HIV-1 RNA concentration and cognitive performance
in a cohort of HIV-positive people
"Among these HIV-1-sero-positive individuals with mainly minor HIV-1-associated cognitive deficits and not receiving HAART, no association between CSF or blood concentration of HIV-1 RNA and cognitive performance could be found"....."The failure to identify a relationship between viral load and severity of cognitive impairment is consistent with the notion that the neurotoxic effect of HIV-1 is caused by inflammatory mediators and a neuroinflammatory cascade initiated by the viral infection but then acting independently of CSF viral concentration"
11 July 2007 - Volume 21 - Issue 11 - p 1415-1422
Vitiello, Benedetto; Goodkin, Karl; Ashtana, Deshratn; Shapshak, Paul; Atkinson, J Hampton; Heseltine, Peter NR; Eaton, Elaine; Heaton, Robert; Lyman, William D
From the aNational Institute of Mental Health, Bethesda, Maryland, USA
bDepartment of Psychiatry and Behavioral Sciences, University of Miami Miller School of Medicine, Miami, Florida, USA
cDepartment of Psychiatry, University of California San Diego, San Diego, California, USA
dDepartments of Medicine and Psychology, University of Southern California, Los Angeles, California, USA
eChildren's Hospital of Michigan, Detroit, Michigan, USA.
*P.N.R.H. is currently Medical Director of Global Businesses Beckman Coulter, Inc., Brea, California, USA.
Objective: To determine whether higher viral concentrations in the cerebrospinal fluid (CSF) and/or peripheral blood were associated with greater severity of cognitive impairment in HIV-1-seropositive subjects with cognitive-motor impairment.
Methods: Cognitive performance measurements and viral load were obtained from HIV-1-seropositive individuals with cognitive-motor impairment entering a clinical trial before the introduction of highly active antiretroviral therapy (HAART). CSF viral load (UltraSensitive Roche HIV-1 Monitor test with detection limit of 50 copies/ml) was available from 179 patients, and peripheral (plasma or serum) viral load from 111 patients. Of these patients, 62% met the 1993 Centers for Disease Control (CDC) criteria for AIDS, and 19% had clinically significant cognitive impairment (i.e., global deficit score ≥ 0.5). Possible associations between viral load and cognitive scores were examined with general linear regression models with and without adjustment for age, education, study site, antiretroviral use, CD4 cell count, and CDC stage.
Results: The mean CSF viral load was 2.83 log10/ml ± 0.94 (SD) (undetectable in 19.5%). Mean peripheral viral load was 4.11 log10/ml ± 0.90 (SD). No statistically significant associations emerged between either CSF or peripheral viral load and the global deficit score, or any of the seven cognitive domain deficit scores.
Conclusions: Among these HIV-1-sero-positive individuals with mainly minor HIV-1-associated cognitive deficits and not receiving HAART, no association between CSF or blood concentration of HIV-1 RNA and cognitive performance could be found. These results suggest that the severity of HIV-1-associated cognitive impairment is not directly related to concurrent viral concentration in the CSF or the peripheral blood.
Human immunodeficiency virus type 1 (HIV-1) enters the brain early in the course of the infection and can cause neuro-cognitive impairment, ranging from mild deficits to frank dementia [1-3]. In particular, deficits in speed of information processing, visuospatial ability, abstraction and flexibility of thinking, working memory, learning and retention, and/or motor performance have been documented [4,5]. Even minor cognitive/motor deficits can have a negative impact on functioning and survival [6-8].
The concentration of HIV-1 RNA (viral load) in the cerebrospinal fluid (CSF) was found to be correlated with brain tissue viral load, thus supporting the validity of CSF viral load as a marker of brain infection . The relationship between concentration of HIV-1 RNA in the cerebrospinal fluid (CSF) and cognitive impairment has been the object of previous reports with differing outcomes as to whether an association may exist [4,9-11].
