The utility of cerebrospinal fluid (CSF) markers of viral replication (HIV viral load [VL]) and immune activation (β-2 microglobulin [β2m]) in the diagnosis of AIDS dementia complex (ADC) has been demonstrated mainly in the era before the availability of highly active antiretroviral therapy (HAART).1,2 In patients who are on HAART, however, the relation between CSF markers and neurologic status seems altered.2,3 ADC seems to be inactive or partially active as a consequence of HAART, and nonelevated CSF markers could delineate inactive ADC in patients receiving HAART. Therefore, the aim of our study was to investigate whether low CSF VL and β2m levels were associated with inactive ADC using neuropsychologic performance as an indicator of ADC stability.
Data were derived from the abacavir (ABC) ADC trial,3 which was originally undertaken to evaluate the benefit of this drug on the neurologic status of ADC patients. The entry criteria were: (1) stage 1 or 2 (mild to moderate) ADC4 supported by neuropsychologic impairment at least 1.5 SDs below normal in at least 2 neuropsychologic domains from the chosen test battery, (2) absence of confounding neurologic disease, and (3) stability on current antiretroviral therapy (ART) for a minimum of 8 weeks before study entry. For this subanalysis of the utility of CSF markers in identifying inactive ADC, patients had to have completed the baseline and retest neuropsychologic assessment after 12 weeks and to have baseline CSF HIV VL as well as β2m levels available. Seventy-eight patients fulfilled these criteria. Details of demographic, clinical treatment, and laboratory data are displayed in Table 1.
Neuropsychologic assessment comprised the evaluation of 5 cognitive domains: verbal memory (Rey Auditory Verbal Learning Test), fine motor coordination (Grooved Pegboard), complex attention/psychomotor speed (Trail Making part A and part B, Symbol Digit Modalities Test, Cal Cap choice, and sequential Reaction Time), and language (Verbal Fluency).
Cerebrospinal Fluid Markers: Viral Load and β-2 Microglobulin
CSF VL was measured at baseline and was extracted, amplified, hybridized, and detected using NASBA HIV-1 QT technology (Organon Teknika), with a detection limit of 100 copies/mL. β2m concentrations were measured using commercially available kits, and results were obtained for all patients except 2.
CSF VL and β2m were log10 transformed. Neuropsychologic scores were transformed into standard z scores using the Multicenter AIDS Cohort Study (MACS) reference data corrected for age and education.5 A summary neuropsychologic z (NPZ) score was computed as the mean of the 9 individual NPZ scores at baseline (NPZ baseline) and at week 12 (NPZ week 12). The relation between CSF markers and the summary NPZ score was examined with the Pearson correlation. Neuropsychologic performance change was explored with the Reliable Change Index (RCI) developed by Jacobson and Truax.6 The RCI can be interpreted as a standardized score; therefore, an RCI >1.96 occurs in <5% of cases. Using 3 different cutoffs, we defined 6 different groups. Forty-one patients had a baseline CSF VL <100 copies/mL (group CSF VL <100), and 37 had a baseline CSF VL >100 copies/mL (group CSF VL >100 log10 mean [SD] = 3.3 [0.83] copies/mL). Forty-three patients had a baseline CSF β2m <2.2 mg/mL (group CSF β2m <2.2), and 33 had a baseline CSF β2m >2.2 mg/mL (group CSF β2m >2.2 log10 mean [SD] = 3.3 [0.95] mg/mL). Twenty-nine patients had CSF markers below both cutoffs (group low CSF markers), and 47 had a CSF VL or CSF β2m above cutoffs (group high CSF markers). The proportion of significant neuropsychologic performance change (with patients stratified as unchanged or changed in a first analysis and then stratified as improved, stable, or deteriorated in a second analysis) according to the RCI was calculated and compared among the groups with χ2 and Fisher exact tests, respectively, in 3 analyses: CSF VL <100 versus CSF VL >100 copies/mL, CSF β2m <2.2 versus CSF β2m >2.2 mg/mL, and CSF low markers versus CSF high markers. Positive and negative values were calculated to evaluate the diagnostic efficiency of the CSF cutoff markers on inactive ADC (stable neuropsychologic performance) and active ADC (deteriorated neuropsychologic performance). Linear regression analyses were separately conducted with log10 CSF VL and log10 CSF β2m as the predictors and the RCI as the outcome variable. Because level of education differed on some subgroups, multiple linear regression analyses with forced entry were conducted with education as a second predictor. The error rate required on all analyses for significance was set at 2-tailed P < 0.05.
When considering the groups according to CSF VL, β2m, and combined CSF marker global cutoffs (Table 2), we found that demographics and ABC use frequency did not differ significantly, except for education. For the groups with CSF VL <100 and >100 copies/mL (χ2 = 15.27, df = 2, P < 0.000) and the groups with low and high combined CSF markers (χ2 = 9.52, df = 2, P < 0.009), the frequency of college and greater than college education varied inversely.
Log10 CSF VL and log10 β2m were significantly correlated (r = 0.41, P < 0.000). Log10 CSF VL and summary NPZ scores at baseline (r = −0.29, P < 0.009) and week 12 (r = −0.39, P < 0.000) were significantly associated. Log10 β2m and summary NPZ scores at baseline (r = −0.19, P < 0.09) and week 12 (r = 0.20, P < 0.08) showed only a trend. Group neuropsychologic performance changes are presented in Table 3.
