In this cross-sectional analysis from the CHARTER cohort, poorer ARV penetration, as estimated by lower CPE ranks, was associated with detectable CSF VL, whether expressed continuously (log10
c/mL) or categorically (≤50 c/mL or <50 c/mL). This relationship was independent of the systemic potency of the regimen (as estimated by plasma VLs), ARV adherence, current CD4 count, total number of ARV drugs, and type and duration of current regimen. Because CSF VL suppression is associated with improvement among individuals who have HIV-associated neurocognitive impairment,12
the CPE scoring system may be clinically relevant, particularly in impaired individuals. Thus, this study’s findings provide further evidence that penetration of ARV drugs into the CNS is important for CSF HIV viral suppression and demonstrates successful use of a new metric designed to consider multiple factors related to ARV effectiveness in the CNS.
Univariate analyses demonstrated that CPE ranks explained 12% of the variance in CSF VLs. This modest correlation may reflect the cross-sectional design of this study, which cannot fully account for time-dependent variables, such as time to virologic suppression. Other potential contributors are the highly positively skewed distribution of CSF VLs in CHARTER (83% undetectable), which is similar to other cohorts with a high prevalence of ARV use, and the importance of other predictors, such as plasma VL, in determining CSF VL. Because of the skewed distribution, the predictive ability of the CPE rank may be more accurately demonstrated by the categorical analysis, which indicated that a CPE rank lower than the median (1.5) was associated with nearly double (odds ratio = 1.88) the odds of having detectable CSF VL.
Because ARV therapy does not immediately suppress VL to undetectable levels, detectable CSF VLs might simply be due to insufficient time receiving therapy. In previous clinical trials,13
3 months of combination ARV therapy has been shown to be sufficient to suppress plasma VL levels to undetectable in most individuals. To address this potential confound, treatment duration (categorized as at least 3 months vs less than 3 months) was included in a multivariate model predicting detectable CSF VL. This model demonstrated that lower CPE ranks remained a significant predictor independent of the effects of treatment duration.
Numerous studies have assessed the relationships between penetration of ARV drugs into the CNS and either neurocognitive response or CSF VL (). These studies used varying definitions of CNS penetration and diverse outcome measures, making direct comparisons difficult. For example, to assess penetration, some studies focused on individual drugs (eg, zidovudine or indinavir), others counted the number of penetrating ARV drugs, and others directly measured ARV concentrations in CSF. Clinical outcome measures in these studies were even more diverse and included CSF VL, clinical staging of impairment (eg, Memorial Sloan-Kettering AIDS Dementia Complex), neuropsychological testing (either brief or comprehensive batteries), magnetic resonance spectroscopy, or evoked potentials. Even when studies were prospectively accrued, they were uncontrolled or had CNS response as only a secondary objective. Not surprisingly, these varying methods yielded inconsistent results, which have failed to resolve the controversy about the importance of CNS penetration. Despite these caveats, larger studies (ie, those with ostensibly greater power) and those that were prospective (vs retrospective) or controlled (included an internal comparison) were more likely to identify that penetration was associated with improved CNS outcomes.
Studies Comparing Effects of CNS Penetrating/Nonpenetrating ARV Drugs on CSF Viral Load
In our recently published article,12
we examined the CNS effectiveness of ARV regimens that contained CNS penetrating drugs. We found that subjects who took greater numbers of CNS penetrating drugs showed significantly greater reduction in CSF VL. Subjects who attained viral suppression in CSF demonstrated greater neurocognitive improvement than those who did not. The current analysis confirms the prior finding that CNS targeted therapy is important for reducing CSF viral load while using a more detailed approach to describing penetration. This analysis improves on the previous analysis because the study sample is larger (467 in the current analysis vs 31 in the prior analysis) and more generally representative of the HIV-infected clinic population and because the ARV regimens analyzed are more representative of contemporary prescribing practices.
The CPE rank is intended to be a practical approach to estimate the CNS effectiveness of ARV drug regimens taken by HIV patients in the United States. An inherent limitation of the CPE rank approach is the paucity of available information on ARV penetration into the CNS. Because ARV drug concentrations and virologic suppression cannot be directly measured in brain tissue, surrogate markers are used instead, such as chemical characteristics and CSF drug levels. However, the availability of data on ARV drug levels in CSF varies, significantly limiting the ability to make direct comparisons between different drugs. To deal with these limitations in available data, we incorporated an additional class of information into the CPE rank: the degree to which a drug, in previous clinical studies, independently reduced CSF VL or improved neurocognitive impairment, ie, its effectiveness in the CNS. Although the validity of suppression of HIV levels in CSF as a surrogate for treatment of the brain is unknown, considerable evidence supports it.
This initial analysis focused on the relationship between CPE ranks and CSF VL, but future analyses from the CHARTER cohort will examine the relationship between CPE ranks and neurocognitive performance. A more powerful method to assess this relationship would be a clinical trial that randomizes subjects to receive regimens with either low or high penetration. Observational analyses can be confounded by a variety of difficult-to-control factors, such as the severity, duration, and reversibility of neurocognitive impairment and the potential influence of the impairment itself on the type of regimens prescribed (eg, the impaired individuals might be more likely to receive more convenient, but less well-penetrating regimens). Another limitation of this study should be noted. Many of the study subjects were recruited from HIV clinics at North American tertiary care centers, which may limit the generalizability of the findings to other populations.
In this observational study, ARV regimens were selected by the subjects and their care providers. Thus, the low CPE ranks (median 1.5) probably reflect current clinical practice. These low values indicate that the penetration of many clinical regimens can be considerably improved, which could, in turn, have favorable neurologic effects. For example, a 1-unit increase in the CPE rank was associated with a 2.5-fold increase in the odds of an undetectable CSF VL. If the penetration estimate of an ARV regimen could be improved by 2 units (eg, from 1.0 to 3.0, which is clinically feasible), the odds of reaching an undetectable CSF VL would improve by 6-fold. Thus, our data suggest that the CNS effectiveness of ARV regimens could be substantially improved for a significant proportion of those receiving ARV therapy. Previous studies have suggested that improving the treatment of CNS HIV infection can benefit neurological outcomes and reduce overall disability due to neurocognitive impairment in HIV infection12