Review
`
`Lancet Neurol 2006; 5: 158–70
`
`Department of Neurology,
`Cleveland Clinic Foundation,
`Cleveland, OH, USA
`(R A Bermel MD); and Center for
`Neurological Imaging, Partners
`Multiple Sclerosis Center,
`Departments of Neurology and
`Radiology, Brigham and
`Women’s Hospital, Harvard
`Medical School, Boston, MA,
`USA (Rohit Bakshi MD)
`
`Correspondence to:
`Dr Rohit Bakshi, Brigham &
`Women’s Hospital, Harvard
`Medical School, 77 Avenue Louis
`Pasteur, HIM 730 Boston,
`MA 02115, USA
`rbakshi@bwh.harvard.edu
`
`The measurement and clinical relevance of brain atrophy in
`multiple sclerosis
`
`Robert A Bermel, Rohit Bakshi
`
`Brain atrophy has emerged as a clinically relevant component of disease progression in multiple sclerosis.
`Progressive loss of brain tissue bulk can be detected in vivo in a sensitive and reproducible manner by MRI.
`Clinical studies have shown that brain atrophy begins early in the disease course. The increasing amount of data
`linking brain atrophy to clinical impairments suggest that irreversible tissue destruction is an important
`determinant of disease progression to a greater extent than can be explained by conventional lesion assessments.
`In this review, we will summarise the proposed mechanisms contributing to brain atrophy in patients with
`multiple sclerosis. We will critically discuss the wide range of MRI-based methods used to quantify regional and
`whole-brain-volume loss. Based on a review of current information, we will summarise the rate of atrophy among
`phenotypes for multiple sclerosis, the clinical relevance of brain atrophy, and the effect of disease-modifying
`treatments on its progression.
`
`Historical background
`Brain-volume loss was reported as a component of
`multiple sclerosis in early descriptions of the pathology.
`In 1938, Robert Carswell discussed multiple sclerosis in
`an article on atrophy in his Atlas of Pathology.1 In 1963,
`German pathologist Eduard Rindfleisch reported focal
`atrophy of brain tissue in his description of the
`perivascular nature of multiple sclerosis lesions.2 With
`advances
`in
`technology,3
`structural neuroimaging
`provides increasingly sensitive methods of monitoring
`brain atrophy in vivo. In this review we will focus on
`brain atrophy, referring the readers to separate reviews
`on atrophy of the optic nerve and spinal cord in multiple
`sclerosis.4–8
`
`Pathogenesis
`Although brain atrophy is probably an endpoint of
`irreversible tissue loss in multiple sclerosis, it is not
`pathologically specific. The underlying mechanisms for
`brain atrophy are diverse and complex. Some proposed
`mechanisms are directly related to the multifocal
`inflammatory disease process, whereas others are
`indirectly related to or are independent from traditional
`measures of overt lesions.9
`
`Overt lesions
`Multiple sclerosis lesions are dynamic and progress
`through distinct stages on MRI scans. Lesions that are
`enhanced with intravenous-gadolinium contrast on T1-
`weighted images represent active areas of inflammation
`and dysfunction of the blood–brain barrier.10,11 Most
`gadolinium-enhancing
`lesions become permanent
`hyperintense
`lesions on T2-weighted
`images and
`represent a range of pathological changes (eg, oedema,
`gliosis, inflammation, demyelination, remyelination,
`and axonal loss). About half of gadolinium-enhancing
`lesions will ultimately persist as severe hypointensity on
`T1-weighted images (T1 black holes).11 These persistent
`lesions represent irreversible tissue destruction and
`axon loss.10,12,13
`
`Focal-tissue loss within overt white-matter lesions is
`probably a major contributor to brain atrophy due to loss
`of myelin and axonal density.10,14 But is there a relation
`between lesion load and brain atrophy, even remote to
`lesion development? One group studied 29 patients with
`a clinically
`isolated demyelinating syndrome over
`14 years.15 Progression of T2-hyperintense lesion load in
`the first 5 years after disease onset was the best predictor
`of whole-brain atrophy 14 years later, suggesting that
