DTI Correlates of Cognition in Conventional MRI of Normal-Appearing Brain in Patients with Clinical Features of Subacute Combined Degeneration and Biochemically Proven Vitamin B12 Deficiency

BACKGROUND AND PURPOSE: Vitamin B12 deficiency may cause neural injury that results in cognitive deficits. The main purpose of our study was to evaluate morphometric and microstructural changes in the brain and relate them to cognition in subacute combined degeneration of the spinal cord and patients with biochemically deficient vitamin B12. MATERIALS AND METHODS: Fifty-one patients were recruited and underwent nerve-conduction velocity tests and routine hematologic examinations. Serum vitamin B12 and homocystine levels were also measured. All patients and 46 age- and sex-matched controls underwent cervical spine and brain MR imaging along with cognition tests. MR imaging included conventional scans and DTI. Voxel-based morphometry was performed for determining the WM and GM volumes, based on T1-weighted images. DTI measures that included fractional anisotropy, ADC, radial diffusivity, and axial diffusivity were determined by using tract-based statistics. RESULTS: None of the patients showed any abnormality on conventional MR imaging. No significant changes in GM and WM volumes were observed in patients compared with controls. Significant reductions in the fractional anisotropy and an increase in ADC and radial diffusivity values were observed in multiple brain regions in patients compared with controls. These changes were confirmed on the region-of-interest analysis. Neuropsychological scores were significantly different in patients compared with controls and showed significant correlation with fractional anisotropy and radial diffusivity in a few brain regions. CONCLUSIONS: Microstructural changes are seen in WM regions on DTI in patients with vitamin B12 deficiency and correlate with cognition scores. DTI can be used for objective assessment of microstructural changes in the brain in vitamin B12 deficiency.

B vitamins contribute to CNS development and proper functioning by acting as a cofactor in numerous catalytic reactions in the human body that are required for the synthesis and functioning of neurotransmitters and myelination. A deficiency of vitamin B 12 may result in injury to the neural tissue. A limited number of clinical studies in children demonstrated a correlation between vitamin B 12 deficiency and cognition. 1,2 Studies in elderly subjects suggest that vitamin B 12 deficiency is associated with cognitive decline and may contribute to Alzheimer dementia, 3,4 whereas others have failed to demonstrated an increased risk. 5,6 Subjects with B 12 deficiency may also show changes in the posterolateral column of the spinal cord on MR imaging. Clinical symptoms relating to neuropathy and spinal cord involvement, referred to as subacute combined degeneration (SACD) of the spinal cord, are common in adult subjects with B 12 deficiency. Changes in the brain parenchyma on MR imaging have been sporadically reported, with poor sensitivity. 7,8 In a cross-sectional study on an elderly population, a reduction in brain volume was observed with B 12 deficiency. 9 There are isolated reports of brain demyelination in patients with SACD, which may or may not show resolution following vitamin B 12 replacement. 10,11 Elderly populations with vitamin B 12 deficiency are reported to show whole-brain atrophy and white matter damage. 12,13 On the basis of a number of case reports, brain atrophy is also a pathologic feature in infants with vitamin B 12 deficiency. [14][15][16] Advanced MR imaging-based modalities such as DTI and MR spectroscopy showed abnormalities in a variety of diffuse neurologic disorders, whereas conventional MR imaging findings appeared normal. We hypothesized that patients with clinical symptoms of SACD and biochemical evidence of vitamin B 12 deficiency will have an associated cognitive decline and microstructural alterations in brain WM on DTI, even when conventional MR imaging findings appear normal. To verify this hypothesis, we performed whole-brain DTI and cognitive assessment in patients who presented with clinical signs of SACD and a biochemical deficiency of vitamin B 12 . The DTI measures were correlated with neuropsychological test scores. We also performed voxel-based morphometry analysis for volumetric changes in GM and WM. To the best of our knowledge, this is the first study to quantify the microstructural changes in normal-appearing brain on MR imaging in patients with SACD and biochemically confirmed vitamin B 12 deficiency.

MATERIALS AND METHODS
Patients with clinical features of peripheral neuropathy suspected of having spinal cord involvement and clinically labeled as having SACD were included in this study. Fifty-one patients who met the above criteria were recruited (38 men, 13 women; mean age, 34.6 Ϯ 12.2 years; age range, 18 -58 years). Forty-six age-and sex-matched healthy controls (not on any medication and not known to have any disease) were also included in the study (33 men and 13 women; mean age, 31.1 Ϯ 8.0 years; age range, 18 -53 years). No significant differences in age and sex were observed between the patient group and healthy controls.