Both brain and peripheral sources contribute to the HIV-1 concentration in the CSF . In AIDS, patients with severe cognitive impairment were found to have higher CSF viral load than those who were cognitively intact or had only minor neurological signs [9,10]. In HIV-1-positive patients without AIDS, no association was reported between CSF viral load and cognitive impairment . The relationship between CSF viral concentration and milder cognitive impairment among HIV-1 positive patients does, however, deserve further examination in other samples as gradually increasing HIV-1 concentrations in the CSF would be expected to be accompanied by increasingly severe degrees of cognitive impairment.
Using baseline neurocognitive assessments and CSF specimens collected as part of a publicly funded clinical trial of D-ala-peptide T-amide (DAPTA) in HIV-1 seropositive patients with cognitive impairment , we examined the possible association between CSF viral load and cognitive performance. The DAPTA clinical trial was conducted during the years 1991-1994, before the introduction of highly active antiretroviral therapy (HAART). Based on existing literature [9,10], we hypothesized that higher concentrations of HIV-1 in the CSF would be associated with greater impairment in neurocognitive performance.
We found no association between degree of cognitive impairment and concentration of HIV RNA in the CSF or periphery in this sample of HIV-1-positive individuals who had been referred for cognitive impairment and found to have deficits on screening tests of cognitive performance. These assessments were obtained prior to the introduction of HAART, which decreases viral load and can improve neurocognitive functioning [17-20].
Neither global scores of performance nor cognitive domain scores were significantly correlated with CSF viral load, regardless of the presence of AIDS. Furthermore, no association between viral load and cognitive performance was found in either the subgroup with a CD4 cell count below 200 cells/µl or the subgroup with CD4 cell count equal or above 200 cells/µl.
The failure to identify a relationship between viral load and severity of cognitive impairment is consistent with the notion that the neurotoxic effect of HIV-1 is caused by inflammatory mediators and a neuroinflammatory cascade initiated by the viral infection but then acting independently of CSF viral concentration [21,22].
Recently, HIV-1 viral load has been found to be only a weak predictor of disease progression at the individual level . The wide inter-subject variability in the reaction to the viral infection can contribute to the inability to find a correlation between viral load and clinical manifestations, including cognitive impairment as well.
The finding of a single significant correlation between CSF (but not peripheral) viral load and scores on the number of perseverative responses on the Wisconsin Card Sorting Test may deserve further attention. Although it emerges from exploratory analyses, the finding may suggest that selected specific tests of executive functioning can be more sensitive indexes of brain HIV-I infection than global scores of cognitive performance. This result, however, could also reflect spurious significance due to multiple testing.
Patients with CD4 cell count below 200 cells/µl had higher blood, but not CSF, viral load, thus indicating that HIV-1-induced immunological dysfunction is linked to greater viral concentrations in the periphery but not necessarily accompanied by increased CSF viral load.
These results must be considered in the light of the limitations of this study and the characteristics of the participating subjects. These analyses were conducted on a sample of convenience from a clinical trial conducted prior to the HAART era. The participants in this trial had HIV-1-associated cognitive difficulties and documented impairment in cognitive performance that was not apparently explained by pre-infection conditions, non-HIV illnesses or substance abuse . In fact, the aim of the clinical trial was to test a potential neuroprotective treatment for HIV-1-infected subjects with mild cognitive difficulties. Thus, a main limitation of this analysis was the rather restricted range of cognitive impairment, with most of the participants having only mild cognitive impairment. For this reason, these results do not exclude the possibility that a correlation between CSF viral load and cognitive performance may exist among more severely impaired patients. In addition, this study did not address the issue of whether HIV-1-associated cognitive deficits correlate with the presence and concentration of inflammatory mediators in the CSF.
In conclusion, no associations between measures of cognitive impairment and level of viral load in the CSF emerged from extensive analyses of this database from a large sample of HIV-1-seropositive patients. These results suggest that the severity of HIV-1-associated cognitive impairment is not proportionally related to the quantity of virus in the CSF or the periphery.
