = +0.5 point or more	<5.5 DP = +1 ⩾5.5 DP = +0.5
2	B/L EDSS <6, DP = +1 point or more; B/L EDSS ⩾6, DP = +0.5 point or more	<6 DP = +1 ⩾6 DP = +0.5
3	all B/L EDSS, DP = +1 point or more	All +1
4	B/L EDSS = 0, DP = +1.5 point or more; B/L EDSS >0, DP = +1 point or more	0 DP = +1.5 >0 DP = +1
5	B/L EDSS <5, DP = +1 point or more; B/L EDSS ⩾5, DP = +0.5 point or more	<5 DP = +1 ⩾5 DP = +0.5
6	B/L EDSS = 0, DP = +1.5 point or more; B/L EDSS >0, DP = +1 point or more; B/L EDSS ⩾5, DP = +0.5 point or more	0 DP = +1.5 >0 DP = +1 ⩾5 DP = +0.5
7	B/L EDSS = 0, DP = +1.5 point or more; B/L EDSS >0, DP = +1 point or more; B/L EDSS ⩾5.5, DP = +0.5 point or more	0 DP = +1.5 >0 DP = +1 ⩾5.5 DP = +0.5
8	B/L EDSS = 0, DP = +1.5 point or more; B/L EDSS >0, DP = +1 point or more; B/L EDSS ⩾6.5, DP = +0.5 point or more	0 DP = +1.5 >0 DP = +1 ⩾6.5 DP = +0.5
9	B/L EDSS = 0, DP = +3 point or more; B/L EDSS >0, DP = +2 point or more	0 DP = +3 >0 DP = +2

DP, disability progression; EDSS, Expanded Disability Status Scale.

Methods and results of cross-sectional and follow-up studies

The methods and results for the 37 cross-sectional studies are described in Table 6. The patient characteristics are summarized in Table 7. Twenty-nine studies investigated cross-sectional associations between baseline BA measured by NBV, NBPV or WBV and baseline EDSS. Of these 29 studies, 26 observed a significant association (p < 0.05) but 2 of them had mixed results with different methods giving a different result.^56,80 The three studies failing to observe an association were small (n = 29, 38 and 95),^61,81,82 as were the studies with mixed results (n = 20 and 61).^56,80 The correlations measured by Spearman’s rho, Pearson’s r, Kendall’s tau were weak for all studies (0.25–0.5) and the coefficients from regression models tended to be small.⁸³

Table 6.

Description of analysis and outcomes of cross-sectional and follow up studies reporting associations between global brain volume and EDSS scores (raw EDSS score or categorized as DP).

