Ventricular enlargement in MS

MRI studies of patients with multiple sclerosis (MS) have reported the presence and progression of brain atrophy occurring to a greater extent and at a higher rate than that observed in age-matched healthy controls.^1–3 Increasing atrophy has been observed at all stages of disease, from the time of onset with a clinically isolated syndrome (CIS) suggestive of MS through the typical relapsing remitting phase to the secondary and primary progressive stages of disease.^4–10 Accelerated loss of brain tissue in MS reflects loss of axons, neurons, and myelin, and brain atrophy measurement might be a marker of disease progression.^11,12 A correlation between increasing brain atrophy and a greater degree of neurologic or cognitive impairment has been reported.^13,14 Brain atrophy measures are frequently used as outcome measures in treatment trials.^15–19

Although atrophy is well documented at all stages of MS, it is less clear whether the rate and location of tissue loss is constant or if it varies over time and between clinical subgroups. Knowledge of the rate and timing of disease-related atrophy may have implications for its use as a surrogate marker in different patient subgroups.¹⁰ A plausible hypothesis is that there will be greater atrophy in secondary progressive MS (SPMS) than relapsing remitting MS (RRMS) because increasing axonal loss is considered to be an important mechanism of progressive disability per se. It is difficult to address this hypothesis from existing studies because they have used a large variety of MR acquisition sequences and image analysis methods that make a direct comparison difficult.

In this study, we report on the 1 year rate of ventricular enlargement in MS from cohorts of patients with early and established RRMS and SPMS who were scanned using the same MR sequences and in whom the same analysis procedure was used. We chose ventricular enlargement (VE) as a measure of brain tissue loss as it has been proven sensitive in detecting change in previous studies of all the MS subgroups studied^1,5 and because it could be measured using a relatively simple, semi-automated and highly reproducible technique.^1,5 We also studied changes in T2 and T1 hypointense lesion load in order to evaluate their extent of change and relationship with the tissue loss measure in each of the clinical subgroups. Persistent T1 hypointense lesions exhibit more axonal loss than T1-isointense lesions and a subsidiary hypothesis was that there would be a larger increase in T1 hypointense lesion volume in SPMS.

Methods.

Patients, aged between 17 and 50 years, who presented to our clinics at Moorfields Eye Hospital or the National Hospital for Neurology and Neurosurgery with CIS suggestive of MS less than 3 months from the onset of symptoms were recruited. Appropriate investigations including full blood count, erythrocyte sedimentation rate, autoantibody screen, syphilis serology, vitamin B12, and MRI brain were used to exclude other diagnoses. Fifty-six patients were recruited consecutively from January 1996 until June 1999. A CIS was defined as a single event of acute onset in the CNS suggestive of demyelination, e.g., unilateral optic neuritis, brainstem, and partial spinal cord syndromes. Individuals with a history of previous neurologic symptoms suggestive of demyelination were excluded. Other diagnoses were excluded by appropriate investigations. VE has previously been described in this group of patients at 1 year.^4,5 Clinical examination, MRI, and Expanded Disability Status Scale (EDSS)²⁰ were performed at each visit. Diagnosis of MS was based on the new McDonald criteria.²¹ The cohort was subdivided into patients with a diagnosis of MS (designated early MS for between group comparisons) and CIS at 1year.

The MRI scans of 64/79 patients with relapsing MS from the previously published placebo arm of a randomized double-blind placebo-controlled trial of natalizumab were analyzed.²² The remaining patients did not have the appropriate electronic data. Inclusion criteria were an EDSS between 2 and 6.5,²⁰ Poser criteria clinically definite or laboratory supported definite MS,²³ either relapsing remitting RRMS or SPMS, a history of at least two relapses within the previous 2 years, a minimum of three lesions on T2-weighted brain MRI, and ages 18 to 65. Patients were subdivided according to their diagnosis of RRMS or SPMS. Patients were excluded if they received immunosuppressive of immunomodulatory treatments within the past 3 months or experienced a relapse or received systemic corticosteroids within the past 30 days.²²

