Assessing combinatorial effects of HIV infection and former cocaine dependence on cognitive control processes: A high-density electrical mapping study of response inhibition.

Abstract

Stimulant drug use in HIV+ patients is associated with poor personal and public health outcomes, including high-risk sexual behavior and faster progression from HIV to AIDS. Inhibitory control--the ability to withhold a thought, feeling, or action--is a central construct involved in the minimization of risk-taking behaviors. Recent neuroimaging and behavioral evidence indicate normalization of inhibitory control processes in former cocaine users as a function of the duration of drug abstinence, but it is unknown whether this recovery trajectory persists in former users with comorbid HIV. Here, we investigate the neural correlates of inhibitory control in 103 human subjects using high-density EEG recording as participants performed a Go/NoGo response inhibition task. Four groups of participants were recruited, varying on HIV and cocaine-dependence status. Electrophysiological responses to successful inhibitions and behavioral task performance were compared among groups. Results indicate persistent behavioral and neurophysiological impairment in HIV+ patients’ response inhibition despite current abstinence from cocaine. Analysis of task performance showed that HIV+ abstinent cocaine-dependent participants demonstrate the lowest performance of all groups across all metrics of task accuracy. Planned comparisons of electrophysiological components revealed a main effect of scalp site and an interaction between HIV-status and scalp site on N2 amplitudes during successful inhibitions. Analysis of the P3 time region showed a main effect of scalp site and an interaction between HIV-status and cocaine dependence. These results suggest synergistic alterations in the neurophysiology of response inhibition and indicate that abstinence-related recovery of inhibitory control may be attenuated in patients with HIV.

1. Introduction

Individuals with a substance use disorder, including cocaine dependence (CD), who also have a diagnosis of human immunodeficiency virus infection (HIV+), typically show poor health outcomes compared to HIV+ individuals who do not use recreational drugs. These outcomes include hastened transition from HIV to AIDS (Baum et al., 2009), increased cognitive impairment (Meade et al., 2011), and decreased medication adherence (Meade et al., 2011). One potential mechanism that may account for this difference is that exposure to cocaine increases blood brain barrier (BBB) permeability to bacterial and viral pathogens (Dhillon et al., 2008), placing active cocaine users who are also HIV+ at significantly greater risk of developing HIV-related opportunistic infections affecting the central nervous system (Lucas et al., 2006). Although the introduction of highly-active antiretroviral medication (HAART) has greatly reduced the impact of HIV on the central nervous system, neurocognitive impairment is still observable in more than half of HIV+ individuals (Heaton et al., 2010). Given that neurocognitive test performance has been shown to predict retention in chemical dependency treatment programs (Aharonovich et al., 2006), neurocognitive impairment in HIV+ individuals represents a significant hurdle to treatment completion and subsequent sobriety.

Active cocaine users, both with and without HIV infection, show clear deficits in response inhibition (RI) and other measures of executive function instrumental in maintaining drug abstinence (Bettcher et al., 2016; Koob and Volkow, 2016; Madsen et al., 2010; Volkow et al., 2005). Functional magnetic resonance imaging (fMRI) has shown hypoactivation during RI in active cocaine users in the right, middle and inferior frontal gyri, bilateral insula and anterior cingulate, as well as the precentral gyrus: canonical nodes of the response inhibition circuit (RIC) (Garavan and Hester, 2007; Garavan et al., 2008; Liu et al., 2019; Spechler et al., 2016). This reduced activity in people who are or were recently CD in regions necessary for error detection and adaptive behavioral control recovers as a function of time after the beginning of abstinence (Connolly et al., 2012; Morie et al., 2014b). However, the combined impact of HIV+ serostatus and CD on executive function and RI specifically are key questions that have not yet been addressed. In particular, it is unknown whether the previously observed recovery of RI processes in CD abstinent individuals, is also observed in individuals who are HIV+ following abstinence.

Here, we employed high-density electrophysiological recordings (256 channels) during a well-established RI task to explore the neural and behavioral correlates of addiction recovery in HIV+ individuals. We used a “Go/NoGo” task whereby participants were instructed to press a button upon presentation of novel pictorial stimuli (“Go” trials), but to withhold/inhibit the button press if the same stimulus was presented twice in a row (“NoGo” trials). Because of the intimate link between drug craving and relapse (Sinha, 2013), a subset of the stimulus tokens also included images of cocaine and cocaine-related paraphernalia. Our primary dependent measures were behavioral indices of task performance and the amplitude of the N2 and P3 components of the event-related potential (ERP). A substantial literature has demonstrated the association of these components with RI processes (Bokura et al., 2001b; Pfefferbaum et al., 1985; Roberts et al., 1994), and both components show a characteristic increase in amplitude during NoGo trials relative to Go trials (Groom and Cragg, 2015). The N2 is a frontally-generated negative potential which occurs 200–300 ms after stimulus presentation and is associated with processes related to the detection of response conflict (Groom and Cragg, 2015). The P3, a positive potential appearing between 300–500 ms after stimulus onset, shows maximal amplitude over the central scalp (Groom and Cragg, 2015) and is associated with the exercise of inhibitory control (Kok et al., 2004).

Prior research has demonstrated associations between NoGo N2/P3 components, task performance, and clinical variables. For example, delayed N2 onset was associated with increased risk of developing mild neurocognitive impairment in ageing adults (Howe, 2014; Polich et al., 2000). However, only one study to date has assessed alterations in N2 amplitude as a function of HIV status, finding no significant amplitude differences between HIV+ and HIV− subjects (Polich et al., 2000). Further, preliminary data indicate that P3 amplitude may correlate with behavioral performance, with low P3 amplitude associated with an increased error rate on NoGo trials (Yin et al., 2016). In line with this, ERP studies of individuals with HIV have consistently shown delayed P3 onset times and reduced fronto-central P3 amplitudes (Bauer, 2011; Fernández-Cruz and Fellows, 2017), findings thought to reflect gray matter atrophy (Dekaban and Sadowsky, 1978). Although a substantial literature indicates that active cocaine use is associated with reduced fronto-central N2 and P3 amplitudes (Morie et al., 2014a; Sokhadze et al., 2008), these deficits have been shown to ameliorate following abstinence from cocaine (Bauer, 2001; Morie et al., 2014b).

