Altered White Matter Integrity in Human Immunodeficiency Virus-Associated Neurocognitive Disorder: A Tract-Based Spatial Statistics Study

Subjects

This prospective study was approved by the Institutional Review Board and informed consent was obtained from all patients and HC. Between December 2012 and September 2014, 22 HIV patients (10 with HAND and 12 without HAND [HIV-IC]) underwent DTI and standardized NP tests (4) within a one-month interval. Patients with comorbid conditions that could influence cognitive status and/or daily functioning were excluded, according to the criteria described elsewhere (4, 6). We also recruited 11 age- and sex-matched HC without objective cognitive impairment.

Assessment of Cognitive Status

To assess the cognitive status, all the HIV patients underwent NP tests designed to represent six cognitive domains: 1) speed of information processing (Korean version of Wechsler Adult Intelligence Scale [K-WAIS] digit symbol subtest and trail making test part A [TMT A]); 2) memory including learning and recall (Korean version of auditory verbal learning test [K-AVLT] and complex figure test); 3) abstraction/executive function (Wisconsin Card Sorting Test [WCST], K-WAIS similarity subtest, and trail making test part B [TMT B]); 4) attention/working memory (K-WAIS digit span subtest); 5) sensory perception/motor skills (Grooved Pegboard Test); and 6) verbal/language (K-WAIS vocabulary subtest). The scores of TMT A, TMT B, and the Grooved Pegboard Test represent time in seconds spent to complete a given task and higher scores denote poor performance. The scores of the other tests are represented as either a standard score or age-scaled score and higher scores denote better performance. Cognitive status was classified as impaired when the scores were greater than 1 standard deviation below the demographically adjusted normative mean from published results with HIV-negative subjects (13, 14, 15, 16, 17). Frascati criteria (6) were used in the outpatient clinic to clinically diagnose HAND.

Statistical Analyses for Demographics and NP Test Results

Age and years of education were compared between the three groups using ANOVA and the Kruskal-Wallis test. The two-sample t test and Mann-Whitney test were performed to compare NP test results between the HAND and HIV-IC groups. All the statistical analyses were performed using SPSS version 20.0 (IBM Corp., Armonk, NY, USA). P < 0.05 was considered statistically significant.

Image Acquisition

All the scans were acquired using a 3T scanner (MAGNETOM Trio Tim, Siemens, Erlangen, Germany) with a 32-channel head coil. Head motion was minimized with restraining foam pads provided by the manufacturer. An axial diffusion tensor single-shot echo-planar imaging acquisition was performed with the following parameters: 60 non-collinear, non-coplanar directions using b = 3000 s/mm² with 7 baseline images without diffusion weighting; number of excitation = 1; field of view = 230 mm; echo time = 109 ms; repetition time = 10000 ms; flip angle = 90°; slice thickness = 2.3 mm; in-plane resolution = 2.3 × 2.3 mm²; and 72 axial slices. Total acquisition time was 11 minutes 40 seconds.

Post-Processing of DTI Data and TBSS Analysis

All the DTI data were processed using the Functional Magnetic Resonance Imaging of the Brain (FMRIB) Software Library (FSL) program (http://www.fmrib.ox.ac.uk/fsl/). Motion artifacts and eddy current distortions were first corrected by normalization of each directional volume to the non-diffusion-weighted volume (b0) using the FMRIB Linear Image Registration Tool with 6 degrees of freedom. The diffusion tensor was calculated using simple least-squares fit of the tensor model, and FA, MD, AD, and RD maps were computed for each voxel using standard methods for the DTIFIT program in FSL. All the FA images were aligned to the standard FMRIB58_FA template, which was provided by the FSL program using a nonlinear registration algorithm implemented in the TBSS package. The FA images were then averaged to create a mean FA skeleton. Each subject's aligned FA images were projected onto this skeleton by filling each voxel on the skeleton with the maximum FA values from a plane perpendicular to the local skeleton structure (18). To exclude voxels of adjacent gray matter or cerebrospinal fluid, a threshold FA value of 0.2 was chosen. The MD, AD, and RD maps were also processed with the same method used for the FA maps by applying the nonlinear registration algorithm and projecting them onto the mean FA skeleton. Voxel-wise statistical analysis of individual skeleton images was performed using nonparametric permutation-based two-sample t tests to compare the DTI values for each group pair. The null distribution was constructed over 5000 permutations.

The threshold-free cluster enhancement approach with two-dimensional parameter settings was used for control over the multiple comparison correction. All the results derived from TBSS analysis were considered significant with a family-wise error-corrected p < 0.05. To identify WM tracts, two WM atlases within FSL (http://fsl.fmrib.ox.ac.uk/fsl/data/atlas-descriptions.html), the Johns Hopkins University (JHU) WM tractography atlas and the International Consortium of Brain Mapping-DTI WM labels atlas, were used.

