Gray Matter Atrophy Is Related to Long-Term Disability in Multiple Sclerosis

Leonora K. Fisniku, MRCP,1,2 Declan T. Chard, PhD,1,2 Jonathan S. Jackson, MSci,1,2

Valerie M. Anderson, BSci,1,2 Daniel R. Altmann, PhD,1,3 Katherine A. Miszkiel, MRCP,4Alan J. Thompson, PhD,1,5 and David H. Miller, MD1,2

Objective: To determine the relation of gray matter (GM) and white matter (WM) brain volumes, and WM lesion load, withclinical outcomes 20 years after first presentation with clinically isolated syndrome suggestive of multiple sclerosis (MS).Methods: Seventy-three patients were studied a mean of 20 years from first presentation with a clinically isolated syndrome (33of whom developed relapsing-remitting MS and 11 secondary-progressive MS, with the rest experiencing no further definiteneurological events), together with 25 healthy control subjects. GM and WM volumetric measures were obtained from three-dimensional T1-weighted brain magnetic resonance images using Statistical Parametric Mapping 2.Results: Significant GM (p � 0.001) and WM atrophy (p � 0.001) was seen in MS patients compared with control subjects.There was significantly more GM, but not WM atrophy, in secondary-progressive MS versus relapsing-remitting MS (p �0.003), and relapsing-remitting MS versus clinically isolated syndrome (p � 0.001). GM, but not WM, fraction correlated withexpanded disability status scale (rs � �0.48; p � 0.001) and MS Functional Composite scores (rs � 0.59; p � 0.001). WMlesion load correlated with GM (rs � �0.63; p � 0.001), but not with WM fraction. Regression modeling indicated that theGM fraction explained more of the variability in clinical measures than did WM lesion load.Interpretation: In MS patients with a relatively long and homogeneous disease duration, GM atrophy is more marked thanWM atrophy, and reflects disease subtype and disability to a greater extent than WM atrophy or lesions.

Ann Neurol 2008;64:247–254

Magnetic resonance imaging (MRI)–detectable whitematter (WM) lesions are usually seen early in relapseonset multiple sclerosis (MS), and in people who de-velop a clinically isolated syndrome (CIS) suggestive ofMS, they are associated with conversion to clinicallydefinite MS,1,2 although they predict subsequent dis-ability only to a limited degree.3,4

Brain atrophy is also seen from clinical disease onsetin MS5; it is prominent in the later stages of the disease,and is more marked in secondary progressive (SP) com-pared with relapsing-remitting (RR) phenotypes ofMS,6,7 although the relative influence of disease pheno-type and disease duration on such atrophy is uncertain.

From pathological studies extensive cortical damagehas been observed predominantly in progressive formsof MS, suggesting that GM pathology may be an im-portant determinant of irreversible disability.8 Al-though whole-brain atrophy has been well explored,the advent of new MRI acquisition and analysis toolsnow makes it possible to determine the relative extent

of both GM and WM atrophy. Recent work investi-gating the progression of tissue-specific atrophy, mea-sured using methods based on Statistical ParametricMapping (SPM) segmentations, after first presenta-tion with a CIS showed significantly greater GMcompared with WM atrophy in those patients whodeveloped clinically definite MS within 3 years.9 Fur-thermore, in patients with early RRMS, GM atrophyover 2 years was more rapidly progressive than WMatrophy.10 These studies suggest that progressive GMatrophy occurs early in the clinical course of MS, andin the case of CIS, is of direct and immediate clinicalrelevance. Although some studies have detected pre-dominantly GM atrophy,9,11–14 not all have; indeed,some observed mostly WM atrophy,15 and it remainsto be definitively determined which tissue is most af-fected at any given stage of the disease, particularly inthe longer term.

The relation between WM lesions and brain atrophyalso remains unclear, with current evidence suggesting

From the 1Nuclear Magnetic Resonance Research Unit; 2Depart-ment of Neuroinflammation, Institute of Neurology, UniversityCollege London; 3Department of Neuroradiology, National Hospi-tal for Neurology and Neurosurgery; 4Department of Brain Repairand Rehabilitation, Institute of Neurology, University College Lon-don; and 5Medical Statistical Unit, London School of Hygiene andTropical Medicine, London, United Kingdom.Received Mar 4, 2008, and in revised form Apr 9, 2008. Acceptedfor publication Apr 11, 2008.

Published online in Wiley InterScience (www.interscience.wiley.com).DOI: 10.1002/ana.21423

Address correspondence to Dr Fisniku, NMR Research Unit, Insti-tute of Neurology, Queen Square, London WC1N 3BG, UnitedKingdom. E-mail: l.fisniku@ion.ucl.ac.uk

ORIGINAL ARTICLE

© 2008 American Neurological Association 247Published by Wiley-Liss, Inc., through Wiley Subscription Services

a partial discordance between these pathological mani-festations of MS16 both in cross-sectional and longitu-dinal studies.9,11,12,17–21 The suggestion that there is atleast a partial discordance between T2 lesion load(T2LL) and atrophy measures during the evolution ofMS is supported by the observation that, althoughdisease-modifying therapies such as �-interferon arerelatively effective in preventing new WM lesion for-mation, their effect in reducing atrophy has been mod-est,22,23 and in some studies, not evident at all.24,25

With this context, the primary objective of this studywas to estimate GM and WM volumes in a cohort ofCIS patients followed-up 20 years from clinical diseaseonset, and to assess the relation between these measuresof tissue-specific atrophy, clinical course, and disability,in particular, investigating the hypothesis that GM atro-phy will correlate better with clinical disease severity.Secondary objectives were as follows: (1) to evaluate therelation of GM and WM volumes with T2LL, and (2)to investigate the relative contributions of GM and WMvolumes and T2LL to disability.