We hypothesized that individuals with a history of cocaine dependence and HIV infection (CD+/HIV+) would show persistent RI deficits despite current cocaine abstinence and that these deficits would be substantially greater than those seen in individuals with either CD or HIV alone. We further hypothesized that these RI deficits would be exacerbated in CD+ patients compared to CD− patients, specifically when participants were directed to withhold a response to cocaine-related images. Differential activation patterns between CD+/HIV+ and CD+/HIV− patients would support our thesis of persistent impairment in the neural mechanisms involved in recovery of inhibitory control following cocaine abstinence in individuals with comorbid HIV.

2. Methods

2.1. Participants

A total of 103 participants were included in this study, comprising 37 healthy controls (CD−/HIV−), 20 CD−/HIV+, 28 CD+/HIV−, and 18 CD+/HIV+ individuals. They were recruited via flyers posted in local community centers and clinics, direct referrals from clinicians, internet postings, word of mouth, and local research participant registries. Table 1 outlines the demographic composition of the sample. At an initial visit, participants were administered the SCID-V-CV psychiatric interview to test for the presence of past and current psychiatric conditions. Individuals with a history of substance use (CD+/HIV− and CD+/HIV+) met diagnostic criteria for Cocaine Use Disorder as determined through the Structured Clinical Interview for the DSM-IV (SCID-IV) Module E(First, 2014). Self-reported duration of abstinence, years of drug use, and years of education were also recorded. Because of the high rate of comorbidity of other drug usage in substance dependent individuals, participants were not excluded if they had abused drugs other than cocaine as long as cocaine was identified as the drug of choice. Abstinence was verified in all subjects via urinalysis and self-report. HIV laboratory measures, including current CD4 count, viral load, and HAART medication regime, were obtained from medical records. All HIV+ participants were receiving HAART.

Table 1: Demographics.

Sample characteristics by study group.

	HIV−		HIV+		Statistic	Effect Size	p value
	CD− (n = 37)	CD+ (n = 28)	CD− (n = 20)	CD+ (n = 18)
Demographic Characteristics
Sex Assigned at Birth, N Male : N Female	14 : 23	13 : 15	14#: 6	10 : 8	χ2(3) = 5.745	Ф = 0.236	0.125
Age in years, M (SD)	42.18 (12.63)	41.69 (9.79)	46.03 (13.3)	49.05 (9.30)	F(3,99) = 2.04	η2 = 0.058	0.113
Race, N					χ2(6) = 25.73	Ф = 0.500	0.000*
African American	5	7	9	12
Caucasian	25	21	10	6
Asian	7	0	1	0
Education in years, M (SD)	16.46 (2.85)	12.75 (1.58)	13.40 (2.44)	12.17 (2.38)	F(3,99) = 19.76	η2 = 0.375	0.000*
THC positive urine, N	3	2	7	2	χ2(3) = 9.246	Ф = 0.314	0.041*
Current Smoker, N	2	22	5	13	χ2(3) = 45.16	Ф = 0.662	0.000*
Cocaine use characteristics
Lifetime use in years, M (SD)		16.14 (10.11)		18.28 (9.89)	t(44) = −0.705	Cohen’s D = −0.213	0.485
Duration of abstinence, M (SD)		0.821 (0.914)		0.895 (1.09)	t(44) = −0.247	Cohen’s D = −0.075	0.806
HIV Characteristics
HIV viral load <20 copies, N			18	14	χ2(1) = 1.064		0.395
Current CD4 cell count, M (SD)			773 (602)	756 (271)	t(36) = 0.115	Cohen’s D = −0.038	0.909
Years since diagnosis, M (SD)			14.26 (10.87)	16.18 (8.46)	t(36) = −0.604	Cohen’s D = −0.196	0.550

^#1 participant assigned male at birth has transitioned to female

Exclusion criteria were as follows: (1) A diagnosis of current Major Depressive Disorder, Bipolar I, or Schizophrenia as assessed via the Structured Clinical Interview for the DSM-V Clinical Version (SCID-V-CV) (First, 2013), (2) A history of head injury or loss of consciousness for > 30 minutes, (3) Any self-reported history of neurological disorders or brain pathology, (4) A positive urine screening for any drug of abuse, excluding marijuana/THC. Participants in the non-substance dependent groups (CD−/HIV− and CD−/HIV+) were additionally excluded for any lifetime history of any substance use disorder. 3 CD−/HIV−, 2 CD+/HIV−, 7 CD−/HIV+, and 2 CD+/HIV+ subjects tested positive for THC on urinalysis.

High-density EEG recordings were obtained while participants performed a RI task for 102 out of 103 participants. For 1 subject, EEG recordings were not obtained due to hairstyle. Of the 102 subjects with EEG data, 6 participants were excluded from data analysis due to poor signal quality. Of the 6 total participants excluded from EEG analysis, 2 were CD−/HIV−, 2 were CD+/HIV−, 1 was CD−/HIV+, and 1 was CD+/HIV+. EEG data from a total of 96 participants and behavioral data from all 103 participants were submitted to further analysis.

All participants provided written and informed consent to the experimental procedures and were compensated $15/hour for participation. Participants with a history of cocaine dependence were compensated in the form of gift cards to local stores. This study was approved by the research participants review board at the University of Rochester (STUDY1081) and conformed to the tenets of the Declaration of Helsinki.

