ORIGINAL ARTICLE
`
`Gray Matter Atrophy in Multiple Sclerosis:
`A Longitudinal Study
`
`Elizabeth Fisher, Ph.D.,1 Jar-Chi Lee, M.S.,2 Kunio Nakamura, B.S.,1 and Richard A. Rudick, M.D.3
`
`Objective: To determine gray matter (GM) atrophy rates in multiple sclerosis (MS) patients at all stages of disease, and to
`identify predictors and clinical correlates of GM atrophy.
`Methods: MS patients and healthy control subjects were observed over 4 years with standardized magnetic resonance imaging
`(MRI) and neurological examinations. Whole-brain, GM, and white matter atrophy rates were calculated. Subjects were cate-
`gorized by disease status and disability progression to determine the clinical significance of atrophy. MRI predictors of atrophy
`were determined through multiple regression.
`Results: Subjects included 17 healthy control subjects, 7 patients with clinically isolated syndromes, 36 patients with relapsing-
`remitting MS (RRMS), and 27 patients with secondary progressive MS (SPMS). Expressed as fold increase from control subjects,
`GM atrophy rate increased with disease stage, from 3.4-fold normal in clinically isolated syndromes patients converting to
`RRMS to 14-fold normal in SPMS. In contrast, white matter atrophy rates were constant across all MS disease stages at
`approximately 3-fold normal. GM atrophy correlated with disability. MRI measures of focal and diffuse tissue damage accounted
`for 62% of the variance in GM atrophy in RRMS, but there were no significant predictors of GM atrophy in SPMS.
`Interpretation: Gray matter tissue damage dominates the pathological process as MS progresses, and underlies neurological
`disabillity. Imaging correlates of gray matter atrophy indicate that mechanisms differ in RRMS and SPMS. These findings
`demonstrate the clinical relevance of gray matter atrophy in MS, and underscore the need to understand its causes.
`
`Ann Neurol 2008;64:255–265
`
`Recent imaging and pathology studies have demon-
`strated that multiple sclerosis (MS) affects both cere-
`bral gray matter (GM) as well as white matter (WM).
`Characteristic findings in GM include focal regions of
`demyelination, activated microglia, apoptotic neu-
`rons, and atrophy of cortical and deep GM struc-
`tures.1– 4 Focal GM lesions are difficult to detect us-
`ing conventional imaging because of low contrast and
`small lesion size. However, GM atrophy can be reli-
`ably measured from standard magnetic resonance im-
`ages
`(MRIs). Quantitative analysis of MRIs has
`shown that GM tissue volumes are lower in MS pa-
`tients than in healthy age-matched control subjects,
`and that GM atrophy begins early in the course of
`disease.5–10 Prior studies of GM atrophy in MS pa-
`tients have been designed to be cross-sectional, short-
`term longitudinal, or limited to a particular disease
`stage. The evolution of GM atrophy over the course
`of MS and how it relates to progression of disability
`have not been fully described. An important unan-
`
`swered question is whether GM atrophy occurs as a
`direct result of GM pathology, or whether it is sec-
`ondary to tissue damage within WM lesions.
`The objective of this study is to characterize MS-
`related GM atrophy in a real-world setting. Patients
`with clinically isolated syndromes
`(CIS),
`relapsing-
`remitting MS (RRMS), and secondary progressive MS
`(SPMS), together with age- and sex-matched healthy
`individuals who served as concurrent control subjects,
`were entered into a prospective longitudinal study.
`This article addresses the pattern of brain tissue loss in
`these patients over the course of 4 years. The correla-
`tions between GM tissue loss and other MRI measures
`of tissue damage, and between GM tissue loss and clin-
`ical worsening were investigated.
`
`Subjects and Methods
`Subjects
`Patients were recruited from the Mellen Center for Multiple
`Sclerosis Treatment and Research at Cleveland Clinic. Age-
`
`From the 1Department of Biomedical Engineering; 2Quantitative
`Health Sciences Lerner Research Institute; and 3Mellen Center for
`Multiple Sclerosis Treatment and Research, Neurologic Institute,
`Cleveland Clinic, Cleveland, OH.
`Received Feb 15, 2008, and in revised form Mar 28. Accepted for
`publication May 12, 2008.
`This article includes supplementary materials available via the Inter-
`net
`at http://www.interscience.wiley.com/jpages/0364-5134/supp-
`mat
`
`InterScience
`in Wiley
`2008,
`23,
`July
`online
`Published
`(www.interscience.wiley.com). DOI: 10.1002/ana.21436
`
`Address correspondence to Dr Fisher, Department of Biomedical
`Engineering ND20, Cleveland Clinic Foundation, 9500 Euclid Av-
`enue, Cleveland, OH 44195. E-mail: fishere@ccf.org
`
`© 2008 American Neurological Association
`Published by Wiley-Liss, Inc., through Wiley Subscription Services
`
`255
`
`MYLAN PHARMS. INC. EXHIBIT 1097 PAGE 1
`
`

`
`and sex-matched healthy control subjects were recruited by
`inviting the research subjects with MS to invite a spouse or
`friend into the study. The study was reviewed and approved
`by the Cleveland Clinic Institutional Review Board (IRB
`study numbers 2612 and 3709), and each study subject re-
`viewed and signed an informed consent document. A finan-
`cial payment of $50 per study visit was given to each patient
`and control subject for their participation.