Clinical Assessment
Neurologic examinations were performed by a neurologist to assess the severity of impairment in patients with vitamin B 12 deficiency. The patients had detailed nerve-conduction velocity tests and biochemical analysis for serum vitamin B 12 and homocysteine levels. The diagnosis of vitamin B 12 deficiency was based on low serum vitamin B 12 levels (Ͻ200 pg/mL). 17 The serum homocysteine level was measured by an enzymatic method. 18 Routine hematologic examinations that included hemoglobin levels and red cell mean corpuscular volume were also performed. Healthy controls were evaluated for serum vitamin B 12 levels and neurologic and cognitive status; however, they did not undergo nerveconduction velocity tests and other biochemical analyses. All patients who showed vitamin B 12 deficiency and had clinical features of SACD underwent MR imaging of both the cervical spine and brain. All patients with normal brain MR imaging findings with or without imaging changes in the cervical spine underwent cognitive testing.
Neuropsychological tests were performed on both patients and controls by an experienced neuropsychologist. This battery included the Trail-Making test, number connection tests A and B, and figure connection tests A and B as well as the performance subset of the modified Wechsler Adult Intelligence Scale (modified for the population), which included picture completion, digit symbol, block design, picture arrangement, and object assembly. These tests evaluate visuospatial capacity and visuomotor speed.
The Trail-Making test assesses the visual motor coordination, concentration, attention, mental speed, and memory alteration. 19 In the number connection and figure connection tests A and B, lower scores represent better performance, whereas in the Wechsler Adult Intelligence Scale, a higher score represents a better performance.
This study protocol was approved by the Institutional Ethics Committee. Informed written consent was obtained from each subject.

MR Imaging Analysis
Structural Analysis. Voxel-based morphometry analysis was performed by using SPM8 (Wellcome Department of Imaging Neuroscience, London, UK; http://www.fil.ion.ucl.ac.uk/spm) to determine possible changes in the GM and WM volumes. Before processing, the images were visually inspected for possible artifacts. For tissue segmentation, a customized template was built from the sample by using a nonlinear registration algorithm (DARTEL toolbox in SPM8). 20 This template was registered to the Montreal Neurological Institute (MNI) template for group comparisons. The Jacobian determinants from the normalization procedure were used to modulate the voxelbased morphometry data to preserve the WM and GM volumes. Individual GM and WM images were smoothed with an isotropic Gaussian kernel of 6-mm full width at half maximum before statistical analysis. Global volumes of GM and WM were assessed from segmented images by using the VBM8 toolbox in SPM8 after correcting for age and sex.

Diffusion Tensor Image Processing
The Diffusion Toolbox software tool in the FMRIB Software Library (FSL; http://www.fmrib.ox.ac.uk/fsl/fdt/index.html) was used for calculating the DTI indices fractional anisotropy (FA), ADC, axial diffusivity (AD), and radial diffusivity (RD). The DWI was corrected for eddy current-induced distortions and minor head movements by using affine registration to the reference B0 images. The Brain Extraction Tool was used for extracting the brain. 21

Tract-Based Spatial Statistics and Voxelwise Analysis
Voxelwise analysis of FA was performed by using tract-based spatial statistics (TBSS), 22 part of the FSL package. 23 Individual skeletonized FA maps were aligned to the MNI 152 template by using the Nonlinear Registration Tool in FMRIB. 24 Each subject's aligned FA map was then projected onto this skeleton, and the voxelwise general linear model was applied by using permutationbased nonparametric testing, corrected for multiple comparisons. Using the same registration parameters from the FA maps, we also spatially transformed ADC, RD, and AD maps to the MNI space.

Region-of-Interest Analysis
In addition to the TBSS analysis, the region-of-interest analysis was also performed on those regions that were observed to be significantly different on the TBSS analysis. Elliptic regions of interest of sizes varying from 25 to 50 mm 2 were placed on different regions in the FA maps of patients and controls.