Author year	Study design	Global BV measurement MRI software *BA definition	Statistical analysis method and cross-sectional association with EDSS	Significant cross-sectional association (p < 0.05	Disease progression definition (see Table 5)	Baseline BV and follow-up EDSS association	Significant follow up association	Follow up in years
Hidalgo de la Cruz,^84,85 2021 and 2022	Prospective cohort and cross sectional baseline analysis	NBV SIENAX	Baseline NBV correlated with baseline EDSS (Spearman’s rho = −0.29, p = 0.008)	Yes	Reaching EDSS = 3,4 or 6 at 1 year follow-up	No correlation between baseline NBV and DP by random forest regression. PBVL at 1 year = −0.46 (−1.01 to −0.06) with no significant change in EDSS, Wilcoxon signed rank test p = 0.7)	No	1
Bernabéu-Sanz,⁸⁶ 2021	Prospective cross-sectional study	NBPV BRAIM/SPM8	NBPV correlated with EDSS (Spearman’s rho or Pearson’s r = −0.47, p = 0.009)	Yes	n/a	n/a	n/a	n/a
Chiang,⁸⁰ 2021	Cross sectional analysis of prospectively recruited patients	NBV SIENAX	An atrophy-based functional network alteration model was strongly correlated (r = 0.68; p = 0.002) with EDSS. NBV not associated with EDSS r = 0.26, p = 0.31 adjusted for disease duration, age and sex	Mixed	n/a	n/a	n/a	n/a
Haider,^87,88 2021	30 year follow up of prospective cohort	NBPV Atlas-based segmentation	n/a	n/a	n/a EDSS score used	NBPV was not associated with 30 year EDSS β = 0.17 (−0.04 to 0.5) by linear regression adjusted for age and gender	No	30
Rocca,⁷¹ 2021	Cross sectional analysis of retrospective cohort	NBV SIENAX	Multivariable linear model found baseline NBV to be associated with baseline EDSS (β = −0.13, p = 0.001) adjusted for age, sex, site, treatment and phenotype	Yes	n/a	n/a	n/a	n/a
Bakshi,⁸⁹ 2020	Retrospective analysis of two merged prospective studies	NBPV SIENAX	Baseline NBPV correlated with baseline EDSS (Pearson’s r = −0.24, −0.36 to −0.10; p = 0.001)	Yes	n/a EDSS score used	Baseline NBPV correlated with EDSS at 5 years (Pearson’s r = −0.31, −0.43 to −0.18; p < 0.001) NBPV correlated with 5 year change in EDSS (Pearson’s r = −0.15, −0.28 to −0.01; p = 0.04); By multivariable linear regression adjusted for T2LV, age and sex β = −2.37 −4.7 to −0.02; p = 0.05.	Yes	5
Filippi,⁹⁰ 2020	Cross-sectional analysis of prospectively recruited patients	NBV Icobrain *BA defined relative to healthy controls	NBV significantly correlated with EDSS (Spearman’s rho = −0.30, p < 0.02)	Yes	n/a	n/a	n/a	n/a
Hänninen,⁹¹ 2020	Retrospective analysis of clinical cohort	NBPV cNeuro/SIENAX/SIENA *BA defined relative to OASIS cohort⁸³	NBPV used to categorize into zero brain atrophy and has atrophy. At baseline zero atrophy group had significantly higher EDSS (2, 0–6.5) compared to atrophy group (4, 2–6), Wilcoxon rank-sum test p = 0.04	Yes	0 DP = +1.5 >0 DP = +1 ⩾5.5 DP = +0.5	5 year DP more likely in people with WB atrophy relative to no atrophy at baseline. OR (maximum likelihood estimate) = 4.3, 95% CI 1.1–16.1)	Yes	5
Vidal-Jordana,⁹² 2020	Cross-sectional analysis of a longitudinal and prospectively recruited ongoing cohort of patients	NBPV SPM8	Multivariable logistic regression adjusted for MRI scanner found NBPV associated with categorized EDSS < 3 vs ⩾3 EDSS at baseline (partial Pearson’s r = −0.47; p < 0.001)	Yes	n/a	n/a	n/a	n/a
Beadnall,⁵⁹ 2019	Cross-sectional analysis of prospective cohort	NBV SIENAX Icobrain	Baseline NBV by SIENAX associated with baseline EDSS, Kendall’s τ = −0.15, p = 0.04; and by Icobrain (Kendall’s τ = −0.15, p = 0.03)	Yes	n/a EDSS score used	Baseline NBV by SIENAX associated with follow-up EDSS, Kendall’s τ = −0.269, p < 0.001); and by Icobrain; Kendall’s τ = −0.236, p = 0.001	Yes	1.05
D’hooghe,⁹³ 2019	Retrospective record review	NBV MSmetrix	Baseline NBV correlated with baseline EDSS (Pearson’s r = −0.46 p ⩽ 0.01)	Yes	0 DP = +1.5 >0 DP = +1 ⩾5.5 DP = +0.5	Kaplan–Meier survival analysis shows no significant difference in time to DP between quartiles of baseline NBV, p = 0.15	No	3
Hänninen,⁹⁴ 2019	Repeated cross-sectional analysis of prospective cohort at baseline and 2 years	NBPV cNeuro/SIENA/SIENAX	Baseline NBPV correlated with baseline EDSS (Pearson’s r = −0.387, p = 0.009)	Yes	n/a EDSS score used	Baseline NBPV correlated with 2 year follow-up EDSS (Pearson’s r = −0.354, p = 0.016)	Maybe	2
Megna,⁹⁵ 2019	Retrospective record review	NBPV SPM8	Ordinal regression analysis adjusted for age, sex and disease duration showed borderline association of baseline NBPV with baseline EDSS (β = 0.09, p = 0.013)	Yes	n/a	n/a	n/a	n/a
Rusz,⁹⁶ 2019	Cross-sectional analysis of previous study data	NBPV Freesurfer	Baseline NBPV correlated with baseline EDSS (Spearman’s rho = −0.45, p < 0.001)	Yes	n/a	n/a	n/a	n/a
Hemond,⁵⁶ 2018	Prospective cohort	SIENAX-NBV SPM8-NBPV	Baseline NBPV (SPM8) correlated with baseline EDSS (Pearson’s r = −0.29, p = 0.02; adjusted for age and gender r = −0.26, p = 0.04) but not significant for NBV (SIENAX, Pearson’s r = −0.22, p = 0.08; adjusted for age and gender r = −0.19, p = 0.15)	Maybe	n/a	n/a	n/a	n/a
Eshaghi,²⁷ 2018	Retrospective analysis of prospective cohort	NBV Freesurfer	n/s	n/s	0 DP = +1.5 >0 DP = +1 ⩾6.5 DP = +0.5	Cox regression found baseline NBV was not associated with DP at 2.4 year follow-up.	No	2.41
Lee,⁸¹ 2018	Cross-sectional	NBV Freesurfer	t-tests and multiple regression did not find a significant correlation between baseline NBV and baseline EDSS	No	n/a	n/a	n/a	n/a
Nakamura,⁹⁷ 2018	Cross-sectional analysis of pooled baseline data from phase II trials	NBV SIENAX	Baseline NBV correlated with baseline EDSS in Japanese (Spearman’s rho = −0.33, p = 0.002) and Caucasians (Spearman’s rho = −0.48, p < 0.0001)	Yes	n/a	n/a	n/a	n/a
Rasche,⁹⁸ 2018	Post hoc cross-sectional analysis of 3 observational studies	NBV SIENAX	Linear regression adjusted for age and sex found baseline NBV to be associated with baseline EDSS for RRMS patients (β = −0.164, p < 0.0001)	Yes	n/a	n/a	n/a	n/a
Ruggieri,⁹⁹ 2018	Cross-sectional analysis of 2 combined clinical studies	NBV SIENAX	Baseline NBV correlated with baseline EDSS (Pearson’s r = −0.26, p < 0.01)	Yes	n/a	n/a	n/a	n/a
Burgetova,¹⁰⁰ 2017	Cross-sectional	NBPV ScanView	Baseline NBPV associated with baseline EDSS for RRMS patients (Spearman’s rho = 0.281, p = 0.01); No association for SPMS (Spearman’s rho = −0.001, p = 0.995)	Yes	n/a	n/a	n/a	n/a
D’Ambrosio,⁸² 2017	Cross-sectional clinical study	NBV SIENAX	Baseline NBV not correlated with baseline EDSS (Spearman’s rho = −0.01, p = 0.9)	No	n/a	n/a	n/a	n/a
Damjanovic,¹⁰¹ 2017	Cross-sectional analysis of MAGNIMS cohort	NBV SIENAX	Baseline NBV was correlated with baseline EDSS by mixed-effect model adjusted for age and site (β = −0.33, p = 0.01)	Yes	n/a	n/a	n/a	n/a
Hynčicová,¹⁰² 2017	Prospective cross-sectional	NBPV SIENAX	Baseline NBPV correlated with baseline EDSS (Spearman’s rho = −0.38, p = 0.01)	Yes	n/a	n/a	n/a	n/a
Khan,¹⁰³ 2017	Recruited from open-label extension trial for cross-sectional study at 20 year follow-up	WBV SPM5	Baseline WBV significantly correlated with baseline EDSS (Spearman’s rho = −0.514, p < 0.05; univariate logistic regression models using EDSS cut-point of 2 showed OR = 0.19 (0.06, 0.64), p = 0.008)	Yes	n/a	n/a	n/a	n/a
Moccia,¹⁰⁴ 2017	Retrospective cohort	NCSFV. Increase in NCSFV indicates decrease in NBPV SPM8	n/a	n/a	10 year change in EDSS Reaching EDSS = 4	Ordinal logistic regression showed baseline NCSFV was not associated with 10 year change in EDSS, unadjusted p = 0.81; adjusted p = 0.69 (age, gender, disease duration, baseline EDSS, annualized relapse rate, MRI protocol and treatment). Cox regression showed NCSFV not associated with reaching EDSS = 4	No	10
Uher,¹⁰⁵ 2017	Retrospective analysis extension of clinical trial	NBPV SIENAX	n/a	n/a	Primary 0 DP = +1.5 >0 DP = +1 Secondary 0 DP = +3 >0 DP = +2	Cox proportional hazard models adjusted for age, gender and disease duration found baseline NBPV significantly associated with DP at 12 years, HR = 0.90, p = 0.05). PBVL from baseline to year 1 was not associated with DP at 12 years (HR = 0.87, p = 0.16)	Mixed	12.2
Chu,¹⁰⁶ 2016	Prospective cohort	NBPV SIENAX	Baseline NBPV correlated with baseline EDSS, Spearman’s rho = −0.43, p = 0.027 using field strength 1.5T and rho = −0.49, p = 0.011 using 3T	Yes	n/a	n/a	n/a	n/a
Daams,¹⁰⁷ 2015	Cross-sectional analysis of prospective cohort	NBV SIENAX	Multivariate general linear model adjusted for age and sex showed baseline NBV significantly associated with baseline EDSS −0.26, p < 0.001	Yes	n/a	n/a	n/a	n/a
Granberg,¹⁰⁸ 2015	Cross sectional analysis of prospective cohort	BV Freesurfer	Baseline BV correlated with baseline EDSS (Spearman’s rho = −0.45, p = 0.001) adjusted for disease duration, age and sex	Yes	n/a	n/a	n/a	n/a
Granberg,¹⁰⁹ 2015	Cross-sectional and follow-up study	NBPV Freesurfer	Baseline NBPV had borderline association with baseline EDSS (Spearman’s rho = −0.320, p = 0.017) adjusted for disease duration, age and sex	Yes	n/a EDSS score used	Baseline NBPV not associated EDSS at 9-year follow-up (r = −0.33, p = 0.2) adjusted for disease duration, age and sex	No	17
Radue,⁵⁸ 2015	Post hoc cross-sectional baseline analysis of 3 clinical trial populations	NBV SIENAX	Baseline NBV significantly associated with baseline EDSS. Pearson’s r = −0.2 to −0.35 (3 cohorts separately), p < 0.001	Yes	n/a	n/a	n/a	n/a
Lavorgna,¹¹⁰ 2014	Longitudinal follow-up of cross-sectional study	NCSFV. Increase in NCSFV indicates decrease in NBPV n/s	n/a	n/a	<6 DP = +1 ⩾6 DP = +0.5 Reach EDSS = 4 Time to reach EDSS = 4	Logistic regression showed NCSFV was not a significant predictor of DP. Mann–Whitney U test showed higher NCSFV at baseline was associated with reaching EDSS = 4 (16% vs 13.7%; p < 0.001). Cox regression found no significant association between NCSFV and time to reach EDSS = 4	Mixed	9
Maghzi,⁶¹ 2014	Post hoc cross-sectional analysis of trial data	NBPV SIENAX	Baseline NBPV not correlated with baseline EDSS, Spearman’s rho = 0.07, p = 0.68	No	n/a EDSS score used	Baseline NBPV not correlated with 3 year EDSS, Spearman’s rho = −0.002, p = 0.73	No	3
Filippi,¹¹¹ 2013	Follow-up of observational cohort	NBPV n/s	n/a	n/a	<6 DP = +1 ⩾6 DP = +0.5	Baseline NBPV not associated with 13 year DP by univariate logistic regression. CS 82% vs CW 81%, p = 0.72	No	13.3
Vågberg,¹¹² 2013	Retrospective cross-sectional analysis of clinical data	NBPV SyMap	Baseline NBPV associated with baseline EDSS Pearson’s r = 0.32, p < 0.001). Significance remained after linear regression adjusted for age and disease duration	Yes	n/a	n/a	n/a	n/a
Zivadinov,⁷⁹ 2013	Prospective follow-up study	NBV SIENAX	n/s	n/s	0 DP = +1.5 >0 DP = +1	Baseline NBV significantly associated with 5-year DP (1489 vs 1524 ml, Student’s t-test, p = 0.003)	Yes	5

BPV, brain parenchymal volume; BV, brain volume; CS, clinically stable; CW, clinically worsened; DP, disability progression; EDSS, Expanded Disability Status Scale; HR, hazard ratio; MRI, magnetic resonance imaging; NBPV, Normalized brain parenchymal volume; NBV, normalized brain volume; NCSFV, normalized cerebrospinal fluid volume; OASIS, Open Access Series of Imaging Studies; PBVL, percentage brain volume loss; SIENAX, Structural Image Evaluation, using Normalization of Atrophy-Cross-Sectional; WBV, whole brain volume.