MethodsSubjectsThis report is based on 20-year follow-up data of a cohortwho had clinical and MRI assessments at approximately5-yearly intervals after presenting with a CIS suggestive ofMS.3,4 Clinical status was documented at the 20-yearfollow-up in 107 patients,4 of whom 75 had an MRI exam-ination, with data from two patients excluded (one who de-veloped cerebrovascular disease and one who did not com-plete the scanning protocol). The remaining 73 patients arethe subject of this report.

Clinically definite MS was diagnosed on clinical groundsalone.26 Disability was assessed using the expanded disabil-ity status scale (EDSS)27 and MS functional composite(MSFC) scores.28 The clinical course of MS (RRMS orSPMS) was defined by Lublin and Reingold criteria.29

Those clinically definite MS patients with an EDSS � 3were defined as benign MS. Patients were studied a mean(standard deviation [SD]) of 20 [1.5] range, (18 –27) yearsafter the CIS (49 women and 24 men; mean age, 51.4[7.2] years); 29 were still classified as CIS (mean diseaseduration, 20.4 [2.06] years; mean age, 51.5 [8.4] years), 33had developed RRMS (mean disease duration, 19.7 [1.1]years; mean age, 51 [6.1] years) and 11 SPMS (mean dis-ease duration, 19.8 [0.68] years; mean age, 52 [7.3] years).The median EDSS was 2.5 (range, 0 – 8) for all patientsand 3.25 (range, 1– 8) for MS patients only. Three patientswere receiving disease-modifying treatments. MRI was alsoperformed in 25 healthy control subjects (14 women and11 men; mean age, 41.7 [7.7] years).

The study was approved by the National Hospital forNeurology and Neurosurgery and Institute of NeurologyJoint Research Ethics Committee. All study participants gavewritten informed consent.

Image Acquisitions and ProcessingWhole-brain MRI was performed on a 1.5-Tesla GE Signascanner (General Electric, Milwaukee, WI) as follows: (1) two-dimensional, dual-echo proton density (TE, 17 milliseconds)and T2 (TE, 103 milliseconds)-weighted fast spin-echo (repe-tition time [TR], 2,000 milliseconds; 28 � 5mm slices; fieldof view, 24 � 18cm; in-plane resolution of 1.1 mm); and (2)three-dimensional, axial, T1-weighted, inversion-prepared, fastspoiled gradient recall (TR, 10.9 milliseconds; TE, 4.2 milli-seconds; inversion time, 450 milliseconds; 124 � 1.5mm slic-es; imaging matrix, 256 � 160, interpolated to a final in-planeresolution of 1.1mm). An experienced neuroradiologist(K.A.M.), blinded to clinical details, identified lesions on hardcopies of the proton density–weighted images, with referenceto the T2-weighted images. This was then used as a referencefor contouring of the lesions on the proton density–weighteddigital images, using a semiautomated local thresholding tech-nique implemented in the image display package DispImage(Plummer, Department of Medical Physics and Bioengineer-ing, University College London, London, United Kingdom).30

Then a computer program summed all the individual lesionvolumes (calculated as surface area of each lesion multiplied byslice thickness), and T2LLs were generated.

Segmentation of the axial, three-dimensional, T1-weightedimages into WM, GM, and cerebrospinal fluid was performedusing SPM2 (Statistical Parametric Mapping; Wellcome De-partment of Cognitive Neurology, Institute of Neurology,London), following a previously described method11 (softwareavailable free to the research community at www.nmrgroup.io-n.ucl.ac.uk/atrophy). The processing parameters for SPM2were set to 0.01 for the bias correction and 30 for the biascutoff. WM and GM fraction volumes (GMF) relative to totalintracranial volume were derived, corrected for lesion misclas-sification as GM.11 The tissue masks were inspected by anexperienced operator, and no significant segmentation errorswere detected.

To assess the robustness of results obtained using SPM2,we reprocessed our data using SIENAX (Structural ImageEvaluation, using Normalization, of Atrophy for cross-sectional measurement), a fully automated technique, toobtain the normalized GM and WM volumes.31 SIENAXmethodology and results are provided in an Appendix.

Statistical AnalysesGroup comparisons of the brain tissue volumes were per-formed using linear regression with group indicator and ageand sex covariates. To assess the associations between thebrain volume measurements, T2LL, and disability (EDSSand MSFC and its components), we used Spearman’s rankcorrelation.

