Cocaine use is associated with cerebral white matter hyperintensities in HIV disease

Abstract Background White matter hyperintensities (WMH), a marker of cerebral small vessel disease and predictor of cognitive decline, are observed at higher rates in persons with HIV (PWH). The use of cocaine, a potent central nervous system stimulant, is disproportionately common in PWH and may contribute to WMH. Methods The sample included of 110 PWH on antiretroviral therapy. Fluid‐attenuated inversion recovery (FLAIR) and T1‐weighted anatomical MRI scans were collected, along with neuropsychological testing. FLAIR images were processed using the Lesion Segmentation Toolbox. A hierarchical regression model was run to investigate predictors of WMH burden [block 1: demographics; block 2: cerebrovascular disease (CVD) risk; block 3: lesion burden]. Results The sample was 20% female and 79% African American with a mean age of 45.37. All participants had persistent HIV viral suppression, and the median CD4+ T‐cell count was 750. Nearly a third (29%) currently used cocaine regularly, with an average of 23.75 (SD = 20.95) days in the past 90. In the hierarchical linear regression model, cocaine use was a significant predictor of WMH burden (β = .28). WMH burden was significantly correlated with poorer cognitive function (r = −0.27). Finally, higher WMH burden was significantly associated with increased serum concentrations of interferon‐γ‐inducible protein 10 (IP‐10) but lower concentrations of myeloperoxidase (MPO); however, these markers did not differ by COC status. Conclusions WMH burden is associated with poorer cognitive performance in PWH. Cocaine use and CVD risk independently contribute to WMH, and addressing these conditions as part of HIV care may mitigate brain injury underlying neurocognitive impairment.


Introduction
][16][17] As persons with HIV (PWH) now have nearly average life expectancies, 18 normal aging processes and comorbid conditions likely contribute to the development and expression of NCI.
5][26] The histopathology of WMHs is heterogeneous, including myelin pallor, demyelination and axonal loss, and mild gliosis. 27,28The majority of WMH result from chronic ischemia associated with cerebral small ª 2023 The Authors.Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association.This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
0][31][32] While the biological mechanisms leading to WMH are not fully understood, possible factors include chronic hypoperfusion, increased permeability of the blood-brain barrier, vascular endothelial dysfunction, and inflammatory responses. 33][36][37][38] Aging and cardiovascular disease (CVD) risk factors such as hypertension, diabetes, and smoking are the primary predictors of WMH in the general population 39,40 and in PWH specifically. 20,22,41,424][45] WMH in PWH may also reflect legacy effects of neuronal injury that occurred during untreated phases of disease.One study reported that longer time spent with a CD4 + T-cell count below 500 cells/lL (a marker of immunosuppression) was predictive of total WMH load, 22 although most studies have found no relationship between HIV disease characteristics and WMH. 20,23,42,46,47In PWH, WMH have been associated with worse global cognitive functioning, although the specific domains affected varied across studies. 22,41,42ocaine use is disproportionately common in PWH and may contribute to neuronal injury.The most recent surveillance by the CDC of adults with HIV reported past year prevalence rates of 5% for non-injection cocaine and 3% for crack cocaine. 48By comparison, the past-year prevalence of use among US adults was 2% for any form of cocaine, including crack cocaine. 491][52] Acute cocaine exposure causes a sudden transient increase in blood pressure, which can rupture leaky and weak vessels.4][55] However, the contribution of cocaine on WMH burden in PWH has received little attention.One study found that a history of cocaine use disorder was not associated with increased risk for WMH load in PWH, 20 although the recency and frequency of cocaine use were not reported.
WMH can be measured quantitatively using noninvasive magnetic resonance imaging (MRI).Specifically, T2-weighted fluid-attenuated inversion recovery (FLAIR) imaging is the industry-standard sequence for delineating WMHs.The total WMH volume is an important determinant of their clinical relevance. 56,57The current study quantified WMH burden in a sample of PWH who were treated with cART and had sustained HIV suppression.Our specific aims were to (1) investigate the independent contribution of cocaine use to WMH, after accounting for age and common CVD risk factors, (2) examine the correlation of WMH volume to cognitive function, and (3) explore the possible association of peripheral biomarkers of immune activation and inflammation to WMH load.We hypothesized that PWH who used cocaine regularly would have a greater WMH burden, and that WMH burden would correlate negatively with global cognitive function.We further expected that an inflammatory cascade would mediate the relationship between cocaine use and WMH burden.