In patients with CIS, brain MRI was performed using a 1.5 Tesla GE scanner. Patients were imaged on three occasions: at baseline, which in all instances was within 3 months of the onset of symptoms, approximately 3 months later (volumes not included in this study), and approximately 1 year later. Proton density/T2-weighted fast spin-echo (FSE) sequences (repetition time [TR] 3,200 msec, effective echo time [TE] 15/95 msec) and a T1-weighted spin echo sequence (TR 600 msec, TE 14 msec) were acquired in each patient, with 3 mm contiguous, axial slices. The matrix used was 256 × 256 with a field of view of 24 cm. An IV bolus of 0.1 mmol/kg gadolinium DTPA was given 5 to 7 minutes prior to the commencement of image acquisition.

During the screening phase (month −1), immediately before treatment (month 0 to 5), 1 month following the last treatment (month 6), and during safety follow-up at months 9 and 12, proton density (PD) and T2-weighted dual echo (FSE) were acquired using TR = 2.5 to 3.3 seconds; TE = 20–40/80–100 msec in patients with RR and SP MS. T1-weighted spin echo sequences (TR 500 to 700, TE 10 to 20 msec) were acquired in each patient pre gadolinium (months −1, 0, 6, 12) and post gadolinium (−1, 0, 1, 2, 3, 4, 5, 6, 9, and 12). For each sequence 46 axial oblique, contiguous, 3-mm-thick slices were obtained using a matrix of 256 × 256, field of view of 25 cm, and one or two excitations.

All MRI studies performed at the individual centers were archived onto hard copy film and electronic media and transported to the MRI analysis transferred to the MRI Analysis Centre (Institute of Neurology, University College London, Queen Square, London, UK).

The hard copies were analyzed by a clinical rater with one neuroradiologist working by consensus, who roughly outlined all T1-weighted Gd enhancing, T2 hyperintense, and T1 hypointense lesions. In a parallel process, follow-up MRI studies were then compared with the baseline scans in order to identify all new gadolinium enhancing, T2 hyperintense, and T1 hypointense lesions.

The ventricles were measured on baseline and at 1 year of follow-up T1-weighted post gadolinium scans using the MIDAS interactive image analysis package by a single observer blinded to patient details in a previously described method.^4,5,24 The observer was blinded to the order of scans (baseline or 1 year follow-up). The initial step was segmentation of the whole brain using a semiautomated interactive morphologic technique with the image intensity threshold for the boundary between the CSF and brain set at 60% mean signal intensity. This method was also used to calculate brain volumes. The inferior cut-off was taken at the lowest point of the cerebellum. Ventricular volume consisted of the lateral ventricles and temporal horns but excluded the third and fourth ventricles. This was measured using a semiautomated seed placing technique, involving voxels with an image intensity of less than 60% of the mean. The reproducibility of the ventricular measurement technique was assessed by measuring and re-measuring ventricular volumes after 7 days by a single observer blinded to patient details. The mean coefficient of variation was 0.89% prior to starting image analysis.

All lesions marked on the hard copy were outlined on the computer image using a semiautomated local thresholding technique, or, if the lesion could not be outlined satisfactorily by this approach, manual outlining was performed.²⁵ After all the lesions had been outlined, the clinical rater checked for consistency of lesion identification from hard copy to computer generated image. Any lesions missed or incorrectly marked on either the hard copy or computer generated image were outlined. A computer program then summed all the individual lesion volumes (calculated as the surface area of each lesion multiplied by the slice thickness [3 mm]) and T2 hyperintense and T1 hypointense lesion volumes were generated and stored in a specially constructed database.

Regarding statistical analysis, between group comparisons of change in ventricular volume from baseline and 1 year scans were made using the general linear model univariate analysis of variance. Ventricular enlargement was input as the dependent variable and the MS subgroup (early MS, RRMS, and SPMS) was input as a fixed factor with age, disease duration, sex input as covariates. Within group comparisons of the baseline and 1 year follow-up ventricular volumes were made using the Wilcoxon signed ranks test.

Results.

VE between group comparisons.

Using the univariate regression model with VE over 1 year input as dependent variable, allowing for age there were differences in the VE among the three MS groups (p = 0.048) (table 3 and figure). Parameter estimates revealed significant differences between SPMS vs RRMS and early MS vs SPMS. There was no difference in VE between early MS and RRMS.

View this table:

• View inline
• View popup
• Download powerpoint

Table 3 Multiple regression with ventricular volume as dependent variable, MS type as fixed factor, and age as covariate

• Download figure
• Open in new tab
• Download powerpoint

Figure. Ventricular, T2, and T1 volume changes over 1 year in each patient subgroup.

Lesion load measurements are shown in table E-1 (available on the Neurology Web site at www.neurology.org) and lesion regression analyses in table E-2.

In the group of patients presenting with a CIS, 40/56 had an abnormal MRI brain scan at baseline (12/26 who remained CIS and 28/30 who developed early MS). All 64 patients with RR and SPMS had T2 lesions on their baseline scan. Baseline median T2 lesion volume in early MS was 1.6 mL, in RRMS was 7.3 mL, and in SPMS was 16.6 mL (0.4 mL in CIS).

Using the univariate regression model with T2 lesion volume change and T1 hypointense lesion volume change over 1 year input as dependent variables, allowing for age there were no significant differences in the T2 lesion volume change among the three MS groups. There were significant differences in the T1 hypointense lesion volume change among the three MS groups. Parameter estimates revealed significant differences in T1 hypointense lesion volume change between SPMS vs early RRMS (table E-2 and figure).

Discussion.

Using ventricular enlargement as a measure of tissue loss, this study investigated the rate of cerebral atrophy occurring in patients with early relapse onset MS, established RRMS, and SPMS, and also investigated the relationship of T2 and T1 hypointense lesion volume changes with clinical subgroups and the atrophy measure.

This study confirms that progressive atrophy occurs at all stages of MS (mean VE for all MS subgroups 1.1 mL and median 0.5 mL). Compared to patients with diagnosis of CIS at 1 year, there was more VE over 1 year in all MS groups (early relapse-onset MS, established RRMS, and SPMS). In addition, there was significantly more atrophy occurring in the SPMS group when compared with either early relapse-onset MS or established RRMS. Although the SPMS group was older, the analysis allowed for an effect of age per se and the greater rate of VE in the SPMS group remained significant. In contrast, no differences in the rate of VE were noted between the group of patients with early MS and those with RRMS. These observations suggest that the rate of atrophy is fairly constant during the RR phase of MS and that there is accelerated atrophy associated with the development of secondary progression.

While the finding of a greater rate of atrophy in SPMS is not surprising, the mechanisms for this are less clear. One factor could be that the larger load of demyelinated white matter lesions are more susceptible to lose further axons, either from a loss of trophic support by myelin or due to a greater susceptibility to neurotoxicity from activated microglia and other inflammatory mediators. The greater accumulation of T1 hypointense lesions indicates that there is probably more axonal loss accumulating in focal white matter lesions in SPMS.²⁶ There may also be mechanisms for neuroaxonal loss arising in the normal appearing white and gray matter that are more prominent in SPMS. The presence of more extensive pathologic abnormality in such tissues in SPMS has been inferred from studies using other quantitative MR measures such as the magnetization transfer ratio or mean diffusivity.^27,28

VE was significantly related to lesion load in early MS and established RRMS, especially T1 hypointense lesions. The increased association between T1 lesions (as compared with T2 lesions) and atrophy confirms other study reports.^29–33 Lesions with persistent T1 hypointensity have more axonal loss.²⁶ As lesions develop in association with axonal loss—the latter is probably mediated by several mechanisms both during and following acute inflammation^34,35—it is expected that measurable tissue loss will follow. This may explain the relationship observed between baseline lesion load and subsequent VE in the early relapse-onset and RR cohorts.

In contrast, the lack of significant relationship between lesions and VE in SPMS suggests that the mechanisms of VE are less dependent on lesion load in later disease. Atrophy may reflect variable rates of axonal loss in chronic lesions, which is due to factors unrelated to lesion load per se, or—as discussed above—it may be influenced to a greater extent by pathologic processes occurring in the normal appearing white and gray matter.³⁶ Gray matter atrophy has been reported in several recent studies of patients with MS, and may in part relate to the occurrence of focal gray mater lesions that can be abundant on careful neuropathologic examination^37,38 but almost never seen on conventional MRI.

This study of ventricular enlargement at various stages of disease is a retrospective analysis of prospectively acquired data from two separate studies. The patients recruited with CIS suggestive of MS and early relapsing remitting MS of 1 year duration are clearly younger, with shorter disease duration and lower EDSS than the patients with established RRMS and SPMS. A limitation with this study is that there is no age-matched healthy control group. However, the CIS-only group—in whom no increase in VE was observed—provided a meaningful pathologic control for comparison because 1) it is clinically relevant to compare the findings in CIS patients who have an early conversion to MS vs those who do not, and 2) a majority of the CIS-only group (14/26) had normal brain MRI, and long-term follow-up studies indicate that only about one-fifth of CIS subjects with normal MRI go on to develop MS.³⁹ Although there were age differences between the clinical subgroups, the inclusion of age as a covariate in the analysis make it unlikely that age per se has contributed to the significant subgroup differences observed. Furthermore, the mean VE in our whole MS cohort (VE = 1.0 mL [interquartile range –0.129 to 1.3]; mean age 38.7 years, range 17 to 62) was comparable with that reported in another study of 26 MS patients of similar age (mean VE = 1.6 mL [interquartile range 0.8 to 3.4]; mean age 44.7 years, range 27 to 65) and greater than that reported from the same study in 26 healthy controls who were also of a similar age (mean VE = 0.3 mL [interquartile range –0.1 to 0.8]; mean age 47 years, range 30 to 59).¹

There were differences in recruitment of the patient cohorts: the patients with CIS and early relapse-onset MS were recruited as part of a natural history study and all their MRI scans were acquired at a single center, whereas the patients with established RR and SP MS were part of the placebo arm of a phase II multicenter trial. However, both cohorts had clinical demographic features and T2 lesion volumes that are very typical for their stage of MS and the MRI acquisition protocols were similar—strict quality control was also applied to all scans and electronic data from the multicenter study to ensure that they complied fully with the prespecified trial scanning protocol. Also, although the observer who measured the ventricular size was aware that the electronic data were from the multicenter trial or the CIS cohort, she was not aware of the scan acquisition order nor of any clinical details beyond their being in either the natural history or trial cohorts: the observer was blinded to the subgroups in the trial cohort in whom significant differences in VE emerged (SPMS patients showing more VE than RR subgroups).

A limitation of the study is that only VE is used as a measure of atrophy and therefore gives no insight into whether atrophy is occurring in gray matter, white matter, or both. There are limitations to inferring the rate of atrophy in SPMS given the time constraints of the study material, which allowed us to look at each group (from early MS to SPMS) at two time points 1 year apart. Ideally, in a much longer study a large group with CIS would be followed through to progression to SPMS in order to fully understand the evolution of atrophy. Finally regional variations in patterns of brain atrophy have been investigated recently and differences noted according to clinical phenotype.¹⁰ A strength of the present study is that it includes three distinctive MS subgroups in terms of disease duration, course, and disability and thus provides further information on MS subgroup patterns of atrophy.

Acknowledgment

The authors acknowledge Mike Panzara, Allison Hulme, Omar A. Khan, William A. Sheremata, Lance D. Blumhardt, Michele A. Libonati, Allison J. Willmer-Hulme, and the International Natalizumab Multiple Sclerosis Trial Group.