2.2. Procedures

Participants completed a phone pre-screening questionnaire to assess eligibility. Eligible participants were scheduled for an in-person visit during which subjects completed the SCID-5 interview, questionnaires, and neurocognitive testing, including the Wechsler Adult Test of Intelligence Working Memory subsection (Wechsler, 1955). This battery consisted of three subsections to evaluate the ability attend to information presented verbally and manipulate information in short term memory. The primary outcome measure from this subsection known as “Working Memory Index” (WMI), represents a proxy measure for IQ. Participants returned to the lab on an additional visit for the EEG recording. Before and after the EEG recording session, participants were administered the Cocaine Craving Questionnaire – Brief (CCQ-brief), a 10-item questionnaire to evaluate craving on a visual-analog scale (Tiffany et al., 1993). CCQ-brief data from before the EEG were not available for 2 CD−/HIV− subjects. CCQ-brief data from after the EEG were not available from the same 2 CD−/HIV− subjects, and 1 additional CD−/HIV+ subject.

2.3. Response Inhibition Task

Participants were asked to respond quickly to the onset of pictorial stimuli with a button press (Go trial) while withholding a response when a stimulus was repeated (NoGo trial). A visual illustration of the task is provided in Figure 1.

Figure 1: Go/NoGo Task.

1A. Illustration of Go/NoGo RI task. Picture of postive valence, neutral valence, and drug-related content are presented in sequence. Participants are instructed to withold a button press response to repeated images. Between blocks of emotional images, participants are presented with a distractor task. 1B. Task performance (D’) in the All Image condition as a function of participant group. 1C. Correlation between working memory IQ (WAIS − WMI) and task performance (D’) (r = 0.497, p < 0.001).

Stimuli were images from the International Affective Picture System (IAPS) (Lang et al., 2008), images of cocaine and related paraphernalia, and the letters “X” and “Y”. “Positive” and “Neutral” images from the IAPS, a database of emotionally evocative photographs, were chosen. The 38 positive images depicted families, animals, food, and sporting events, with a mean valence and arousal rating of 7.34 and 5.07 respectively on a normative scale of 1 to 9. The 38 neutral images depicted faces, objects, and abstract patterns and had a mean valence and arousal rating of 5.13 and 3.13 respectively. 38 images depicting cocaine, cocaine paraphernalia, and individuals using cocaine were picked from a stock image database (www.shutterstock.com).

All stimuli were presented centrally against a gray background and subtended a visual angle of 8.25 degrees horizontally and 6.0 degrees vertically. Stimuli were presented pseudorandomly for a duration of 60 ms, followed by a gray screen containing a central fixation cross presented for 940 ms. Each 30-second block of positive, neutral or cocaine-related images was preceded by a brief rest period lasting 2 seconds. Following the rest period, subjects were presented with 10 seconds of a distractor task in which the letters “X” and “Y” alternated serially at a frequency of 1.2Hz. Participants were asked to withhold a button press response when the same letter appeared twice in a row. The purpose of the distractor task was to disrupt cognitive processes related to the preceding block.

Participants completed a total of 540 trials per condition (Positive, Neutral, Cocaine, and XY distractor task). Of these 540 trials, 78 NoGo trials (14.4%) in each condition required the participant to withhold a button press response. Each image block of 30 trials contained between 2 and 8 NoGo trials (mean = 4.3, Standard Deviation (SD) = 1) and each XY block of 10 trials contained between 0 and 2 NoGo trials (mean = 1.4, SD = .6). NoGo trials could occur at any time within a block except on the first trial of each block or directly after another NoGo trial. Across all conditions, participants completed a total of 2160 trials. Participants were permitted to take a break every 2.2 minutes. The total task length not including breaks was 39.6 minutes.

2.4. Electrophysiological data collection and analysis

Participants were seated in a dimly lit, sound-attenuated, electrically shielded room (Industrial Acoustics Company, Bronx, New York). All participants completed one mandatory practice block before the experiment began to ensure understanding of the task. The RI task was controlled by custom software written in Presentation (Neurobehavioral systems) and presented on an Acer Predator Z34 LCD monitor (San Jose, California). Participants were seated approximately 95 cm away from the center of the monitor. High-density continuous EEG was acquired from 256 surface electrode sites using the BioSemi Active II system. Data were digitized online at 512Hz and referenced to the Common Mode Sense (CMS) active electrode. Facial electrodes were positioned bilaterally on the temple, upper mastoid, and lower mastoid for a total of six external electrodes (i.e. not incorporated into the electrode cap).

Offline analyses were conducted using custom MATLAB scripts leveraging EEGlab functions (Delorme and Makeig, 2004). Data were high-pass filtered at 0.01 Hz and low-pass filtered at 40 Hz using a Chebychev Type 2 spectral filter. Several artifact and channel rejection procedures were employed. Bad channels were detected automatically based on joint-probability, kurtosis, and covariance and were verified by visual inspection of topographic plots. A channel was defined as “bad” when recorded data from that electrode exceeded more than 3 SD from all other electrodes (joint-probability, kurtosis) and 2 SD from neighboring channels (joint-probability, covariance). Rejected channels were spherical spline interpolated.

Next, data were re-referenced offline to the common average, including the interpolated channels. For all participants, EEG epochs (−100 to 900 ms post-stimulus onset) were constructed offline using a baseline of - 100 to 0 ms. An automatic artifact rejection threshold of 75uV was used to eliminate trials containing blinks, eye movements, or electrical artifacts. An additional artifact rejection threshold was calculated based on an array of maximum amplitudes for each trial (the largest absolute value recorded in a given epoch across all channels). Epochs containing values > 3 SD from the median of this array of maximum values were removed.

“Go” trials on which the participants responded successfully were defined as Hits. “NoGo” trials on which participants correctly withheld their response were defined as Correct Rejections (CR). Trials on which participants responded incorrectly were excluded from the analysis. For each participant, epochs were averaged separately for each of three conditions (Positive, Neutral, Cocaine) and for each of two trial types (Hits and CR). An additional “All Image” condition was constructed by averaging Cocaine, Positive, and Neutral trials for Hits and CR separately. Grand average waveforms for each condition (Positive, Neutral, Cocaine, All Image) for each trial type (Hits and CR) were then generated for each participant group by averaging individual participant waveforms.

