The majority of older PWH are currently between the ages of 50 to 65, with a much smaller percentage over the age of 65. However, aging trends in the HIV population are predicted to continue . Additionally, age-associated physical comorbidities appear 5-10 years earlier in PWH , and, there is evidence of premature brain aging . Due to the neurotoxic effects of HIV and ART, as well as medical comorbidities and possible accelerated brain aging, PWH also may have less brain reserve to compensate for accumulating neurodegenerative pathology. Therefore, cognitive deficits indicative of aMCI could appear earlier in PWH compared to HIV-negative peers. Taken together, examining PWH in mid-life is advantageous as it could identify those with early signs of aMCI when interventions may be particularly efficacious. After excluding participants that did not meet inclusion/exclusion criteria as detailed below, the study included 92 PWH between the ages of 45 to 68 years old. All participants underwent at least one structural MRI scan between 2008 and 2010, comprehensive neuropsychological, neuromedical, and neuropsychiatric evaluation, as well as a blood draw. Most participants completed at least one follow-up neuropsychological, neuromedical, and neuropsychiatric study visit occurring in 6-month intervals. Participants were drawn from five participating sites: Johns Hopkins University, Mt. Sinai School of Medicine, University of California San Diego, University of Texas Medical Branch, and University of Washington. All CHARTER study procedures were approved by local Institutional Review Boards, and all participants provided written informed consent. UC San Diego IRB approval was sought for the current study, and it was determined by the IRB that this study was exempt. The CHARTER study aimed to recruit PWH to reflect the geographic and sociodemographic diversity of PWH around university-affiliated treatment centers in the U.S.; thus, CHARTER inclusion criteria were minimal and did not exclude participants with comorbidconditions that may impact cognitive function.
To determine the extent to which non-HIV related comorbidities have contributed to neurocognitive impairment,rolling tables developmental and medical histories of each participant were determined by Dr. R. K. Heaton and re-reviewed by an independent CHARTER clinician investigator. Participants with severe “confounding” comorbidities, as defined by Frascati criteria , were excluded from this project. Severe “confounding” comorbid conditions include comorbidities that could sufficiently explain neurocognitive deficits and thus preclude a HAND diagnosis. During clinician review, time course of comorbidities in relation to HIV and cognitive decline as well as the severity of comorbidities were considered when making comorbidity classification determination. Comorbid conditions that were reviewed and considered include history of neurodevelopmental disorders , cerebrovascular events , systemic medical comorbidities , non-HIV neurological conditions , and substance-related comorbidities . This comorbidity classification system has been shown to have excellent inter-rater reliability . The decision to exclude confounding comorbidities was further supported by a recent CHARTER paper showing that those with severe “confounding” comorbidities had worse brain integrity, but those with moderate comorbidities had fairly equivalent brain abnormalities as those with mild comorbidities . Additionally, CHARTER recruited a wide range of ages. To study the effect of aging with HIV, the age range for the current study was restricted to participants that were aged 45 or older at the time of the MRI scan. Additionally, one participant was excluded from the study given that their T1 structural MRI scan did not yield usable data .Tests of memory in the CHARTER study included the Hopkins Verbal Learning Test – Revised and the Brief Visuospatial Memory Test-Revised .
The HVLT-R and the BVMT-R include three learning trials, a longdelay free recall trial in which participants are asked to recall the stimuli previously presented, and a recognition trial in which participants are presented both target stimuli and non-target stimuli and asked if stimuli were presented in the learning trials. The delayed recall raw score is the total number of words correctly recalled during the long-delay free-recall trial. A recognition discrimination raw score was calculated by subtracting false positives from the total number of true positives. Note, this score is reflective of recognition discriminability, but this will be referred to simply as “recognition” throughout the text. Both the HVLT-R and BVMT-R have six alternate forms to attempt to correct for practice effects. Raw recognition scores were converted to Z-scores that account for demographic variables using normative data from the HNRP . Given that practice effect correction was not available for recognition and participants had a varying number of previous administrations, number of prior neuropsychological evaluations was included as a covariate in statistical analyses examining recognition. Raw delayed recall scores were converted to T-scores that account for demographic variables and practice effects using normative data from the HNRP. HVLT-R and BVMT-R recognition Z-scores were averaged to create a recognition composite. HVLT-R and BVMT-R delayed recall T-scores were averaged to create a delayed recall composite. Test-retest reliability estimates of the and HVLT-R recognition ranges from r = 0.27 – 0.40 and delayed recall ranges from r = 0.36 – 0.39. HVLT-R recognition and delayed recall show adequate convergent validity with other tests of verbal memory . The BVMT-R recognition and delayed recall trial have been shown to have adequate convergent validity with other tests of visual memory . Recognition and delayed recall were initially examined continuously rather than dichotomously splitting participants into impaired versus unimpaired groups. Examining recognition and delayed recall continuously is advantageous because it increases variability and more subtle differences observed in mid-life may not be captured by diagnostic cut-points.