Materials and Methods
The data in this report were collected as part of a larger project on neuroimaging and immunological features of NCI in PWH. 58The study was approved by the Duke Health Institutional Review Board, and all participants provided written, informed consent.

Participants
PWH aged 18-59 years were recruited from infectious diseases clinics in the Raleigh/Durham area of North Carolina, USA.Participants met the following criteria: stable ART for >1 year, nadir CD4 cell count of ≤350 cells/mm 3 or HIV infection of ≥5 years; and sustained viral suppression (plasma HIV RNA < 200 copies/mL) for >1 year.Exclusion criteria included perinatal HIV infection; <9 years of education; BMI > 40; English non-fluency or illiteracy; severe learning disability; severe head trauma with loss of consciousness >30 min and persistent functional decline; serious neurological disorders not caused by HIV, including stroke; acute CNS infection or chronic CNS infection with residual symptoms; severe mental illness or acute psychiatric symptoms; systemic autoimmune disease; inpatient hospitalization within 30 days; use of an immunomodulatory medication, steroid, or antibiotic within 30 days; contraindications to MRI; and other underlying or uncontrolled medical illness with altered cognition.
The sample was stratified by cocaine history.A positive history (COC+) was defined by lifetime history of regular use and ≥12 days of use in the past 90 days or a positive urine screen, while a negative history (COCÀ) was defined as no regular use in the past 10 years, no use in the past 90 days, and a negative urine drug screen.Alcohol, cannabis, and nicotine use were allowed, and controlled substances such as benzodiazepines were permitted if prescribed.For other illicit drugs of abuse, participants were excluded for: a positive urine drug screen; >2 days of use within 30 days; >2 years of regular use within the past 20 years; and any regular use in the past 5 years.

Substance use assessment
The Addiction Severity Index-Lite (ASI-L), a semistructured interview, was administered to characterize lifetime substance use and associated problems. 59Module E of the Structured Clinical Interview for DSM-5 (SCID-5) was used to assess substance use disorder symptoms. 60requency of use in the past 90 days for cocaine, marijuana, and other substances was facilitated with Timeline Follow-Back methodology. 61,62A nine-panel urine toxicology screen that tested for cocaine, amphetamine, barbiturates, benzodiazepine, methamphetamine, opioids, methadone, marijuana, and oxycodone assessed recent drug use.

Clinical data
Participants provided a release for the study team to obtain copies of their healthcare records.This information was used to confirm no exclusionary conditions, and to abstract relevant clinical data.HIV disease characteristics included date of diagnosis, nadir and current CD4 + T-cell counts, current antiretroviral regimen, and most recent plasma HIV viral load.The following clinical diagnoses were coded as present or absent: diabetes, hypertension, hyperlipidemia, and any prior CVD such as coronary artery disease and cardiomyopathy.Body mass index (BMI) was calculated from weight and height measured at last clinical visit.Participants reported frequency of smoking in the past 3 months.As in prior studies, 21,42,63 a composite CVD risk score was calculated as the sum of the following risk factors: diabetes, hyperlipidemia, hypertension, CVD diagnosis, obesity (BMI ≥ 30), and current smoking.Total scores ranged from 0 to 6.

Neuropsychological testing was administered by
Bachelor's-level staff member under the supervision of a licensed clinical psychologist.The 60-min battery of tests spanned seven domains of cognitive function: 1. Attention: Paced Auditory Serial Addition Task-50total number correct  72 Using published normative data, 64,65,73

