16 HIV+ women with a history of cocaine dependence (COC+/HIV+) and 21 healthy females participated in the study. Abstinence from illegal drugs was determined by self-report corroborated by a urine test on the scan date. Patients were recruited from clinics associated with the Einstein-Montefiore Center for AIDS research, who verified past cocaine dependence and HIV+ status. Healthy controls were recruited from a university-wide database for research study participation. Potential participants completed an in-person screening and returned on a different day for scanning. Groups differed in age [t₍₃₄₎ = 3.91, p < 0.001], years of education [t₍₃₄₎ = 4.76, p < 0.001], and ethnicity (Fisher’s exact count data, p < 0.001) (Table 1). As age (21) and years of education (22), key components of socioeconomic status (23), are known to impact brain structure, both variables were included as covariates in all subsequent analyses. Since no research has demonstrated ethnicity-related WM differences between African-American, Caucasian, and/or Hispanic individuals, ethnicity was not included as a covariate. Exclusion criteria for the study were as follows: (1) head trauma resulting in loss of consciousness for longer than 30 min, (2) the presence of any past or current brain pathology, (3) a positive urine screening for any illegal drugs, and (4) the presence of any contraindication to MRI scanning. Healthy controls were excluded for any self-reported history of psychiatric illness or a history of drug or alcohol dependence. Due to high rates of psychiatric comorbidity in HIV+ individuals, COC+/HIV+ were excluded for the presence of an untreated psychiatric condition. 3 COC+/HIV+ participants had a diagnosis of major depressive disorder (one with comorbid generalized anxiety disorder), and one had a diagnosis of bipolar disorder.

Participants with a history of cocaine dependence completed the cocaine subsection of the Kreek–McHugh–Schluger–Kellogg (KMSK) Scale to further characterize cocaine dependence during the most severe period of use (24). Because of the high rates of alcohol and other drug comorbidity among COC+/HIV+ populations, patients were not excluded if they had abused other drugs or alcohol prior to the onset of their cocaine abstinence as long as cocaine was identified as their primary drug of choice based on referral from addiction treatment centers, the KMSK, and verbal agreement. 2 COC+/HIV+ individuals had past comorbid alcohol abuse, and 1 COC+/HIV+ individual had comorbid past heroin dependence.

COC+/HIV+ participants were abstinent from cocaine for an average of 460 ± 381 weeks (range: 8.7–1,043 weeks). Years of self-reported cocaine use was collected from 11 of the 15 participants included in the final analysis; patients used cocaine for an average of 11.94 ± 6.53 years (range: 3–22). All COC+/HIV+ participants were receiving highly active antiretroviral therapy. 1 COC+/HIV+ patient had an additional diagnosis of hepatitis C. Current CD4 count was available for 13 out of the 15 women included in the final sample. These patients had an average current CD4 count of 609 ± 265 cells/μL (range: 197–1,097). Viral load data were available for 12 out of the 15 HIV+ participants included in the final sample, and 66% of these patients had a current viral load below 75 copies/mL (range: undetected—25,027). No COC+/HIV+ patients had any AIDS-defining clinical conditions. This study was approved by the Albert Einstein College of Medicine institutional review board. Written, informed consent was obtained from all participants in accordance with the tenants of the Declaration of Helsinki.

Impulsive was assessed through self-administration of the Barratt Impulsiveness Scale (BIS-11). BIS-11 scores were available for 14 COC+/HIV+ participants and 10 healthy controls (25). This measure contains 30 questions aimed to assess personality traits associated with risk taking and impulsivity. Subsections include questions pertaining to attention impulsiveness, motor impulsiveness, and non-planning impulsiveness.

MRI was performed at the Magnetic Resonance Research Center at the Albert Einstein College of Medicine using a 3.0 T Freewave Achieva MRI Scanner (Philips Medical Systems, Best, The Netherlands) with a 32-Channel SENSE RF head coil (3 MHz per channel bandwidth; 80 MHz per channel all digital sampling rate). A T1-weighted, sagittal 3D MPRAGE acquisition over a 240 mm field of view with 240 × 240 in-plane matrix and 1 mm isotropic resolution, TR = 8.2 s/TE = 3.7 ms, SENSE factor = 2 (3 min) was acquired. Diffusion-weighed images were obtained using single shot echo-planar imaging (EPI) along 32 independent, non-collinear diffusion sensitizing directions at b = 800 s/mm² (TR = 7,600 ms, TE = 56 ms, voxel size: 2 mm × 2 mm × 2 mm, FOV = 256 mm², in-plane imaging matrix = 128 mm × 127 mm, number of slices = 70, and SENSE factor = 2.5). Two additional images with no diffusion weighting (b = 0) were also collected. Total DTI scan time was 4 min·26 s.