To assess the relative contribution of the WM and GMvolume loss and T2LL to accrued disability, we used ordinallogistic regression (for EDSS) and linear regression (forMSFC). Both EDSS (categorized as follows: �1.5; �1.5,and �3; �3 and �6; �6) and MSFC (as a continuous vari-able) were modeled as response variables, with tissue vol-umes, lesion load, age, and sex as covariate predictors. Lesionload was log-transformed to improve normality before inclu-sion in the regression models; where the T2LL was zero (10subjects), the log volume was given a value of 0.01 to in-

248 Annals of Neurology Vol 64 No 3 September 2008

clude these subjects. Changing the EDSS category intervals,or the small value given for the log volume where the T2LLwas zero, did not materially change the results.

MRI covariates were entered together and removed singlyby manual backward stepwise exclusion until all model predic-tors were significant at p � 0.1. Age and sex were added tothe final models but omitted if the adjusted coefficients wereboth nonsignificant and not materially different from unad-justed coefficients. Models were applied to the whole cohort ofpatients and the MS subgroup separately.

The data were analyzed using SPSS 11 (SPSS, Chicago,IL) and Stata 9.2 (Stata Corporation, College Station, TX).Statistical significance was taken at p � 0.05.

ResultsTissue-Specific Volumes and Clinical SubgroupsTissue-specific volumes were significantly lower in MSpatients and MS subgroups (RRMS and SPMS) versuscontrol subjects (Tables 1 and 2). Significant GM andWM atrophy was seen in MS patients compared withcontrol subjects. There was significantly more GM at-rophy, but not WM atrophy, in SPMS versus RRMSand RRMS versus CIS. There was significantly greaterGM atrophy, but not WM atrophy, in those (nonbe-nign) MS patients with an EDSS � 3 (22 patients)compared with those (benign) MS patients with anEDSS � 3 (22 patients). There were no significant dif-ferences for any of the volume measurements betweenthe control subjects and those remaining classified as aCIS after first presentation.

Magnetic Resonance Imaging Measures and DisabilityGMF correlated significantly with EDSS and MSFCfor all patients and for the MS subgroup alone (Table3). WM fraction volumes showed no such correlations.T2LL also correlated with EDSS (rs � 0.49; p �0.001) and MSFC (rs � �0.53; p � 0.001) for allpatients, as well as in the MS subgroup (rs � 0.38,p � 0.009; and rs � �0.42, p � 0.005, respectively).

Correlations of Lesions with Gray and WhiteMatter VolumesT2LL correlated significantly with GMF (rs � �0.63;p � 0.001) but not with WM fraction volumes (rs ��0.15; p � 0.19) for the whole cohort of patients andfor the MS subgroup only (rs � �0.66, p � 0.001,and rs � �0.18, p � 0.22, respectively).

Predicting DisabilityFor the whole cohort of patients, only GMF and log-transformed T2LL independently predicted EDSS cat-egory, with GMF the stronger predictor: There was anestimated 64% (p � 0.001) reduction in the odds ofhaving greater disability per 1 SD greater GMF, and a52% odds reduction (p � 0.05) per 1 SD greater log-transformed T2LL.

Only GMF independently predicted disability as mea-sured by MSFC scores; there was an estimated 0.61 in-crease (p � 0.001) in MSFC per 1 SD greater GMF.

Restricting regression models to the MS subgroupof patients, only GMF independently predicted dis-ability, whether EDSS or MSFC: There was a 59%( p � 0.007) reduction in the odds of being in moresevere EDSS category per 1 SD greater GMF, andthere was a 0.67 increase ( p � 0.001) in MSFC per1 SD greater GMF.

The findings using SIENAX measured tissue vol-umes were similar to those obtained using SPM2 (seethe Appendix for further details).

DiscussionThis study builds on previous work,20,32,33 character-izing tissue-specific brain atrophy in a group of peo-ple with MS or CIS who have a uniquely long andhomogeneous disease duration (approximately 20years). It has allowed an exploration of the associa-tions and role as predictors of MRI measures, tissue-specific (GM and WM) atrophy and WM lesion load,with clinical phenotype and disability, relatively freeof confounding by variability in disease duration.

In this cohort of patients, both GM and WM at-rophy was seen in MS patients compared with controlsubjects, and the extent of GM atrophy was greaterthan that of WM atrophy in keeping with some pre-vious studies.9,11–14 Furthermore, there was signifi-cantly more GM, but not WM, atrophy in SPMSversus RRMS, and RRMS versus those remaining CIS

Table 1. Mean and Median (Standard Deviation) ofBrain Volume Measurements

Group (n) GMF Mean;Median (SD)

WMF Mean;Median (SD)

Controlsubjects(25)

0.51; 0.52 (0.01) 0.29; 0.29 (0.01)

All patients(73)

0.49; 0.49 (0.03) 0.28; 0.28 (0.01)

CIS (29) 0.50; 0.50 (0.02) 0.28; 0.28 (0.01)MS (44) 0.47; 0.48 (0.03) 0.28; 0.28 (0.01)RRMS (33) 0.48; 0.49 (0.02) 0.28; 0.28 (0.01)Benign MS(22)a

0.49; 0.49 (0.02) 0.28; 0.28 (0.01)

NonbenignMS (22)b

0.46; 0.46 (0.30) 0.28; 0.27 (0.01)

SPMS (11) 0.45; 0.45 (0.03) 0.27; 0.27 (0.01)aBenign multiple sclerosis (MS) � expanded disability statusscale (EDSS) � 3.bNonbenign MS � EDSS � 3.GMF � gray matter fraction; WMF � white matter fraction;CIS � clinically isolated syndrome; MS � multiple sclerosis(RRMS and SPMS); RRMS � relapsing-remitting MS;SPMS � secondary progressive MS.