Table 7.

Patient characteristics for cross-sectional and follow-up studies reporting associations between global brain volume and EDSS scores (raw EDSS score or categorized as DP) and other study data collected including treatment, measures of physical disability and regional brain atrophy.

Author, year	%CIS/RRMS/PMS	% PPMS	Sample size (MS only)	%female	Mean age (baseline)	EDSS (baseline)	Disease duration, years (baseline)	Country	Treatment at baseline	Other reported measures of physical disability (extra to EDSS)	Regional measures of brain atrophy reported
Hidalgo de la Cruz^84,85 2021 and 2022	0/70/30	14	86	59	45 (±12.7)	3 (1.5–5.5)	13 (6.3–19)	Italy, Switzerland	GA/IFNB-1a/DMF 40% NAT/FTY/TFM 25% AZA/CYC/MTX/RTX 9% No DMT 26%	None	nGMV CTh
Bernabéu-Sanz,⁸⁶ 2021	0/100/0	0	30	70	41 (24–63)	2 (0–2.5)	9.5 (±6.3)	Spain	IFNB-1a 47% GA 17% DMF 17% NAT 10% FTY 3% None 7%	None	GMV WMV Putamen vol Caudate vol Thalamic vol Hippocampal vol
Chiang,⁸⁰ 2021	0/100/0	0	20	60	36 (±9)	4 (1–8)	7.9 (±5.5)	USA	n/s	None	nGMV nWMV
Haider,^87,88 2021	33/43/24	0	63	68	61 (±7.1)	n/a 30-year follow-up	31 (±1)	UK	DMT 14% No DMT 86%	T25FW 9HPT	TVW MEDW
Rocca,⁷¹ 2021	9/57/35	11	398	59	47 (18–77)	3.5 (0–8.5)	14.1 (0.1–46)	NL, UK, Spain, Germany, Italy, Switzerland	DMT 61% No DMT 39%	None	nGMV nWMV
Bakshi,⁸⁹ 2020	0/100/0	0	200	74	40 (±8.4)	1.2 (±1.1) (0–5.5)	7.7 (±6.8)	USA	n/s	None	None
Filippi,⁹⁰ 2020	0/100/0	0	56	68	41 (±11.2)	3.25 (±2.5) (1–5.5)	10 (±8.5)	Slovakia	IFNB-1a 21% GA 25% TFM 9% DMF 5% FTY 14% NAT 23% ALEM 2%	None	nGMV nWMV
Hänninen,⁹¹ 2020	0/41/59	0	59	68	41 (±10) to 53 (±8)	2–4 (0–6.5)	5 (±6) to 20 (±11)	Finland	DMT 49% No DMT 51%	None	GMV WMV Putamen vol Hippocampal vol Thalamic vol nucleus Caudatus vol Globus pallidus vol
Vidal-Jordana,⁹² 2020	0/100/0	0	80	76	38 (±10.2)	2 (±1.5)	8.1 (±8.8)	Spain	DMT 47% No DMT 53%	None	WMF GMF
Beadnall,⁵⁹ 2019	1/97/2	0	102	80	38 (±9)	2 (IQR 1.88)	7.4 (±7.4)	Australia	DMT 72%	None	None
D’hooghe,⁹³ 2019	0/90/10	<10	470	68	46 (±11)	4.2 (±2.1) (0–9)	14.6 (±9.9)	Belgium	Immunotherapy 82%	None	nGMV nWMV nDGMV nCGMV
Hänninen,⁹⁴ 2019	0/43/57	0	60	68	36 (±7.5)/53 (±7.3)	1.2/4.5 (0–6.5)	0.8/20 (0.7–35.7)	Finland	DMT 58%	None	GMV WMV Putamen vol Hippocampal vol Thalamic vol nucleus Caudatus vol Globus pallidus vol
Megna,⁹⁵ 2019	0/100/0	0	241	60	34 (±8)	2.4 (±0.8) (1–5)	4.1 (±5)	Italy	IFNB-1a 61%	None	GMV
Rusz,⁹⁶ 2019	2/79/19	7	123	75	44 (±11)	3.6 (±1.4) (1–6.5)	14.4 (±7.6)	Czech Republic	n/s	9HPT T25FW	WMF, GMF
Hemond,⁵⁶ 2018	7/84/10	2	61	69	41 (±8.6)	1.6 (±1.7)	8.3 (±7.2)	USA	DMT 80%	T25FW	None
Eshaghi,²⁷ 2018	21/58/21	10	1214	64	33–48 CIS 33 ± 8, RRMS 38.2 ± 9. SPMS 48.2 ± 9.8, PPMS 48.5 ± 10.1	0–6 CIS 1 (0–4.5), RRMS 2 (0–7), SPMS 6 (2.5–9.0), PPMS 5 (2–8)	1–15, CIS 0.4 ± 1.4, RRMS 6.7 ± 7.3, SPMS 15.6 ± 9.9, PPMS 6.8 ± 5.9	Europe^a	DMT CIS 20% DMT RRMS 49% DMT SPMS 41% DMT PPMS 6%	None	CGM, GMV, DGMV, GMV of temporal, parietal, occipital, frontal lobes, cerebellum
Lee,⁸¹ 2018	0/100/0	0	29	59	41 ± 9.4	2 (IQR 1.5–4)	6 (IQR 2.5–11)	China	Immunotherapy 71%	None	nGMV nWMV NBsV
Nakamura,⁹⁷ 2018	0/100/0	0	341	72	34 ± 8	2 ± 1.7	7.4 ± 6.4	Canada, Japan, Europe^b	DMT or Immunotherapy 57%	None	nCGMV nDGMV nTV nWMV
Rasche,⁹⁸ 2018	0/100/0	0	59	60	CIS 34 ± 7, RRMS 38 ± 10	1.5 (0–3)	CIS 2.3 ± 1.4; RRMS 4.1 ± 2.7	Germany	n/s	9HPT SMWS	nGMV, nWMV nTV
Ruggieri,⁹⁹ 2018	n/s	n/s	96	67	42 ± 11	3 (1–6)	8.8 ± 4.6	Italy	n/s	Postural sway	None
Burgetova,¹⁰⁰ 2017	0/83/17	17	144	60	46–49 ± 7	2.5 ± 2.5	12–13 ± 11	Czech Republic	n/s	None	CCV
D’Ambrosio,⁸² 2017	0/76/24	0	95	60	45 ± 11	2.5 (1–8)	14.6 (0–42)	Italy	n/s	9HPT	nGMV nWMV
Damjanovic,¹⁰¹ 2017	0/100/0	0	62	65	40 ± 9	2 (0–6)	6–12 7 centres	Europe^a	n/s	None	nGMV, nCGMV nWMV
Hynčicová,¹⁰² 2017	100/0/0	0	51	57	32 ± 9	1.6 ± 0.6	0.5 (0–0.9)	Czech Republic	n/s	None	nGMV nWMV
Khan,¹⁰³ 2017	0/100/0	0	39	67	56 ± 6	3.6 ± 2.5	27 ± 4.7	USA	GA 100%	AI	CTh GMV WMV
Moccia,¹⁰⁴ 2017	0/100/0	0	149	62	33 ± 7.2	1.8 ± 0.4 (1–3)	4.2 ± 2.8	Italy	DMT 100%	None	GM:NAWMV ratio GMf WMf NAWMf
Uher,¹⁰⁵ 2017	0/100/0	0	177	78	31 ± 8	1.9 ± 0.9 (0–3.5)	5 ± 5.2	Czech Republic	IFNB-1a 100% NAT/FTY/RTX/ALEM 25%	None	GMf WMf Tf CCf
Chu,¹⁰⁶ 2016	ns/⩽85/15e	⩽15	26	69	43 (21–55)	2 ± 2 (0–6.5)	9.6 ± 8.6	USA	n/s	T25FW	None
Daams,¹⁰⁷ 2015	0/64/36	11	195	67	53 ± 9.6	4 (1–8)	20 ± 7	NL	n/s	9HPT T25FW MSWS	nGMV nWMV nBstV CTh
Granberg,¹⁰⁸ 2015	0/43/57	8	37	70	42 ± 10	4.5 (0–8)	11 ± 8.5	Sweden	DMT 65%	None	GMf WMf CCI CCA CCV
Granberg,¹⁰⁹ 2015	13/10/0	0	23	78	42 ± 10	4.5 (0–8)	11 ± 8.5	Sweden	DMT 65%	None	GMf WMf nCCA
Radue,⁵⁸ 2015	0/100/0	0	3635	n/s	n/s	n/s	n/s	Multiple countries^c	n/s	None	None
Lavorgna,¹¹⁰ 2014	0/100/0	0	241	67	45 (22–77)	2.3 (0–4.5)	8.5 (1–35)	Italy	n/s	None	GMf WMf change
Maghzi,⁶¹ 2014	CIS/RRMS n/s	0	38	71	36 ± 9	2 (0–5.5)	7.5 ± 4.9	USA	Riluzole 50% IFNB-1a 100%	9HPT T25FW	nGMV nNAWMV
Filippi,¹¹¹ 2013	29/44/27	0	48	63	25–38	2.5 (IQR 1–4.5) (1–5)	4–5	Italy	Immunotherapy 100%	None	GMf WMf Tf
Vågberg,¹¹² 2013	0/64/36	10	99	69	18–64	2.5 (0–8)	9	Sweden	n/s	None	None
Zivadinov,⁷⁹ 2013	0/100/0	0	180	78	29 (±7.8) to 32 (±7.9)	1.8 (±0.8) to 2 (±1.0)	4 (±4.3) to 6 (±5.8)	USA, CZ, Australia	IFNB-1a 33% IFNB-1a/AZA 32% IFNB-1a/AZA 35%	None	nGMV, nWMV nLVV nCV nSDGMV