Statistical Analysis
The differences in the neuropsychological scores between patients and controls were analyzed with independent t tests by using the FIG 1. Differences in DTI measures between patients and controls. Statistical maps show voxels that exhibit differences in DTI parameters in patients versus controls (red and yellow colors, according to the lower and higher significance, respectively; dark blue and light blue colors, according to the lower and higher significance, respectively). FA is significantly decreased, and ADC and RD are increased in the patient group compared with controls. Differences are widespread and evident in various WM regions. All WM tracts are overlaid on a 1-mm standard image in MNI 152 (TBSS analysis, 2-sample, P Ͻ .05, threshold-free cluster enhancement corrected). Statistical Package for Social Sciences software, Version 16.0l (IBM, Armonk, New York). The correlation between the neuropsychological score and DTI-derived indices of WM was based on the Pearson coefficient. All statistical analyses were based on a 2-tailed test with an ␣ level of Ͻ.05 for statistical significance. Morphologic differences between the patients and control subjects were estimated by using an independent-samples t test at the voxel level within the general linear model framework of statistical parametric mapping. Comparison between patients and control subjects was made for 2 different contrasts, corresponding to an increase (patient Ͼ controls) or decrease (patient Ͻ controls) in GM and WM volumes. A false discovery rate at P Ͻ .05 was used. Differences were considered significant if the cluster size was Ͼ15. 25 The general linear model was applied across all subjects to identify the brain regions in which the patient group showed significant differences in FA, ADC, AD, and RD relative to the healthy control group. The correlation analyses were performed to study the relationship between neuropsychological scores and each of the DTI-derived indices FA, ADC, AD, and RD in the WM by using neuropsychological scores as regressors in the framework of a general linear model. The effects of age and sex were regressed out in these models. DTI-derived maps were included in a nonparametric permutation-based group model by using "Randomize" in FSL. An independent t test was used to determine the differences in FA, ADC, RD, and AD values obtained from the region-of-interest analysis between patient and control groups.

Clinical Assessment
The mean hemoglobin level in patients with vitamin B 12 deficiency was 10.95 Ϯ 2.28 g/dL (range, 6.8 -16 g/dL), the mean serum vitamin B 12 concentration was 145.26 Ϯ 42.07 pg/mL (range, 26 -199 pg/ mL), the mean cell volume was 103.6 Ϯ 12.26 fL (range, 81.3-122.3 fL), and the serum homocysteine level was 19.84 Ϯ 7.32 mol/L (range, 7.9 -41.29 mol/L). The mean serum vitamin B 12 concentration in healthy controls was 330.22 Ϯ 135.22 pg/mL (range, 255-678 pg/mL). All patients had gait disturbance, sensory disturbance, mental impairment, and neuropathy. A total of 11.8% of patients presented clinically with pyramidal tract damage. Differences were observed in all the neuropsychological scores between patients and controls (Table).

Conventional MR Imaging
MR imaging of the cervical spine showed T2 hyperintensity in the posterior spinal cord in 7 patients and diffuse hyperintensity of the cervical cord in 1 patient. The remaining 43 patients did not show any abnormality in the cervical cord on conventional MR imaging.

Voxel-Based Morphometry Analysis
The voxel-based morphometry analysis of the 3D-T1WI did not show a significant difference in the cerebral GM and WM volumes between patients and controls.

TBSS Analysis
TBSS analysis showed significantly reduced FA values in patients compared with controls in a number of WM regions, which included the frontal, parietal, and temporal lobes and the entire corpus callosum and its associated fibers (Fig 1). The patient group also had a widespread increase in ADC and RD values, predominantly in the right hemisphere tracts (right Ͼ left) and the corpus callosum.

Correlation Analysis of DTI Indices with Neuropsychological Scores
Correlation maps showed a significant positive correlation between the Digit Symbol scores and FA values (Fig 2A) and a significant negative correlation between the number connection test and FA values (Fig 2B) in many WM bundles (P corrected Ͻ .05). RD values were found to correlate negatively with Digit Symbol scores ( Fig 3A) and positively with number connection test scores (Fig 3B). ADC and AD did not show any significant correlation  with neuropsychological scores. The mean FA and RD values were extracted subject by subject from correlation maps from regions showing significant correlation between DTI indices and neuropsychological scores. Thereafter, Pearson correlation coefficients were obtained by using SPSS in the Digit Symbol test, with FA (r ϭ 0.63, P Ͻ .001 in patients; r ϭ 0.56, P Ͻ .001 controls) and RD (r ϭ Ϫ0.54, P Ͻ .001 in patients; r ϭ Ϫ0.53, P Ͻ .001 in controls); and the number connection test with FA (r ϭ Ϫ0.56, P Ͻ .001 in patients; r ϭ Ϫ0.22, P ϭ .14 in controls) and RD (r ϭ 0.57, P Ͻ .001 in patients; r ϭ 0.403, P Ͻ .005 in controls) to verify the results obtained on correlation maps.