`Patients with MS met the International Panel criteria,11
`and each had a cranial MRI scan demonstrating lesions con-
`sistent with MS. Patients were classified as RRMS if they
`had two or more discrete relapses with significant neurolog-
`ical recovery in the prior 3 years, and as SPMS if they expe-
`rienced continued deterioration for at least 6 months, with
`or without superimposed relapses in a patient who had a
`prior history of at least two relapses. Patients with CIS had
`an episode of neurological dysfunction typical for an initial
`MS presentation (eg, optic neuritis,
`transverse myelitis,
`brainstem syndromes). Conversion of CIS patients to clini-
`cally definite MS was based on a clinical relapse. Disease
`therapy with interferon-␤, glatiramer acetate, methotrexate,
`or azathioprine was allowed. Age- and sex-matched healthy
`control subjects were required to have a normal neurological
`examination, a normal brain MRI, and no history of symp-
`toms suggestive of MS. Patients and healthy control subjects
`were excluded if they received corticosteroid therapy within 2
`months, were on bimonthly corticosteroid pulses, required
`therapy for hypertension, or had a history of transient isch-
`emic attack or stroke, heart disease, pulmonary disease, dia-
`betes, or chronic renal insufficiency.
`
`Visit Schedule
`Healthy control subjects were evaluated annually; CIS and
`MS patients were evaluated biannually. Demographic infor-
`mation (sex, age, racial background, educational status, and
`birthplace) and MS disease history (date and nature of first
`symptom, date of diagnosis, clinical pattern of disease from
`study onset, and clinical pattern of disease in the year before
`study entry) were recorded at baseline. At each visit, clinical
`assessments
`included Kurtzke Extended Disability Status
`Scale (EDSS), timed ambulation, 9-hole peg test, 3-second
`Paced Auditory Serial Addition Test, relapse history, and
`medications. The timed ambulation, 9-hole peg test, and
`Paced Auditory Serial Addition Test scores were transformed
`into the Multiple Sclerosis Functional Composite (MSFC)
`by normalization to a published MS reference group.12 At
`each study visit, subjects underwent standardized MRI exam-
`inations (see later for descriptions). For subjects requiring
`corticosteroids for relapses, study visits were postponed until
`at least 6 weeks after steroid therapy to avoid confounding
`effects. Disease worsening among the MS subjects was de-
`fined in two ways: (1) subjects who progressed to a more
`severe stage of MS (eg, CIS patients who converted to
`RRMS; RRMS patients who converted to SPMS) were con-
`sidered worse; and (2) subjects whose conditions worsened
`by 1.0 EDSS point (or 0.5 point for those who started the
`study with EDSS ⬎ 5.0) sustained for two consecutive
`6-month visits were considered worse.
`
`256 Annals of Neurology Vol 64 No 3
`
`September 2008
`
`Magnetic Resonance Imaging Examinations
`Images were acquired on a 1.5-Tesla magnetic resonance
`scanner and consisted of a T2-weighted, fluid-attenuated in-
`version recovery image (FLAIR), proton density-weighted
`images acquired with and without a magnetization transfer
`pulse for calculation of magnetization transfer ratio (MTR),
`and T1-weighted images acquired before and after injection
`of
`standard-dose gadolinium godopentetate dimeglumine
`(Gd-DTPA) (T1 and T1gad). The details of the image ac-
`quisitions are provided in Supplementary Table 1. (www.
`mrw.interscience.wiley.com/suppmat/0364-5134/suppmat/
`ana.21436.html)
`MRIs were analyzed to calculate brain parenchymal frac-
`tion (BPF), gray matter fraction (GMF), white matter frac-
`tion (WMF), T2 lesion volume (T2LV), T1 hypointense le-
`sion volume (T1LV), gadolinium-enhancing lesion volume,
`mean MTR of normal-appearing brain tissue (NABT MTR),
`and mean lesion MTR relative to normal-appearing tissue
`(lesion MTR ratio). All software for MRI analysis was devel-
`oped at the Cleveland Clinic Department of Biomedical En-
`gineering.