To ascertain time and regions-of-interest (ROI) effects, we generated a grand average CR waveform consisting of data from all participants (Figure 2A). Based on an extensive literature showing maximal N2 and P3 amplitudes over fronto-central sites (Patel and Azzam, 2005), we inspected the CR grand average of trials from the All Image condition from a cluster of midline frontal electrodes centering around electrode Fz in order to determine the N2/P3 peak latency. The maximal N2 amplitude occurred at 230 ms post-stimulus onset, and the maximal P3 amplitude occurred at 488 ms post-stimulus onset. This peak latency was used to define a 30 ms time window for the N2 (215 – 245 ms) and a 50 ms time window for the P3 (463 – 513 ms). A five-electrode cluster around the midline-frontal channel Fz, and a five-electrode cluster encircling the midline-parietal channel Pz served as ROIs. For every participant, average amplitudes within the N2 and P3 time windows were extracted from each of these two ROIs.

Figure 2: Experiment 2 – Effect of CD and HIV on N2 and P3 Amplitude.

2A. A 30 ms time-window for the N2 component and a 50 ms time-window for the P3 component were constructed based on peak amplitudes from the All Image CR trace consisting of data from all subjects at electrode Fz. For each individual, average amplitude within each time-window was extracted and compared. 2B. Individual data (thin lines) and group average (thick lines) amplitude of N2 component in the All Image condition are shown for HIV-negative (left) and HIV+ (right) subjects at anterior and posterior ROIs. 2C. Mean amplitude of N2 component at each ROI in the All Image condition. Error bars represent 95% confidence intervals. 2D. Individual data (thin lines) and mean P3 amplitude (thick line) in the All Image condition for each participant group at each ROI. 2E. Mean P3 amplitude in the All Image condition at each ROI. Error bars represent 95% confidence intervals. 2F. Table of N2 and P3 amplitudes given as Mean (SD) for each group and ROI.

Select, pairwise comparisons were chosen to isolate the effect of CD, HIV, and the interaction between CD and HIV on the neurophysiology of RI. To test the effect of CD on the neurophysiology of RI, we compared CR traces from CD+/HIV− to CD−/HIV− (both HIV− groups). To understand the effect of HIV on RI, we compared CR traces from CD−/HIV− to CD−/HIV+ (both CD− groups). To examine the effect of HIV on RI in CD+ participants, we compared CR traces from CD+/HIV− to CD+/HIV+. Finally, to examine the effect of CD on RI in HIV+ participants, we compared CR traces from CD−/HIV+ to CD+/HIV+.

2.5. Topographical Voltage Maps

Scalp topographic maps represent interpolated voltage distributions derived from 256 scalp measurements. These interpolated, 2D topographic maps quantify the average amplitude for each participant group within the time windows of the N2 and P3 components. Maps were rendered using EEGlab.

2.6. Response Inhibition task performance data

To examine behavioral performance during RI, several measures were considered. Reaction times (RT) to “Go” stimuli were extracted for each participant for each condition. To examine post-error slowing, RT values from Go trials directly following commission error trials—also known as False Alarms (FAs)—were subtracted from RT values from Go trials directly following CR trials. An additional measure of error-related behavioral adjustment was computed to account for instances of failure to respond to a Go trial directly following an FA. This measure, which we have termed “Post-Error Task Abort,” quantifies the percentage of post-FA Go trials on which participants did not respond with a Hit. Task accuracy was quantified through Hit rate (percentage of Go trials correctly eliciting a response), CR rate (percentage of NoGo trials correctly eliciting no response), and D’, a measure of signal detection sensitivity (Green and Swets, 1966). For each participant, D’ was computed by subtracting the z-score of their Hit rate from the z-score of their FA rate as given by the following formula: d’ = z(FA) − z(Hit).

2.7. Statistical Analyses

Statistical analyses were conducted using IBM SPSS Statistics 26.0, 2020. For demographic data, 1-way ANOVAs were used to test for between-group differences among continuous variables (e.g. age, years of education). Chi-Square tests were used to identify between-group differences among categorical variables (e.g. race, sex). For analyses including categorical variables containing fewer than 5 observations per cell, Fishers Exact Test was used. Independent-sample T-tests were used when data for only two groups was present (e.g. duration of cocaine usage). For analyses of RI task performance, WAIS-WMI scores, and RT measures, 2-way ANOVAs with the between-subjects factors “HIV” and “CD” were employed to examine main and interaction effects.

To examine group differences in the amplitude and topography of N2 and P3 components, we fit one univariate General Linear Model (GLM) for each component of interest for a total of two omnibus models. CD (CD+ and CD−) and HIV (HIV+ and HIV−) were included as between-subject factors, and ROI (anterior vs. posterior) was included as a within-subjects factor. 2-way interaction terms between CD, HIV, and ROI were included in the model, in addition to a 3-way interaction term between ROI × CD × HIV. Because advancing age is known to alter the topography of the components of interest (Friedman et al., 1993), age was included as a covariate.

To characterize the relationship between EEG measures and behavioral task performance, Pearson correlation coefficients were computed separately for each ERP outcome measure and task performance measure (D’, CR rate, FA rate, Hit Rate). To evaluate the relationship between task performance and executive function, we correlated WAIS-WMI scores with d’.

2.8. Exploratory Statistical Cluster Plots

A secondary exploratory analysis was also performed to fully explore the richness of these high-density data. We computed statistical cluster plots (SCP) using MATLAB to test the entire data matrix for putative effects, a method which has been effectively used by our group and others in post hoc analyses as a means to generate pointed follow-up hypotheses (De Sanctis et al., 2014; Molholm et al., 2002). Pointwise, 1-way ANCOVAs were calculated to examine differences in CR traces between participant groups while controlling for Age. F-values for the main effect of participant Group were plotted as intensity values to efficiently summarize and facilitate identification of differences between groups in the onset and general topographic distribution of activation. The x-axis of the SCP represents time post-stimulus onset in milliseconds, and the y-axis represents individual electrodes. We are aware that conclusions based on SCPs could potentially result in a high rate of false-positive errors due to the large volume of tests performed. To account for this issue, only timepoints for which the ANCOVA main or interactive effect exceeded the 0.05 p-value criterion for at least 8 consecutive data points (15.6 ms) were considered significant.

