The study included 59 HIV-seropositive (HIV) and 69 HIV-seronegative healthy control (HC) participants. HIV serostatus was documented by ELISA and confirmed by Western blot test. Participants were recruited from Icahn School of Medicine at Mount Sinai in New York, NY, United States and The Miriam Hospital in Providence, RI, United States. PLWH were recruited from affiliated clinics and local community groups; HC were recruited from community (e.g., local ads, word of mouth). Inclusion criteria were: age between 21 and 70 years, right-handed, completed ≥8 years of education, native English speaker. All participants obtained a score of ≥24 points on the Mini-Mental State Exam (MMSE). To avoid confounds that could potentially impact performance on cognitive tests, we excluded participation on the basis of reported history of developmental or learning disability; uncorrected abnormal vision; major psychiatric illness (e.g., schizophrenia, bipolar disorder); neurological illness affecting the central nervous system (e.g., stroke, progressive multifocal leukoencephalopathy); and traumatic head injury with a loss of consciousness of >10 min. Substance use exclusion criteria included reported current alcohol dependence; use of heroin/opiates or any intravenous drug within the past 6 months; use of cocaine within the past month; and positive urine toxicology at the time of assessment (cocaine, opiates, methamphetamine, amphetamine, benzodiazepine, barbiturates, methadone, and oxycodone). Individuals with substance use history were excluded as described above to enhance internal validity. One investigator (USC) oversaw all procedures. The research was approved by the Institutional Review Boards at the Icahn School of Medicine at Mount Sinai and The Miriam Hospital. All participants gave their informed written consent and received $125 compensation for their time upon completion of the study.

All participants provided their alcohol and drug use histories, which were quantified using the Kreek-McHugh-Schluger-Kellogg scale (KMSK; Kellogg et al., 2003). There were three subscales characterizing lifetime consumption of alcohol (KMSK-A), cocaine (KMSK-C), and opiates (KMSK-O). The Wechsler Test of Adult Reading (WTAR) estimated premorbid intellectual function (Wechsler, 2001); scaled scores were derived using published normative data. WTAR data were unavailable for four participants (one HC, three HIV). Participants were assessed for HCV infection, defined as positive HCV antibody. HIV-disease characteristics collected from PLWH participants included HIV disease duration, nadir CD4 levels (i.e., the lowest ever CD4 T-cell count), and antiretroviral (ARV) use via self-report and verified against the medical record. Current CD4 levels and plasma HIV viral loads (HIVL) were obtained from medical records. HIVL was log 10 transformed to normalize the distribution. Nadir CD4 and current HIVL data were not available for one and three PLWH participants, respectively. All PLWH participants were prescribed ARV medications. See Table 1 for group characteristics.

Early life stress exposure was determined using a self-report measure known as the Early Life Stress Questionnaire (ELSQ; Cohen et al., 2006a). The ELSQ assesses the occurrence of 17 adverse childhood events (ACE; e.g., severe family conflict, physical abuse, emotional abuse, and neglect) prior to the age of 18 (Table 2). Participants who reported ≥3 adverse events were characterized as having high ELS and participants that reported fewer than three ACEs were characterized as low ELS, consistent with prior studies (Clark et al., 2012). Using these criteria, our HIV and HC groups were classified as follows: HC Low-ELS (N = 36), HC High-ELS (N = 33), HIV Low-ELS (N = 26), and HIV High-ELS (N = 33; Table 1).

The N-back is a computerized working memory task that was used to determine RT-IIV. This task is described in detail in a prior report (Clark et al., 2018). Briefly, the task entails a series of letters presented visually, one every 2 s. Several N-back conditions were presented (e.g., 0-back, 1-back, and 2-back) during functional magnetic resonance imaging (fMRI; data not examined in the current report). Reaction time data during the 1-back condition were examined based on prior observations of ELS-related (Clark et al., 2018) and HIV-related (Chang et al., 2001) effects among PLWH. During the 1-back, participants indicate (yes/no) whether the letter presented is the same as, or different from, the letter presented immediately prior (Figure 1). As described previously (Clark et al., 2018), the N-back was completed during an magnetic resonance imaging (MRI) session as part of a broader study of ELS-related effects among PLWH. A total of 64 1-back trials were administered. Participants who demonstrated accuracy levels below chance or responded to less than 80% of 1-back trials were not included, resulting in the exclusion of 16 participants (six HC Low-ELS, four HC High-ELS, two HIV Low-ELS, and four HIV High-ELS) from the sample described herein. Those excluded did not differ significantly with respect to age (p = 0.991), gender (p = 0.544), or estimated premorbid intelligence (p = 0.356) when compared to the sample described herein.

We calculated a nonparametric measure of signal detection, A′ (Snodgrass and Corwin, 1988), which accounts for both omission and commission errors, and thus characterizes performance during the 1-back. A′ ranges from 0 to 1, with scores ≥0.5 indicating performance that is at or above chance. For each participant, mean RT latency was calculated using data from all correct 1-back trials. To assess RT-IIV, we calculated the coefficient of variation (CoV) for each participant using the following formula: CoV = standard deviation across all correct 1-back trials/mean RT (Stuss et al., 2003). Therefore, CoV is measure of RT variability which controls for mean RT, where higher scores indicate greater variability.