Fisniku et al: GM Atrophy and Disability in MS 249

patients. It should be noted that GM atrophy has notbeen a universal finding in MS, and that a definitiveconsensus on the location and timing of brain atro-phy has yet to be reached; however, a significantnumber of recent studies suggest that GM atrophy isa consistent finding throughout the clinical course ofMS, seemingly mirroring clinical status.9,10,32,34 Theapparent discrepancy in some previous studies mayrepresent a combination of cohort-related and techni-cal factors. Although there is no universally acceptedgold standard method for measuring GM and WMvolumes, the SPM-based approach has provided con-sistent findings in several previous studies,9 –11,32 andin this study, the robustness of the results obtainedusing SPM-based methods have been consolidated bysimilar findings with another widely used segmenta-tion method (SIENAX technique; see Appendix).

Given the relatively homogeneous disease durationand age distribution of the clinical subgroups in-cluded in this work, the association of GM atrophywith clinical status is not explained by these factors;

rather, the findings suggest a direct link between GMatrophy and clinical disease severity.

Differential tissue-specific atrophy in MS may be par-tially explained by variable degrees of inflammatory ac-tivity in WM and GM,35,36 with relatively greater com-pensation of cell loss by inflammatory infiltrates andedema in WM compared with GM. Differential inflam-matory noise in the volumetric measures may also leadto greater attenuation of WM compared with GM asso-ciations with clinical parameters. However, it may be ex-pected that eventually atrophy, if progressive, wouldreach a magnitude where it would no longer be dis-guised by inflammatory interference; given this, ourobservations in MS patients with relatively long dis-ease duration suggest that WM atrophy is truly lessprogressive than that of GM, and not simply the re-sult of compensation by, and short-term fluctuationsassociated with, inflammation. In addition, althoughwe detected no clear evidence of an association be-tween WM atrophy and disability, 50% of MS pa-tients in this cohort had a benign clinical course, and

Table 3. Correlations of Brain Volume Measurements with Clinical Features

rs (p)

EDSS (n �73)a (44b)

MSFC (n � 67)a

(41b)Z-PEG (n � 70)a

(42b)Z-WALK (n � 68)a

(40b)Z-PASAT (n � 68)a

(42b)

GMFa �0.48 (<0.001) 0.56 (<0.001) 0.59 (<0.001) �0.40 (0.001) 0.27 (0.026)GMFb �0.41 (0.005) 0.55 (<0.001) 0.44 (0.003) �0.49 (0.001) 0.32 (0.038)WMFa �0.20 (0.086) 0.03 (0.784) 0.16 (0.176) �0.11 (0.337) �0.07 (0.537)WMFb �0.11 (0.443) 0.10 (0.526) 0.28 (0.071) �0.09 (0.560) �0.04 (0.761)

aAll patients.bMultiple sclerosis (MS) subgroup only.rs � Spearman’s rank correlation coefficient; EDSS � expanded disability status scale; MSFC � multiple sclerosis functional compositescore; GMF � gray matter fraction; WMF � white matter fraction.

Table 2. Age- and Sex-Adjusted Mean Difference between Patient Subgroups and Control Subjects

Group Comparisons GMF WMF

Adjusted Mean Difference (95% CI) p Adjusted Mean Difference (95% CI) p

MS-control subjects �0.027 (�0.041 to (�0.014)) <0.001 �0.009 (�0.017 to (�0.001)) 0.017MS-CIS �0.028 (�0.039 to (�0.017)) <0.001 �0.003 (�0.010 to 0.003) 0.318SPMS-control subjects �0.046 (�0.063 to (�0.028)) <0.001 �0.013 (�0.024 to (�0.002)) 0.018RRMS-control subjects �0.021 (�0.035 to (�0.008)) 0.002 �0.008 (�0.017 to (�0.001)) 0.042SPMS-CIS �0.046 (�0.062 to (�0.030)) 0.001 �0.006 (�0.016 to 0.003) 0.179RRMS-CIS �0.022 (�0.033 to (�0.010)) <0.001 �0.002 (�0.009 to 0.004) 0.540SPMS-RRMS �0.024 (�0.040 to (�0.008)) 0.003 �0.004 (�0.014 to 0.005) 0.361Benign-nonbenign MSa 0.022 (0.004–0.040) 0.01 0.004 (�0.004 to 0.013) 0.328CIS-control subjects 0.001 (�0.013 to 0.015) 0.089 �0.006 (�0.015 to 0.002) 0.142

Benign MS � expanded disability status scale (EDSS) � 3; nonbenign MS � EDSS � 3.GMF � gray matter fraction; WMF � white matter fraction; CI � confidence interval; MS � multiple sclerosis (RRMS and SPMS);CIS � clinically isolated syndrome; SPMS � secondary progressive MS; RRMS � relapsing-remitting MS.