3. Results

3.1. Participant Characteristics

Table 1 summarizes the sample by group. The groups were comparable on age, sex, HIV− characteristics, and cocaine use characteristics, but there were significant difference in race (χ²(6) = 25.73, Φ = 0.500, p < 0.001), years of education (F(3,99) = 19.76, η² = 0.375, p < 0.001), percentage of current smokers (χ²(3) = 45.16, Φ = 0.662, p < 0.001) and number of THC+ urine samples between groups (χ²(3) = 9.246, Φ = 0.314, p = 0.041). T-tests between both HIV+ groups showed no significant difference in current CD4 cell count, HIV viral load, or duration of infection. T-tests between both CD+ groups showed no significant difference in duration of abstinence or duration of cocaine usage.

3.2. Cocaine Craving Questionnaire – Brief

Compared to CD− participants, CD+ participants showed significantly elevated cocaine craving both at baseline (t(98) = −2.995, p = 0.003) and following task completion (t(96) = −4.295, p < 0.001). Within the CD+ group, HIV− participants showed elevated craving relative to HIV+ participants at baseline (t(44) = 2.139, p = 0.038), but not following task completion (t(44) = 0.790, p = 0.433).

3.3. Response Inhibition Task Performance and Neurocognitive Assessment

CD+/HIV+ participants demonstrated the lowest behavioral task performance of all participant groups for all image conditions across all metrics of task accuracy (Figure 1B, Table 2). Using the sensitivity index (d’), 2-way ANOVAs revealed significant main effects of HIV and CD across all task conditions. For Hits, results showed statistically significant main effects of CD and HIV across all conditions, with all patient groups showing significantly reduced hit rate relative to CD−/HIV− healthy control participants. For CRs, results indicated a statistically significant main effect of HIV in Positive and All Image conditions, with all patient groups showing significant reductions in the rate of CRs compared to healthy control participants. There were no statistically significant differences in FA-rate as a function of image valance or participant group.

Table 2: Experiment 1 – Neurocognitive Testing and Task Performance.

d’, Hit rate, CR rate and FA rate (Mean (SD)) presented for each participant group and each image valance condition. Shaded cells indicate a statistically significant result. Working Memory Index (WMI) derived from the Wechsler Adult Test of Intelligence is also given.

	HIV−		HIV+		CD Main Effect			HIV Main Effect			HIV by CD interaction
	CD− (n = 37)	CD+ (n = 28)	CD− (n = 20)	CD+ (n = 18)
					F(1,99)	η2	p	F(1,99)	η2	p	F(1,99)	η2	p
WMI	105 (13.1)	89.4 (10.3)	89.5 (13.6)	82.9 (13.0)	18.473	0.157	0.000	18.187	0.155	0.000	2.888	0.027	0.092
All Image
D’	2.58 (0.97)	2.04 (1.07)	1.99 (1.24)	1.49 (1.11)	5.450	0.052	0.022	6.534	0.062	0.012	0.009	0.000	0.927
Hit %	96.6 (4.00)	92.5 (8.18)	93.3 (7.08)	84.8 (12.8)	15.167	0.133	0.000	11.458	0.104	0.001	1.822	0.018	0.180
CR %	65.8 (19.7)	59.0 (19.2)	56.1 (24.3)	51.5 (21.3)	1.750	0.017	0.189	4.046	0.039	0.047	0.066	0.001	0.797
FA %	33.5 (19.2)	39.1 (18.7)	42.9 (23.6)	43.1 (21.3)	0.493	0.005	0.484	2.538	0.025	0.114	0.431	0.004	0.513
Positive
D’	2.66 (1.03)	2.03 (1.04)	1.97 (1.30)	1.53 (1.10)	5.74	0.053	0.020	6.926	0.065	0.010	0.166	0.002	0.684
Hit %	96.6 (4.13)	92.3 (8.29)	93.05 (7.74)	84.7 (13.3)	14.446	0.127	0.000	11.212	0.102	0.001	1.450	0.014	0.231
CR %	66.0 (19.6)	54.6 (25.0)	59.7 (18.6)	52.1 (19.7)	1.135	0.011	0.289	4.945	0.048	0.028	0.226	0.002	0.634
FA %	33.1 (19.3)	38.5 (19.1)	44.3 (24.4)	42.0 (20.3)	0.140	0.001	0.709	3.050	0.030	0.084	0.833	0.008	0.364
Neutral
D’	2.67 (1.03)	2.10 (1.13)	2.04 (1.28)	1.60 (1.22)	4.677	0.045	0.033	5.853	0.056	0.017	0.069	0.001	0.793
Hit %	96.8 (4.04)	92.3 (9.53)	93.8 (6.59)	85.2 (12.7)	15.239	0.133	0.000	9.108	0.084	0.003	1.559	0.015	0.215
CR %	66.8 (20.6)	59.5 (19.6)	55.8 (25.4)	52.8 (23.6)	1.305	0.013	0.256	3.872	0.039	0.052	0.225	0.002	0.637
FA %	32.6 (20.0)	38.4 (18.6)	43.0 (24.7)	41.4 (23.7)	0.217	0.002	0.642	2.366	0.023	0.127	0.717	0.007	0.399
Cocaine
D’	2.50 (0.97)	2.04 (1.15)	2.04 (1.26)	1.42 (1.19)	5.524	0.053	0.021	5.494	0.053	0.021	0.114	0.001	0.737
Hit %	96.3 (4.26)	92.9 (7.44)	93.1 (7.30)	84.8 (13.0)	13.546	0.120	0.000	12.577	0.113	0.001	2.429	0.024	0.122
CR %	64.5 (20.5)	58.0 (22.1)	57.9 (24.0)	49.7 (22.3)	2.687	0.026	0.104	2.742	0.027	0.101	0.036	0.000	0.850
FA %	34.6 (20.0)	40.5 (21.4)	41.2 (23.1)	45.7 (21.6)	1.405	0.014	0.239	1.794	0.018	0.184	0.027	0.000	0.870

2-way ANOVAs revealed no statistically significant differences in RT to Hit trials as a function of CD, HIV, or their combination (Table 3). A main effect of HIV on Post-Error Slowing was seen in the Positive and All Image conditions. A main effect of CD on ‘Post-Error Task Abort’ was observed in all conditions except Neutral, and a main effect of HIV was observed in the Positive and All Image conditions.