MRI data acquisition and processing
All MRI scans were performed at Duke University Hospital using a 3.0T GE Discovery MR750 scanner with an 8channel head coil.High-resolution T1-weighted images were recorded using a spoiled echo sequence (voxel size = 1 mm 3  Cortical reconstruction and volumetric segmentation of the T1-weighted images was done with the FreeSurfer image analysis suite version 6.0 to obtain estimated total intracranial volume (eTIV). 74WMH were detected using the Lesion Segmentation Toolbox (LST) version 3.0.0,7][78][79] Bias-corrected FLAIR images were first coregistered to normalized T1-weighted anatomical images, which were segmented into cerebrospinal fluid, gray matter, and WM maps using SPM12 (http://www.fil.ion.ucl.ac.uk/spm/software/spm12).This information was combined with the coregistered FLAIR image to provide a lesion belief map for each class of tissue using the automated lesion growth algorithm (LGA) of LST.These maps were thresholded with a lesion probability threshold (j) to generate an initial binary lesion map that was subsequently grown along voxels that appear hyperintense on the FLAIR image.To define the optimal threshold, T1 and FLAIR images from 11 randomly selected participants were manually segmented by a neuroradiologist, blinded to COC status, to establish ground truth lesion masks.A range of lesion volumes (1.34-24.3mL; M = 6.51,SD = 8.33) was included in the training set.These same cases were then segmented at j = 0.10, j = 0.20, and j = 0.30 using the LGA algorithm.Comparing the dice similarity coefficients (DSC) between the manually segmented images and the LGAgenerated lesion belief maps, it was determined that a j = 0.10 was the optimal threshold (average DSC = 0.32; 100% specificity, 29% sensitivity).This was confirmed via visual inspection of the segmentation results.1][82] The remainder of the FLAIR data were then processed with LGA to derive whole brain WMH volumes for each participant.The WMH volume in milliliters (mL) was normalized to the eTIV to account for inter-individual variations in brain size. 83These values were log10 transformed for analyses.As suggested by prior studies, 84,85 lesion load was categorized based on total WMH volume: minimal (<1 mL), low (≥1 through <5 mL), moderate (5-15 mL), and high (>15 mL). Figure 1 shows group frequency maps for the low, moderate, and high WMH lesion load classifications.

Peripheral inflammatory biomarkers
Non-fasting blood samples were collected by peripheral venipuncture immediately following the MRI or on a subsequent day.Plasma and serum were purified using standard procedures, and then, supernatant was pipetted into aliquots and stored at À80°C until time of assay.A selected set of 10 biomarkers representing various components of the inflammatory cascade were measured using commercially available enzyme-linked immunosorbent assay (ELISA) kits.All analytes were measured in duplicate following manufacturers' instructions.Serum samples were assayed on the Luminex MagPix instrument using commercially available multiplex kits from EMD Millipore (Burlington, MA) to measure concentrations of IL-4, IL-6, IL-8, IL-13, and TNFa (Milliplexâ Human High Sensitivity T Cell) and MPO (Milliplexâ Human CVD Magnetic Bead Panel 2).Single ELISA assays were used for the following analytes: MCPI/CCL2 (serum; Human Uncoated; Thermo Fisher Scientific, Waltham, MA, USA), IP-10/CXCL10 (plasma; LEGENDplex TM ; Biolegend, San Diego, CA, USA), sCD14 (plasma; Quantikineâ, R&D Systems, Minneapolis, MN, USA), and sCD163 (plasma; Quantikineâ; R&D Systems).Laboratory personnel were blinded as to the demographic and COC status of the participants.Four people were missing blood data, and one person was excluded because values were >3 SD above the mean on >3 biomarkers.The values for the cytokines/chemokines were log10 transformed to improve normality.

Data analysis
Descriptive statistics were used to characterize the sample on demographic, HIV disease, CVD risk characteristics, and inflammatory biomarkers.We used two-tailed independent samples t-tests, Mann-Whitney U-tests, chisquared tests, and Fisher's exact tests, as appropriate, to compare COC+ and COCÀ participants.A hierarchical multiple regression model was conducted to identify the independent association of cocaine use on WMH, after accounting for common risk factors.Variables were entered in a series of blocks in the following order: demographics (age in years, biological sex, race), clinical factors (CVD risk score, nadir and current CD4 + count, years since HIV diagnosis), and substance use (alcohol, marijuana, and cocaine; yes/no).A partial correlation was used to examine the strength of the relationship between WMH load and global cognitive function (T-score) controlling for age.Finally, a series of analysis of covariance (ANCOVA) models examined the association of WMH burden to each biomarker.Lesion load was the betweensubjects factor defined as minimal, low, moderate, or high, controlling or age.All analyses were conducted in SPSS 28.0.