Data were preprocessed using the ExploreDTI Software (v 4.8.6)¹ (26). In one step, the data were corrected for eddy current-induced geometric distortions, subject motion, and EPI by coregistration and resampling to each subjects T1-weigthed anatomical image (27). The B-matrix rotation was also performed in this step to reorient the data appropriately (28). After visual inspection of the data, one COC+/HIV+ participant was removed due to excessive motion artifacts in the B0 image which persisted after motion correction. Therefore, 15 COC+/HIV+ and 21 healthy control subjects were included in the analysis. Diffusion-based measures of fractional anisotropy (FA), radial diffusivity (RD), axial diffusivity (AD), and MD were extracted for each participant.

Voxelwise statistical analyses were carried out using tract-based spatial statistics from FMRIB’s Software Library² (29, 30). Following preprocessing, FA maps from ExploreDTI were aligned to a 1 mm × 1 mm × 1 mm standard space target image (FMRIB58_FA) using FMRIB’s non-linear image registration tool. The target image was then aligned to Montreal Neurological Institute (MNI) template space using an affine transformation. Each subjects’ FA images were transformed to MNI space by combining the non-linear transformation to the target FA image with the affine transformation to the MNI template in one step to avoid resampling the image twice. A mean FA map was created by averaging all subjects’ FA images; this map was then thinned (threshold of 0.2) to generate a mean FA skeleton representing the center of the tracts common to all groups. Each subject’s aligned FA images were then projected onto this FA skeleton for voxelwise analyses. The same non-linear transformation and FA skeleton projection procedure used for FA were applied to the other diffusion measures (RD, AD, and MD).

Group comparisons for FA, MD, AD, and RD diffusion measures were conducted using the FMRIB randomize toolbox. Ten thousand permutations were used to estimate a null distribution for each contrast (control > COC+/HIV+ and COC+/HIV+ > control) with age and years of education included as covariates. The results were thresholded at p ≤ 0.05 and corrected for multiple comparisons across voxels using the threshold-free cluster enhancement option in randomize (31). To further quantify these results, the percentage of “affected” voxels (those showing significant between-group differences in diffusion metrics) was calculated for each major WM tract. WM tracts were identified using John’s Hopkins University (JHU) DTI-based WM atlas. The mean FA skeleton was masked with JHU atlas-derived regions of interests (ROI). The percentage of affected voxels in each ROI was then computed by dividing the number of WM voxels showing significant between-group differences by the total number of WM voxels in each ROI.

The average FA and RD value from WM voxels were also extracted from cognitive control-related WM tracts for correlations with BIS-11 score. Subjects’ FA skeletons were masked with the JHU atlas-derived ROI for the forceps minor and bilateral uncinate fasciculus. For each subject, the average FA or RD value of voxels within these masks were extracted.

Between-group differences in demographics were examined using two-tailed independent samples t-tests conducted in R statistical computing software (32). Pearson’s correlations between diffusion measures, CD4 count and duration of abstinence were performed. Partial correlations with mean FA and RD values from the forceps minor and bilateral uncinate fasciculus were performed using the R package “ppcor” with age and years of education included as covariates (33).

To measure possible between-group differences in brain volume, total estimated intracranial volume was calculated based on T1-weighted images using FreeSurfer³ (34). A two-tailed independent samples t-test was performed.

COC+/HIV+ subjects scored significantly higher on the BIS-11 measure of self-reported impulsivity compared with healthy controls [t₍₂₂₎ = 3.86, p < 0.001].

Statistically significant between-group differences were seen diffusely in WM tracts for FA and RD diffusion measures such that decreased FA and increased RD were found in COC+/HIV+ patients compared with controls (Figure 1; Table 2). No WM regions showed decreased FA or increased RD in controls compared with COC+/HIV+. There were no statistically significant voxelwise differences between groups for diffusion measures of MD and AD.

The percentage of affected voxels calculation showed that the between-group differences spanned all major WM tracts for FA and most WM tracts for RD (Table 3). The tract with the greatest percentage of voxels showing significant between-group differences was the forceps minor (FA: 75.6%, RD: 59.7%). For FA, the region with the lowest percentage of voxels showing significant differences was the right cingulum (31.6%). For RD, there were no significant between-group differences on the left corticospinal tract.

There was no significant difference in estimated total intracranial volume between groups [t₍₃₄₎ = 1.055, p = 0.30], therefore it was not included as a covariate in any analyses.

To investigate the relationship between diffusion metrics and disease measures, the average FA and RD value across all WM voxels was extracted. The average FA and RD values were correlated with duration of abstinence (FA: r = 0.18, p = 0.51; RD: r = −0.19, p = 0.47) and CD4 count (FA: r = −0.06, p = 0.84; RD: r = 0.11, p = 0.72). No significant correlations were found.

Further analyses examined the association between impulsivity and DTI metrics specifically within cognitive control-related brain circuitry. Partial correlations between BIS-11 impulsivity score and the mean FA and RD in the forceps minor (FA: r = −0.39, p = 0.074; RD: r = 0.34, p = 0.12), left uncinate fasciculus (FA: r = −0.31, p = 0.16; RD: r = 0.27, p = 0.22), and right uncinate fasciculus (FA: r = −0.36, p = 0.10; RD: r = 0.37, p = 0.09) showed no statistically significant correlations.