`The whole brain was segmented and BPF was calculated
`from FLAIR images as described previously.13,14 The FLAIR
`images were also used to segment T2 lesions as described
`later. GM voxels were segmented automatically from the T1
`images using a new method that combines an intensity-based
`probability map and two types of regional probability maps
`(K. Nakamura and E. Fisher, unpublished data). First, the
`FLAIR image was registered to the T1 image, and the whole-
`brain and T2 lesion masks were applied to mask out non-
`brain and lesion voxels from the T1 image. The intensity-
`based GM probability map was
`calculated
`using
`unsupervised clustering (modified fuzzy c-means)15 applied
`to the masked T1 image. An anatomic GM probability map
`was derived from a brain atlas coregistered to the patient’s
`MRI. Lastly, an individualized GM morphological probabil-
`ity map was created using the brain surface and lateral ven-
`tricles as landmarks to define regions that have a high like-
`lihood for GM. The three probability maps were then
`combined to create a final GM map. An example of the GM
`segmentation is shown in Figure 1. The lesion-masking step
`and the use of the regional probability maps effectively pre-
`vent the misclassification of lesions and partial volume voxels
`as GM, a common problem with GM segmentation algo-
`rithms.16
`GM volume was calculated from the final GM map using
`a three-compartment model to account for partial-volume ef-
`fects. The measured GM volume was then adjusted to cor-
`rect for an artifact related to the use of fuzzy c-means. It was
`determined in a separate study that as T2 lesions enlarge,
`there is a reduction in the measured GM volume caused by
`slight shifts in the probabilities assigned to voxels in the in-
`tensity range between GM and WM. This results in an un-
`wanted, but consistent and linear, dependence of GM vol-
`umes on T2LVs. The GM volumes were adjusted to account
`for this technical issue as follows: adjusted_GM_volume ⫽
`measured_GM_volume ⫹ 0.26*T2LV. GM fraction was cal-
`culated as the final adjusted GM volume divided by the vol-
`ume within the outer contour of the brain (the same volume
`as the denominator for BPF). WMF was calculated as BPF
`
`MYLAN PHARMS. INC. EXHIBIT 1097 PAGE 2
`
`

`
`Table 1. Baseline Characteristics
`
`Characteristics
`
`HC
`(n ⴝ 17)
`
`CIS (n ⴝ 7)
`
`CIS3 MS
`(n ⴝ 8)
`
`RRMS
`(n ⴝ 28)
`
`RR3 SPMS
`(n ⴝ 7)
`
`SPMS
`(n ⴝ 19)
`
`41.6 (8.1)
`[32–56]
`
`44.9 (10.1)
`[27–53]
`
`36.1 (7.1)
`[26–49]
`
`39.7 (8.4)
`[17–53]
`
`42.2 (7.0)
`[33–50]
`
`49.7 (7.4)
`[39–65]
`
`10 (59)
`
`6 (86)
`
`5 (63)
`
`23 (82)
`
`4 (57)
`
`14 (74)
`
`NA
`
`0.28 (0.07)
`
`0.51 (0.70)
`
`6.7 (5.1)
`
`18.5 (11.1)
`
`17.4 (5.0)
`
`NA
`
`0.86 (0.85)
`
`1.19 (0.37)
`
`2.0 (1.5)
`
`4.79 (1.58)
`
`5.39 (1.34)
`
`0.55 (27)
`
`0.36 (0.37)
`
`0.51 (0.46)
`
`0.39 (0.62)
`
`⫺0.34 (0.78)
`
`⫺1.04 (1.49)
`
`NA
`
`2.5 (2.3)
`
`7.6 (8.3)
`
`20.7 (17.8)
`
`42.4 (24.1)
`
`43.8 (26.0)
`
`NA
`
`0.15 (0.22)
`
`0.29 (0.48)
`
`1.74 (2.49)
`
`8.17 (6.89)
`
`8.73 (9.00)
`
`Mean agea
`(SD) [range],
`yr
`Female sex, n
`(%)
`Mean
`symptom
`durationb
`(SD), yr
`Mean EDSS
`scoreb (SD)
`Mean MSFCa
`(SD)
`Mean T2
`lesion
`volumec (SD),
`ml
`Mean T1
`lesion
`volumed
`(SD), ml
`Gd⫹
`Mean NABT
`MTRa (SD)
`Mean lesion
`MTR ratio
`(SD)
`Mean BPFe
`(SD)
`Mean GMFf
`(SD)
`Mean WMFg
`(SD)
`HC ⫽ healthy control subjects; CIS ⫽ patients who had a clinically isolated syndrome and did not meet the criteria for a diagnosis of
`clinically definite multiple sclerosis (MS) over the course of 4 years; CIS3 MS ⫽ CIS patients who converted to clinically definite MS
`over the course of 4 years; RRMS ⫽ relapsing-remitting MS patients (throughout study); RR3 SPMS ⫽ relapsing-remitting MS
`patients who converted to secondary progressive MS over the course of 4 years; SPMS ⫽ secondary progressive MS patients (throughout
`study); SD ⫽ standard deviation; EDSS ⫽ Expanded Disability Status Scale; MSFC ⫽ Multiple Sclerosis Functional Composite; Gd⫹
`⫽ patients with gadolinium-enhancing lesions; NABT ⫽ normal-appearing brain tissue; MTR ⫽ magnetization transfer ratio; BPF ⫽
`brain parenchymal fraction; GMF ⫽ gray matter fraction; WMF ⫽ white matter fraction.