Germany, Italy, NL, Spain, Switzerland, UK (MAGNIMS).

Denmark, Finland, France, Germany, Italy, Poland, Portugal, Spain, Switzerland, UK.

Argentina, Australia, Austria, Belgium, Brazil, Canada, Egypt, France, Germany, Greece, Hungary, Italy, Republic of Korea, Portugal, Spain, Switzerland, UK, USA.

AI, ambulation index; ALEM, alemtuzumab; AZA, azathioprine; CCA, corpus callosum area; CCf, corpus callosum fraction; CCI, corpus callosum index; CCV, corpus callosum volume; CGM, cortical grey matter; CIS, clinically isolated syndrome; CTh, cortical thickness; CYC, ‎cyclophosphamide; DGM, deep grey matter (thalamus, putamen, globus pallidus, caudate and amygdala); DMF, dimethyl fumarate; DMT, disease modifying therapy; EDSS, Expanded Disability Status Scale; FTY, fingolimod; GA, Glatiramer acetate; GMf, grey matter fraction; GMV, grey matter volume; 9HPT, 9 hole peg test; IFNB-1a, interferon beta-1a; IQR, interquartile range; LVV, lateral ventricle volume; MAGNIMS, Magnetic Resonance Imaging in Multiple Sclerosis; MEDW, medullary width; MS, multiple sclerosis; MSWS, MS waking scale; MTX, methotrexate; NAT, natalizumab; NAWMV, normal appearing white matter volume; nBstV, normalized brain stem volume (includes Midbrain, Pons, Medulla oblongata); nCGMV, normalized cortical grey matter volume; nCV, normalized cortical volume; nDGMV, normalized deep grey matter volume; nGMV, normalized GMV; nLVV, normalized LVV; nSDGMV, normalized sub-cortical deep grey matter volume (includes Caudate nucleus, Putamen, Globus pallidus, Thalamus, Hippocampus, Nucleus accumbens, Amygdala); nWMV, normalized white matter volume; PPMS, primary progressive multiple sclerosis; RRMS, relapsing-remitting multiple sclerosis; RTX, rituximab; Tf, thalamic fraction; TFM, teriflunomide; TV, Thalamic volume; SPMS, secondary progressive multiple sclerosis; TVW, third ventricle width; T25FW, timed 25 feet walk; WMf, white matter fraction; WMV, white matter volume.

Fifteen studies investigated the association between baseline BA and follow-up EDSS or DP. Of these 15 studies, 5 observed a significant association (p < 0.05),^{59,79,89,91,94} 8 did not and 2 had mixed results according to the DP outcome definition¹⁰⁵ or the follow-up time.¹¹⁰

Sixteen of the 37 cross-sectional studies reported NBV (a registration-based method), 17 studies reported NBPV (a proportion-based method) and 2 studies reported BV or whole BV but were unclear if this was normalized.^103,108 Two studies reported normalized cerebrospinal fluid volume (NCSFV) which is directly related to NBPV. NBPV is the sum(WMV&GMV)/sum(WMV, GMV&CSFV); therefore, as CSFV increases NBPV decreases. Two studies defined BA as part of the analysis, one referenced to healthy controls⁹⁰ and the other to the Open Access Series of Imaging Studies cohort.^91,113

The majority of cross-sectional studies used SIENAX (20), followed by SPM5 or SPM8 (6), Freesurfer (5), Icobrain (2), ScanView (1), SyMap (1), MSmetrix (1), BRAIM (1), cNeuro (1) and 2 did not state the image analysis software used. Some studies used more than one type of image analysis software. The majority of the study populations reported treatment with DMT (23/37, 62%) but many studies did not report treatments (14/37, 38%). Most studies also reported measures of regional atrophy (30/37, 81%). Some studies reported additional measures of physical disability, specifically 9HPT, T25FW, SMWS, postural sway and ambulation index (10/37, 27%). Of the 29 cross-sectional studies investigating baseline associations, 2 dichotomized EDSS scores and 27 used EDSS score directly. Nine studies used regression analysis, and the most frequently used measure of correlation was Spearman’s rho (12/29) or Pearson’s r (8/29). Many studies adjusted for combinations of age, gender, disease duration, phenotype and scanner (11/29). Of the 15 follow-up studies 9 categorized EDSS scores for analysis, 6 according to DP (definition 2, 4, 7–9 in Table 5) and 3 in other ways, such as reaching EDSS = 4, or time to reach EDSS = 4. Ten follow-up studies used regression or survival analysis adjusting for combinations of age, gender, T2 lesion volume and disease duration.

Summary of longitudinal findings

The analytical approaches varied between studies making between study comparisons difficult other than the observation of an association, or not. However, for those studies amenable to comparisons, that is, reported associations between annual or annualized PBVC and EDSS, the outcomes were consistent as listed below (Table 3).

0.13% BVL per year associated with a 1 point increase in EDSS.¹¹

0.4% BVL per year distinguished patients with zero mean annual change in EDSS from those with a mean annual EDSS increase of 0.14.⁴⁹

1.0% BVL per year in patients with DP versus 0.28% without DP.⁷¹

1.1% BVL per year in patients with DP versus 0.8% without DP or 0.7% with disability improvement.⁶⁷

1.26% BVL per year with DP and 0.19% without DP.⁶⁶

Discussion

Main findings

Our main finding is that the majority (19/23) of longitudinal studies observed a significant association (p < 0.05) between PBVL and change in EDSS or DP defined by EDSS score. Looking more closely at the four studies failing to observe an association between PBVL and change in EDSS or DP, all were inconclusive, possibly as a consequence of being underpowered. Our second main finding is that the majority of cross-sectional studies observed a significant association between NBV or NBPV and EDSS scores (26/29), although two studies reported mixed results.^56,80 The three studies failing to observe any association between baseline NBV^81,82 or NBPV⁶² and EDSS were all small studies and potentially underpowered.