`lesions partly contribute to the development of global
`atrophy. However, most studies examining the relation
`between T2-hyperintense
`lesions and whole-brain
`atrophy reported that lesion load accounts for at most
`10% of the variance in atrophy in patients with
`relapsing-remitting multiple sclerosis, and has limited
`predictive value for the subsequent development of
`atrophy.16–18 An increasing amount of data show that
`whole-brain
`grey-matter
`volume
`is moderately
`correlated with lesion volume, suggesting that remote
`effects of white-matter lesions are partly responsible for
`grey-matter
`degeneration.16,19–23
`However,
`this
`association might not extend to atrophy in deep grey-
`matter nuclei, in which atrophy is poorly related to such
`lesions.24
`the
`lesions partly predict
`Gadolinium-enhancing
`development of whole-brain atrophy in early relapsing-
`remitting multiple sclerosis in some studies.14,25–33 One
`study26 reported that the number of enhancing lesions
`was moderately correlated with ventricular enlargement
`in 16 patients with relapsing-remitting multiple
`sclerosis. Three other studies of patients with similar
`characteristics lend support to this association.27–29 Two
`groups showed that ring-enhancing lesions might be
`particularly predictive of brain atrophy.26,30 However,
`results from two studies17,31 suggest that gadolinium-
`enhancing lesions do not predict whole-brain atrophy.
`Overall, the relation between enhancing lesion volume
`and brain atrophy seems to disappear in later disease
`stages.32 Even with near complete pharmacological
`suppression of gadolinium-enhancing
`lesions
`in
`
`158
`
`http://neurology.thelancet.com Vol 5 February 2006
`
`MYLAN PHARMS. INC. EXHIBIT 1096 PAGE 1
`
`

`
`Review
`
`patients with secondary progressive multiple sclerosis,
`brain atrophy continues to progress.33
`The association between T1-black holes and brain
`atrophy however, is unclear.14,17,34,35 One study34 recorded a
`relation between T1-black holes and supratentorial brain
`atrophy (r=0·48) in a cross-sectional study of patients
`with relapsing-remitting multiple sclerosis
`(n=55).
`Results from our group supported this finding in 78
`patients with multiple sclerosis (r=0·35),35 and we also
`reported an association between subcortical atrophy and
`volume of T1-black holes.36 This finding contrasts with
`studies showing that the volume of T1-black holes at
`baseline does not predict
`the development of
`longitudinal whole-brain atrophy in relapsing-remitting
`multiple sclerosis.17
`Are multiple sclerosis lesions and brain atrophy linked,
`or are they unique results of different pathological
`processes? On the basis of the above results, there seems
`to be a partial link between lesions and brain atrophy.
`One key mechanism linking MRI lesion load to brain
`atrophy is the remote effect of axonal injury on neuronal
`loss (Wallerian degeneration). A study of the pathology37
`has shown the abundance of transected axons in the brain
`of patients with multiple sclerosis (more than 11 000 per
`mm3 of lesional tissue). Wallerian degeneration can be
`detected at the earliest stage of multiple sclerosis (in
`patients with a clinically
`isolated demyelinating
`syndrome).38 However, the slight predictive value of
`lesion volumes suggests that other mechanisms might be
`contributing to volume loss.14 There has been a recent
`effort to correlate diffuse damage in the normal-
`appearing brain tissue with brain atrophy. One group
`used MR spectroscopy to quantify N-acetyl aspartate as a
`marker of neuronal integrity and showed that decreased
`whole-brain N-acetyl
`aspartate preceded
`global-
`
`parenchymal loss in 42 patients with relapsing-remitting
`multiple sclerosis.39 Another group, measuring whole-
`brain N-acetyl aspartate and diffusivity, showed these
`markers
`(but not any
`lesion-load volumes) were
`associated with whole-brain atrophy.40 These studies
`provide evidence to show that in addition to overt lesion
`formation, global dysfunction or degeneration of neurons
`might be occuring, and perhaps preceding brain atrophy.