`Significant differences between groups ( p ⬍ 0.05): aCIS3 MS versus SPMS; b(CIS, CIS3 MS, RRMS) versus (RR3 SPMS, SPMS);
`c(CIS) versus (RRMS, RR3 SPMS, SPMS); d(CIS, CIS3 MS) versus (RR3 SPMS, SPMS); e(HC, CIS, CIS3 MS) versus
`(RR3 SPMS, SPMS); f(HC, CIS, CIS3 MS) versus (RR3 SPMS, SPMS), RR versus SPMS; g(HC, CIS, CIS3 MS, RRMS) versus
`SPMS.
`
`NA
`35.6 (0.75)
`
`14.3%
`35.5 (0.98)
`
`37.5%
`35.9 (0.81)
`
`17.9%
`35.3 (1.01)
`
`28.6%
`34.6 (0.35)
`
`15.6%
`34.35 (1.19)
`
`NA
`
`0.89 (0.12)
`
`0.92 (0.05)
`
`0.92 (0.04)
`
`0.88 (0.03)
`
`0.90 (0.05)
`
`0.862 (0.012)
`
`0.861 (0.008)
`
`0.855 (0.023)
`
`0.840 (0.027)
`
`0.810 (0.02)
`
`0.801 (0.04)
`
`0.554 (0.015)
`
`0.551 (0.010)
`
`0.555 (0.016)
`
`0.537 (0.018)
`
`0.519 (0.017)
`
`0.528 (0.032)
`
`0.308 (0.011)
`
`0.309 (0.009)
`
`0.300 (0.017)
`
`0.304 (0.016)
`
`0.291 (0.015)
`
`0.280 (0.016)
`
`minus GMF. The accuracy and reproducibility of the GM
`segmentation method were evaluated in a separate study.
`The mean GM volume errors were determined to be 1.2%
`when assessed with BrainWeb (http://www.bic.mni.m-
`cgill.ca/brainweb/) and 3.1% when compared with manual
`tracings. In a scan-rescan evaluation consisting of nine pa-
`tients imaged three separate times over 2 weeks, the mean
`coefficient of variation for GMF was 1.1% (K. Nakamura
`and E. Fisher, submitted).
`
`Baseline T2 lesions were automatically segmented in
`brain-masked FLAIR MRIs using a modified version of the
`Iterated Conditional Modes algorithm.17 The T2 lesion
`mask was used to guide the automated segmentation of T1
`hypointensities and gadolinium-enhancing lesions in coregis-
`tered T1 and T1gad images, as described previously.18 Seg-
`mentation of T2 lesions in the follow-up images utilized the
`baseline lesion mask and a registration and subtraction
`method to detect both persistent and new T2 lesions. Lesion
`
`Fisher et al: Gray Matter Atrophy in MS
`
`257
`
`MYLAN PHARMS. INC. EXHIBIT 1097 PAGE 3
`
`

`
`Results
`One hundred six research subjects were enrolled in the
`study. Nineteen of these subjects (18%) discontinued
`the protocol for various reasons; this report provides
`information on the 87 subjects who remain in the pro-
`tocol. These subjects were observed for a mean of 4.1
`years (range, 3.4 – 4.8 years). At each study visit, the
`disease category for subjects with MS was assessed de
`novo, without reference to the disease category previ-
`ously assigned. Eight of 15 subjects who initially en-
`tered with a diagnosis of CIS transitioned to RRMS; 7
`of 35 initially categorized as RRMS transitioned to
`SPMS; and 1 of 20 initially categorized as SPMS was
`classified at every visit
`subsequent
`to baseline as
`RRMS. Of the eight CIS patients who converted to
`RRMS, five had new T2 lesions on the year 4 MRI,
`whereas only one of the seven CIS patients who did
`not convert to RRMS based on clinical criteria had a
`new T2 lesion.
`Table 1 shows baseline characteristics for each sub-
`group, as classified at the final visit. As expected, com-
`pared with RRMS group, the average SPMS patient
`was significantly older, had greater symptom duration,
`greater EDSS score, lower MSFC score, greater T2LV
`and T1LV, lower NABT and lesion MTR score, and
`lower BPF, GMF, and WMF. Compared with RRMS
`patients who did not progress to SPMS, patients who
`converted had longer symptom duration, greater base-
`line EDSS, more EDSS change, greater baseline T2LV
`and T1LV, and lower baseline NABT MTR, BPF,
`GMF, and WMF. Table 2 shows the mean changes
`over 4 years for each subgroup. RRMS patients who
`converted to SPMS had greater changes in T1LV,
`BPF, GMF, and WMF than the RRMS patients who
`did not convert.
`(BPF,
`in fractional brain volumes
`The changes
`WMF, and GMF) are plotted in Figure 2 for each sub-
`group. Whole-brain atrophy rates were similar in stable
`CIS patients and healthy control subjects, but steadily
`increased as disease severity increased. Increasing atro-
`phy was driven entirely by increasing rates of GM at-
`rophy. Expressed as a fold increase compared with the
`concurrently studied healthy control subjects, GMF
`change was 3.4-fold greater than normal in patients
`converting from CIS to RRMS, 8.1-fold greater in
`RRMS patients, 12.4-fold greater in patients convert-
`ing from RRMS to SPMS, and 14-fold greater in
`SPMS patients. The GM atrophy rates in the com-
`bined set of RRMS patients (CIS3 RRMS and RRMS
`stable)
`and
`combined
`set
`of
`SPMS
`patients
`(RRMS3 SPMS and SPMS) were significantly greater
`than GM atrophy rate in healthy control subjects ( p ⫽
`0.05 and p ⫽ 0.005, respectively). In contrast, WM
`atrophy rates were similar in all disease categories, at
`approximately threefold greater than in healthy control
`subjects.