As far as we are aware, the previous reviews on this topic were published in 2016 but only cover the literature up to 2013.^23,48 These reviews both observe a consistent association, but there is a range of approaches to measuring regional and global atrophy as well as a range of disability outcome measures. A substantial body of literature has grown over the last decade, and this review provides increasing evidence of a consistent association between PBVL and change in EDSS or DP, and BA measured by NBV or NBPV and EDSS scores, that is, between objective and clinically meaningful measurements. Furthermore, this review has made some progress in quantifying the magnitude of the association between PBVL and EDSS in PwMS. Two prospective studies observed a consistent and significant association between PBVL and DP (−3% at 4 years and −3.8% 5 years)^65,78 and three retrospective analyses also had similar findings, for example, 1.0%,⁷¹ 1.1%⁶⁷ and 1.26%⁶⁶ annual PBVC with DP.

Implication of multiple definitions of DP

Most studies were predominantly RRMS diagnoses (due to studies with >20% PPMS patients being excluded) and typically had mean EDSS scores around 2. The range of baseline EDSS scores was 0–9. The variation in definition of DP (Table 5) may be a substantial source of bias for this review. (However, the 8 longitudinal studies reporting change in EDSS and the 29 cross-sectional studies reporting baseline EDSS would not be affected by this bias.) For example, 8/15 longitudinal studies using DP included people with EDSS of zero at baseline yet 5/15 studies defined DP as a 1.5 point change from baseline, while 10/15 defined DP as a 1 point change. At other end of the EDSS range 10/15 studies varied the baseline EDSS score between ⩾5, ⩾5.5 or ⩾6, where a 0.5 point increase rather than a 1 point increase defined DP. Patients with EDSS score below 5 are fully ambulatory and the functional systems scores are the main determinants of EDSS. Given that ambulation is the main determinant of change in EDSS when scores are above 5, and scores are less sensitive to change in patients with higher EDSS scores⁵³, categorizing EDSS into DP might be a more useful outcome measure than change in EDSS. However, if the definitions are not aligned it will lead to misclassification and difficulty in comparing studies. Altogether, there were nine definitions of DP in the papers included in this review and future studies would benefit from a universally agreed definition of DP. The European Medicines Agency recommends 1 point on the EDSS scale with a baseline EDSS score less than or equal to 5.5 and 0.5 points in an EDSS score over 5.5.⁵³

Statistical methods and covariates

Most studies reported PBVL over a time period rather than annualized. This is probably because BVL doesn’t progress linearly,¹¹⁴ nonetheless over a short time period it is a helpful way to compare study outcomes. Some statistical analyses (Tables 3 and 6) assumed a linear relationship but not all. Most studies used models that adjusted for various combinations of sex, age, education, disease phenotype, disease duration, study cohort, lesion volume, DMT or other treatment and MRI field strength. We hypothesize that DMTs mostly have an impact on PBVL or NBV or NBPV which in turn affects EDSS or DP rather than directly affecting EDSS via another mechanism. This is speculative but as none of the studies have consistent usage of DMT or other treatments, we have to be pragmatic and suggest that the best approach is to report associations both with, and without adjustment, for treatments. If our hypothesis is correct and treatments are affecting EDSS mainly, or solely, through modifying BVL then models adjusted for treatment or unadjusted will provide similar results. Most studies reported both univariate and multivariate associations, and these were generally in agreement (Tables 3 and 6). Ideally a study would adjust for variables associated with EDSS, change in EDSS or DP but not variables solely affecting PBVL, NBV or NBPV.

Regional atrophy considered in the studies

We only considered global BA in this review but 44/58 included studies also considered regional atrophy, often as the main focus of the study, plus a further 39 studies were identified in screening that only addressed regional atrophy. The quantity of studies identified by the literature search pointed to the need for a separate review to consider the association between EDSS and regional BA. Different regions of the brain show different degrees of atrophy relative to global BA.¹¹⁵

The most frequently reported regional measures were the corpus callosum area or volume, cortical thickness, grey matter (GM) and deep GM volumes, white matter volume, lateral ventricle volumes, thalamic volumes and third ventricle width (Tables 3 and 6). BA in the deep GM can be detected earlier than whole BA⁶² and the mid-sagittal cross sectional area of the corpus callosum seems to be particularly affected by MS.^116
–118 Atrophy in the thalamus appears to proceed faster than in other brain regions.^27,98,119 Volume loss in the GM is more related to neuro-axonal loss and neuronal shrinkage than demyelination.^120,121 This may make GM matter changes less susceptible to confounding that might occur when using PBVL, NBV or NBPV due to active inflammation, vasogenic oedema, dehydration and gliosis, or decreased by anti-inflammatory drugs.^{12,24,122,123} All the studies in this review addressed potential confounding by pseudo-atrophy by excluding PwMS with recent relapse or use of anti-inflammatory drugs and observing changes over at least 1 year.

Baseline BA as a predictor of future DP

Most studies found no association between baseline NBV or NBPV and DP at follow-up. NBV or NBPV at a single point in time may reflect the prior BA rate but the future atrophy rate may differ. Many studies adjusted for age, gender and disease duration, but disability and BA at follow-up may be also influenced by treatment and baseline EDSS.

Limitations of the review

Although we find evidence of a consistent association between BVL and EDSS, we are unable to compile an estimated effect size because of the wide range of methodological approaches. None of the longitudinal studies had the primary aim of measuring the association between BVL and EDSS. Furthermore, most studies were retrospective analyses of data collected for other purposes. Quality is generally evaluated with respect to the primary outcome of the study. The best way to estimate the effect size using these data would be a meta-analysis of the original data. We could potentially divide the studies into groups, where DP happened to be equivalent by coincidence of the definition and the baseline EDSS, but this would make several small groups and is unlikely to be meaningful.

Conclusion

Our review of 58 studies consistently finds an association between PVBL, NBV or NBPV and EDSS scores or DP defined by EDSS score. It supports the MAGNIMS recommendation that global BA should be included as a secondary outcome in therapeutic clinical trials.⁵¹ We further recommend that a consensus on the definition of DP is agreed to allow for more accurate quantification of the magnitude of the relationship between DP defined by EDSS scores and MS going forward. Baseline BVL or NBPV does not appear to reliably predict future DP or EDSS score.

Footnotes

Appendix

Acknowledgements

Medical writing and editorial support was provided by OPEN Health.

Declarations

ORCID iDs

Robert Zivadinov

Alexander Keenan

Susan Jill Stocks

References

Browne

Chandraratna

Angood

, et al. Atlas of Multiple Sclerosis 2013: a growing global problem with widespread inequity. Neurology 2014; 83: 1022–1024.

Richards

Sampson

Beard

, et al. A review of the natural history and epidemiology of multiple sclerosis: implications for resource allocation and health economic models. Health Technol Assess 2002; 6(10): 1–73.

Kister

Bacon

Chamot

, et al. Natural history of multiple sclerosis symptoms. Int J MS Care 2013; 15(3): 146–158.

Brownell

Hughes

JT.

The distribution of plaques in the cerebrum in multiple sclerosis. J Neurol Neurosurg Psychiatry 1962; 25: 315–320.

Losseff

Wang

Lai

, et al. Progressive cerebral atrophy in multiple sclerosis. A serial MRI study. Brain 1996; 119(Pt 6): 2009–2019.

Rudick

Fisher

Lee

, et al. Use of the brain parenchymal fraction to measure whole brain atrophy in relapsing-remitting MS. Multiple Sclerosis Collaborative Research Group. Neurology 1999; 53(8): 1698–1704.

Vollmer

Signorovitch

Huynh

, et al. The natural history of brain volume loss among patients with multiple sclerosis: a systematic literature review and meta-analysis. J Neurol Sci 2015; 357(1–2): 8–18.

Giovannoni

Butzkueven

Dhib-Jalbut

, et al. Brain health: time matters in multiple sclerosis. Mult Scler Relat Disord 2016; 9: S5–S48.