Region-of-Interest Analysis
Region of interest-based analysis confirmed the TBSS results. The results are summarized in Fig 4. AD values did not show a significant change between patients and controls either on the TBSS or region-of-interest analysis.

DISCUSSION
This study shows widespread changes in the cerebral WM in patients with vitamin B 12 deficiency in all the DTI metrics, except AD, indicating altered WM microstructure in multiple regions in these patients. These results were confirmed in the region-of-interest analysis. In addition, all these patients showed significant cognitive decline over the controls. The cognitive scores correlated with DTI measures in various brain regions. Our findings suggest that the microstructure changes in WM are quite widespread in the brain, even when patients showed clinical symptoms related to the spinal cord. We did not observe any volume changes in WM and GM on voxel-based morphometry analysis, suggesting that atrophy is not a major pathologic component in these patients.
FA is widely considered a robust measure of WM organization, and a number of studies have reported abnormalities in FA across both psychotic and affective illnesses. 26 Disruptions in WM organization (reflected in reductions in FA) can result from various mechanisms, including demyelination as well as axonal loss and lacks pathologic specificity. However, the other DTI-derived measures, such as RD and AD, are thought to reflect myelin and axonal integrity, respectively. 27 While still controversial, some published literature suggests that RD is a more sensitive measure of myelin integrity, while AD reflects axonal integrity. 28,29 In the current study, we observed increased RD and unchanged AD, suggesting that demyelination, not axonal loss, is the major pathologic substrate in B 12 deficiency. Scalabrino and Veber 30 have demonstrated, in the rat model, that vitamin B 12 deficiency damages myelin and causes myelin vacuolation and reactive astrocytosis in the CNS. In a recent study, Minn et al 31 reported that patients with SACD have progressive degeneration in the following sequential order: lower spinal cord, cervical spinal cord, peripheral nerve/optic nerve, and, finally, the brain. Our data do not support this view and show that B 12 deficiency causes demyelination in the brain even when it may manifest clinically with SACD.
In a prospective study on 50 patients with vitamin B 12 deficiency, Reynolds 32 has shown cognitive impairment or an affective disorder in one-third of patients. Minn et al 31 did not find any symptoms of dementia in any of their patients who presented with SACD. However, in the current study, all 51 patients showed sig- nificant cognitive deficits on neuropsychological tests compared with the age-and sex-matched controls, suggesting that a complete battery of tests may be needed to detect cognitive impairment in these patients. Dementia and cognitive changes have been typically reported in elderly subjects with B 12 deficiency. In contrast, we detected cognitive deficits in our relatively young patient cohort.
Cognitive decline has been observed at all age groups in patients with vitamin B 12 deficiency. Sensitive neuropsychological tests have shown an association with vitamin deficiency. 33 In the current study, we also observed a decline in visuospatial and performance skills in these patients. Another group has reported cognitive decline in patients between 35 and 50 years of age with vitamin B 12 deficiency, which improved following its supplementation. 34 Most studies in the geriatric population show conflicting results. Some authors have found a strong causal relationship between cognitive decline and vitamin B 12 deficiency, while others have described the relationship as mere coincidence in the geriatric population. 3,5,12 The mechanism for cognition decline is the accumulation of methylmalonic acid secondary to nonactivation of methylmalonyl-CoA mutase, which is myelinotoxic. 35 We also observed a strong correlation of the neuropsychological scores with FA, ADC, and RD changes in some brain regions on DTI.

CONCLUSIONS
Microstructural changes in various brain regions are demonstrated on DTI metrics in patients with SACD, and these are associated with abnormal neuropsychological scores and show a correlation with various specific brain regions. The imaging technique may be of value in the objective assessment of the brain changes in B 12 deficiency.
Disclosures: Ponnada Narayana-UNRELATED: Grants/Grants Pending: National Institutes of Health,* Department of Defense. **Money paid to the institution.