`
`Fig 1. Example of gray matter (GM) segmentation results. (A)
`Fluid-attenuated inversion recovery (FLAIR) image used for
`segmentation of brain versus nonbrain structures and segmen-
`tation of T2 lesion versus normal-appearing brain tissue. (B)
`T1-weighted image used for segmentation of GM versus white
`matter. Arrow indicates a T1-hypointense lesion with similar
`intensity to GM that would be potentially misclassified. (C)
`Final GM segmentation results. Note that lesion is not classi-
`fied as GM.
`
`segmentation results were visually verified and semiautomat-
`ically corrected, using interactive software. Partial-volume ef-
`fects were accounted for in the calculation of all brain and
`lesion volumes using either two- or three-compartment mix-
`ture models as appropriate.19
`MTR images were calculated from the proton density im-
`age pair acquired with and without an MT pulse.20 The
`mean MTR of NABT and mean lesion MTR were calculated
`from voxels included within the brain and T2 lesion masks,
`respectively. To ensure a consistent measure of the degree of
`abnormality of MTR within lesions, we normalized the
`mean lesion MTR by the mean MTR of normal-appearing
`WM (lesion MTR ratio).
`
`Statistical Analysis
`Baseline and on study changes were assessed and compared
`among disease subgroups and healthy control subjects using
`analysis of covariance. For categorical variables, a ␹2 test was
`performed. Spearman’s rank correlation was used for corre-
`lations between GM atrophy and clinical disability. Pearson’s
`correlation was used to assess correlations between GMF and
`WMF, and between GMF and age.
`Multiple regression models were developed for whole-
`brain, GM, and WM atrophy using a set of baseline MRI
`predictors and their 4-year changes. The regression models
`were developed on 2 datasets: 36 RRMS patients, including
`both the CIS3 RRMS and the stable RRMS groups, and 27
`SPMS patients, including both the RRMS3 SPMS and the
`SPMS groups. For each regression model, we followed a two-
`step process. First, we utilized bootstrap bagging method for
`predictor selection.21,22 This method used automated for-
`ward stepwise selection to identified MRI predictors to in-
`clude in 1,000 bootstrapped samples, which were 100% the
`size of the original dataset. All variables that were significant
`predictors in more than 50% of the bootstrap runs were re-
`tained. In the final regression models, only selected predic-
`tors were included. Adjusted R2 is reported for the final
`models. All analyses were performed using SAS version 8.2
`(SAS Institute, Cary, NC).
`
`258 Annals of Neurology Vol 64 No 3
`
`September 2008
`
`MYLAN PHARMS. INC. EXHIBIT 1097 PAGE 4
`
`

`
`Table 2. On-Study Changes
`Characteristics HC (n ⴝ 17)
`
`CIS (n ⴝ 7)
`
`CIS3 MS
`(n ⴝ 8)
`
`RRMS
`(n ⴝ 28)
`
`RR3 SPMS
`(n ⴝ 7)
`
`SPMS
`(n ⴝ 19)
`
`NA
`
`NA
`
`32.5 (47.4)
`
`89.6 (24.6)
`
`77.6 (41.1)
`
`87.2 (21.7)
`
`52.8 (42.0)
`
`0.50 (1.32)
`
`0.63 (1.71)
`
`⫺0.036 (0.78)
`
`0.64 (0.69)
`
`0.08 (0.38)
`
`0.23 (0.14)
`
`0.49 (0.39)
`
`0.34 (0.42)
`
`0.13 (0.28)
`
`⫺1.44 (2.50)
`
`0.12 (0.71)
`
`NA
`
`NA
`
`NA
`
`⫺0.26 (0.23)
`
`0.54 (1.56)
`
`1.34 (3.21)
`
`1.06 (2.92)
`
`0.64 (5.28)
`
`0.007 (0.21)
`
`⫺0.06 (0.15)
`
`0.84 (1.06)
`
`1.88 (1.54)
`
`1.01 (1.37)
`
`14.3%
`
`50.0%
`
`35.7%
`
`100%
`
`36.8%
`
`% Time on
`DMTa
`Mean ⌬EDSS
`score (SD)
`Mean
`⌬MSFCb
`(SD)
`Mean ⌬T2
`lesion volume
`(SD), ml
`Mean ⌬T1
`lesion
`volumec (SD),
`ml
`Gd⫹ during
`studyd
`Mean
`⌬NABT
`MTR (SD)
`Mean ⌬lesion
`MTR (SD)
`Mean BPF
`%⌬/yeare
`(SD)
`Mean GMF
`%⌬/year (SD)
`Mean WMF
`%⌬/year (SD)
`HC ⫽ healthy control subjects; CIS ⫽ patients who had a clinically isolated syndrome and did not meet the criteria for a diagnosis of
`clinically definite multiple sclerosis (MS) over the course of 4 years; CIS3 MS ⫽ CIS patients who converted to clinically definite MS
`over the course of 4 years; RRMS ⫽ relapsing-remitting MS patients (throughout study); RR3 SPMS ⫽ relapsing-remitting MS
`patients who converted to secondary progressive MS over the course of 4 years; SPMS ⫽ secondary progressive MS patients (throughout
`study); DMT ⫽ disease-modifying therapy; EDSS ⫽ Expanded Disability Status Scale; MSFC ⫽ Multiple Sclerosis Functional
`Composite; ⌬ ⫽change; Gd⫹ ⫽patients with gadolinium-enhancing lesions; NABT ⫽ normal-appearing brain tissue; MTR ⫽
`magnetization transfer ratio; BPF ⫽ brain parenchymal fraction; GMF ⫽ gray matter fraction; WMF ⫽ white matter fraction.