ROI-Based Analysis of DTI Values and Correlation Analysis with NP Tests

The group comparison results were significant only between the HAND and control groups on TBSS analysis, which might be attributed to the relatively small number of subjects in our cohort. Therefore, we performed an ROI-based analysis to define group differences between the HAND and HIV-IC groups, our main study goal. In this analysis, we defined masks on areas which revealed significantly different diffusion measures (FA, MD, and RD) between the HAND and HC groups. Then, we subdivided WM into 20 tracts identified by the JHU WM tractography atlas (Supplementary Fig. 1 in the online-only Data Supplement), which is comprised of the left/right (L/R) anterior thalamic radiation (ATR), L/R corticospinal tract (CST), L/R cingulum (cingulate gyrus), L/R cingulum (hippocampus), forceps major, forceps minor, L/R inferior fronto-occipital fasciculus (IFOF), L/R inferior longitudinal fasciculus (ILF), L/R superior longitudinal fasciculus (SLF), L/R uncinate fasciculus, and L/R SLF temporal part using the semi-automated method. Subsequently, we created tract-specific masks with brain regions that showed significant differences between the HAND and HC groups on TBSS analysis overlapping with the ROI of each atlas. As a result, we divided the FA masks into the right CST, right SLF, and right SLF temporal part; MD masks into bilateral ATR, bilateral cingulum (cingulate gyrus), bilateral CST, forceps major and minor, bilateral IFOF, bilateral ILF, bilateral SLF, bilateral SLF temporal part, and left uncinate fasciculus; and RD masks into 19 parts except for the cingulum (hippocampus). We extracted the mean DTI values from each tract-specific ROI projected onto each subject's skeletonized map and compared the mean values between groups using ANOVA, a post-hoc analysis with the two-sample t test and Kruskal-Wallis test, and a post-hoc analysis with the Mann-Whitney test applying a corrected p value by the Bonferroni method according to the results of the normality test, respectively. Additionally, we performed trend analyses of DTI values in the ROI between the groups using the Jonckheere-Terpstra test under the hypothesis that HC, HIV-IC, and HAND were all components of a single disease spectrum (with altered WM representing the most severe parameter in the HAND group). To define the WM tract associated with the cognitive decline in HIV patients, we performed a correlation analysis between DTI values of a particular WM tract and NP test scores showing significant differences between the HAND and HIV-IC groups using Pearson's and Spearman's tests for normally and non-normally distributed data, respectively. All statistical analyses were performed using SPSS version 20.0 (IBM Corp.).

Demographics

All the subjects enrolled in this study were male, without any significant differences in age (53.0 ± 2.3 years, 52.5 ± 2.0 years, and 56.0 ± 2.6 years; p = 0.699) or years of education (12.4 ± 0.4 years, 12.8 ± 1.0 years, and 13.2 ± 1.3 years; p = 0.958) between the HC, HIV-IC, and HAND groups. All patients were on stable highly active anti-retroviral therapy and had CD4 > 200 cells/mL and viral load < 20 copies/mL except for one patient in the HAND group who had a viral load of 52.2 copies/mL. Although the HAND group required a longer mean time to complete TMT A (p = 0.032), only one HAND patient met the criteria of impaired information processing speed. On the other hand, compared with the HIV-IC group, the HAND group showed cognitive dysfunction associated with memory including learning and recall (p = 0.010), abstraction/executive function (p = 0.002), and sensory perception or motor skills (p = 0.008). Patients in the HAND group had lower memory scores on K-AVLT sub-tests, lower executive function in WCST and TMT B, and higher motor dysfunction in the Grooved Pegboard Test than patients in the HIV-IC group. The NP test results are briefly summarized in Table 1.

Table 1 Neuropsychological Status of HIV-IC and HAND Click for larger imageClick for full tableDownload as Excel file

TBSS Results

Compared with the HC group, the HAND group showed decreased FA in the right corona radiata including CST (Fig. 1) and increased MD in the bilateral superior and posterior corona radiata, right anterior corona radiata, bilateral SLF and ILF, body and splenium of corpus callosum, bilateral posterior limbs and retrolenticular parts of internal capsules, bilateral posterior thalamic radiations, bilateral cerebral peduncles, bilateral sagittal striata, and bilateral cerebellar peduncles and medial lemnisci (Fig. 2). For RD, the HAND group showed increased values in the bilateral corona radiata, bilateral SLF and ILF, corpus callosum, bilateral internal capsule, bilateral thalamic radiations, bilateral superior and middle cerebellar peduncles, and bilateral medial lemnisci (Fig. 3). AD values were not significantly different between the HAND and HC groups. DTI values did not significantly differ between the HAND and HIV-IC groups, nor did they differ between the HIV-IC and HC groups.