Lycklama à Nijeholt

. Reduction of brain volume in MS. MRI and pathology findings. J Neurol Sci 2005; 233(1–2): 199–202.

10.

Bermel

Bakshi

The measurement and clinical relevance of brain atrophy in multiple sclerosis. Lancet Neurol 2006; 5(2): 158–170.

11.

Ghione

Bergsland

Dwyer

, et al. Brain atrophy is associated with disability progression in patients with MS followed in a clinical routine. AJNR Am J Neuroradiol 2018; 39(12): 2237–2242.

12.

Riley

Azevedo

Bailey

, et al. Clinical applications of imaging disease burden in multiple sclerosis: MRI and advanced imaging techniques. Expert Rev Neurother 2012; 12(3): 323–333.

13.

Brex

Jenkins

Fox

, et al. Detection of ventricular enlargement in patients at the earliest clinical stage of MS. Neurology 2000; 54(8): 1689–1691.

14.

Dalton

Brex

Jenkins

, et al. Progressive ventricular enlargement in patients with clinically isolated syndromes is associated with the early development of multiple sclerosis. J Neurol Neurosurg Psychiatry 2002; 73(2): 141–147.

15.

Dalton

Chard

Davies

, et al. Early development of multiple sclerosis is associated with progressive grey matter atrophy in patients presenting with clinically isolated syndromes. Brain 2004; 127(Pt 5): 1101–1107.

16.

Gasperini

Paolillo

Giugni

, et al. MRI brain volume changes in relapsing-remitting multiple sclerosis patients treated with interferon beta-1a. Mult Scler 2002; 8(2): 119–123.

17.

Ingle

GT.

Primary progressive multiple sclerosis: a 5-year clinical and MR study. Brain 2003; 126(11): 2528–2536.

18.

Molyneux

Kappos

Polman

, et al. The effect of interferon beta-1b treatment on MRI measures of cerebral atrophy in secondary progressive multiple sclerosis. European Study Group on Interferon beta-1b in secondary progressive multiple sclerosis. Brain 2000; 123(11): 2256–2263.

19.

Coles

Wing

Molyneux

, et al. Monoclonal antibody treatment exposes three mechanisms underlying the clinical course of multiple sclerosis. Ann Neurol 1999; 46(3): 296–304.

20.

Zivadinov

Sepcic

Nasuelli

, et al. A longitudinal study of brain atrophy and cognitive disturbances in the early phase of relapsing-remitting multiple sclerosis. J Neurol Neurosurg Psychiatry 2001; 70(6): 773–780.

21.

Fisher

Rudick

Cutter

, et al. Relationship between brain atrophy and disability: an 8-year follow-up study of multiple sclerosis patients. Mult Scler 2000; 6(6): 373–377.

22.

Andravizou

Dardiotis

Artemiadis

, et al. Brain atrophy in multiple sclerosis: mechanisms, clinical relevance and treatment options. Auto Immun Highlights 2019; 10(1): 7.

23.

Vollmer

Huynh

Kelley

, et al. Relationship between brain volume loss and cognitive outcomes among patients with multiple sclerosis: a systematic literature review. Neurol Sci 2016; 37(2): 165–179.

24.

Anderson

Fox

Miller

DH.

Magnetic resonance imaging measures of brain atrophy in multiple sclerosis. J Magn Reson Imaging 2006; 23(5): 605–618.

25.

Filippi

Evidence for widespread axonal damage at the earliest clinical stage of multiple sclerosis. Brain 2003; 126(2): 433–437.

26.

Filippi

Rovaris

Inglese

, et al. Interferon beta-1a for brain tissue loss in patients at presentation with syndromes suggestive of multiple sclerosis: a randomised, double-blind, placebo-controlled trial. Lancet 2004; 364(9444): 1489–1496.

27.

Eshaghi

Prados

Brownlee

, et al. Deep gray matter volume loss drives disability worsening in multiple sclerosis: deep gray matter volume loss. Ann Neurol 2018; 83(2): 210–222.

28.

Artemiadis

Anagnostouli

Zalonis

, et al. Structural MRI correlates of cognitive function in multiple sclerosis. Mult Scler Relat Disord 2018; 21: 1–8.

29.

Benedict

RHB

Weinstock-Guttman

Fishman

, et al. Prediction of neuropsychological impairment in multiple sclerosis: comparison of conventional magnetic resonance imaging measures of atrophy and lesion burden. Arch Neurol 2004; 61(2): 226.

30.

Benedict

RHB

Bruce

Dwyer

, et al. Neocortical atrophy, third ventricular width, and cognitive dysfunction in multiple sclerosis. Arch Neurol 2006; 63(9): 1301.

31.

Benedict

RHB

Ramasamy

Munschauer

, et al. Memory impairment in multiple sclerosis: correlation with deep grey matter and mesial temporal atrophy. J Neurol Neurosurg Psychiatry 2009; 80(2): 201–206.

32.

Calabrese

Agosta

Rinaldi

, et al. Cortical lesions and atrophy associated with cognitive impairment in relapsing-remitting multiple sclerosis. Arch Neurol 2009; 66(9): 1144–1150.

33.

Eijlers

AJC

Dekker

Steenwijk

, et al. Cortical atrophy accelerates as cognitive decline worsens in multiple sclerosis. Neurology 2019; 93(14): e1348–e1359.

34.

Rudick

Fisher

Lee

, et al. Brain atrophy in relapsing multiple sclerosis: relationship to relapses, EDSS, and treatment with interferon β-1a. Mult Scler 2000; 6(6): 365–372.

35.

Engl

Tiemann

Grahl

, et al. Cognitive impairment in early MS: contribution of white matter lesions, deep grey matter atrophy, and cortical atrophy. J Neurol 2020; 267(8): 2307–2318.

36.

Furby

Hayton

Altmann

, et al. A longitudinal study of MRI-detected atrophy in secondary progressive multiple sclerosis. J Neurol 2010; 257(9): 1508–1516.

37.

Furby

Hayton

Anderson

, et al. Magnetic resonance imaging measures of brain and spinal cord atrophy correlate with clinical impairment in secondary progressive multiple sclerosis. Mult Scler 2008; 14(8): 1068–1075.

38.

Hildebrandt

Hahn

Kraus

, et al. Memory performance in multiple sclerosis patients correlates with central brain atrophy. Mult Scler 2006; 12(4): 428–436.

39.

Jasperse

Vrenken

Sanz-Arigita

, et al. Regional brain atrophy development is related to specific aspects of clinical dysfunction in multiple sclerosis. Neuroimage 2007; 38(3): 529–537.

40.

Magon

Tsagkas

Gaetano

, et al. Volume loss in the deep gray matter and thalamic subnuclei: a longitudinal study on disability progression in multiple sclerosis. J Neurol 2020; 267(5): 1536–1546.

41.

Mineev

Prakhova

Il’ves

, et al. Characteristics of neurological and cognitive status in patients with multiple sclerosis in relation to the location and volumes of demyelination foci and the severity of brain atrophy. Neurosci Behav Physiol 2009; 39(1): 35–38.

42.

Mowry

Beheshtian

Waubant

, et al. Quality of life in multiple sclerosis is associated with lesion burden and brain volume measures. Neurology 2009; 72(20): 1760–1765.

43.

Janardhan

Bakshi

Quality of life and its relationship to brain lesions and atrophy on magnetic resonance images in 60 patients with multiple sclerosis. Arch Neurol 2000; 57(10): 1485–1491.

44.

Marrie

Fisher

Miller

, et al. Association of fatigue and brain atrophy in multiple sclerosis. J Neurol Sci 2005; 228(2): 161–166.

45.

Tauhid

Chu

Sasane

, et al. Brain MRI lesions and atrophy are associated with employment status in patients with multiple sclerosis. J Neurol 2015; 262(11): 2425–2432.

46.

Jacobsen

Zivadinov

Myhr

, et al. Brain atrophy and employment in multiple sclerosis patients: a 10-year follow-up study. Mult Scler J Exp Transl Clin 2020; 6(1): 2055217320902481.

47.