250 Annals of Neurology Vol 64 No 3 September 2008

it is conceivable that larger cohort with more severedisability (eg, EDSS score � 7) might exhibit moreWM atrophy; further work is required to explore thispossibility. Considered overall, our findings suggestthat measures of GM atrophy will be more usefulthan WM volume in natural history studies or treat-ment trials, for example, in a study of potentiallyneuroprotective agents, although serial studies shouldfurther investigate the relation between longitudinalGM volume and clinical changes.

GM, but not WM volume measurements, corre-lated with clinical disability (EDSS, MSFC, and itscomponents). Although T2LL correlated significantlywith disability, GMF was a better predictor of disabil-ity when included in the regression models. Whereasnoting the caveats about inflammatory noise discussedpreviously, these data suggest that GM atrophy hasmore clinical relevance in the long term than eitherlesion load or WM atrophy in people with MS, beingmore closely related to long-term disability and clin-ical course. This study’s findings consolidate and ex-tend the observation made in several previous studiesthat MRI markers of GM involvement correlate morestrongly with measures of physical disability thanWM lesion load.8,32,33,37

The amount of tissue loss in MS probably representsa balance between several pathological processes: irre-versible neuronal and axonal loss, myelin loss, and re-versible neuroaxonal atrophy, on the one hand, withpartial compensation by inflammation-associated cellu-lar infiltrates, and cellular (including axonal38) and in-terstitial edema on the other. With regard to the mech-anisms of brain atrophy, there may be: (1) antegradeand retrograde neuroaxonal tract degeneration associ-ated with focal WM inflammatory lesions,39 with apotentially significant delay between axonal demyelina-tion and subsequent neuroaxonal degeneration; and(2) a more widespread process directly targeting neu-rons, myelin (including cortical demyelination35,36),and glia.

GM (but not WM) volume measurements correlatedwith WM lesion load, which is in keeping with otherstudies.9–11,32 This correlation may reflect secondary de-generation from WM lesions to GM. That the degree ofcorrelation is only moderate suggests that processes in-dependent of WM lesions are also contributing to GMatrophy in MS. One such explanation might be thatGM demyelinating lesions, although not visible on con-ventional MRI, are commonly found at autopsy.35,36

Our findings emphasize that further research to eluci-date pathogenic mechanisms in MS should focus onGM as well as WM pathology.

When considering the significance of the findings ob-served in our study, it is important to take into accounta few limitations. First, neither WM lesion volume nortissue-specific brain atrophy measurement is pathologi-

cally specific. WM lesions on T2-weighted MRI maycontain variable amounts of inflammation, demyelina-tion, edema, and axonal loss. The brain volume mea-surements, although affected by the same factors, arethought to be more specifically weighted toward neuro-degeneration. Second, our brain volume measurementdata are cross sectional and do not provide any directinformation on the temporal evolution of atrophy, in-formation that can be gathered using only serial MRIdata. We therefore cannot determine whether the atro-phy observed in this study occurred immediately beforeor many years before this study. Although the patientswere scanned at earlier time points,3 there has been amajor scanner hardware upgrade since then, rendering itdifficult to directly compare measurements from earlierscanning with that obtained at 20 years. Third, some ofthe more disabled patients were not able to be scanned,so our data are relatively biased toward a less disabledsubset of the patients previously studied.4 Fourth, spinalcord involvement makes an important contribution tolocomotor disability in MS and was not included in thisinvestigation. Finally, with the SPM-based methods,misclassification of lesions or nonbrain tissue as GMmay lead to a relative underestimation of the apparentmagnitude of GM disease effects; however, correctionfor lesion misclassification was performed, and qualityassurance review of the scans found no additional signif-icant segmentation errors; thus, there should not havebeen significant misclassification effects.

Notwithstanding these caveats, the study clearlyfound that in MS patients with a relatively long andhomogeneous disease duration (approximately 20years), GM atrophy is greater than WM atrophy, andreflects disease subtype and disability. It also helps tounderstand why a limited relation between WM lesionsand disability in MS has been evident in many previ-ous MRI clinical studies of both natural history andtherapeutic intervention, and highlights a need to bet-ter understand and monitor GM pathology in MS.

Appendix: Gray and White Matter VolumesMeasured Using SIENAX and Their Relationshipwith Clinical Subgroups and DisabilitySIENAX MethodologySIENAX was used to obtain the normalized (per subjecthead size) GM and WM brain volumes (NGMV andNWMV). In brief, SIENAX first extracts brain and skullvoxels from the input MR data, using the Brain Extrac-tion Tool (www.fmrib.ox.ac.uk/fsl). The brain image isthen affine-registered to standard space brain and skullimages, derived from the MNI152 standard space refer-ence set, with the skull registration used to determinethe head size normalization factor. Next, tissue type seg-mentation, with partial volume estimation, is performedto calculate the total volume of brain tissues, including

Fisniku et al: GM Atrophy and Disability in MS 251

separate estimates of volumes of GM, WM, and ventric-ular CSF. The estimated volumes for a subject then aremultiplied by the normalization factor to obtain NGMVand NWMV (normalized CSF and whole brain volumeswere also obtained; data not presented here).