We examined neuropsychiatric symptoms known to be both elevated among PLWH (Clark et al., 2017) and associated with RT-IIV performance (Clark et al., 2018). To examine neuropsychiatric symptoms, we created a neuropsychiatric symptom composite index score to quantify overall psychological difficulty. This index included: The Center for Epidemiological Studies-Depression Scale (CESD; Radloff, 1977), Perceived Stress Scale (PSS; Cohen et al., 1983), and Posttraumatic Stress Disorder Checklist—Civilian (PCLC; Weathers et al., 1993), which assessed current levels of depression, stress, and posttraumatic stress disorder (PTSD) symptoms, respectively. PCLC scores were not available for three PLWH participants. Consistent with prior studies (Clark et al., 2017, 2018), for each participant, z-scores were calculated for each measure based on the HC sample mean for that measure. Next, the three z-scores were averaged to generate a composite index score for each participant. Hence, the index signifies overall degree of neuropsychiatric difficulty reported across all cognitive domains assessed where higher index scores indicate greater global neuropsychiatric difficulty.

Group differences on demographic variables and neuropsychiatric index scores were assessed using one-way ANOVAs, independent-samples t-tests, chi-square, and Fisher’s exact tests. Our primary goals were 2-fold. First, we examined the independent and combined effects of HIV status and high-ELS exposure on RT-IIV using a two-way ANCOVA with factors of HIV status (HIV, HC) and ELS status (High, Low). Covariates were included to control for demographic variables that differed significantly between groups and also correlated significantly (as assessed by Pearson correlations) with RT-IIV (i.e., age). Partial eta-squared (η_p²) was used as an indicator of effect size, where values of 0.01, 0.06, and 0.14 indicate small, medium, and large effects, respectively (Cohen, 1988). Our second goal was to explore whether RT-IIV is sensitive to psychosocial (degree of ELS exposure) and biological (HIV-disease characteristics) risk factors known to impact brain and cognition. As such, we used linear regression to examine the relation between ELS exposure, as measured by number of ACEs reported, and RT-IIV accounting for covariates as defined above. Separate analyses were conducted for the HIV and HC groups to examine associations in each group independently. Likewise, in the HIV group, we used linear regression to examine associations between HIV-disease characteristics and RT-IIV for those disease characteristics that demonstrated significant univariate associations with RT-IIV (Pearson correlations, p < 0.050).

Table 1 includes demographic data for all groups, including group means and statistics. The four groups were well matched on several demographic variables including estimated premorbid intelligence and 1-back accuracy (A; ps > 0.050); however, groups differed significantly with respect to age, education, and lifetime drug use histories for alcohol (KMSK-A) and cocaine (KMSK-C; Table 1). We observed a significant correlation between age and RT-IIV in the HC Low-ELS group only [r(36) = −0.341, p = 0.042]. RT-IIV did not correlate significantly with education, alcohol use history, nor cocaine use history in any group (ps > 0.050). Accordingly, only age is included as a covariate in RT-IIV analyses. Regarding HIV-disease characteristics, the HIV Low-ELS and HIV High-ELS groups were well matched with respect to nadir and current CD4 levels, current HIVL, and length of HIV infection (ps > 0.050; Table 1).

Group mean neuropsychiatric composite index scores and statistics are reported in Table 1. Results from the one-way ANOVA revealed significant differences across groups [F(3,124) = 5.06, p = 0.002], with higher symptom levels observed in both High-ELS groups relative to the Low-ELS groups (Table 1) as indicated by Fisher’s LSD post-hoc tests (all ps ≤ 0.050). More specifically, the HC High-ELS group reported greater neuropsychiatric symptoms than the HC Low-ELS (p = 0.007) and HIV Low-ELS (p = 0.050) groups. Similarly, the HIV High-ELS group reported greater neuropsychiatric symptoms than the HC Low-ELS (p = 0.001) and HIV Low-ELS (p = 0.013) groups. Neuropsychiatric composite scores did not correlate significantly with RT-IIV in any group (ps > 0.050) and thus were not included as a covariate in RT-IIV analyses.

All groups exhibited a mean response rate of >97% during the 1-back (Table 1). Groups did not differ significantly in mean RT or 1-back accuracy (A′; Table 1). By contrast, results from the ANCOVA, examining RT-IIV, covarying for age (p < 0.001), revealed a significant HIV-ELS interaction [F(1,123) = 5.21, p = 0.024, η_p² = 0.04]. In this model, the main effects of HIV and ELS were non-significant (p > 0.050). Follow-up analyses revealed that the HIV-ELS interaction was driven by significant and trend-level elevations in the HIV High-ELS group relative to all other groups [HIV High-ELS versus: HC Low-ELS, t(67) = 1.95, p = 0.055; HC High-ELS, t(64) = 2.54, p = 0.013; and HIV Low-ELS, t(57) = 2.39, p = 0.020; Figure 2]. All additional group comparisons were non-significant (ps > 0.050).

We examined the association between degree of ELS exposure and RT-IIV using linear regressions conducted in the HIV and HC groups separately. In the HIV group, we observed a significant association between ELS exposure and RT-IIV [R² = 0.08; F(1,58) = 5.16, p = 0.027; β = 0.007, (t = 2.27, p = 0.027)]. By contrast, the association between ELS exposure and RT-IIV in the HC sample was non-significant (p = 0.854).

Among the HIV group, RT-IIV correlated significantly with nadir CD4 [r(58) = −0.26, p = 0.049] and HIVL [r(56) = 0.29, p = 0.031] as well as with HIV-disease duration at a trend level [r(59) = 0.24, p = 0.062]. Correlations with current CD4 levels were non-significant [r(59) = −0.17, p = 0.207]. Results from the linear regression revealed a trend-level association of RT-IIV to HIVL [β = 0.014, (t = 1.85, p = 0.070)] and a non-significant association with nadir CD4 (p = 0.124).