The cutoff CSF VL <100 copies/mL had a sensitivity of 53% and specificity of 100% for inactive ADC (stable neuropsychologic change). The positive predictive value was 100%, and the negative predictive value was 6%. The cutoff CSF β2m <2.2 mg/L had a sensitivity of 55% and specificity of 50% for inactive ADC. The positive predictive value was 97%, and the negative predictive value was 3%. The low CSF combined markers had a sensitivity of 37% and specificity of 100%. The positive predictive value was 100%, and the negative predictive value was 4.5%.
A linear regression model showed that log10 CSF VL was significantly associated with the NP performance change as measured by the RCI [R 2 = 0.064, F(1,76) 5.2, P < 0.025; Fig. 1]. Education did not make a significant contribution to the model (P < 0.07) when entered as a second predictor. A linear regression model showed that a log10 β2m was not associated with the neuropsychologic performance change as measured by the RCI [R 2 = 0.003, F(1,74) 0.23, P < 0.62]. When level of education was entered as a second predictor, the model remained nonsignificant.
We investigated whether normal levels of CSF markers related to productive HIV infection and immunologic activation would define inactive ADC. Using frequency and predictive value analyses, we found that an undetectable CSF VL (<100 copies/mL) and normal β2m (<2.2 mg/mL) do not indicate inactive disease independently or in combination. Linear regression analyses demonstrated that only CSF VL was predictive of neuropsychologic performance change but that the magnitude of the predictive model was weak.
Despite the fact that, to our knowledge, this is the largest ADC ART trial reported to date, the power of our study is significantly hampered by the fact that few subjects exhibited a change in their neuropsychologic performance. Given the low number of such events, we may have been unable to detect a difference in marker levels that was present. When we considered a more liberal cutoff to define change (RCI = 1.64, P < 0.10) or considered RCI as a continuous variable, there was still no trend toward a change in the CSF.
There are several explanations for the results depending on whether the CSF is still a window on the events in the brain. If the CSF still does reflect brain events, the period of observation was too short, the markers were too insensitive, or both. The observation period may well have been too short; however, to date, an observation period of 5 months would have been considered adequate. At entry, the patients had to have been stable on their HAART regimen for at least 2 months so that by week 12 of the study, patients would have been on HAART for at least 5 months. Because the addition of ABC to the regimen of half the patients did not lead to any significant improvement, it can be considered that the whole group was on stable HAART for 5 months. In the pre-HAART era, significant elevation of CSF markers increased the risk of developing ADC at 6 months.1 The degree of elevation in the raised CSF marker groups was comparable to the pre-HAART data.7 Indeed, CSF β2m varied between 2.47 and 4.13 mg/L (group elevated β2m >2.2 mg/mL). Results from the pre-HAART era1,2 demonstrated that for a CSF β2m varying between 2.6 and 5 mg/L, 71% of patients progressed to ADC stage 1 and 21% progressed to ADC stage 2. It therefore seems likely that HAART has “dampened” viral and immunologic activity in the CSF in ways that are not captured by CSF HIV VL or β2m.8,9
It is also possible that the CSF markers, specifically the CSF HIV VL lower limit of 100 copies/mL, were too insensitive. Studies with a CSF VL detection limit down to 80 copies/mL have not demonstrated a better sensitivity to cognitive change in ADC, however.8,9 The alternate explanation, namely, that the CSF no longer reflects brain disease, seems unlikely. This would potentially mean that HAART is working more effectively in the brain than in the CSF. This is improbable, given the observations that the degree of CSF penetration of an antiretroviral drug seems to correlate with its efficacy in ADC treatment.1 It thus seems likely that the period of observation and the sensitivity of the markers explain the results.
The explanation as to why some patients continued to improve despite having normal CSF markers at baseline is likely related to “reparative” processes at work. Indeed, neurocognitive improvement has been shown to occur after 3 years.10
In conclusion, further studies are necessary to determine better CSF and central nervous system (CNS) markers of active and inactive ADC. The clinical value of CSF VL and β2m should be considered with caution in HAART-treated ADC patients, at least in the 6 months after HAART initiation.
They also thank Matthew Law at the National Centre in HIV Epidemiology and Clinical Research, University of New South Wales, Sydney, Australia, for his statistical support.
1. Brew BJ. AIDS dementia complex. In: HIV Neurology
. Oxford: Oxford University Press; 2001:53-90.
2. McArthur JC, Haughey N, Gartner S, et al. Human immunodeficiency virus-associated dementia: an evolving disease. J Neurovirol
3. Brew BJ, Brown SJ, Catalan J, et al. Safety and efficacy of abacavir (ABC, 1952) in AIDS dementia complex (Study CNAB 3001) [abstract 561/32192]. Presented at: XII International AIDS Conference; 1998; Geneva.
4. Price RW, Brew BJ. The AIDS dementia complex. J Infect Dis
5. Selnes OA, Jacobson L, Machado AM. Normative data for a brief neuropsychological screening battery. Percept Mot Skills
6. Jacobson NS, Truax P. Clinical significance: a statistical approach to defining meaningful change in psychotherapy research. J Consult Clin Psychol
7. Brew BJ, Dunbar N, Pemberton L. Cerebrospinal fluid concentrations of β2-microglobulin and neopterin predict the development of AIDS dementia complex. J Infect Dis
8. McArthur JC, McDermott MP, McClernon D, et al. Attenuated CNS infection in advanced HIV/AIDS with combination antiretroviral therapy. Arch Neurol
9. Sevigny JJ, Albert SM, McDermott MP, et al. Evaluation of HIV RNA and markers of immune activation as predictors of HIV-associated dementia. Neurology
10. Tozzi V, Balestra P, Galgani S, et al. Changes in neurocognitive performance in a cohort of patients treated with HAART for 3 years. J Acquir Immune Defic Syndr