However, when examining linear regression analyses from aim 1, the recognition analyses did not meet all assumptions for linear regression . Therefore, recognition was dichotomized into an impaired recognition group and an unimpaired recognition group for all analyses. Processing speed and psychomotor T-scores were used to examine processing speed and psychomotor performance . Raw scores from individual tests were converted to T-scores that adjust for the effect of age, sex, education, race/ethnicity, and practice effects using center-specific normative data. The T-scores from all tests in the domain are then averaged to obtain a domain T-score . The Wide Range Achievement Test-III , which has been shown to be a measure of premorbid verbal IQ in PWH , was reported to characterize the sample. Participants completed a standardized CHARTER neuromedical evaluation at each study timepoint. HIV serostatus was determined by enzyme-linked immunosorbent assay with a confirmatory Western Blot. The following HIV disease characteristics were collected from most participants at each visit: 1) current CD4 count measured via flow cytometry; 2) nadir CD4 measured via a combination of self-report and medical records; 2) CDC HIV staging; 3) HIV RNA in plasma measured by ultra-sensitive PCR ; 4) estimated duration of HIV disease collected via self-report; and 5) current ART regimen. Comorbid medical conditions , diabetes, hypertension, hyperlipidemia) were determined by self-report or taking medication for the condition. Comorbid psychiatric and substance use conditions were determined with the Composite International Diagnostic Interview , which is consistent with the DSM-IV. Additional details on the standardized CHARTER neuromedical assessment can be found in Heaton et al. . Additionally, CHARTER participants also have APOE genotype data for additional information). APOE genotype was dichotomized into APOE e4+ and APOE e4- . FreeSurfer version 7.1.1 was used to obtain cortical thickness and subcortical volume measures for several regions of interest , with a similar approach as earlier CHARTER work . After FreeSurfer processing, all T1 scans were visually inspected; in addition to the one participant excluded from all analyses as described above, one participant’s hippocampi were very overestimated, and therefore their hippocampal data were excluded from analyses. Neocortical thickness regions of interest included medial temporal lobe structures , prefrontal , and primary motor cortical areas. Specific structures were analyzed separately. Left and right volumes or cortical thicknesses for these regions of interest were averaged. In post hoc analyses,cannabis grow supplies if there were significant findings for the average region of interest then the left and right regions were examined separately to examine laterality. The differences in scanner from site to site was corrected for by regressing scanner from the data, given that differences between scanners have been well-documented in prior CHARTER work .
Differences in head size was accounted for by including estimated total intracranial vault volume as a covariate in volumetric data. Mean cortical thickness was included as a covariate in cortical thickness analyses. Additionally, age was included as a covariate to adjust for the normal differences of age on the brain. Five inflammation biomarkers were examined in this study. All inflammatory biomarkers have been found to be elevated in the context of HIV and aMCI . Plasma for biomarker assays was collected via routine venipuncture and EDTA vacuum tubes from all participants. All plasma biomarkers were measured using commercially available, multiplex, bead-based immunoassays according to manufacturer protocols; CRP was plated on a separate immunoassay given that it required a different dilution than other plasma biomarkers. Biomarker precision was ensured by assaying specimens in duplicate and repeating measurements with coefficients of variation greater than 20% or outliers that were more than 4standard deviations from the mean. Additionally, 10% of all assays were repeated to ensure batch consistency. The concentrations of these biomarkers typically have skewed distributions; therefore, the data were log-transformed prior to statistical analysis. Logistic regression was used for dichotomous recognition analyses . Multi-variable linear regression was used for continuous outcomes in aims 1b, 1c, and part of 1d. Primary predictors were tested separately. Age and imaging covariate were included as covariates in every model. The number of prior neuropsychological evaluations was included as a covariate in recognition models. Additional covariates , comorbidities, HIV disease characteristics, APOE status were selected by evaluating the bivariate relationships between potential covariates and outcomes. If a potential covariate was significantly associated with an outcome at p<0.10 it was then entered as a covariate in the model. Given the number of possible additional covariates, these additional covariates were only retained in the full model if the covariate remained associated with the outcome at p<0.10. Power analysis was conducted using GPower . These analyses were powered to detect medium effect sizes , with a two-tailed a = 0.05, and up to 5 covariates. Current CDC guidelines recommend immediately initiating ART and maintaining an undetectable viral load . Despite the fact that only 80% of PWH are engaged in care and 57% of PWH in the United States are virally undetectable , there is a trend towards examining PWH who are virally suppressed and on ART particularly in studies examining biological processes such as inflammation and neuroimaging . Therefore, post hoc analyses examining delayed recall, processing speed, psychomotor skills excluding participants that were not ideally treated for HIV disease , had a detectable viral load were excluded. Additionally, given the significant effects of methamphetamine on the brain , participants who had a current methamphetamine use disorder were also excluded in post hoc analyses . Dichotomous recognition models were not re-examined given that, with these exclusions, only 7 participants were impaired on the recognition composite. This aim utilized multi-level modeling to examine recognition and delayed recall across follow-up visits. Outcomes were examined separately. The “lme4 version 1.1-30” R package was used to conduct mixed-effects regressions . Mixed-effects logistic regression models were used to examine dichotomous recognition as the outcome. Models examining continuous delayed recall used linear mixed effects models. Analyses included a random intercept and a random effect for years since baseline . A cross-level interaction was used to test if baseline medial temporal lobe structure is associated with longitudinal recognition impairment or decline in delayed recall. Between-persons covariates included: age at baseline, imaging covariate, and covariates identified in aim 1. Power analysis was conducted using RMASS2 , and observed attrition was accounted for in these estimates. These analyses were found to be powered to detect small-to-medium effect sizes , with a two-tailed a = 0.05. Multi-level modeling was selected because it uses all available data and gives heavier weight to participants with more waves of data; thus, this methodology can account for participants that may have missed a follow-up visit and samples that have a differing number of follow-up assessments.