As summarized in Table 1, the sample (N = 179) consisted of predominantly white, high-school-educated men in their late 30s, with AIDS, and mean CD4 cell count of 219.9 ± 196.3 (SD). CSF viral load measurements were not available for 35 of the 214 HIV-1 seropositive subjects participating in the DAPTA clinical trial. These 35 subjects did not significantly differ from the 179 subjects with viral load measurements as to sex, age, education, race/ethnicity, Centers for Disease Control and Prevention (CDC) clinical diseases stage, CD4 cell count, history of drug abuse, severity of cognitive impairment, use of antiretrovirals or duration of their use. Likewise, no statistically significant differences for the foregoing measures were detected between subjects with and without peripheral viral load.
There was no difference among the three clinical trial sites in baseline CSF viral load (P = 0.78), or between UCSD and UM in baseline serum viral load (USC collected plasma). The within-subject correlation between CSF and serum viral load was Spearman r = 0.38 (N = 55, P = 0.004) and between CSF and plasma viral load was r = -0.04 (N = 55, P = 0.78). The subgroup of subjects with CD4 cell count below 200 cells/µl had higher peripheral viral load [N = 60, 4.40 ± 0.63 (SD) log10 copies/ml] than the group with CD4 cell count equal or above 200 cells/µl [N = 51, 3.75 ± 0.94 log10 copies/ml, t = 4.09, degrees of freedom (df) = 109, P < 0.001]. No significant difference in CSF viral load was, however, found between the group with CD4 cell count below 200 cells/µl (N = 91, 2.7 ± 1.00 94 log10 copies/ml) and the group with CD4 cell count equal or above 200 cells/µl (N = 88, 2.9 ± 0.87 log10 copies/ml, t = 1.56, df = 177, P = 0.12).
Viral load vs. cognitive performance
The cognitive test scores and deficit scores of the study sample are provided in Tables 2 and 3. The results of the GLM testing for associations between viral load and neuro-cognitive deficit scores are summarized in Table 4. No statistically significant association was found between viral load, either in the CSF or the periphery, and global cognitive deficit score, or between viral load and any of the seven cognitive domain deficit scores (Fig. 1). The result was unchanged after controlling for possible effects of age, education, CD4 cell count, antiretroviral use, or CDC HI-1 infection stage. Similar analyses using the global cognitive z score as a measure of cognitive performance  did not show any significant association with CSF (P = 0.43) or peripheral (P = 0.73) viral load.
Exploratory GLM analyses were similarly conducted searching for a possible association between CSF or peripheral viral load and each of the 23 test scores that constituted the cognitive battery. The only statistically significant association was between higher CSF viral load and greater Wisconsin Card Sorting Test-number of perseverative responses (beta = 3.49, SE = 1.11, P < 0.01).
Of the 179 subjects with CSF viral load measurements, 91 (51%) had a CD4 cell count below 200 cells/µl). When the entire sample was dichotomized based on CD4 cell count below 200 cells/µl vs. equal or above 200 cells/µl, no significant correlation was found between CSF viral load and global or cognitive domain scores within each CD4 cell count subgroup.
Of the 179 subjects with CSF viral load measurements, 34 (19%) had a global cognitive deficit score of 0.5 or greater, indicating clinically significant impairment, whereas most (N = 145 or 81%) had a global deficit score below 0.5. CSF viral load was not different between these two groups (P = 0.40). Of the 34 subjects with deficit score ≥ 0.5, 18% (N = 6) had undetectable CSF viral load (i.e., < 50 copies/ml), as compared with 20% (N = 29) of the 145 less impaired subjects.
The analyses reported herein were based on data and specimens collected as part of the baseline assessments for a clinical trial of DAPTA in HIV-1-associated cognitive-motor impairment. The methods and results of this trial have already been published [13,14]. Briefly, this was a publicly funded, multisite, randomized, placebo-controlled trial to evaluate the potential neuro-cognitive protective effects of DAPTA in HIV-1 seropositive patients. The trial was conducted at the University of Southern California (USC), principal investigator (PI): P. N. R. Heseltine, MD; University of Miami (UM), PI: Karl Goodkin, MD, PhD; and University of California at San Diego (UCSD), PI: J. Hampton Atkinson, MD. The study was approved by the Institutional Review Board at each site. Study participants were enrolled between 1 March 1991 and 30 June 1992 at USC (N = 99), and between 1 April 1993 and 30 March 1994 at UM (N = 60) and UCSD (N = 56).