Rao

Martin

Huelin

, et al. Correlations between MRI and information processing speed in MS: a meta-analysis. Mult Scler Int 2014; 2014: 975803.

48.

Rojas

Patrucco

Miguez

, et al. Brain atrophy in multiple sclerosis: therapeutic, cognitive and clinical impact. Arq Neuropsiquiatr 2016; 74(3): 235–243.

49.

De Stefano

Airas

Grigoriadis

, et al. Clinical relevance of brain volume measures in multiple sclerosis. CNS Drugs 2014; 28(2): 147–156.

50.

Grassiot

Desgranges

Eustache

, et al. Quantification and clinical relevance of brain atrophy in multiple sclerosis: a review. J Neurol 2009; 256(9): 1397–1412.

51.

Sastre-Garriga

Pareto

Battaglini

, et al.; MAGNIMS Study Group. MAGNIMS consensus recommendations on the use of brain and spinal cord atrophy measures in clinical practice. Nat Rev Neurol 2020; 16(3): 171–182.

52.

Kurtzke

JF.

Rating neurologic impairment in multiple sclerosis: an Expanded Disability Status Scale (EDSS). Neurology 1983; 33(11): 1444–1452.

53.

Meyer-Moock

Feng

Maeurer

, et al. Systematic literature review and validity evaluation of the Expanded Disability Status Scale (EDSS) and the Multiple Sclerosis Functional Composite (MSFC) in patients with multiple sclerosis. BMC Neurol 2014; 14(1): 58.

54.

Smith

Zhang

Jenkinson

, et al. Accurate, robust, and automated longitudinal and cross-sectional brain change analysis. Neuroimage 2002; 17(1): 479–489.

55.

Despotović

Goossens

Philips

MRI segmentation of the human brain: challenges, methods, and applications. Comput Math Methods Med 2015; 2015: 450341.

56.

Hemond

Chu

Tummala

, et al. Whole-brain atrophy assessed by proportional- versus registration-based pipelines from 3T MRI in multiple sclerosis. Brain Behav 2018; 8(8): e01068.

57.

Moher

Shamseer

Clarke

, et al.; PRISMA-P Group. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement. Syst Rev 2015; 4(1): 1.

58.

Radue

Barkhof

Kappos

, et al. Correlation between brain volume loss and clinical and MRI outcomes in multiple sclerosis. Neurology 2015; 84(8): 784–793.

59.

Beadnall

Wang

Van Hecke

, et al. Comparing longitudinal brain atrophy measurement techniques in a real-world multiple sclerosis clinical practice cohort: towards clinical integration? Ther Adv Neurol Disord 2019; 12: 1756286418823462.

60.

Sprenger

Kappos

Radue

, et al. Association of brain volume loss and long-term disability outcomes in patients with multiple sclerosis treated with teriflunomide. Mult Scler 2020; 26(10): 1207–1216.

61.

Maghzi

Revirajan

Julian

, et al. Magnetic resonance imaging correlates of clinical outcomes in early multiple sclerosis. Mult Scler Relat Disord 2014; 3(6): 720–727.

62.

Chu

Kim

Tauhid

, et al. Whole brain and deep gray matter atrophy detection over 5 years with 3T MRI in multiple sclerosis using a variety of automated segmentation pipelines. PLoS One 2018; 13(11): e0206939.

63.

Sastre-Garriga

Tur

Pareto

, et al. Brain atrophy in natalizumab-treated patients: a 3-year follow-up. Mult Scler 2015; 21(6): 749–756.

64.

von Gumberz

Mahmoudi

Young

, et al. Short-term MRI measurements as predictors of EDSS progression in relapsing-remitting multiple sclerosis: grey matter atrophy but not lesions are predictive in a real-life setting. PeerJ 2016; 4: e2442.

65.

Uher

Horakova

Bergsland

, et al. MRI correlates of disability progression in patients with CIS over 48 months. Neuroimage Clin 2014; 6: 312–319.

66.

Ciampi

Pareto

Sastre-Garriga

, et al. Grey matter atrophy is associated with disability increase in natalizumab-treated patients. Mult Scler 2017; 23(4): 556–566.

67.

Ghione

Bergsland

Dwyer

, et al. Disability improvement is associated with less brain atrophy development in multiple sclerosis. AJNR Am J Neuroradiol 2020; 41(9): 1577–1583.

68.

Dekker

Eijlers

AJC

Popescu

, et al. Predicting clinical progression in multiple sclerosis after 6 and 12 years. Eur J Neurol 2019; 26(6): 893–902.

69.

Uher

Krasensky

Malpas

, et al. Evolution of brain volume loss rates in early stages of multiple sclerosis. Neurol Neuroimmunol Neuroinflamm 2021; 8(3): e979.

70.

De Stefano

Stromillo

Giorgio

, et al. Establishing pathological cut-offs of brain atrophy rates in multiple sclerosis. J Neurol Neurosurg Psychiatry 2016; 87(1): 93–99.

71.

Rocca

Valsasina

Meani

, et al. Association of gray matter atrophy patterns with clinical phenotype and progression in multiple sclerosis. Neurology 2021; 96(11): e1561–e1573.

72.

Koch

Mostert

Repovic

, et al. MRI brain volume loss, lesion burden, and clinical outcome in secondary progressive multiple sclerosis. Mult Scler 2022; 28(4): 561–572.

73.

Pontillo

Cocozza

Di Stasi

, et al. 2D linear measures of ventricular enlargement may be relevant markers of brain atrophy and long-term disability progression in multiple sclerosis. Eur Radiol 2020; 30(7): 3813–3822.

74.

Zivadinov

Horakova

Bergsland

, et al. A serial 10-year follow-up study of atrophied brain lesion volume and disability progression in patients with relapsing-remitting MS. AJNR Am J Neuroradiol 2019; 40(3): 446–452.

75.

De Stefano

Giorgio

Battaglini

, et al. Reduced brain atrophy rates are associated with lower risk of disability progression in patients with relapsing multiple sclerosis treated with cladribine tablets. Mult Scler 2018; 24(2): 222–226.

76.

Jeffery

Di Cantogno

Ritter

, et al. The relationship between the rate of brain volume loss during first 24 months and disability progression over 24 and 48 months in relapsing MS. J Neurol 2016; 263(2): 299–305.

77.

Sormani

De Stefano

Francis

, et al. Fingolimod effect on brain volume loss independently contributes to its effect on disability. Mult Scler 2015; 21(7): 916–924.

78.

Jacobsen

Hagemeier

Myhr

, et al. Brain atrophy and disability progression in multiple sclerosis patients: a 10-year follow-up study. J Neurol Neurosurg Psychiatry 2014; 85(10): 1109–1115.

79.

Zivadinov

Bergsland

Dolezal

, et al. Evolution of cortical and thalamus atrophy and disability progression in early relapsing-remitting MS during 5 years. Am J Neuroradiol 2013; 34(10): 1931–1939.

80.

Chiang

Feng

Romero

, et al. Disruption of the atrophy-based functional network in multiple sclerosis is associated with clinical disability: validation of a meta-analytic model in resting-state functional MRI. Radiology 2021; 299(1): 159–166.

81.

Lee

Mak

HKF

Chiu

, et al. Differential brainstem atrophy patterns in multiple sclerosis and neuromyelitis optica spectrum disorders. J Magn Reson Imaging 2018; 47(6): 1601–1609.

82.

D’Ambrosio

Pagani

Riccitelli

, et al. Cerebellar contribution to motor and cognitive performance in multiple sclerosis: an MRI sub-regional volumetric analysis. Mult Scler 2017; 23(9): 1194–1203.

83.

Akoglu

User’s guide to correlation coefficients. Turk J Emerg Med 2018; 18(3): 91–93.

84.

Hidalgo de la Cruz

Valsasina

Gobbi

, et al. Longitudinal cortical thinning progression differs across multiple sclerosis phenotypes and is clinically relevant: a multicentre study. Mult Scler 2021; 27(6): 827–840.

85.

Hidalgo de la Cruz

Valsasina

Meani

, et al. Differential association of cortical, subcortical and spinal cord damage with multiple sclerosis disability milestones: a multiparametric MRI study. Mult Scler 2022; 28(3): 406–417.