Relationship of SIENAX Tissue Volumes withClinical SubgroupsSIENAX yielded NGMV results consistent with theGMF obtained by SPM2. For the NWMV, there wereno significant differences in any group comparisons(Appendix Tables 1 and 2).

Relationship of SIENAX Tissue Volumes withMeasures of Clinical Impairment and DisabilityNGMV correlated significantly with EDSS and MSFCfor all patients, and the MS subgroup only (Table 3).

NWMV correlated significantly only with EDSS andnot with MSFC (Table 3).

Correlations between SIENAX Tissue Volumes andLesion LoadT2LL correlated significantly with NGMV (rs � �0.57;p � 0.001) but not with NWMV (rs � �0.14; p �0.22) for all patients and for the MS subgroup only:rs � �0.67; p � 0.001 and rs � �0.10; p � 0.50,respectively.

Models to Predict Disability with SIENAX TissueVolumes and Lesion LoadOn the regression models, only NGMV independentlypredicted disability whether EDSS or MSFC, for thewhole cohort of patients and for the MS subgrouponly. For the whole cohort of patients, there was a

Table 1. Normalized Brain Volume Measurements in Controls and Clinical Subgroups

Groups (Number ofPatients)

NGMV, Mean (ml);Median (SD)

NWMV Mean (ml);Median (SD)

Controls (25) 910; 916 (40.4) 663; 669 (33.0)CIS (29) 885; 888 (40.8) 658; 657 (36.0)All patients (73) 858; 870 (60.7) 653; 652 (35.4)CDMS (44) 840; 853 (65.4) 650; 649 (35.0)RRMS (22) 856; 865 (60.4) 649; 649 (33.7)Benign MSa (22) 874; 885 (61.8) 659; 660 (31.0)Nonbenign MS (22) 806; 798 (50.3) 640; 643 (36.8)SPMS (11) 795; 793 (61.3) 652; 646 (40.6)

aBenign MS � Expanded Disability Status Score (EDSS) �3; nonbenign MS � EDSS �3.CDMS � clinically definite multiple sclerosis; CIS � clinically isolated syndrome; NGMV � normalized gray matter volume;NWMV � normalized white matter volume; RRMS � relapsing-remitting multiple sclerosis; SD � standard deviation; SPMS �secondary progressive multiple sclerosis.

Table 2. Age- and Sex-Adjusted Mean Difference between Patient Subgroups and Controls

GroupComparisons

NGMV NWMV

Adjusted Mean Difference(ml), 95% CI)

p Value Adjusted Mean Difference(ml), (95% CI)

p Value

MS-controls �56.52 (�85.05 to �27.99) �0.001 �14.11 (�33.99 to 5.76) 0.162MS-CIS �47.60 (�71.64 to �23.57) �0.001 �8.62 (�25.37 to 8.11) 0.309SPMS-controls �96.63 (�133.91 to �59.35) �0.001 �11.12 (�38.43 to 16.17) 0.420RRMS-controls �43.49 (�71.98 to �15.00) 0.003 �15.08 (35.95 to 5.78 0.155SPMS-CIS �87.06 (�121.02 to �53.11) �0.001 �5.68 (�30.55 to 19.18) 0.651RRMS-CIS �33.92 (�58.48 to �9.37) 0.007 �9.64 (27.63 to 8.33) 0.290SPMS-RRMS �53.14 (�86.08 to �19.47) 0.002 3.95 (�20.70 to 28.62) 0.751Benign-nonbenign MSa 65.25 (30.05 to 100.44) 0.001 17.69 (�3.30 to 38.70) 0.096CIS-controls �8.91 (�39.51 to 21.68) 0.564 �5.48 (�26.80 to 15.82) 0.610

aBenign MS � Expanded Disability Status Score (EDSS) �3; nonbenign MS � EDSS � 3CI � confidence interval; CIS � clinically isolated syndrome; MS � multiple sclerosis (all cases); NGMV � normalized gray mattervolume; NWMV � normalized white matter volume; RRMS � relapsing-remitting multiple sclerosis; SPMS � secondary progressivemultiple sclerosis.

252 Annals of Neurology Vol 64 No 3 September 2008

79% (p � 0.001) reduction in odds of being in a moresevere EDSS category per 1 SD higher NGMV, andthere was a 0.53 increase (p � 0.001) in MSFC per 1SD higher NGMV.

In the MS subgroup, there was a 69% (p � 0.001)reduction in odds of being in a more severe EDSS cat-egory per 1 SD higher NGMV, and there was a 0.48increase (p � 0.001) in MSFC per 1 SD higherNGMV.

SummaryIn this study, SIENAX and SPM-based methods gen-erally provided similar results when investigating therelationship of GM and WM volumes with clinicalcourse, clinical features, and T2LL. However, WM-segmentation obtained by SIENAX appeared to beless accurate compared with SPM2, because deep graymatter was not as clearly segmented, and misclassifi-cation of WM lesions was not adjusted for, perhapsexplaining why there were no differences between theNWMV (derived from SIENAX) between MS pa-tients and controls, while differences were observed inWMF (derived from SPM2).