Table 3: Experiment 1 – Reaction Time:

Reaction time, post-error slowing, and ‘post-error task abort’ presented for each participant group and each image valence condition. Values indicate Mean (SD). Shaded cells indicate significant results.

	HIV−		HIV+		CD Main Effect			HIV Main Effect			HIV by CD interaction
	CD− (n = 37)	CD+ (n = 28)	CD− (n = 20)	CD+ (n = 18)
					F(1,99)	η2	p	F(1,99)	η2	p	F(1,99)	η2	p
All Image
RT	409 (62.2)	425 (80.6)	410 (74.5)	441 (79.5)	2.454	0.025	0.120	0.291	0.003	0.591	0.267	0.003	0.606
Post-Error Slowing	64.7 (70.9)	80.7 (87.3)	109 (65.5)	108.3 (74.9)	0.239	0.002	0.626	5.423	0.052	0.022	0.302	0.003	0.584
Post-Error Task Abort	9.64 (11.5)	12.5 (10.3)	11.5 (10.4)	21.3 (18.9)	6.046	0.058	0.016	4.280	0.041	0.041	1.872	0.019	0.174
Positive
RT	403 (61.2)	413 (75.5)	400 (71.7)	432 (76.2)	2.152	0.021	0.146	0.340	0.003	0.561	0.585	0.006	0.446
Post-Error Slowing	65.6 (74.6)	75.1 (98.6)	113 (74.9)	118 (90.8)	0.174	0.002	0.678	6.740	0.064	0.011	0.017	0.000	0.898
Post-Error Task Abort	7.78 (11.0)	11.7 (11.0)	12.3 (12.7)	23.3 (22.1)	6.880	0.065	0.010	8.016	0.075	0.006	1.539	0.015	0.218
Neutral
RT	402 (61.8)	417 (77.6)	402 (72.7)	439 (83.7)	3.135	0.031	0.080	0.653	0.007	0.421	0.844	0.008	0.361
Post-Error Slowing	68.2 (82.5)	85.2 (87.0)	96.2 (70.8)	106 (78.9)	1.191	0.012	0.278	2.813	0.028	0.097	0.221	0.002	0.639
Post-Error Task Abort	11.1 (15.3)	14.5 (13.4)	9.73 (9.72)	15.47 (11.3)	2.875	0.028	0.093	0.006	0.000	0.938	0.187	0.002	0.666
Cocaine
RT	423 (66.4)	440 (82.5)	427 (81.1)	451 (80.4)	1.722	0.017	0.192	0.220	0.002	0.640	0.069	0.001	0.793
Post-Error Slowing	68.4 (78.2)	92.6 (90.1)	121 (73.7)	92.7 (95.3)	0.016	0.000	0.901	2.358	0.023	0.128	2.346	0.023	0.12.9
Post-Error Task Abort	10.2 (11.9)	12.1 (12.2)	12.0 (10.5)	21.2 (22.4)	4.164	0.040	0.044	3.851	0.037	0.053	1.838	0.018	0.178

No statistically significant HIV by CD interaction effects were observed in any condition for any metric of task accuracy or RT, indicating that the combinatorial effect of CD and HIV on behavioral response inhibition is additive, rather than the synergistic.

3.4. Electrophysiological Data

3.4.1. N2 and P3 Components

To examine potential combinatorial effects of CD and HIV on the neurophysiology of RI, mean amplitude values for each participant were extracted and compared using GLM with the factors of ROI (anterior vs. posterior), CD status (CD+ vs. CD−), and HIV status (HIV+ vs. HIV−) while controlling for the effect of participant age (Figure 2A). The GLM for the N2 component revealed a significant main effect of ROI (F(1,183) = 134.742, η² = 0.424, p < 0.001), and an interaction between ROI and HIV (F(1,183) = 4.880, η² = 0.026, p = 0.028) (2B, 2C). The GLM for the P3 component revealed a significant main effect of ROI (F(1,183) = 18.877, η² = 0.094, p < 0.001) and a trend toward an interaction between CD × HIV (F(1,183) = 2.976, η² = 0.016, p = 0.086) (2D, 2E), with CD+/HIV+ showing reduced P3 amplitude at both anterior and posterior sites relative to all other participant groups in the time windows specified. 2F shows the mean and standard deviation of N2 and P3 amplitude as a function of subject group and ROI.

3.4.2. Electrophysiological Waveforms

Figure 3A shows the mean ERP waveforms from the All Image condition for CR trials measured from the anterior cluster of electrodes centered on Fz (top), and the posterior cluster centered on electrode Pz (bottom). 3B shows a topographic representation of the average voltage in CR traces in the time regions of the peak N2 (left column) and P3 (right column) components. In conjunction with omnibus GLM indicating an interaction between HIV and ROI on N2 amplitude, the topographic plots provide a visual representation of an anterior topographic shift in the N2 component. To identify temporal and spatial regions showing synergistic/non-additive changes in RI neurophysiology as a function of CD and HIV, we constructed an SCP (3C). Colored regions indicate the F-values of the interaction term between CD and HIV in a 2-way ANCOVA controlling for the effect of participant age. Statistically significant interactions were seen primarily over frontal, central, and parietal regions primarily between 50–200 ms (frontal, parietal) and between 400–550 ms (central). To further characterize the observed interactions, CR traces from electrodes showing statistically-significant interactions were extracted from each group and provided for illustrative purposes (3D).