Sample characteristics
The sample included 32 COC+ and 78 COCÀ participants (Table 1).They were primarily male at birth (84%) and African American (72%) and had a mean age of 45.37 years (SD = 9.34).Compared to participants in the COCÀ group, COC+ participants were  1637 significantly older and less educated, and they were more likely to identify as African American.Participants had been diagnosed with HIV disease for a mean of 14.60 years (SD = 7.60), with no difference by COC status, but COC+ participants had significantly lower nadir and current CD4 T-cell counts compared to COCÀ participants.Participants had been on ARV regimens for 12.33 years on average (SD = 6.82).Current ARV regimens were similar across the groups, with the majority (72%) being on an integrase strand transfer inhibitorbased combination and 24% being on a regimen containing a protease inhibitor.The number of CVD risk factors ranged from 0 to 5, with a mean of 1.56 (SD = 1.15), and there was no difference by COC status.However, COC+ were more likely to currently smoke cigarettes, while COCÀ were more likely to have a BMI in the obese range.
Participants in the COC+ group had been using cocaine regularly for an average of 18.28 years (SD = 10.70).The majority (66%) reported smoking as their primary route of administration; all others reported nasal administration.In the 90 days prior to enrollment, they had used cocaine on an average of 23.75 days (SD = 20.95).While the overall prevalence of use was high for current alcohol (75%) and cannabis (48%), COC+ participants were significantly more likely than COCÀ participants to have used these substances (both P < 0.05).
Step 2 added clinical factors.As expected, CVD risk score was associated with higher WMH volume.None of the HIV clinical variables were predictive.Step 3 added the substance use factors, and cocaine was associated with higher WMH volume.In the final model, age, CVD risk, and cocaine use were independent predictors of WMH burden.

Exploratory correlations of inflammatory biomarkers and WMH burden
After accounting for age, the COC group had comparable levels for all of the inflammatory markers except IL-13 (Table 3).Table 4 summarizes the results of the ANCOVA models comparing biomarker values across WMH categories.Controlling for age, concentrations of IP-10 and MPO were significantly different across the four categorizations.Specifically, participants who were categorized as having moderate or high lesion burden had the highest concentrations of IP-10 and the lowest concentration of MPO.However, the group differences did not pass family   wise error correction for multiple comparisons.Since IP-10 and MPO did not differ by COC status, we did not pursue mediation analysis.