The Nuclear Magnetic Resonance Research Unit, University CollegeLondon is supported by the Multiple Sclerosis Society of Great Brit-ain and Northern Ireland. This work was undertaken at UniversityCollege London Hospital/University College London, which re-ceived funding from the Department of Health’s National Institutefor Health Research Biomedical Research Centers funding scheme.

We thank Dr P. Brex for setting up the database, Dr D. Tozer forhelp with SIENAX analyses, C. Benton and R. Gordon for perform-ing the MRI scans, and the subjects who participated in this study.

References1. Polman CH, Reingold SC, Edan G, et al. Diagnostic criteria

for multiple sclerosis: 2005 revisions to the “McDonald Crite-ria.” Ann Neurol 2005;58:840–846.

2. Swanton JK, Rovira A, Tintore M, et al. MRI criteria for mul-tiple sclerosis in patients presenting with clinically isolatedsyndromes: a multicentre retrospective study. Lancet Neurol2007;6:677–686.

3. Brex PA, Ciccarelli O, O’Riordan JI, et al. A longitudinal studyof abnormalities on MRI and disability from multiple sclerosis.N Engl J Med 2002;346:158–164.

4. Fisniku LK, Brex PA, Altmann DR, et al. Disability and T2MRI lesions: a 20-year follow-up of patients with relapse onsetof multiple sclerosis. Brain 2008;131:808–817.

5. Brex PA, Jenkins R, Fox NC, et al. Detection of ventricularenlargement in patients at the earliest clinical stage of MS.Neurology 2000;54:1689–1691.

6. Miller DH, Barkhof F, Frank JA et al. Measurement of atrophyin multiple sclerosis: pathological basis, methodological aspectsand clinical relevance. Brain 2002;125:1676–1695.

7. Dalton CM, Miszkiel KA, O’Connor PW, et al. Ventricularenlargement in MS: one-year change at various stages of disease.Neurology 2006;66:693–698.

8. Pirko I, Lucchinetti CF, Sriram S, et al. Gray matter involve-ment in multiple sclerosis. Neurology 2007;68:634–642.

9. Dalton CM, Chard DT, Davies GR, et al. Early developmentof multiple sclerosis is associated with progressive grey matteratrophy in patients presenting with clinically isolated syn-dromes. Brain 2004;127:1101–1107.

10. Tiberio M, Chard DT, Altmann DR, et al. Gray and whitematter volume changes in early RRMS: a 2-year longitudinalstudy. Neurology 2005;64:1001–1007.

11. Chard DT, Griffin CM, Parker GJ, et al. Brain atrophy in clin-ically early relapsing-remitting multiple sclerosis. Brain 2002;125:327–337.

12. De SN, Matthews PM, Filippi M, et al. Evidence of early cor-tical atrophy in MS: relevance to white matter changes and dis-ability. Neurology 2003;60:1157–1162.

13. Quarantelli M, Ciarmiello A, Morra VB, et al. Brain tissuevolume changes in relapsing-remitting multiple sclerosis: cor-relation with lesion load. Neuroimage 2003;18:360 –366.

14. Sastre-Garriga J, Ingle GT, Chard DT, et al. Grey and whitematter atrophy in early clinical stages of primary progressivemultiple sclerosis. Neuroimage 2004;22:353–359.

15. Ge Y, Grossman RI, Udupa JK, et al. Brain atrophy inrelapsing-remitting multiple sclerosis: fractional volumetricanalysis of gray matter and white matter. Radiology 2001;220:606–610.

Table 3. Correlations of Brain Volume Measurements with Clinical Features

NormalizedBrain

Volumes

rs (p )

EDSSN � 73a

(44b),

MSFCN � 67a

(41b)

9HPTN � 70a

(42b)

Time-WalkedN � 68a

(40b)

PASATN � 68a

(42b)

aNGMV �0.62 (�0.001) 0.50 (�0.001) 0.62 (�0.001) �0.37 (0.002) 0.21(0.086)

bNGMV �0.53 (�0.001) 0.41 (0.007) 0.53 (�0.001) �0.45 (0.003) 0.15(0.325)

aNWMV �0.31 (0.006) 0.16 (0.179) 0.13 (0.260) �0.30 (0.011) 0.19(0.118)

bNWMV �0.28 (0.058) 0.16 (0.305) 0.26 (0.097) �0.21 (0.183) 0.21 (0.75)aAll patients.bCDMS patients only.CIS � clinically isolated syndrome; CDMS � clinically definite multiple sclerosis (all cases); RRMS � relapsing-remitting multiplesclerosis; SPMS � secondary progressive multiple sclerosis; NGMV � normalized gray matter volume; NWMV � normalized whitematter volume; 9HPT � 9 hole peg test; PASAT � Paced Auditory Serial Addition Task Scores.

Fisniku et al: GM Atrophy and Disability in MS 253

16. Anderson VM, Fox NC, Miller DH. Magnetic resonance im-aging measures of brain atrophy in multiple sclerosis. J MagnReson Imaging 2006;23:605–618.