Participants were HIV-1-seropositive patients referred for complaints of impaired neuro-cognitive functioning. Positive HIV-1 serostatus was verified using an enzyme-linked immunosorbent assay (Abbott Co., Chicago, Illinois, USA). Positive samples were repeated, and doubly positive samples were confirmed by protein immunoblot (Western blot). To be eligible for the study, patients must have had impaired performance on a nine-test screening neurocognitive battery, which included the American version of the Nelson Adult Reading Test (number of errors by years of education), Trail-making Test-B (time to complete), Stroop Test (time for C-W interference), Paced Auditory Serial-Addition Task (number correct, series 1), Sternberg Memory Scanning Test (mean reaction time for correct trials), Benton Visual Retention Test (number of correct answers), Repeatable Episodic Memory Test (total correct, trials 1-3), and Grooved Pegboard (time with nondominant hand). Cognitive impairment was defined as a score below the corresponding population-based norms by at least 1.5 SD on at least two tests or below 2.5 SD on one test .
To participate, patients were also required not to have taken any antiretroviral medication for at least 4 weeks prior to study entry (33% of the cases were in this category), or to have had stable antiretroviral treatment for at least 12 weeks before study entry (67% of the cases). Antiretrovirals included zidovudine, didanosine, and zalcitabine (Table 1). No patient had received HAART.
Reasons for exclusion from the study were: presence of active AIDS-defining opportunistic infections; Kaposi's sarcoma or other malignant neoplasm requiring treatment; need for more than two blood transfusions per month; current alcohol or substance dependence or abuse during the previous 3 months, current or recent (within the past 6 months) DSM-III-R axis one psychiatric disorder; history of psychotic disorder or bipolar mania; history of mental retardation or learning disability; or treatment with psychoactive agents within 4 weeks or within 8 weeks for long-acting agents (e.g., fluoxetine).
Assessments of neurocognitive performance
Before starting treatment in the DAPTA trial, participants were tested on a comprehensive battery of cognitive tests yielding a total of 23 scores and grouped a priori into the following seven cognitive domains: verbal fluency, visuospatial ability, abstract thinking, speed of information processing, working memory, learning and retention, and motor performance (Table 2) . These cognitive domains are consistent with recent efforts to conceptualize and measure neuropsychological performance in HIV and non-HIV disease [5,15].
For each test score, a deficit score was computed using z or T scores from available population norms [13,16]. A z (or corresponding T) score equal or greater than -1.0 corresponded to a deficit score of 0, a z score between -1.1 and -1.5 to a deficit score of 1, a z score between -1.6 and -2.0 to a deficit score of 2, a z score between -2.1 and -2.5 to a deficit score of 3, a z score between -2.6 and -3.0 to a deficit score of 4, and a z score below -3.0 to a deficit score of 5. An average ('global') deficit score of 0.5 or greater was taken as indicative of clinically significant impairment .
The mean deficit score from the 23-test battery constituted the global deficit score. Likewise, mean deficit scores were computed for each of the seven cognitive domains. Two scores from California Verbal Learning Test (long delay and percentage of recall consistency), which were included in the global deficit score, were not included in the domain score because only a single representative measure from each test was used. The global mean z score, computed by averaging all the individual test z scores , was also analyzed for further exploratory purposes.