Discussion
This study investigated the contribution of ongoing cocaine use to WMH in persons with treated HIV disease who had sustained viral suppression.Our principal finding is that cocaine use in PWH was associated with greater WM lesion burden, even after accounting for age and CVD risk factors.Moreover, higher WM lesion burden was correlated with poorer cognitive function, underscoring the role of vascular processes in the development of NCI in PWH independent of COC status.While chronic HIV disease may contribute to the development of cerebral small vessel disease, it appears that age, CVD risk, and cocaine are robust predictors of WMH with high relevance for PWH.Finally, circulating neurotoxic chemokines linked to endothelial functioning were associated with WM lesion burden, suggesting a potential role of inflammatory processes in the pathogenesis of cerebrovascular insults in the context of chronic HIV infection.This study was specifically designed to test the effects of cocaine on health outcomes in PWH, comparing individuals with ongoing cocaine use to those with no history of substantive cocaine use.Our results support the independent role of ongoing cocaine use on burden of WMH.A recent study did not find an association between cocaine use and WMH in PWH, but groups were defined by a past history of cocaine use disorder rather than current use. 20Acute cocaine administration inhibits catecholamine reuptake at sympathetic nerve terminals, with potent vasoconstrictive effects. 86This persistent hypoperfusion can result in ischemic damage to WM. 87,88 Future research is needed to empirically test whether recovery from cocaine use disorder mitigates or reverses the effects of past cocaine use on WMH burden.
Consistent with the extensive literature on cerebral small vessel disease, 39,89 CVD risk was strongly associated with WMH.Interestingly, cocaine use was unrelated to overall CVD risk score, suggesting that CVD risk did not mediate the relationship between cocaine and WMH burden.COC+ participants were more likely to smoke cigarettes but also less likely to have a BMI in the obese range, balancing out overall risk score.However, there was a trend for COC+ participants to have a higher prevalence of hypertension.While the link between chronic cocaine use and CVD risk is well established, 90,91 several factors likely contributed to our results.First, the prevalence of CVD risk factors was high overall, as HIV disease is associated with increased risk. 92,93Second, our scale did not account for severity of each risk factor (e.g., controlled versus uncontrolled disease).Third, a history of a cerebrovascular events, such as stroke, was an exclusion criterion for the study, and persons who used cocaine were more likely to be excluded for this reason.While our exclusions may limit the generalizability of findings with respect to prevalence of CVD risk, it allowed us to better isolate the effects of cocaine on WMH.
Despite prior reports that HIV disease confers an elevated risk of WMH, [19][20][21][22][23] the pathophysiology remains uncertain.Consistent with other studies, 20,23,46,47 we found no relationship between WMH and HIV disease characteristics, such as time since diagnosis and CD4 counts of immune function.Given our inclusion criteria, study participants were fairly homogeneous in terms of their HIV disease; all had a diagnosis for >5 years and/or a low nadir CD4 suggestive of long-standing disease and were on combination ARV with viral suppression for >1 year.COC+ participants did have a history of more severe immunosuppression and less robust immune reconstitution, but these differences did not account for the effects of cocaine on WMH.
At the cellular level, cerebral small vessel disease is believed to result from endothelial dysfunction and subsequent exposure of the arterial vessel wall, making it susceptible to thrombosis. 35,94Although our analyses were exploratory, we found that concentrations of IP-10, a chemokine implicated in the induction of leukocyte migration to the site of inflammation, 95,96 were associated with higher WMH burden.IP-10 is believed to inhibit endothelial healing, 97 and it has been suggested as a marker of thrombosis. 98,99We also found that MPO, an inflammatory enzyme, was lower in participants with higher WMH burden.MPO is expressed following acute neuronal injury, triggering the production of the bactericide hypochlorous acid, a strong oxidant that can cause local tissue damage and amplify the inflammatory cascade. 100,101Elevated MPO levels are associated with inflammation and oxidative stress and are predictive of CVD risk, including thrombosis and stroke. 102One study of stroke-free adults found that elevated MPO was associated with greater burden of WMH. 103However, an analysis of >1700 adults enrolled in the Framingham Heart Study found that WMH and silent cerebral infarcts were associated with lower MPO while cerebral microbleed was associated with higher MPO. 38The authors suggested that different inflammatory pathways may be involved in the pathogenesis of ischemic versus hemorrhagic MRI markers.The mechanism underlying the lower levels of circulating MPO in persons with higher WMH is uncertain but is consistent with the hypothesis that MPO deficiency, via a diminished antioxidant response, may promote ischemic neuronal injury. 104,105Since there were no differences in the levels of these inflammatory biomarkers, the potential mechanisms do not appear to mediate the effects of cocaine on WMH.
Strengths of this study include our use of comprehensive assessments to characterize the sample and rigorous eligibility criteria related to both substance use and HIV disease to minimize confounds and maximize power.
Our study also has some limitations.First, we did not include an HIV-negative comparison group because the goal of the larger project was to identify neural correlates of NCI in PWH.Second, the sample size of 110 is relatively modest, but the study was designed specifically to isolate the effects of cocaine on HIV-related outcomes.With clearly delineated groups, our power to detect differences attributed to cocaine use was strengthened.Third, we used a research-domain automated pipeline for lesion segmentation, as manual delineation of WM lesions is labor intensive with considerable inter-and intra-rater variability. 106While the LGA method has been found to be valid and robust, [107][108][109] our own data suggested high specificity but low sensitivity, so we may have underestimated WMH burden in our sample.Furthermore, the lesion segmentation toolbox we used did not quantify the location of the WMH.Visual inspection of the FLAIR images revealed that most large lesions were periventricular, consistent with prior studies in HIV disease, 20 but deep WMH were also common although generally smaller.Recent studies suggest that subclassifying WMH according to location and severity may reveal more specific information about NCI. 110Fourth, our study did not assess all possible relevant outcomes and constructs that may contribute to WMH burden.For example, we only examined WMH, rather than the full-spectrum of cerebral small vessel disease.While our analysis did evaluate some social determinants of health outcomes, there are likely a host of other important socioeconomic factors that our analysis did not consider, such as food insecurity and poverty.Finally, our crosssectional analysis does not allow for causal inferences, and we cannot exclude the possibility of residual confounding.
Our principal finding is that PWH who use cocaine have higher WM lesion burden, even after accounting for CVD risk factors, that contribute to cognitive impairment.While the relative burden of WMH remained modest for most participants, these lesions represent only the end of a continuous spectrum of WM injury and neurodegeneration. 111,112As the sample was relatively young, with a mean age of 45, these WMH may portend future cerebrovascular events, vascular dementia, and frailty as PWH age.Clinically, as we observed, even small lesion load can predict cognitive impairment in middleage cognitively healthy adults. 113Longitudinal studies can help identify patients with progressive changes in brain structure and integrity, and ultimately determine the clinical significance of WMH observed in PWH.In sum, given the deleterious role of cocaine use in promoting neural injury and NCI in PWH, our results reinforce the importance of substance use screening for all patients receiving HIV care and referral to treatment and support services when indicated to minimize additional WM injury that may exacerbate NCI.
raw scores for each test were converted to demographically corrected Tscores (M = 50, SD = 10).Domain T-scores were computed by averaging the T-scores for the tests within each domain, and a global T-score was computed by averaging the 7 domain T-scores.