17. Sailer M, Fischl B, Salat D, et al. Focal thinning of the cerebralcortex in multiple sclerosis. Brain 2003;126:1734–1744.

18. Paolillo A, Pozzilli C, Gasperini C, et al. Brain atrophy inrelapsing-remitting multiple sclerosis: relationship with ‘blackholes’, disease duration and clinical disability. J Neurol Sci 2000;174:85–91.

19. Ge Y, Grossman RI, Udupa JK, et al. Brain atrophy inrelapsing-remitting multiple sclerosis and secondary progressivemultiple sclerosis: longitudinal quantitative analysis. Radiology2000;214:665–670.

20. Chard DT, Brex PA, Ciccarelli O, et al. The longitudinal rela-tion between brain lesion load and atrophy in multiple sclerosis:a 14 year follow up study. J Neurol Neurosurg Psychiatry 2003;74:1551–1554.

21. Rudick RA, Lee JC, Simon J, et al. Significance of T2 lesions inmultiple sclerosis: a 13-year longitudinal study. Ann Neurol2006;60:236–242.

22. Filippi M, Rovaris M, Inglese M, et al. Interferon beta-1a forbrain tissue loss in patients at presentation with syndromes sug-gestive of multiple sclerosis: a randomised, double-blind,placebo-controlled trial. Lancet 2004;364:1489–1496.

23. Zivadinov R, Locatelli L, Cookfair D, et al. Interferon beta-1aslows progression of brain atrophy in relapsing-remitting mul-tiple sclerosis predominantly by reducing gray matter atrophy.Mult Scler 2007;13:490–501.

24. Molyneux PD, Kappos L, Polman C, et al. The effect of inter-feron beta-1b treatment on MRI measures of cerebral atrophyin secondary progressive multiple sclerosis. European StudyGroup on Interferon beta-1b in secondary progressive multiplesclerosis. Brain 2000;123(pt 11):2256–2263.

25. Kappos L, Traboulsee A, Constantinescu C, et al. Long-termsubcutaneous interferon beta-1a therapy in patients withrelapsing-remitting MS. Neurology 2006;67:944–953.

26. Poser CM, Paty DW, Scheinberg L, et al. New diagnostic cri-teria for multiple sclerosis: guidelines for research protocols.Ann Neurol 1983;13:227–231.

27. Kurtzke JF. Rating neurologic impairment in multiple sclerosis:an expanded disability status scale (EDSS). Neurology 1983;33:1444–1452.

28. Fischer JS, Rudick RA, Cutter GR, et al. The Multiple SclerosisFunctional Composite Measure (MSFC): an integrated approachto MS clinical outcome assessment. National MS Society ClinicalOutcomes Assessment Task Force. Mult Scler 1999;5:244–250.

29. Lublin FD, Reingold SC. Defining the clinical course of mul-tiple sclerosis: results of an international survey. NationalMultiple Sclerosis Society (USA) Advisory Committee onClinical Trials of New Agents in Multiple Sclerosis. Neurol-ogy. 1996;46:907–911.

30. Sailer M, O’Riordan JI, Thompson AJ, et al. Quantitative MRIin patients with clinically isolated syndromes suggestive of de-myelination. Neurology 1999;52:599–606.

31. Smith SM, Zhang Y, Jenkinson M, et al. Accurate, robust, andautomated longitudinal and cross-sectional brain change analy-sis. Neuroimage 2002;17:479–489.

32. Sanfilipo MP, Benedict RH, Sharma J, et al. The relationshipbetween whole brain volume and disability in multiple sclerosis:a comparison of normalized gray vs. white matter with misclas-sification correction. Neuroimage 2005;26:1068–1077.

33. Carone DA, Benedict RH, Dwyer MG, et al. Semi-automaticbrain region extraction (SABRE) reveals superior cortical anddeep gray matter atrophy in MS. Neuroimage 2006;29:505–514.

34. Sastre-Garriga J, Ingle GT, Chard DT, et al. Grey and whitematter volume changes in early primary progressive multiplesclerosis: a longitudinal study. Brain 2005;128:1454 –1460.

35. Peterson JW, Bo L, Mork S, et al. Transected neurites, apopto-tic neurons, and reduced inflammation in cortical multiple scle-rosis lesions. Ann Neurol 2001;50:389–400.

36. Bo L, Vedeler CA, Nyland H, et al. Intracortical multiple scle-rosis lesions are not associated with increased lymphocyte infil-tration. Mult Scler 2003;9:323–331.

37. Sanfilipo MP, Benedict RH, Weinstock-Guttman B, et al. Grayand white matter brain atrophy and neuropsychological impair-ment in multiple sclerosis. Neurology 2006;66:685–692.

38. Fisher E, Chang A, Fox RJ, et al. Imaging correlates of axonalswelling in chronic multiple sclerosis brains. Ann Neurol 2007;62:219–228.

39. Trapp BD, Peterson J, Ransohoff RM, et al. Axonal transectionin the lesions of multiple sclerosis. N Engl J Med 1998;338:278–285.

254 Annals of Neurology Vol 64 No 3 September 2008