Figure 3: Experiment 2 - Neurophysiology during Response Inhibition.

3A. Grand average waveforms of CR trials for the All Image condition plotted for each participant group. Gray shading surrounding ± 1 standard error. 3B. Topographic plots of voltage for a 30ms time-window surround the peak frontal N2 for each participant group (left column) and a 50ms time-window for the peak P3 (right column). 3C. Statistical cluster plot displaying regions showing statistically-significant interactions between CD and HIV from a 2-way ANCOVA controlling for participant age. Colors indicate the F-statistic of the interaction term. White box indicates electrodes plotted in Figure 3D. 3D.Grand average waveforms averaged across eight fronto-central electrodes showing statistically-significant interactions between CD and HIV based on the SCP in Figure 3C.

Pairwise differences between groups were examined using select comparisons of CR waveforms at electrodes Fz and Pz. To examine the effect of CD on the neurophysiology of RI controlling for the effect of HIV, we compared CR traces from both HIV− participant groups (CD−/HIV− to CD+/HIV−) (Figure 4A). To explore the effect of HIV on RI while controlling for the effect of CD, we compared traces from both CD− groups (CD−/HIV− to CD−/HIV+) (4B). To test whether CD-related changes were exacerbated in individuals who also had HIV, we compared CR traces from both CD+ groups (CD+/HIV− to CD+/HIV+) (4C). Finally, to examine whether HIV-related changes were exacerbated in participants who also had CD, we compared CR traces from both HIV+ groups (CD−/HIV+ to CD+/HIV+) (4D).

Figure 4: Experiment 2 – Effect of CD, HIV, and CD/HIV on Response Inhibition:

4A. Effect of CD on CR waveforms in HIV− participants. 4B. Effect of HIV on CR waveforms in CD− participants. 4C. Effect of HIV on CR waveforms in CD+ participants. 4D. Effect of CD on CR waveforms in HIV+ participants. Gray shading represents ± 1 standard error.

For each comparison, SCPs were constructed showing F-values from a 1-way ANCOVA examining the effect of Group on CR waveforms while controlling for participant age (4A–D). Statistically significant differences in CR traces were observed between CD−/HIV− and CD+/HIV− primarily over central, parietal and occipital regions between 100–150 ms and later between 750 –900 ms post-stimulus onset (4A). Significant differences were observed between CD−/HIV− and CD−/HIV+ groups primarily during the baseline period over parietal regions (4B). Significant differences were observed between CD+/HIV− and CD+/HIV+ over all regions between 150–250 ms post-stimulus onset, and over frontal, central and parietal scalp regions at 450–650 ms (4C). Finally, widespread significant differences were observed between CD−/HIV+ and CD+/HIV+ primarily in the 50–250 ms post-stimulus interval over frontal, central, parietal, and occipital regions, as well as over central and parietal regions in the 450–600 ms, and later over bilateral temporal regions (700–900 ms) (4D).

3.4.3. Effect of Image Valence and Cocaine Cues

To test whether RI neurophysiology is altered by image valance and cocaine-cuing, CR traces for each image valance condition (Positive, Neutral, Cocaine) were plotted separately for each group. Figure S1A–D show waveforms from electrodes Fz and Pz for each group separately, and S1E shows all participants grouped together. There were minimal statistically significant differences in CR traces as a function of image valence in any group except for CD−/HIV+ over the central region 800–900ms post-stimulus onset, and CD−/HIV+ across central, temporal, parietal and occipital regions between 0 – 50ms post-stimulus onset.

3.5. Correlations

To further characterize the relationship between task performance measures (d’, CR rate, FA rate, Hit rate) and EEG amplitude, several exploratory correlations were performed. Statistically-significant correlations were identified between N2 amplitude, d’, FA rate, and CR rate over midline-posterior electrodes. (Figure S2A, S2B).

4. Discussion

This study provides evidence for both additive and interactive impairment in response inhibition CD+/HIV+ individuals relative to those with either condition alone. Planned, between-group comparisons of ERP amplitude during the N2 time period, associated with response conflict detection (Groom and Cragg, 2015), showed an interaction between HIV-status and scalp site, with HIV+ participants showing increased N2 amplitudes relative to HIV−. Planned comparisons of EEG amplitude for the P3 time period showed a trend toward an interaction between HIV and CD. In a whole-brain analysis, interactive changes in the P3 time region were observed primarily over fronto-central scalp sites. Taken together, these results indicate HIV and CD may impact overlapping neural substrates to cause substantial reductions in neurophysiological indices of cognitive control.

It is well established that active users of cocaine experience broad neuropsychological dysfunction in domains of executive processing such as attention, working memory, and declarative memory (Jovanovski et al., 2005; Vonmoos et al., 2013). However, data from our lab and others using methods such as neurocognitive assessment (Vonmoos et al., 2014), neurophysiological recording (Morie et al., 2014b), and neuroimaging (Bell et al., 2011; Bell et al., 2014; Wakim et al., 2017), have shown that these impairments are at least partially reversible following abstinence from cocaine. The current study provides evidence indicating that this same recovery trajectory may not hold true for patients recovering from CD who are also HIV+.

CD+/HIV+ participants showed clearly reduced task accuracy compared to all other groups across every condition tested herein. For measures of signal detection sensitivity—including d’ and Hit-rate—statistically significant decreases in performance across all task conditions as a function of HIV, or both CD and HIV were observed. CD+/HIV+ participants were also more likely than any other group to respond with a “miss” following a commission error, suggesting that CD+/HIV+ participants showed alterations in error processing which could result in difficulty re-initiating task engagement following error. Although CD+/HIV+ demonstrated the lowest performance of all groups across all metrics of task speed and performance, no statistically significant interactions were identified, suggesting that the effect of HIV and CD on behavioral RI may be additive, rather than interactive.