Cerebrospinal fluid and peripheral blood samples
Cerebrospinal fluid and blood were collected before starting treatment in the DAPTA clinical trial. Plasma was stored at USC and serum at UM and UCSD. The samples were sent to a central repository at the Los Angeles Veteran Administration hospital for storage at -70°C. Samples of CSF from 193 participants in the DAPTA trial were available at the central repository. Of these, 179 could be linked to baseline neuro-cognitive assessments. For 111 participants, peripheral blood samples (plasma for 55 subjects from USC, and serum for 56 subjects from UM or UCSD) were also available. These samples were re-coded in order to mask subject identity, site, and time of collection, and were sent, together with blinded repeats and HIV-1-seronegative control samples, to the University of Miami for assays of HIV-1 concentration.
HIV-1 concentration assay
CSF and peripheral viral concentrations were assayed at the University of Miami Miller School of Medicine under a National Institute of Mental Health professional contract to Dr K. Goodkin. Plasma HIV-1 RNA copy number was measured using the UltraSensitive Roche HIV-1 MONITOR Test, version 1.5 (Roche Diagnostics, Branchburg, New Jersey, USA), which was used to quantify HIV-1 RNA copy number in the range of 50 to 75 000 copies/ml. An advantage of version 1.5 is greater sensitivity to non-B clades, which are becoming more common in the USA. For the results in which the upper limit of this assay was reached, the Standard AMPLICOR HIV-1 MONITOR test, version 1.5 (Roche Diagnostics) was used to quantify HIV-1 RNA copy number (range: 400-750 000 HIV-1 RNA copies/ml). In the standard assay, the additional step of centrifugation at 23 600 g for 1 h to concentrate the virus, which is needed in case of the Ultra Sensitive assay, is not required. These tests were performed under licensure approvals obtained from the manufacturer. Both assays employ a competitive reverse transcriptase (RT) PCR methodology and were performed according to the manufacturer's instructions. Samples were processed using a four-step procedure required by the training method provided by Roche Diagnostics: sample processing (HIV-1 concentration, manual RNA extraction, amplification, and combination of RNA and master mix solution), master mix preparation (RT-PCR reagent preparation), amplification (thermal cycling and reaction termination), and detection (plate hybridization, plate incubation, plate wash, and detection). Internal standards were used for serum and CSF.
Five CSF samples from HIV-1 seronegative controls and 18 repeat CSF samples from HIV-1-seropositive individuals were interspersed. All samples were blindly assayed and the identification code was revealed only after viral load measurements were completed. Measurements of HIV-1 RNA concentration were transformed to the logarithm base 10 (log10 or 'log') per ml for data analysis. Samples with undetectable viral load were assigned a log10 = 1.7 (corresponding to the threshold for detection of 50 copies/ml). The five HIV-1-seronegative controls all had viral load copy numbers below 50 HIV-1 RNA copies/ml (i.e., undetectable). The results of the 18 CSF specimens were within 0.5 log10 from each other, that is, within the acceptable limit of variation of this assay.
A possible association between viral load and cognitive performance was tested through the application of generalized linear models (GLM). The primary analysis tested for the presence of an association between global cognitive deficit score and CSF viral load. Secondary associations between CSF viral load and each of the seven cognitive domain deficit scores also were tested. The same analytic models were applied to the peripheral viral load measurements. In order to maximize sample size and statistical power, both the plasma (from USC) and serum (from UM and UCSD) viral measurements were first included in these analyses, but subsequent analyses were controlled for possible site effects.
Analyses were repeated while controlling for age, education, study site, history of drug abuse, antiretroviral treatment, previous use of DAPTA, and CD4 cell count, and CDC stage of HIV-1 infection (including AIDS status). In addition, the models were applied to the global cognitive z score, which was the primary outcome measure used in the DAPTA trial . Zero-order correlations between variables were measured with Pearson's or Spearman's correlation coefficient, as appropriate. Finally, for further exploratory purposes, the GLM was also applied to each of the 23 individual scores.
Owing to the multiple comparisons performed, these results are subject to increased risk of a type I error. In an attempt to control, at least in part, for spurious significance findings, alpha was set at ≤ 0.05 for the primary test with the global deficit score, but at ≤ 0.01 for all the other tests.
Statistical analyses were carried out by Constella Health Sciences under contract with the National Institute of Mental Health.