`Significant differences between groups ( p ⬍ 0.05): aCIS versus (CIS3 MS, RR3 SPMS); bRR3 SPMS versus (CIS, CIS3 MS,
`RR3 SPMS, SPMS); cCIS3 MS versus RR3 SPMS; dRR3 SPMS versus (CIS, RRMS, SPMS); eCIS versus SPMS.
`
`0.74 (1.03)
`
`0.43 (0.59)
`
`⫺0.55 (1.1)
`
`⫺0.29 (0.98)
`
`⫺0.22 (1.02)
`
`⫺0.32 (0.75)
`
`NA
`
`0.032 (0.07)
`
`⫺0.01 (0.04) ⫺0.015 (0.03) ⫺0.012 (0.03) ⫺0.016 (0.03)
`
`⫺0.066 (0.22) ⫺0.003 (0.15) ⫺0.15 (0.14)
`
`⫺0.23 (0.32)
`
`⫺0.35 (0.18)
`
`⫺0.39 (0.31)
`
`⫺0.028 (0.24) ⫺0.028 (0.25) ⫺0.096 (0.23)
`
`⫺0.23 (0.34)
`
`⫺0.35 (0.37)
`
`⫺0.39 (0.50)
`
`⫺0.076 (0.35)
`
`0.11 (0.25)
`
`⫺0.24 (0.29)
`
`⫺0.24 (0.72)
`
`⫺0.33 (0.53)
`
`⫺0.25 (0.49)
`
`Correlations between Atrophy and Age
`The differences in GM atrophy rates between groups
`could not be explained by differences in age. GM at-
`rophy rate was not significantly correlated with age in
`the patients as a whole or in any of the MS sub-
`groups. After adjusting for age, the GM atrophy rate
`in RRMS patients was still significantly greater than
`in healthy control subjects ( p ⫽ 0.048). The age-
`adjusted GM atrophy rate in the SPMS patients was
`also significantly greater than in healthy control sub-
`jects ( p ⫽ 0.019).
`
`Correlations between Atrophy and Disability
`Table 3 shows clinical characteristics and atrophy mea-
`surements for RRMS and SPMS patients according to
`
`whether they had sustained EDSS worsening over the 4
`years. Other than EDSS change, there were no signif-
`icant differences between the subgroup of 13 subjects
`with sustained EDSS worsening compared with more
`stable patients, probably because of the small number
`of subjects in the comparison groups. However, there
`was a consistent trend toward more whole-brain and
`GM atrophy in patients with EDSS progression. Of
`interest, a similar pattern was not observed for change
`in WMF.
`GMF was correlated with clinical disability scores, as
`shown in Figure 3. Correlations were greatest with
`rank correlation coefficient ⫽
`MSFC (Spearman’s
`0.52) and were similar between RRMS and SPMS sub-
`groups. GMF was also moderately correlated with
`
`Fisher et al: Gray Matter Atrophy in MS
`
`259
`
`MYLAN PHARMS. INC. EXHIBIT 1097 PAGE 5
`
`

`
`for almost half the variance in GMF percentage change
`in RRMS. The addition of on-study changes improved
`the model slightly: lower NABT MTR, greater lesion/
`NAWM MTR ratio, and increase in T2LV predicted
`greater GM atrophy and accounted for 62% of the
`variance in GMF percentage change (see Table 5). In
`contrast, there were no baseline MRI measurements or
`on-study MRI changes that were significant predictors
`of GMF changes in the SPMS group.
`In the RRMS group, there were no correlations be-
`tween WMF and GMF at baseline or follow-up. In
`SPMS, however, WMF and GMF were moderately
`correlated at baseline (r ⫽ 0.42; p ⫽ 0.03) and at
`follow-up (r ⫽ 0.46; p ⫽ 0.015). Baseline WMF did
`not predict subsequent change in GMF, and baseline
`GMF did not predict subsequent change in WMF in
`either RRMS or SPMS.