`
`Alternative mechanisms
`The absence of a strong association between lesion load
`and brain atrophy suggests that other mechanisms
`contribute to atrophy. There has been evidence of a
`genetic
`predisposition
`to
`brain
`atrophy
`or
`neurodegeneration in multiple sclerosis. Specifically,
`patients who carry the APOE ⑀4 allele have up to five
`times greater annual rate of brain atrophy than do
`patients who do not have the allele.41 Patients with this
`genotype are also at risk for developing persistent T1-
`black holes.41 Brain atrophy has also been associated
`with damage in grey matter, as measured by T2
`hypointensity of subcortical structures, and is thought to
`represent pathological iron deposition18,42,43 (figure 1).
`
`Localisation of brain atrophy
`Brain atrophy seems to be widespread in multiple
`sclerosis, affecting all brain regions, including all the
`cerebral
`lobes,
`the compact white-matter
`tracts,
`brainstem, and cerebellum.44 Both grey-matter and
`white-matter compartments are affected. Several groups
`have assessed grey-matter versus white-matter volume
`loss using various segmentation tools, and selective
`grey-matter atrophy has been documented in patients
`with a clinically isolated demyelinating syndrome or
`relapsing-remitting multiple
`sclerosis.16,20,21,23,24,45–47
`
`Patient A
`
`Baseline T2 intensity values:
`Mean thalamus=0·44
`Mean caudate=0·48
`Left putamen=0·42
`Right putamen=0·42
`Left globus pallidus=0·28
`Right globus pallidus=0·32
`Mean red nucleus=0·39
`
`Baseline BPF=0·82
`
`Year 2 BPF=0·77
`
`Patient B
`
`Baseline T2 intensity values:
`Mean thalamus=0·59
`Mean caudate=0·63
`Left putamen=0·59
`Right putamen=0·59
`Left globus pallidus=0·44
`Right globus pallidus=0·44
`Mean red nucleus=0·46
`
`Baseline BPF=0·85
`
`Year 2 BPF=0·85
`
`Figure 1: Midthalamic T2-weighted axial images from two placebo patients at baseline
`We did a post-hoc analysis of patients with MS from a 2-year clinical trial of interferon beta-1a (30␮g IM weekly) or placebo.18 We determined deep grey-matter T2
`hypointensity, brain parenchymal fraction (BPF), and total T2-, gadolinium-enhancing-, and T1-lesion volumes. T2 hypointensity was chosen in regression modeling
`as the best predictor of BPF change at the 1 year (R2=0·23, p=0·002) and 2 year (R2=0·33, p⬍0·001) time points after accounting for all MRI variables. Mid-thalamic
`T2-weighted axial images from two placebo patients at baseline are shown, relating visual to quantitative findings, showing a range of T2 intensity and BPF values.
`Patient A (left) had relative baseline T2 hypointensity and atrophy, with marked atrophy progression. Patient B (right) had higher baseline T2 intensity and BPF
`values, with no atrophy progression. Thus, grey matter T2-hypointensity predicts the progression of brain atrophy in patients with early relapsing-remitting multiple
`sclerosis. This predictive effect is seen as early as the first year. We hypothesise that brain atrophy may involve iron deposition as a mediator of neurotoxicity or as a
`disease epiphenomenon. Reproduced with permission from the American Medical Association.18
`
`http://neurology.thelancet.com Vol 5 February 2006
`
`159
`
`MYLAN PHARMS. INC. EXHIBIT 1096 PAGE 2
`
`

`
`Review
`
`matter volume loss, which could have been the result of
`atrophy, was offset by inflammation or oedema-related
`increases in tissue bulk. These data suggest that
`decreases in grey-matter volume can identify progressive
`neurodegeneration more sensitively than can decreases
`in white matter or whole-brain volumes.