ª
2023 The Authors.Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association.

1636 ª 2023
The Authors.Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association.

Figure 1 .
Figure 1.Group frequency maps showing the WMH distributions for participants categorized as having low, moderate, and high WM lesion loads.

Figure 2 .
Figure 2. Stacked bar graphs show the proportion of participants in each COC group categorized by white matter burden.

Figure 3 .
Figure 3.The scatterplot shows the negative correlation between WM lesion burden and global T-score across the full sample.The standardized residuals from the partial correlation controlling for age are plotted.The individual data points are coded by COC status to illustrate that the relationship was similar across the groups.
Information Processing: Trail Making Test Part Anumber of seconds to completion 66 ; Wechsler Adult Intelligence Scale-IV (WAIS-IV) Coding subtest-total number correct 65 ; Stroop Color and Word Test color naming score-total number of items completed 67 3. Learning (immediate recall): Hopkins Verbal Learning Test-Revised (HVLT-R)-total number of words recalled on trials 1-3 68 ; Brief Visuospatial Memory Test-Revised (BVMT-R)-total score for figures recalled on trials 1-3 69 4. Memory (delayed recall): HVLT-R-total number of words recalled on trial 4 68 ; BVMT-R-total score for figures recalled on trial 4 69 5. Executive function: Stroop Color and Word Test interference score-difference between actual and predicted score on the Color-Word trial 67 ; Trail Making Test Part B-number of seconds to completion 66 ; Wisconsin Card Sorting Test-64 (WCST)-total errors 70 6.Verbal fluency: FAS letter fluency-total number of words generated; and category fluency-total number of animals generated 71 7. Motor skills: Grooved Pegboard Test dominant and non-dominant hand-number of seconds to completion 65; WAIS-IV Digit Span subtesttotal number correct65; WAIS-IV Letter-Number Sequencing subtest-total number correct652.

Table 1 .
Sample characteristics by cocaine use status.
INSTI, integrase strand transfer inhibitor; NNRTI, non-nucleoside reverse transcriptase inhibitor; NRTI, nucleoside reverse transcriptase inhibitor.1638 ª 2023 The Authors.Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association.

Table 2 .
Hierarchical linear regression model predicting WM lesion burden.

Table 3 .
Biomarkers by cocaine use status, controlling for age.
ª 2023 The Authors.Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association.

Table 4 .
Biomarker values and association to WMH categorization.
ª 2023 The Authors.Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association.