`
`Discussion
`Many studies have reported brain atrophy in patients
`with MS.23–25 Brain atrophy begins early in MS and
`
`Table 3. Characteristics of Study Subjects by
`Expanded Disability Status Scale Progression Status
`
`Characteristics
`
`Worse
`(n ⴝ 13)
`
`Not Worse
`(n ⴝ 50)
`
`42.51 (7.86)
`11.81 (8.29)
`65.2%
`3.42 (2.66)
`
`42.81 (9.50)
`10.24 (8.59)
`75.2%
`3.25 (2.07)
`
`Clinical Features
`Age
`Symptom duration
`% Time on DMT
`EDSS score at
`baseline
`EDSS score ⌬
`baseline to last
`visit
`MSFC at baseline
`MSFC ⌬ baseline
`to last visit
`Baseline
`BPF
`GMF
`WMF
`Change during
`Study
`⫺0.380 (0.351) ⫺0.263 (0.278)
`BPF % ⌬/year
`⫺0.441 (0.422) ⫺0.243 (0.381)
`GMF % ⌬/year
`⫺0.106 (0.482) ⫺0.294 (0.595)
`WMF % ⌬/year
`DMT ⫽ disease-modifying therapy; EDSS ⫽ Expanded
`Disability Status Scale; ⌬ ⫽change; MSFC ⫽ Multiple
`Sclerosis Functional Composite; BPF ⫽ brain parenchymal
`fraction; NABT ⫽ normal-appearing brain tissue; MTR ⫽
`magnetization transfer ratio; GMF ⫽ gray matter fraction;
`WMF ⫽ white matter fraction.
`
`0.96 (0.52)
`
`⫺0.06 (0.82)
`
`⫺0.49(1.66)
`⫺0.79 (1.94)
`
`⫺0.01(0.98)
`0.18 (0.47)
`
`0.835 (0.034)
`0.534 (0.022)
`0.299 (0.021)
`
`0.824 (0.038)
`0.531 (0.026)
`0.293 (0.019)
`
`Fig 2. Plot of the mean annualized rates of change for white
`matter fraction (WMF), gray matter fraction (GMF), and
`brain parenchymal fraction (BPF) in each subgroup, including
`healthy control subjects (HCs; white bars; n ⫽ 17), patients
`with clinically isolated syndrome (CIS; gray bars) throughout
`the 4-year study (n ⫽ 7), patients who started the study with
`a diagnosis of CIS and converted to relapsing-remitting multi-
`ple sclerosis (CIS3 RRMS; light yellow bars; n ⫽ 8), pa-
`tients with RRMS throughout the study (RRMS; dark yellow
`bars; n ⫽ 28), patients who started the study with a diagno-
`sis of RRMS and progressed to secondary progressive MS
`(RRMS3 SPMS; light orange bars; n ⫽ 7), and patients
`with SPMS (orange bars) throughout the study (n ⫽ 19).
`Error bars indicate standard error of the mean (S.E.M.).
`
`EDSS for the group as a whole (Spearman’s rank cor-
`relation coefficient ⫽ ⫺0.48). Correlations differed be-
`tween the RRMS and SPMS groups for EDSS and for
`the individual components of the MSFC.
`
`Predictors of Atrophy
`MRI predictors of atrophy were different between the
`RRMS and SPMS groups, and for GM atrophy and
`WM atrophy, as shown in Tables 4 and 5. For whole-
`brain atrophy in the RRMS group, lower NABT MTR
`and the presence of gadolinium-enhancing lesions at
`baseline were associated with a greater rate of BPF
`change over the subsequent 4 years, but these variables
`accounted for only 25% of the variance. In the SPMS
`group, 52% of the variance in BPF percentage change
`could be accounted for by baseline MRI measures, in-
`cluding T2LV, lesion MTR, and GMF (see Table 4).
`Changes in MRI measures over the course of the study
`were not predictive of BPF change in the SPMS group
`(see Table 5). However, in the RRMS group, on-study
`changes were strongly predictive of concurrent change
`in BPF, accounting for 72% of the variance in atrophy
`rate in the RRMS patients.
`The most striking differences between the RRMS
`and SPMS groups were for predictors of GM atrophy.
`In the RRMS group, lower NABT MTR, greater lesion
`MTR ratio, and the presence of gadolinium-enhancing
`lesions were associated with greater rates of GM atro-
`phy (see Table 4). These baseline variables accounted
`
`260 Annals of Neurology Vol 64 No 3
`
`September 2008
`
`MYLAN PHARMS. INC. EXHIBIT 1097 PAGE 6
`
`

`
`Fig 3. Plots of gray matter fraction (GMF) at last visit versus various clinical disability scales, including (A) multiple sclerosis func-
`tional composite (MSFC), where the Spearman’s rank correlation coefficient (SRCC) was 0.52; (B) timed 25-foot walk, SRCC ⫽
`⫺0.43; (C) nine-hole peg test (9-HPT), SRCC ⫽ 0.46; (D) 3-second Paced Auditory Serial Addition test (3-second PASAT),
`SRCC ⫽ 0.43; and (E) Expanded Disability Status Scale (EDSS), SRCC ⫽ ⫺0.48. Red dots represent relapsing-remitting MS
`patients; black dots represent secondary progressive MS patients.
`
`continues throughout the course of disease. In patients
`with CIS, brain atrophy is observed in those converting
`to clinically definite MS.26 In RRMS, relapses and
`EDSS scores are only loosely correlated with brain at-
`rophy, but the rate of brain atrophy predicts future
`clinical status.27 Antiinflammatory disease-modifying-
`drugs have been reported to slow brain atrophy during
`the RRMS disease stage,14,28 but not in the SPMS
`stage.29,30 In the few prior studies comparing disease
`categories, whole-brain atrophy rates were similar in
`RRMS and SPMS patients.31–33 In nearly all prior
`studies, the extent of whole-brain atrophy correlated
`better with disability than did lesions.25
`GM atrophy has not been studied extensively. Some
`small studies demonstrated GM atrophy early during
`the disease course.7–10 GM tissue loss is of particular
`importance in MS, because GM makes up more than
`half the total brain parenchyma, tissue damage in GM
`
`is a large component of overall MS disease burden,2,34
`and conventional MRI techniques are not sensitive to
`lesions in GM.35 Therefore, GM atrophy measures are
`the only current method that can be used to quantify
`GM pathology over time in MS patients. The aims of
`this study were to assess GM pathology by determining
`the rate and pattern of GM atrophy across the disease
`severity spectrum in MS, and to determine clinical cor-
`relates and MRI predictors of GM atrophy. MS pa-
`tients with CIS, RRMS, and SPMS were studied con-
`currently with age- and sex-matched healthy control
`subjects over 4 years in an observational protocol, in
`which no attempt was made to standardize disease-
`modifying therapy that may influence brain atrophy.
`Also, brain atrophy results were not used for clinical
`management.
`There were several significant findings. First, whole-
`brain atrophy and GM atrophy accelerated in patients
`
`Fisher et al: Gray Matter Atrophy in MS
`
`261
`
`MYLAN PHARMS. INC. EXHIBIT 1097 PAGE 7
`
`

`
`Table 4. Baseline Predictors of Atrophy
`
`Dependent
`Variable
`
`BPF, %
`⌬/year
`
`GMF, %
`⌬/year
`
`WMF, %
`⌬/year
`
`RRMS
`
`STB
`
`p
`
`Significant
`Predictors
`
`SPMS
`
`STB
`
`p
`
`Adjusted
`R2
`
`0.25
`
`Significant
`Predictors
`
`NABT MTR0
`Gd group0
`
`0.45
`⫺0.29
`
`0.004
`0.059
`
`NABT MTR0
`Lesion MTR0
`Gd group0
`
`0.64
`⫺0.60
`⫺0.28
`
`0.0001
`0.0002
`0.04
`
`0.49
`
`0.12
`
`T2 lesion volume0
`Lesion MTR0
`GMF0
`(none)
`
`⫺0.93 ⬍0.0001
`⫺0.40
`0.027
`⫺0.78
`0.0003
`
`Adjusted
`R2
`
`0.52
`
`0.32
`
`Lesion MTR0
`
`0.38
`
`0.033
`
`⫺0.64
`0.001
`T2 lesion volume0
`⫺0.43
`WMF0
`0.02
`RRMS ⫽ relapsing-remitting multiple sclerosis patients; SPMS ⫽ secondary progressive MS patients; STB ⫽ standardized regression
`coefficients; adjusted R2 ⫽ R2 adjusted for degrees of freedom; BPF ⫽ brain parenchymal fraction; ⌬ ⫽change; NABT ⫽ normal-
`appearing brain tissue; MTR ⫽ magnetization transfer ratio; Gd ⫽ gadolinium; GMF ⫽ gray matter fraction; WMF ⫽ white matter
`fraction.
`
`with more advanced stages of MS. Atrophy rates were
`no different from healthy control subjects in patients
`with CIS who did not convert to MS during the 4-year
`observation. This is consistent with prior reports that
`increased whole-brain atrophy rates were observed in
`CIS patients who progressed to clinically definite
`
`MS.26,36 As shown in Figure 2, annualized rates of
`whole-brain and GM atrophy increased with disease
`stage, from less than 0.2% in CIS patients converting
`to RRMS to almost 0.4% in patients with SPMS. In-
`creasing GM atrophy entirely accounted for increasing
`whole-brain atrophy with advancing disease, as rates of
`
`Table 5. Baseline Plus 4-Year Change Predictors of Atrophy
`
`Dependent
`Variable
`
`BPF, %
`⌬/year
`
`GMF, %
`⌬/year
`
`WMF, %
`⌬/year
`
`RRMS
`
`STB
`
`p
`
`Significant
`Predictors
`
`Significant
`Predictors
`
`Adjusted
`R2
`
`0.72
`
`SPMS
`
`STB