Recent functional neuroimaging data in active cocaine users with comorbid HIV provide support for the idea of interactive impairment in neural function. Meade and colleagues demonstrated using fMRI that cocaine use exacerbates HIV-associated alterations in cognitive control-related brain circuitry during the evaluation of delayed rewards (Meade et al., 2017). Interactive effects between HIV and CD were seen in several regions of the RIC, including the inferior and middle frontal gyri, orbitofrontal cortex (OFC), and precentral gyrus. Notably, these regions show strong overlap with the neural generators of N2 and P3 ERP components (Bokura et al., 2001a; Huster et al., 2010). Current-source imaging implicates the right lateral OFC and mid-cingulate cortex as major generators of the NoGo N2 and P3 components respectively (Bokura et al., 2001a; Huster et al., 2010), and intracranial recordings in the Macaque have clearly localized N2 processes to the ACC (Dias et al., 2006). The anatomical convergence of interactive effects reported by Meade and colleagues, as well as the interactive effects shown in the current study suggest that the identified CD/HIV-related executive impairments have functional implications that extend beyond RI to the broader constructs of decision making and risk evaluation. Further, that these neural changes persist after cessation of cocaine usage supports the thesis that HIV+ patients recovering from CD represent a sub-population of CD patients who may be especially vulnerable to poor clinical outcomes, including early termination of substance use treatment programs and drug relapse.

The pattern of activation observed in CD+/HIV+ individuals—that is, an amplified and more anteriorly distributed N2 accompanied by a decreased and also anteriorly shifted P3—could be reasonably interpreted in terms of neural compensation theory (Bartlett et al., 2005). This theory posits that individuals with reduced cognitive reserve may recruit alternative brain networks not typically used by healthy individuals or hyper-activate existing networks in order to maintain task performance (Bartlett et al., 2005; De Sanctis et al., 2009). As evidenced by substantial impairment in task performance in the CD+/HIV+ group (Table 2, Figure 1B), increased N2 amplitudes at fronto-central sites in CD+/HIV+ relative to CD+/HIV− participants (Figure 4C) and a topographically-shifted P3 component (Figure 2B), it is reasonable to speculate that neural compensation may be at play and, further, may be insufficient to overcome the neurocognitive barrier of comorbid HIV and CD. It is of note that previous reports have shown reduced NoGo fronto-central P3 amplitude with HIV+ serostatus (Bauer, 2011; Fernández-Cruz and Fellows, 2017), whereas in the current study we did not find reduced P3 in participants with HIV who did not also have a history of CD. Here, it was only participants with both HIV and a history of drug use that showed such a reduction, suggesting that accounting for drug use history may be an important factor in studies moving forward.

In regards to our initial hypothesis of RI modulation by drug cuing and emotional content, we did not observe electrocortical or behavioral modulation by image valence in any participant group. While this finding is consistent with data from our cocaine craving questionnaire indicating no difference in self-reported craving from baseline in CD participants, these data contrast with some previous EEG findings. Dunning et al. used a passive-viewing task to investigate sustained attention to drug cues in active and recently abstinent CD patients (Dunning et al., 2011). The authors identified late (400–2000 ms post-stimulus onset) ERP modulation by valence and drug-content. These conflicting findings may be driven by methodological differences in study design. The current study used a 1000ms inter-trial interval, with stimuli presented on screen for just 60 ms, whereas Dunning et al. employed a 2500 ms inter-trial interval and presented stimuli for 2000 ms. While the short image duration in the current study was chosen in order to isolate cognitive components of interest while minimizing tonic visual activation, as well as to ensure appropriate task difficulty, this duration may have been insufficient to stimulate detectable differences in late ERP components associated with salience attribution. As an alternative explanation, participants in the current study were engaging in a cognitively taxing inhibitory control task. It is possible that the image valence content was of secondary importance as participants worked to maintain task accuracy. This finding is consistent with that of a recent study by Luijten and colleagues examining the impact of smoking cues on RI in smokers (Luijten et al., 2011). While they observed lower overall task accuracy in smokers compared to non-smokers, this reduced performance was seen in both the neutral and smoking conditions, consistent with the interpretation that smokers exhibit global deficits in inhibitory control, rather than specific impairment in the presence of salient cues. It is also worth pointing out that, in this study, IAPS images were chosen only from the neutral and positive stimulus sets, and the negatively valanced images were not used. In prior work, it was the negative images from this picture set that evoked the largest ERP modulations under active cognitive control task conditions (De Sanctis et al., 2013).

This study is not without limitations. Although participant groups were matched for age, sex, and clinical variables, there were significant between-group differences in group racial composition, years of education, and current smoking status. However, it should be noted that the prevalence of cigarette smoking in those undergoing treatment for substance use disorders in the United States (76.3%) (Guydish et al., 2011) is markedly higher than in the general population (14% in 2019) (Cornelius et al., 2020). Because age is a known driver of changes in ERP topography, it was included as a covariate on all between-group statistical analysis of neurophysiological data. In addition, despite our large sample size of 103 participants for behavioral data and 96 participants for EEG, this 2 × 2 study design may have been underpowered to detect small effects. Further, since all HIV patients were receiving ART, it was not possible to tease apart the effect of HIV infection and the neurotoxic or neuroprotective effect of HIV treatment.

5. Conclusions

This study provides evidence of persistent inhibitory control deficits in CD+/HIV+ individuals despite current cocaine abstinence. These results indicate that abstinence-related neural, behavioral, and neurocognitive recovery trajectories observed in previous studies may be attenuated in CD individuals who are also HIV+. This suggests that further or more targeted treatment interventions are needed to facilitate positive health outcomes in this vulnerable population.

• Cocaine dependence (CD) is a common comorbidity of HIV infection

• Combined effects of CD and HIV on neural function were assessed with task-based EEG

• Additive effects of HIV and CD were observed on task performance

• Interactive effects on EEG amplitude were observed over fronto-central scalp regions

Abbreviations Table:

CD

Cocaine Dependence

CR

Correct Rejection

EEG

Electroencephalography

ERP

Event-Related Potential

FA

False Alarm

HIV

Human Immunodeficiency Virus

RI

Response Inhibition