86.

Bernabéu-Sanz

Morales

Naranjo

, et al. Contribution of gray matter atrophy and white matter damage to cognitive impairment in mildly disabled relapsing-remitting multiple sclerosis patients. Diagnostics 2021; 11(3): 578.

87.

Haider

Chung

Birch

, et al. Linear brain atrophy measures in multiple sclerosis and clinically isolated syndromes: a 30-year follow-up. J Neurol Neurosurg Psychiatry 2021; 92(8): 839–846.

88.

Haider

Prados

Chung

, et al. Cortical involvement determines impairment 30 years after a clinically isolated syndrome. Brain 2021; 144(5): 1384–1395.

89.

Bakshi

Healy

Dupuy

, et al. Brain MRI predicts worsening multiple sclerosis disability over 5 years in the SUMMIT study. J Neuroimaging 2020; 30(2): 212–218.

90.

Filippi

Vestenická

Siarnik

, et al. Neurofilament light chain and MRI volume parameters as markers of neurodegeneration in multiple sclerosis. Neuro Endocrinol Lett 2020; 41(1): 17–26.

91.

Hänninen

Viitala

Paavilainen

, et al. Thalamic atrophy predicts 5-year disability progression in multiple sclerosis. Front Neurol 2020; 11: 606.

92.

Vidal-Jordana

Pareto

Cabello

, et al. Optical coherence tomography measures correlate with brain and spinal cord atrophy and multiple sclerosis disease-related disability. Eur J Neurol 2020; 27(11): 2225–2232.

93.

D’hooghe

Gielen

Van Remoortel

, et al. Single MRI-based volumetric assessment in clinical practice is associated with MS-related disability: MRI brain volume and MS disability. J Magn Reson Imaging 2019; 49(5): 1312–1321.

94.

Hänninen

Viitala

Paavilainen

, et al. Thalamic atrophy without whole brain atrophy is associated with absence of 2-year NEDA in multiple sclerosis. Front Neurol 2019; 10: 459.

95.

Megna

Alfano

Lanzillo

, et al. Brain tissue volumes and relaxation rates in multiple sclerosis: implications for cognitive impairment. J Neurol 2019; 266(2): 361–368.

96.

Rusz

Vaneckova

Benova

, et al. Brain volumetric correlates of dysarthria in multiple sclerosis. Brain Lang 2019; 194: 58–64.

97.

Nakamura

Gaetano

Matsushita

, et al. A comparison of brain magnetic resonance imaging lesions in multiple sclerosis by race with reference to disability progression. J Neuroinflammation 2018; 15(1): 255.

98.

Rasche

Scheel

Otte

, et al. MRI markers and functional performance in patients with CIS and MS: a cross-sectional study. Front Neurol 2018; 9: 718.

99.

Ruggieri

Fanelli

Castelli

, et al. Lesion symptom map of cognitive–postural interference in multiple sclerosis. Mult Scler 2018; 24(5): 653–662.

100.

Burgetova

Dusek

Vaneckova

, et al. Thalamic iron differentiates primary-progressive and relapsing-remitting multiple sclerosis. AJNR Am J Neuroradiol 2017; 38(6): 1079–1086.

101.

Damjanovic

Valsasina

Rocca

, et al. Hippocampal and deep gray matter nuclei atrophy is relevant for explaining cognitive impairment in MS: a multicenter study. AJNR Am J Neuroradiol 2017; 38(1): 18–24.

102.

Hynčicová

Vyhnálek

Kalina

, et al. Cognitive impairment and structural brain changes in patients with clinically isolated syndrome at high risk for multiple sclerosis. J Neurol 2017; 264(3): 482–493.

103.

Khan

Seraji-Bozorgzad

Bao

, et al. The relationship between brain MR spectroscopy and disability in multiple sclerosis: 20-year data from the U.S. Glatiramer Acetate Extension Study. J Neuroimaging 2017; 27(1): 97–106.

104.

Moccia

Quarantelli

Lanzillo

, et al. Grey:white matter ratio at diagnosis and the risk of 10-year multiple sclerosis progression. Eur J Neurol 2017; 24(1): 195–204.

105.

Uher

Vaneckova

Sobisek

, et al. Combining clinical and magnetic resonance imaging markers enhances prediction of 12-year disability in multiple sclerosis. Mult Scler 2017; 23(1): 51–61.

106.

Chu

Tauhid

Glanz

, et al. Whole brain volume measured from 1.5T versus 3T MRI in healthy subjects and patients with multiple sclerosis: brain volume in MS from 1.5T versus 3T. J Neuroimaging 2016; 26(1): 62–67.

107.

Daams

Steenwijk

Wattjes

, et al. Unraveling the neuroimaging predictors for motor dysfunction in long-standing multiple sclerosis. Neurology 2015; 85(3): 248–255.

108.

Granberg

Bergendal

Shams

, et al. MRI-defined corpus callosal atrophy in multiple sclerosis: a comparison of volumetric measurements, corpus callosum area and index: corpus callosum area, a time-effective biomarker in MS. J Neuroimaging 2015; 25(6): 996–1001.

109.

Granberg

Martola

Bergendal

, et al. Corpus callosum atrophy is strongly associated with cognitive impairment in multiple sclerosis: results of a 17-year longitudinal study. Mult Scler 2015; 21(9): 1151–1158.

110.

Lavorgna

Bonavita

Ippolito

, et al. Clinical and magnetic resonance imaging predictors of disease progression in multiple sclerosis: a nine-year follow-up study. Mult Scler 2014; 20(2): 220–226.

111.

Filippi

Preziosa

Copetti

, et al. Gray matter damage predicts the accumulation of disability 13 years later in MS. Neurology 2013; 81(20): 1759–1767.

112.

Vågberg

Lindqvist

Ambarki

, et al. Automated determination of brain parenchymal fraction in multiple sclerosis. AJNR Am J Neuroradiol 2013; 34(3): 498–504.

113.

LaMontagne

Benzinger

TLS

Morris

, et al. OASIS-3: longitudinal neuroimaging, clinical, and cognitive dataset for normal aging and Alzheimer disease. Radiol Imaging 2019.

114.

Andorra

Nakamura

Lampert

, et al. Assessing biological and methodological aspects of brain volume loss in multiple sclerosis. JAMA Neurol 2018; 75(10): 1246.

115.

Fujimori

Fujihara

Ogawa

, et al. Patterns of regional brain volume loss in multiple sclerosis: a cluster analysis. J Neurol 2020; 267(2): 395–405.

116.

Turner

Ramli

Blumhardt

, et al. Ventricular enlargement in multiple sclerosis: a comparison of three-dimensional and linear MRI estimates. Neuroradiology 2001; 43(8): 608–614.

117.

Liu

Edwards

Gong

, et al. Three dimensional MRI estimates of brain and spinal cord atrophy in multiple sclerosis. J Neurol Neurosurg Psychiatry 1999; 66(3): 323–330.

118.

Pelletier

Suchet

Witjas

, et al. A longitudinal study of callosal atrophy and interhemispheric dysfunction in relapsing-remitting multiple sclerosis. Arch Neurol 2001; 58(1): 105–111.

119.

Ontaneda

Raza

Mahajan

, et al. Deep grey matter injury in multiple sclerosis: a NAIMS consensus statement. Brain 2021; 144(7): 1974–1984.

120.

von Bartheld

Bahney

Herculano-Houzel

The search for true numbers of neurons and glial cells in the human brain: a review of 150 years of cell counting. J Comp Neurol 2016; 524(18): 3865–3895.

121.

Popescu

Klaver

Voorn

, et al. What drives MRI-measured cortical atrophy in multiple sclerosis? Mult Scler 2015; 21(10): 1280–1290.

122.

Zivadinov

Reder

Filippi

, et al. Mechanisms of action of disease-modifying agents and brain volume changes in multiple sclerosis. Neurology 2008; 71(2): 136–144.

123.

Nakamura

Brown

Araujo

, et al. Correlation between brain volume change and T2 relaxation time induced by dehydration and rehydration: implications for monitoring atrophy in clinical studies. Neuroimage Clin 2014; 6: 166–170.