Fig. 1 TBSS analysis of FA maps. Areas in sky blue-blue represent brain regions with significant decrease in FA (FWE-corrected p < 0.05) in HAND group relative to HC. Results are shown overlaid on Montreal Neurological Institute 152-T1 template and mean FA skeleton (green). Left side of image corresponds to right hemisphere of brain. FA = fractional anisotropy, FWE = family-wise error, HAND = HIV-associated neurocognitive disorders, HC = healthy controls, HIV = human immunodeficiency virus, TBSS = tract-based spatial statistics Click for larger imageDownload as PowerPoint slide

Fig. 2 TBSS analysis of MD maps. Areas in orange-red represent brain regions with significant increase in MD (FWE-corrected p < 0.05) in HAND group relative to HC. Green represents mean white matter skeleton of all subjects. MD = mean diffusivity Click for larger imageDownload as PowerPoint slide

Fig. 3 TBSS analysis of RD maps. Areas in orange-red represent brain regions with significant increase in RD (FWE-corrected p < 0.05) in HAND group relative to HC. Green represents mean white matter skeleton of all subjects. RD = radial diffusivity Click for larger imageDownload as PowerPoint slide

ROI-Based Analyses of DTI Values between the Groups

Compared with the HIV-IC group, the HAND group showed significantly lower mean FA in the right SLF mask (p = 0.008). Although there were no significant differences, the mean MD in the left SLF temporal (p = 0.058) masks and the mean RD in the left SLF mask (p = 0.059) showed higher values in the HAND group compared with the HIV-IC group. Compared with the HC group, the HIV-IC group showed significantly altered DTI values mainly in CST, ATR, IFOF, and ILF and the HAND group showed altered DTI values in wider areas (Table 2). In the trend analysis, all but the mean MD in the forceps minor and right cingulum (hippocampus) masks demonstrated significant trends across the groups. All the mean FA values showed an increasing trend from the HC group to the HAND group, while the mean MD and RD values showed a decreasing trend (Fig. 4).

Fig. 4 Bar graph representing mean DTI values in JHU WM tractography-based ROIs that were extracted from TBSS results (MD and RD values were multiplied by 1000 for good visibility). *p < 0.05 in trend analyses. ATR = anterior thalamic radiation, CCG = cingulum (cingulate gyrus), Ch = cingulum (hippocampus), CON = healthy controls, CST = corticospinal tract, DTI = diffusion tensor imaging, FCPM = forceps major, FCPm = forceps minor, HIV-IC = HIV-infected patients with intact cognition, IFOF = inferior fronto-occipital fasciculus, ILF = inferior longitudinal fasciculus, JHU = Johns Hopkins University, L = left, R = right, ROI = region-of-interest, SLF = superior longitudinal fasciculus, SLFTP = SLF temporal part, UF = uncinate fasciculus, WM = white matter Click for larger imageDownload as PowerPoint slide

Table 2 Comparison of Mean DTI Values between Groups in ROI-Based Analysis Click for larger imageClick for full tableDownload as Excel file

Correlation between WM Tract Integrity and Cognitive Function among the HIV Patients

Among the seven NP tests which showed significant differences between the HAND and HIV-IC groups, the K-AVLT total scores and time in seconds required to complete TMT A and B, and the Grooved Pegboard Test showed significant correlation with mean DTI values in tract-specific ROIs showing significant group differences. A lower mean FA value in the right SLF mask was associated with lower performance of TMT A (r = −0.425, p = 0.049) and B (r = −0.455, p = 0.038), the K-AVLT test (r = 0.564, p = 0.006), and the Grooved Pegboard Test (r = −0.475, p = 0.025) (Fig. 5). Among the mean DTI values in the masks showing different trends between the HAND and HIV-IC groups, a higher MD value in the left SLF temporal part was associated with lower performance of TMT A (r = 0.493, p = 0.020).

Fig. 5 Correlation analysis with FA value in right SLF masks and NP test results. Region in pink represents right SLF mask in JHU WM tractography atlas. Inside this mask, regions that showed significantly decreased FA between HAND and HIV-IC groups on TBSS analysis are marked in red and was used as mask for ROI (A). Mean FA value in this ROI was correlated with NP test results. Correlation results are shown graphically in (B). K-AVLT = Korean version of Auditory Verbal Learning Test, NP = neuropsychological, TMT A = trail making test part A, TMT B = trail making test part B Click for larger imageDownload as PowerPoint slide