`Atrophy of the deep grey nuclei also seems to happen
`in multiple sclerosis (figures 2 and 3). Tissue loss from
`the thalamus52,53 and the caudate nucleus24 has been
`measured with MRI
`segmentation. There was
`substantial neuron loss, with neuronal density decreased
`by 22%, in the thalamus of patients with multiple
`sclerosis.52
`This
`pathological
`evidence
`for
`neurodegeneration in the grey matter highlights the
`underlying causes of damage detected non-invasively by
`imaging studies, such as hypometabolism,54 decreased
`neuronal activation,55,56
`increased diffusivity,57
`and
`decreased N-acetyl aspartate.52
`
`Methods used to measure brain atrophy
`Qualitative measures
`from
`identified
`Clinically, brain atrophy can be
`qualitative images by the recognition of an increase in
`cerebrospinal fluid spaces or a reduction in size of
`parenchymal structures compared with the normal
`appearance for age (figure 4). Upon review of serial
`studies, progressive atrophy can be detected by
`comparison of images (figure 5). Such relatively simple
`determinations are easy to implement in routine patient
`care or for semiquantitative analysis,44 and can help to
`assess disease severity and disease progression.
`
`Quantitative two-dimensional measures
`Because of the limited reproducibility and precision of
`visually
`based
`atrophy measures,
`quantitative
`
`I cm
`
`Figure 3: Deep grey-matter damage might be an important component of
`the MS disease process
`Superior view of three-dimensional caudate reconstructions from a patient with
`MS (green) and a normal control (pink), both age 50 years old, normalised for
`head size. Surfaces were reconstructed from volume acquisition coronal MR
`images using a manual parcellation technique, normalised for head size, and co-
`registered using Visualization Toolkit.24
`
`Normal
`n=10
`
`MS
`n=24
`
`8000
`
`7000
`
`6000
`
`5000
`
`4000
`
`3000
`
`Caudate volume (mm3)
`
`Figure 2: Box plot of normalised caudate volumes
`We tested whether caudate atrophy occurs in MS, and whether it correlates with
`conventional MRI or clinical markers of disease progression. Caudate nuclei of
`24 patients with MS and ten age-matched healthy controls were traced,
`normalised, reconstructed, and visualised from high-resolution MRI scans.
`Normalised bicaudate volume was 19% lower in MS vs controls (p⬍0·001), an
`effect that persisted after adjusting for whole-brain atrophy (p⬍0·008).
`Caudate volume did not correlate with total brain T2 hyperintense or
`T1-hypointense lesion load (both p⬎0·05). Reproduced with permission from
`Lippincott, Williams & Wilkins.24
`
`Cortical atrophy happens early in the course of multiple
`sclerosis45,48,49 and can be localised to specific regions of
`the brain. One group measured cortical thickness in
`20 patients with multiple sclerosis and noted the
`presence of focal atrophy in the bilateral frontal and
`temporal cortices early in the disease course, and
`additionally in the motor cortex in patients with more
`advanced disease.49 Evidence
`from other groups,
`however, suggests that atrophy similarly affects both
`grey and white-matter compartments20,50 or preferentially
`affects the white-matter compartment.19 The similarity of
`the patient populations, small sample sizes, and
`differences in techniques might explain, in part, the
`conflicting conclusions of these studies.
`Data show that the grey matter, although affected by
`the disease, has relatively less inflammation51 compared
`with white matter, which is more prominently affected.
`This process could potentially cause increases in white-
`matter volume and mask ongoing atrophy. Consistent
`with this hypothesis, in a 3 year longitudinal study,16
`patients with early-onset multiple sclerosis had large
`decreases in grey-matter volume but small increases in
`white-matter volume. Additionally,
`those with a
`clinically isolated demyelinating syndrome had loss of
`grey-matter volume, but there were no changes in white-
`matter volume. Both groups of patients had large
`increases in white-matter lesions during the observation
`period, prompting the authors to speculate that white-
`
`160
`
`http://neurology.thelancet.com Vol 5 February 2006
`
`MYLAN PHARMS. INC. EXHIBIT 1096 PAGE 3
`
`

`
`Review
`
`Healthy control
`
`RRMS
`
`RRMS
`
`SPMS
`
`BPF 0·89
`
`BPF 0·84
`
`BPF 0·80
`
`BPF 0·70
`
`Figure 4: Whole brain atrophy in MS as measured by a normalised proportional method
`Representative midventricular axial non-contrast T1-weighted MRI scans are shown from age-matched individuals in the sixth decade. Note the progressive decrease
`in brain parenchyma, increase in cerebrospinal fluid spaces, and decrease in brain parenchymal fraction (BPF) among the scans from left to right. The first patient with
`relapsing-remitting multiple sclerosis (RRMS) has an EDSS score of 1·5 and disease duration of 5 years. The next patient with relapsing-remitting multiple sclerosis
`has an EDSS score of 4·0 and disease duration of 10 years. The patient with secondary-progressive multiple sclerosis has an EDSS score of 6·5 and disease duration of
`18 years. Reproduced with permission from the American Society for Experimental NeuroTherapeutics.11
`
`techniques are preferred. Quantitative two-dimensional
`measures of atrophy include linear measures, which can
`be quantified on a single-image section with a distance
`tool on a computer workstation (or even a ruler on
`
`hardcopy films).29,35,36,58–63 Although
`two-dimensional
`measures have the advantage of relative ease of
`implementation
`in the clinical setting, the main
`disadvantage is the absence of reproducibility compared
`
`Figure 5: Progressive brain atrophy in a 41-year-old man with RRMS imaged at baseline and 4 years later
`Non-contrast T1-weighted images show progressive enlargement of the ventricles and subarachnoid spaces consistent with diffuse brain volume loss. Top row is at
`baseline, bottom row is at 4 year follow-up. Reproduced with permission from the American Society for Experimental NeuroTherapeutics.11
`
`http://neurology.thelancet.com Vol 5 February 2006
`
`161
`
`MYLAN PHARMS. INC. EXHIBIT 1096 PAGE 4
`
`

`
`Review
`
`Figure 6: Two-dimensional measures of brain atrophy
`Third ventricle width (A), maximum lateral ventricle width (B; arrows), brain width (C; arrows), corpus callosum area (D), and bicaudate ratio (E; minimum intercaudate distance [solid line] divided by
`the brain width along the same line [dashed line]). Measurement of third ventricular width is done by measuring the width along the anteriorposterior midpoint of the third ventricle. Lateral ventricular
`width is determined along a plane corresponding to the anteroposterior midpoint of the ventricle on an anatomical level from an axial slice at which the septum pellucidum remains thin. Brain width is
`the distance between two points on the cortical surface, measured at the same level and along the same line as the lateral ventricle width. Corpus callosum area is determined by outlining the margins
`of the structure from the best available midsagittal section The bicaudate ratio, also known as the intercaudate nucleus ratio, is measured from an axial section on which the heads of the caudate
`nuclei are most visible and closest together. These two dimensional measures of brain atrophy have shown longitudinal sensitivity to disease progression,29 meaningful correlations with clinical
`findings,29,36 and strong associations with three dimensional measures of whole brain atrophy.64 Both the third ventricle width and bicaudate ratio show associations with cognitive impairment, even
`after accounting for other MRI biomarkers including whole-brain atrophy and lesion load.36,60 A–D reproduced with permission from Lippincott, Williams & Wilkins29
`
`with three-dimensional measures. Figure 6 shows
`examples of two-dimensional measures.
`
`Quantitative three-dimensional measures
`Quantitative measures of whole-brain atrophy, acquired
`by automated or semiautomated methods,64 have
`become the most powerful methods for assessing
`patients with multiple sclerosis over time because of
`their reproducibility, sensitivity, and potential to capture
`global disease effects.65,66
`Segmentation-based techniques include separation of
`intracranial contents
`into parenchymal and non-
`parenchymal classes, thereby arriving at an expression
`of what proportion of space is occupied by brain tissue at
`any given time. The most popular terminology for this
`“proportional method” is brain parenchymal fraction
`(BPF) (figures 4 and 7).17,35,64,67,68 Other examples of
`techniques which use segmentation-based methods
`include 3DVIEWNIX,69 SIENAX,70,
`index of brain
`atrophy,71 whole brain ratio,72 brain to intracranial cavity
`ratio,73 brain
`to
`intracranial volume ratio,74
`fuzzy
`connectedness,75
`the Alfano method,76 MIDAS,77 and
`ILAB4.78 Segmentation-based
`techniques can also
`separately quantify the volume of grey matter, white
`matter,
`and
`cerebrospinal fluid
`compartments
`(figure 7).23 The abnormal signal intensity of multiple
`sclerosis lesions presents a potential source of voxel
`misclassification. Lesions in the white matter tend to be
`misclassified as grey matter,20 and misclassified volumes
`should be corrected to avoid confounding clinical
`correlations.23
`Registration-based techniques are designed to follow
`patients longitudinally over time.79 These methods align
`two serial scans from a patient and find areas of signal-
`intensity change after accounting for changes in head
`position or slice plane. The result is a number such as
`percent brain volume change. Examples of such
`measures include SIENA,80 voxel-based morphometry,81
`
`(figure 7),23,64,67,82,83
`statistical parametric mapping
`template-driven segmentation,84 and brain boundry shift
`integral.79 The advantages of registration-based methods
`are that they are sensitive to longitudinal changes and
`(because they are mostly automated) need little operator
`input and time. The main disadvantage is the need for
`additional steps in image processing.
`Thus, there are a multitude of techniques now
`available to measure brain atrophy. However, the main
`limitation of the use of brain atrophy in clinical trials is
`that different measures can produce conflicting results.
`Differing results might be valid,85 but as a result of the
`low specificity of brain atrophy, techniques can quantify
`different physiological observations.
`
`Atrophy and disease course
`Clinically isolated syndrome
`There is increasing evidence that brain atrophy is not
`restricted to the later progressive stages, but begins in
`the earliest stages of multiple sclerosis.6 In patients with
`clinically isolated demyelinating syndrome, whole-brain
`atrophy is detectable in those who go on to develop
`multiple
`sclerosis.86,87
`Patients who meet
`the
`international panel criteria for multiple sclerosis88 (but
`who have had only a single demyelinating event) had
`detectable ventricular enlargement over the subsequent
`year (median enlargement=0·3 cm3, n=27) compared
`with those who do not meet the criteria (median
`enlargement=0·05 cm3, n=28, p=0·03).77 In another
`study of patients with clinically isolated demyelinting
`syndrome followed up over 3 years, the group that
`developed multiple sclerosis during the study period
`had detectable grey-matter atrophy, but no white-matter
`atrophy.16
`In the placebo group of a randomised
`controlled trial of subcutaneous interferon beta-1a in
`clinically
`isolated
`demyelinating
`syndrome,
`investigators measured a mean decrease in brain-
`parenchymal volume of 1·68% over 2 years.89 In
`
`162
`
`http://neurology.thelancet.com Vol 5 February 2006
`
`MYLAN PHARMS. INC. EXHIBIT 1096 PAGE 5
`
`

`
`Review
`
`Figure 7: Representative slices from a T1-weighted gradient echo series and obtained grey, white, and CSF tissue compartments after automated skull-
`stripping and segmentation by SPM99 for a normal control (top row) and a patient with MS (bottom row)
`Images from left to right are: raw T1-weighted image; white matter only (bright areas); grey matter only (bright areas); CSF only (bright areas) with background brain
`parenchyma (inner dark areas). Note the misclassification of white-matter lesions that have been classified as grey matter. This procedure can be used to quantify the
`volumes separately, or determine residual or proportion-based measures of normalised whole-brain atrophy. BPF=total parenchymal volume divided by total
`intracranial volume. Applying this method we studied 41 patients with MS and 18 age and sex-matched healthy controls. We also measured lesion load (total T1-
`hypointense and FLAIR hyperintense lesion volume), third ventricular width and bicaudate ratio. Disability was assessed by the EDSS and timed 25-foot walk. Patients
`with MS had lower grey matter (⫺3·9%, p=0·003) and total parenchymal volume (⫺3·8%, p=0·003), but only a trend for lower white-matter volume (⫺3·7%,
`p=0·052) relative to normal controls. Grey-matter atrophy was most closely related to clinical status as compared to all other MRI measures (see figure 8 and table for
`more information). Reproduced with permission from Academic Press.23
`
`summary, when brain atrophy occurs in clinically
`isolated demyelinating syndrome it seems to be related
`to the underlying multiple sclerosis disease process, as
`pathological brain atrophy is more common in patients
`who go on to develop multiple sclerosis than in those
`who do not.
`
`Relapsing-remitting multiple sclerosis
`Brain atrophy has been well documented in relapsing-
`remitting multiple sclerosis, and seems to happen
`rapidly in early stages of the disease.6 Overall, the rate of
`brain-parenchymal volume loss in relapsing-remitting
`multiple sclerosis is 0·6–1·35% per year. In the pivotal
`trial of weekly intramuscular interferon-beta 1a, patients
`with early relapsing-remitting multiple sclerosis (mean
`disease duration 6·2 years) and only mild disability
`(mean expanded disability status scale (EDSS] 2·3) had a
`lower whole-brain volume than age-matched normal
`controls (5 SD below the normal control group) and a
`higher annual rate of brain-volume loss than controls
`over 2 years (0·6% per year vs ⬍0·1% per year for
`healthy controls).17 Another study measured brain
`parenchymal volume in 53 patients with early untreated
`relapsing-remitting multiple sclerosis (disease duration
`1–5 years, EDSS⬍5) and reported that volume loss
`averaged 2·7% over 2 years.90 Other groups have
`observed the rates of brain atrophy in either untreated
`patients31,68,79,91–95 or those in clinical trials.74,89,96–100
`
`Secondary-progressive multiple sclerosis
`Studies which have assessed both patients with
`relapsing-remitting multiple sclerosis and those with
`secondary-progressive multiple sclerosis have reported
`that the annual rates of atrophy are similar between the
`two groups, with rates between 0·6% and 0·8% per
`year.6,79,91,98 In a clinical trial of an immunomodulatory
`drug in secondary-progressive multiple sclerosis, the
`placebo group (n=31) had a mean brain-volume loss of
`3·86% over 3 years.32 Other studies showed that brain
`atrophy occurred at a lower rate in secondary-progressive
`multiple sclerosis than in relapsing-remitting multiple
`sclerosis,92,101 suggesting that the most aggressive rates of
`atrophy happen early on
`in
`the disease course,
`coincident with rapid accumulation of lesion burden.33
`
`Primary-progressive multiple sclerosis
`In a multicentre study of progressive multiple
`sclerosis,102 137 patients with primary-progressive
`disease averaged 1·3%
`loss of brain-parenchymal
`volume over 1 year. Another group103
`assessed
`41 patients with primary-progressive multiple sclerosis
`and showed a mean reduction in brain-parenchymal
`volume of 3·7% over 5 years. In the placebo group of a
`therapeutic trial in this type of multiple sclerosis,
`ventricular volume increased over 2 years.104 Therefore
`brain atrophy happens
`in patients with primary-
`progressive multiple sclerosis at a rate greater than in
`
`http://neurology.thelancet.com Vol 5 February 2006
`
`163
`
`MYLAN PHARMS. INC. EXHIBIT 1096 PAGE 6
`
`

`
`Review
`
`Neuropsychological impairment
`Multiple neuropsychological domains are affected by
`multiple
`sclerosis,
`including
`cognition, mood,
`personality, fatigue, and quality of life.108–117 An early
`study118
`identified an association between
`third
`ventricular width and cognition. Multiple subsequent
`studies have confirmed the link between brain atrophy
`and neurobehavioural status.36,48,60,62,74,90,108,110,119–125 Our
`group60 reported that third ventricular width was the
`most significant MRI predictor of cognitive dysfunction
`in patients with multiple sclerosis. In this same study,
`whole-brain atrophy (ie, brain parenchymal fraction)
`was the next significant predictor, and notably both
`atrophy measures predicted more
`variance
`in
`neuropsychological
`function
`than did MRI-lesion
`measures. Another study90 noted a strong correlation
`between whole-brain parenchymal volume (but not
`lesion loads) and cognitive function in early relapsing-
`remitting multiple sclerosis (r=0·51, p=0·0003). Many
`studies suggest
`that subcortical atrophy
`is well
`correlated with cognition, to a greater extent than whole-
`brain atrophy or lesion load.36,60,124 Grey-matter atrophy48
`and lobar atrophy125 have also shown correlations with
`cognitive dysfunction, to a greater extent than seen with
`lesion-load measures.
`From the studies on brain atrophy and neuropsycho-
`logical impairment we conclude that subcortical atrophy
`has a stronge association with general cognitive
`dysfunction, which can be explained by a disruption of
`frontal-subcortical circuits. Regional or cortical atrophy
`measures might provide insight into impairment of
`more specific functional domains in some patients.
`
`Effect of technical or other factors on volume
`measurement
`A prerequisite for the use of brain volume as a surrogate
`marker for irreversible tissue damage and neuro-
`degeneration in patients with multiple sclerosis is the
`need to understand the potential effect of other factors
`such as MRI-related technical error and biological
`factors that can affect brain volume independent of
`tissue destruction.
`
`Effect of MRI protocol and analysis approach
`Changes in the position of the patient’s head or slice
`plane between scans are potential sources of error when
`measuring serial changes in brain volume. Either a
`normalisation or co-registration procedure should
`address these concerns. Normalised measures of whole-
`brain volume (eg, expressed as a ratio to head size) show
`a higher degree of reproducibility and sensitivity than do
`those expressed as absolute volumes and have about
`twice the statistical power.65 This degree of difference
`might improve the ability to detect treatment effects in
`clinical trials.
`In studying the longitudinal reproducibility of brain
`parenchymal fraction, investigators did two serial MRI
`
`those with secondary-progressive multiple sclerosis, but
`not as aggressively as reported in some studies of
`relapsing-remitting multiple sclerosis.6
`
`Clinical correlations
`Physical disability
`Lesion-load measures on MRI are associated only weakly
`with physical disability in cross-sectional studies, which
`has led to the traditional “clinical-imaging paradox” of
`multiple sclerosis.105 Lesions can appear on MRI without
`any progression in physical disability, and conversely,
`patients can progress in disability without the appearance
`of new lesions. Whole-brain atrophy has a stronger, yet
`moderate, imaging association with physical disability,
`and is a stronger predictor of future disability than T1-
`hypointense and T2-hyperintense lesion load.6 Many
`groups have tested the relation between whole-brain
`atrophy and EDSS score. One study showed that the rate
`of progression of brain atrophy over 18 months was
`significantly heightened in patients with deterioration of
`disability, whereas only trends were seen between
`disability change and progression of T2-hyperintense
`lesions.96 Another group reported significant correlations
`between EDSS score and ventricular volume, but no
`association between lesion loads and EDSS.106 Our group
`and others have recorded similar associations between
`brain atrophy and measures of physical disability
`(figure 7, figure 8, table).22,23,35
`the subsequent
`Early atrophy seems
`to predict
`development of physical disability to a better extent than
`lesion-load measures. A longitudinal study107 reported
`that patients with more atrophy during the first 2 years
`had greater disability 8 years later. Furthermore, the
`change in brain parenchymal fraction over the first
`2 years was the best MRI predictor of 8 year disability
`after accounting for the 2 year change in all MRI-lesion
`measures. At 8 years, patients with the largest amount of
`brain atrophy were about four times more likely to have a
`disability level needing assistance with walking, or worse.
`Thus, brain atrophy is emerging as a clinically relevant
`measure of multiple sclerosis disease progression,
`supported by its association with physical disability.
`Spinal cord atrophy seems to show particularly strong
`relations with physical disability, as reviewed separately.4
`
`25-foot timed walk (residual)
`
`EDSS score (residual)
`
`Whole grey-matter volume (residual)
`
`Whole grey-matter volume (residual)
`
`Figure 8: Correlation of whole-brain normalised grey-matter volume with physical disability
`Left: corrected whole grey-matter volume with EDSS scores (partial r=⫺0·46, p=0·004). Right: