Mean normalized cortical volume (NCV) changes in cognitively stable or improving and deteriorating patients. Error bars indicate SD.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Amato MP, Portaccio E, Goretti B, et al. Association of Neocortical Volume Changes With Cognitive Deterioration in Relapsing-Remitting Multiple Sclerosis. Arch Neurol. 2007;64(8):1157–1161. doi:10.1001/archneur.64.8.1157
We previously reported selective decreases of neocortical volumes in patients with early relapsing-remitting (RR) multiple sclerosis (MS) with mild cognitive impairment, with a good correlation between cortical volumes and cognitive measures.
To assess the relevance of gray matter changes over time to changes in cognition in RRMS.
A longitudinal survey after 2.5 years. Each patient underwent a magnetic resonance imaging (MRI) protocol identical to that performed at baseline; cognitive performance was reassessed with the Rao Brief Repeatable Battery of Neuropsychological Tests in Multiple Sclerosis.
Two university MS clinics.
Of 41 patients with RRMS who participated in the original cross-sectional study, 28 were available for the follow-up evaluation (18 women; mean ± SD age, 37.1 ± 8.9 years; mean ± SD MS duration, 7.3 ± 2.9 years; mean ± SD Expanded Disability Status Scale score, 1.8 ± 1.5).
Main Outcome Measures
We measured the percentage of brain volume changes, normalized cortical volume (NCV) changes, and normalized deep gray matter volume changes on conventional T1-weighted MRIs and changes in lesion load on T2-weighted MRIs. The number of tests failed on the Rao Brief Repeatable Battery were used to classify the patients as cognitively deteriorating or stable or improving.
We identified 12 of 28 cognitively deteriorating and 16 of 28 stable or improving patients. These subgroups did not differ in the mean ± SD percentage of brain volume changes (−2.1% ± 1.2% vs −1.3% ± 1.3%; P = .11), normalized deep gray matter volume changes (−2.1 ± 2.8 mL vs −0.6 ± 3.1 mL; P = .60), and changes in lesion load on T2-weighted MRIs (1.9 ± 2.6 mL vs 1.6 ± 2.3 mL; P = .73). However, NCV changes were significantly higher in deteriorating than in stable or improving patients (−43.0 ± 18.9 mL vs −17.8 ± 26.6 mL; P = .007). In deteriorating patients, NCV changes were correlated with performance in a verbal fluency test (r = 0.73; P < .001). In a regression model, only NCV changes were significantly associated with deteriorating cognitive performance (odds ratio, 0.8; 95% confidence interval, 0.7-0.9).
Progressive neocortical gray matter loss is relevant to MS-associated cognitive impairment and may represent a sensitive marker of deteriorating cognitive performance in RRMS.
Cognitive impairment is a core feature of multiple sclerosis (MS), with prevalence estimates ranging from 40% to 65%.1 Deficits typically involve memory, attention, information processing speed, and executive functions. Although there is great intersubject variability, cognitive deficits may be evident from the early stages of the disease and tend to progress over time.2-4
Several studies have associated cognitive dysfunction in MS with white matter (WM) disease. However, correlation with tissue damage as measured by WM lesion load on T2-weighted magnetic resonance imaging (MRI) and magnetization transfer imaging or with metabolic changes as assessed by magnetic resonance spectroscopy has been modest.1 In contrast, the cognitive deterioration seems to increase in parallel with the decrease of brain parenchymal volume rather than with the increase of brain lesion load.5
The recent development of computerized techniques that allow accurate measurements of regional brain volumes has revealed a specific role of gray matter (GM) disease in the development of cognitive impairment in MS.6-9 In particular, in a previous cross-sectional study,6 we reported selective decreases of neocortical volumes in patients with early relapsing-remitting (RR) MS with mild cognitive impairment, with a good correlation between cortical volumes and cognitive measurements.
On the basis of these findings, we hypothesized that the decrease of neocortical volumes over time may represent a sensitive marker of cognitive deterioration in RRMS. Therefore, we performed a 2.5-year follow-up evaluation of neuropsychological and MRI measurements in the patients with MS studied in our cross-sectional study,6 with the aim of assessing the time-dependent nature of GM changes and their importance on the progression of cognitive impairment in MS.
Twenty-eight of the original sample of 41 patients with RRMS6 were available for the follow-up evaluation, which took place after a mean ± SD period of 2.5 ± 1.1 years. The study sample consisted of 18 women and 10 men whose mean ± SD age was 37.1 ± 8.9 years and mean ± SD educational level was 11.2 ± 3.3 years. At time 2, all patients still had an RR disease form, the mean ± SD disease duration was 7.3 ± 2.9 years, and the mean ± SD score on the Expanded Disability Status Scale10 was 1.8 ± 1.5. All 28 patients were relapse free and were not taking steroids for at least 1 month before MRI and neuropsychological assessment. No patient was taking psychoactive drugs or substances that might interfere with neuropsychological performance. At time 2, all 28 patients were treated with interferon beta-1a (Rebif, 22 or 44 μg; mean ± SD treatment duration, 1.8 ± 0.6 years).
Comparing the 28 patients available for the follow-up study with the 13 patients lost at follow-up, we did not find any significant difference in terms of the main demographic and clinical characteristics, with the exception of the disease duration, which was significantly shorter in the group of patients lost at follow-up (5.1 ± 5.8 years vs 7.3 ± 2.9 years; P = .03). Baseline cognitive performance was comparable in the 2 groups, with 7 of 13 (54%) and 16 of 28 (57%) patients classified as cognitively impaired at time 1 (P = .80).
The study was approved by the Ethics Committee of the Faculty of Medicine of the University of Siena, where patients with MS underwent the MRI protocol. An informed consent was obtained from all participating patients at baseline and follow-up evaluations.
For each patient, neuropsychological testing at time 2 was performed within 1 week of MRI examination by the same neuropsychologist who performed the test at time 1 and who was blinded to the MRI results. The neuropsychological performance was reassessed by using the alternate version B of the Rao Brief Repeatable Battery of Neuropsychological Tests in Multiple Sclerosis.11 This battery incorporates tests of verbal memory acquisition and delayed recall (Selective Reminding Test and Selective Reminding Test–Delayed Recall), spatial memory acquisition and delayed recall (10/36 Spatial Recall Test and 10/36 Spatial Recall Test–Delayed), sustained attention, concentration and speed of information processing (Paced Auditory Serial Addition Test at 3 and 2 seconds; Symbol Digit Modalities Test), and verbal fluency on semantic stimulus (Word List Generation). Consistent with baseline evaluation, a test was considered failed when the score was at least 2 SDs below the mean normative values.12 The cognitive performance of the patient was considered deteriorating when the number of tests failed at time 2 was greater than at time 1, improving when it was lower, and stable when the patient failed the same number of tests at both evaluations. Finally, depression was reassessed with the Montgomery and Asberg Rating Scale for Depression.13
All patients were examined on a Philips Gyroscan operating at 1.5 T (Philips Medical Systems, Best, the Netherlands) using the same MRI protocol. This protocol was identical to that performed at baseline14 and included (1) a transverse dual-echo, turbo spin-echo sequence (repetition time, 2075 milliseconds; echo time 1, 30 milliseconds; echo time 2, 90 milliseconds; 256 × 256 matrix; 1 signal average; 250 × 250-mm field of view) that yielded proton density (PD) and T2-weighted images with 50 contiguous 3-mm-thick slices, acquired parallel to the line connecting the anterior and posterior commissures; and (2) transverse T1-weighted images (repetition time, 35 milliseconds; echo time, 10 milliseconds; 256 × 256 matrix; 1 signal average; 250 × 250-mm field of view) that yielded images of 50 contiguous 3-mm-thick slices, oriented to match exactly the PD/T2-weighted images acquisition. No major hardware upgrades were performed on the scanner during the study, and monthly quality assurance sessions confirmed the stability of measurements throughout the study.
Classification of T2-weighted lesion volume (LV) was performed on each patient by a single observer (M.L.S.), who was unaware of the patients' identity, using a segmentation technique based on user-supervised local thresholding. Lesion borders were determined primarily on PD-weighted images, but information from T2- and T1-weighted images was also considered. Such information was considered because the software used (Jim 3.0; Xinapse System, Leicester, England) offered the ability to switch among the PD, T2-weighted, and T1-weighted images, providing the operator with convenient access to the information in both data sets, while defining lesions and facilitating the discrimination of cerebrospinal fluid from the periventricular plaques. The value of total brain LV was calculated by multiplying lesion area by slice thickness. The coefficient of variation mean was 5% or less in serial measurements.
In all patients, longitudinal (2 time points) normalized percentage of brain volume changes (PBVCs) were estimated on T1-weighted images with SIENA (structural image evaluation using normalization of atrophy).14 To estimate changes between the images, SIENA finds all brain surface edge points using tissue-type segmentation and then correlates differentiated 1-dimensional perpendicular profiles taken around the position of these points in both images. This technique gives edge motion to subvoxel accuracy. Thus, the method is relatively insensitive to changes in intensity of tissues from one scan to the next. Brain atrophy is quantified by taking the mean perpendicular edge motion over all edge points and converting this to the PBVC. A variety of validation tests showed the accuracy in measuring the PBVC to be approximately 0.2%.14
Because the cross-sectional version of the SIENA method (SIENAX) allows for regional measures of tissue volume,15 normalized values of cortical GM were separately assessed on both time points as previously described.16 The difference between measurements of 2 time points of the same patients provided changes of normalized cortical volume (NCV) and percentage of changes in NCV. Also, selective measures of deep GM were obtained by using the standard space-based masks that selectively included the GM present in the basal ganglia. In each patient, the difference between these measures at the 2 time points provided the changes in normalized deep GM.
Group analyses were performed using the t test for unpaired samples, the nonparametric Mann-Whitney test, and a χ2 test when appropriate. Relationships between MRI and cognitive variables were assessed using the nonparametric Spearman rank order correlation. Data below the .05 level were considered statistically significant. To assess possible predictors of the patient's cognitive outcome at time 2 (deteriorating or stable or improved), we used a stepwise logistic regression model that included age, educational level, number of tests failed at time 1, PBVC, T2-weighted LV changes, NCV changes, and normalized deep GM changes as possible predictors. The SPSS software version 12.0 running on Windows (SPSS Inc, Chicago, Illinois) was used for the analysis.
On the basis of their neuropsychological performance on the Rao Brief Repeatable Battery of Neuropsychological Tests in Multiple Sclerosis, 12 patients (43%) were classified as deteriorating, 11 as stable (39%), and 5 as improving (18%). The number and type of tests failed at time 2 and comparisons with time 1 are reported in Table 1 and Table 2. Comparing patients with a deteriorating cognitive performance with patients with a stable or improving performance, we did not find any significant difference in terms of the main clinical and demographic characteristics (Table 3). Moreover, 4 patients in the cognitively stable or improving subgroup and 3 in the cognitively deteriorating subgroup (25%) were classified as depressed on the Montgomery and Asberg Rating Scale for Depression.13
When comparing patients with MS and a deteriorating cognitive performance and those with a stable or improving performance, quantitative MRI analysis showed no significant differences in mean ± SD T2-weighted LV changes (1.9 ± 2.6 mL vs 1.6 ± 2.3 mL; P = .73), PBVC (−2.1% ± 1.2% vs −1.3% ± 1.3%; P = .11), and mean ± SD normalized deep GM changes (−2.1 ± 2.8 mL vs −0.6 ± 3.1 mL; P = .60). However, both values of NCV changes (Figure) and percentage of NCV changes were significantly more pronounced in MS patients with a deteriorating cognitive performance than in those with a stable or improving performance (mean ± SD NCV changes: −43 ± 18.9 mL vs −17.8 ± 26.2 mL; P = .007; mean ± SD percentage of NCV changes: −6.2% ± 2.5% vs −2.1% ± 3.6%; P = .002). Also focusing the analysis on the 7 patients who failed 2 or more tests compared with their baseline performance, we confirmed that the percentage of NCV changes was significantly higher in deteriorating than in stable or improving patients (mean ± SD percentage of NCV changes: −4.8% ± 2.2% vs −2.1% ± 3.6%; P = .04), whereas the difference in NCV did not reach the level of statistical significance (mean ± SD NCV changes: −32.9 ± 17.0 mL vs −17.8 ± 26.6 mL; P = .10).
In the whole MS sample, measures of NCV changes showed a trend toward a correlation with the change in the number of tests failed by the patients (r = −0.33; P = .08) and a significant correlation with change of scores in the Word List Generation test (r = 0.57; P < .001). This latter correlation was stronger (r = 0.73; P < .005) when we considered only the subgroup of patients who showed cognitive deterioration during the follow-up period. No other close relationship was found between changes in MRI and cognitive measures. Finally, in a stepwise logistic regression model, NCV change was the only MRI parameter significantly associated with deteriorating cognitive performance (β = −.2; odds ratio, 0.8; 95% confidence interval, 0.7-0.9; P = .04).
This study represents, to our knowledge, the first longitudinal assessment of the contribution of GM disease to the progression of MS-associated cognitive impairment. In our cohort, after a mean follow-up of 2.5 years, 43% of the patients showed cognitive deterioration, whereas the remaining patients were cognitively stable or improved. Previous longitudinal studies have highlighted the great interpatient variability in the cognitive outcome of patients with MS. However, a few well-designed longitudinal studies have consistently shown that nearly one third of patients deteriorate during 2- or 3-year follow-up periods2,4,5,17 and the progression of cognitive dysfunction has been confirmed in a controlled 10-year observation.3 In this study, we considered as deteriorating those patients who failed 1 or more tests compared with their baseline performance. Relying on the number of tests failed to provide an estimate of cognitive deterioration has been previously applied in longitudinal studies.3,18 Moreover, we focused our study on patients with RRMS at early disease stage; therefore, because of the short follow-up and possible practice effects, we expected only minor cognitive changes. This approach was also consistent with our definition of mild cognitive impairment at baseline,6 which required the failure of at least 1 test, thus allowing a direct comparison of the results of the 2 studies.
In our sample, neocortical volume loss over time was significantly more pronounced in patients with a deteriorating cognitive performance than in patients who remained stable or improved. These values showed a good correlation with performance on a verbal fluency test, especially in cognitively deteriorating patients. In contrast, changes in both total brain volumes and T2-weighted LV were not significantly different between cognitively deteriorating or stable or improving patients, and these MRI parameters did not seem to correlate with cognitive scores. Moreover, in a stepwise logistic regression model, changes in neocortical volumes were the only significant correlate of deteriorating cognitive performance. Similar results were also obtained by using a more conservative approach in the definition of cognitive change (ie, failure of 2 or more tests compared with baseline performance). This finding further supports our conclusions on the relevance of GM volume changes to cognitive impairment in patients with early RRMS. Therefore, in our sample, cognitive deterioration seemed to proceed in parallel with the decrease of neocortical volumes.
Several recent pathological and imaging studies19-25 have recognized that neocortical abnormalities are relevant in MS, can be detected even in the earliest stages of the disease, and may be, at least in part, unrelated to WM lesion accumulation.16,23 It is not clear, however, whether the cortical volume loss is associated with a diffuse neurodegeneration of the neocortex21,26-30 or the presence of neocortical inflammatory lesions that are rarely detected by conventional MRI.25,31,32 Independently of the underlying pathological mechanisms, the present data add to prior evidence6,8,9,33 in supporting the relevance of GM disease in MS. They also extend previous information by reporting evidence that MS-related cognitive dysfunction is closely associated with progressive neocortical changes in early RRMS, even after a relatively short follow-up period.
In previous cross-sectional studies,6,8,9,34,35 smaller GM volumes were associated with poorer performances on specific neuropsychological tests. In particular, in one of these studies,8 performance on the Paced Auditory Serial Addition Test was correlated with GM volume in brain regions associated with working memory and executive functions. In the present longitudinal study, we found only a significant correlation between NCV and changes in scores on a verbal fluency test. This finding may be due to the small sample size and limitations of serial administration of neuropsychological tests during a short follow-up period. Moreover, our GM volume estimates did not allow assessment of specific brain regions that may also explain the paucity of correlations found in our study. In this context, studies of larger cohorts of patients during a longer period and the assessment of specific GM regions may improve correlations between neuropsychological and MRI findings.
Interestingly, the results of our study did not show a clear involvement of the deep GM in the pathological process, since we observed comparable deep GM volume decreases in patients with stable or deteriorating MS. However, structural and metabolic abnormalities of the deep GM have been documented in a number of previous MS studies,36-41 and its pathological features have been indicated as a possible factor contributing to MS-related cognitive impairment.42 It is possible that the small sample size used in the present study and the technical difficulties in performing an accurate segmentation in this brain region with the present postprocessing imaging procedures may have contributed to our negative results. Probably, studies on larger patient populations using model-based parcellation or diffusion tractography for a specific evaluation of subcortical structures43 could provide a more appropriate answer to this interesting issue.
In this study, we used quantitative MRI to investigate the underlying pathological basis of cognitive dysfunction in MS. From our results, neocortical atrophy emerges as a significant determinant of progressive cognitive dysfunction to a greater extent than can be explained by conventional WM lesion assessment. Therefore, MS-related cognitive impairment may be, at least in part, due to mechanisms that are not related to focal WM lesion genesis. However, the relationship between brain disease and cognition remains complex, and it likely involves multiple factors, brain structures, and reciprocal connections.44-47 Further research is needed to clarify these complexities.
Correspondence: Maria Pia Amato, MD, Department of Neurology, University of Florence, Viale Morgagni, 85-50134 Florence, Italy (email@example.com).
Accepted for Publication: January 29, 2007.
Author Contributions:Study concept and design: Amato and De Stefano. Acquisition of data: Amato, Portaccio, Goretti, Zipoli, Bartolozzi, Stromillo, Guidi, Siracusa, and De Stefano. Analysis and interpretation of data: Amato, Portaccio, Battaglini, Stromillo, Sorbi, Federico, and De Stefano. Drafting of the manuscript: Amato, Portaccio, Goretti, Zipoli, Battaglini, Bartolozzi, Stromillo, Guidi, Siracusa, and De Stefano. Critical revision of the manuscript for important intellectual content: Amato, Sorbi, and Federico. Statistical analysis: Portaccio and De Stefano. Administrative, technical, and material support: Guidi. Study supervision: Amato, Goretti, Zipoli, Stromillo, Sorbi, and Federico.
Financial Disclosure: None reported.
Create a personal account or sign in to: