Does efavirenz replacement improve neurological function in treated HIV infection?

Objectives The contribution of specific antiretroviral drugs to cognitive function in HIV‐infected people remains poorly understood. Efavirenz (EFV) may plausibly cause cognitive impairment. The objective of this study was therefore to determine whether chronic EFV therapy is a modifier of neurocognitive and neurometabolic function in the setting of suppressive highly active antiretroviral therapy. Methods We performed an open‐label phase IV controlled trial. Adult subjects who were stable on suppressive EFV therapy for at least 6 months were switched to ritonavir‐boosted lopinavir (LPV/r) with no change in the nucleoside reverse transcriptase inhibitor (NRTI) backbone. The following parameters were assessed before and 10 weeks after therapy switch: cognitive function (by CogState® computerized battery); brain metabolites (by proton magnetic resonance spectroscopy); brain activity [by attentional processing task‐based functional magnetic resonance imaging]; and sleep quantity and quality [by sleep diary, Pittsburgh Sleep Quality Index (PSQI) and Epworth Sleepiness Scale]. Results Sixteen subjects completed the study. Despite most subjects (81%) self‐reporting memory problems at baseline, cognitive function, brain metabolites, and brain activity showed no change at 10 weeks after switch. Sleep quality improved on switch off EFV [mean PSQI (standard deviation): EFV, 8.5 (6.5); LPV/r, 5.8 (5.5); mean difference −0.4; 95% confidence interval −6.0 to −0.7]. Conclusions This is the first study to assess the effects of chronic EFV therapy on neurological function in a controlled setting. We conclude that EFV withdrawal is unlikely to result in significant modification of neurocognitive function in otherwise stable HIV‐infected people.


Introduction
Mild cognitive impairment remains common in the highly active antiretroviral therapy (HAART) era, but the reasons for this remain incompletely understood [1]. In particular, the role of specific antiretroviral drugs in mediating cognitive function has received relatively little attention [2]. While it is essential to treat viral replication in the central This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. nervous system (CNS), it is plausible that some antiretroviral drugs also have adverse effects on neuronal function. Efavirenz (EFV) remains one of the most commonly used antiretroviral drugs. It is well recognized that CNS side effects (especially sleep disturbance) are common in the first 6 weeks of EFV therapy and sometimes lead to discontinuation, but generally subside with continued therapy [3,4]. Some recent observational studies have suggested that EFV may additionally be associated with increased rates of cognitive impairment, a notion that is supported by some neurotoxicity data from animal and in vitro studies [5][6][7][8][9][10]. This hypothesis, however, remains controversial and has not been confirmed in a controlled trial.
The aim of this study was therefore to determine whether chronic EFV therapy is a modifier of neurocognitive function in the setting of suppressive HAART.

Participants
We performed a self-controlled, open-label phase IV pilot study. Participants were adult patients (aged 18-65 years) with HIV-1 infection receiving suppressive HAART with an EFV-containing regimen. All observations were performed at a single study site. Subjects were required to have a documented HIV-1 RNA viral load (VL) measurement of < 50 HIV-1 RNA copies/ml within the 4 months preceding study entry and no VL exceeding 200 copies/ ml within 12 months. Subjects were on HAART for at least 12 months and on EFV for at least 6 months. This constraint was to ensure that only chronic effects of EFV were captured. Subjects were ineligible if they had current self-reported recreational drug use or weekly alcohol consumption exceeding 35 units.
All subjects gave written informed consent for participation in the study. The study was approved by the local research ethics committee (12/NE/0071), Newcastle-upon-Tyne Hospitals NHS Foundation Trust R&D approvals committee (5946), and the UK regulatory authority Medicines & Healthcare products Regulatory Agency (MHRA) and registered with the EudraCT trials database (2011-005581-37).

Interventions and measurements
Ritonavir-boosted lopinavir (LPV/r; Kaletra â , AbbVie Inc., North Chicago, IL, USA; twice daily dosing) was used as a comparator drug to assess the effects of EFV removal. All subjects switched from EFV to LPV/r. Participants had study observations performed at baseline and 10 weeks after switch. An additional safety monitoring visit was performed 4 weeks after therapy switch.
A computerized cognitive testing battery (CogState â , CogState Ltd., Melbourne, Australia) was performed comprising six tests [listed with abbreviated test name; cognitive domain(s) measured]: Detection (DET; psychomotor function/speed of processing), Identification (IDN; visual attention/vigilance), One card learning (OCL; visual learning and memory), One back (ONB; attention/working memory), Continuous paired associate learning (CPAL; visual learning and memory), and Groton maze learning (GML; executive function/spatial problem solving). Subjects performed a practice run of each test to minimize training effects. Baseline cognitive impairment was defined with respect to CogState â normative data (excluding CPAL as sufficient quality data are not currently available for this test) [11].
Proton magnetic resonance spectroscopy ( 1 H-MRS) was performed using PRESS (point resolved spectroscopy) volume selection (TR/TE (repetition time / echo time) = 3 s/ 37 ms) on a 3T magnet (Achieva; Philips Medical Systems, Amsterdam, Netherlands) with~8 cm 3 sampled voxels in frontal white matter (FWM), frontal grey matter (FC) and basal ganglia (BG). Metabolite levels of N-acetylaspartate (NAA), choline (Cho) and creatine (Cre) were quantified using the QUEST method in JMRUI software [12]. Concentrations were expressed relative to Cre. NAA is a widely studied measure of neuronal integrity and is frequently abnormal in HIV infection. Cho is frequently altered in inflammation.
Task-based (attentional processing) functional magnetic resonance imaging (fMRI) assessed response to the Stroop paradigm using incongruent visual stimuli [13]. The Stroop test has been shown to be sensitive to psychotropic drug effects including stimulants and sedatives [14][15][16], as well as to EFV use in one previous study [5]. fMRI was performed using a gradient-echo EPI (echo planar imaging) sequence (TR/TE = 1.7 s/30 ms; 27 9 4.0 mm thick slices; 0.5 mm inter-slice gap; 3 mm inplane resolution) with a block design Stroop paradigm. Data were collected in a single fMRI acquisition during which 360 image volumes were acquired. The stimulus consisted of a rest period, 18 interleaved blocks of incongruent and neutral stimuli and a final rest period. Subject responses were monitored using a response box. Analysis of the fMRI scans was performed using standard processing in SPM8 (http://www.fil.ion.ucl.ac.uk/spm/software/ spm8/). Analysis was performed for each subject contrasting the incongruent response against the neutral response. A group analysis was then conducted using within-subject paired comparison of responses at each time-point.
Sleep was assessed by means of a 2-week sleep diary, completion of the Epworth Sleepiness Scale (ESS) (where ESS > 10 indicates excessive daytime sleepiness) and completion of the Pittsburgh Sleep Quality Index (PSQI) (where PSQI > 5 indicates some level of sleep disturbance and > 10 indicates significant sleep disturbance) [17,18].

Statistics
As this was a pilot study, a pragmatic sample size was used [19]. Descriptive statistics and confidence intervals (CIs) are reported. No data replacement or imputation of missing data took place.

Results
Seventeen participants were recruited, of whom 16 (three female and 13 male) completed the study (one subject was withdrawn at the screening visit). The median age was 50.4 years [interquartile range (IQR) 43.2À55.7 years]. The median duration of diagnosed HIV infection was 6.7 years (IQR: 4.2À10.0 years). Subjects had been receiving EFV for a median of 4.5 years (IQR: 4.0À5.8 years). At baseline, 14 subjects were receiving a backbone of tenofovir/ emtricitabine, and the remaining two subjects were receiving tenofovir/raltegravir and abacavir/lamivudine, respectively. The median CD4 lymphocyte count at entry was 660 cells/lL (IQR: 536À737 cells/lL), and the median nadir CD4 count was 237 cells/lL (IQR: 37À299 cells/lL; n = 12 subjects with available data for nadir). The median baseline EFV plasma concentration was 2455 ng/mL (range 818-7197 ng/mL; n = 15 subjects with analysable EFV levels).
Summary statistics for cognitive testing are presented in Table 1.
fMRI data for 14 subjects were analysable. First, the effects of the 'Stroop' task on brain activity were assessed at the baseline visit to ensure that the fMRI protocol employed elicited a measureable and expected effect in our subject group. The analysis assessed activity during the incongruent stimulus, contrasted against that during the neutral stimulus. Three clusters of activation during the task were identified, in an anatomical distribution corresponding to that expected for this test paradigm: BA45/BA8/BA46, BA20/BA7/BA40 and BA7 (BA, Brodmann's area; Table S1). Secondly, a comparison was made of brain activation in response to the incongruent stimulus between the two study time-points. No change in brain activation was observed following switch.
There was no overall change in the ESS for patients on a switch from EFV to LPV/r [mean ESS (SD): EFV, 9.8   Of seven patients with a PSQI > 10 (indicating poor sleep quality) at baseline, four patients had a PSQI of < 10 at follow-up. No subjects had an insomnia phenotype (low ESS < 2 with high PSQI > 10) at either timepoint. As executive function is most typically affected by sleep disturbance, we assessed the data for any relationship between baseline sleep quality and GML score. Mean GML score did not differ in subjects with an abnormal PSQI score (> 10). Furthermore, there was no association between baseline EFV plasma concentration and baseline GML, ESS or PSQI scores or change in these scores after cessation of EFV. HIV-1 RNA VL was checked at the safety monitoring visit 4 weeks after treatment switch. At this point, three subjects (19%) had VL > 50 copies/mL, with values of 69, 107 and 115 copies/mL, respectively. By the follow-up visit, the first two subjects had re-suppressed to < 50 copies/mL, whereas the third subject had a VL of 337 copies/mL. Self-reported adherence was assessed at each of three follow-up visits, and at each visit four subjects reported at least one missed dose per week. At baseline, three subjects (19%) reported diarrhoea (of any severity). Fourteen subjects reported diarrhoea at any time following the switch from EFV to LPV/r (88%) (11 of the 13 not affected at baseline; 85%). There were no serious adverse events (SAEs) recorded.

Discussion
We have explored the hypothesis that chronic EFV therapy is associated with adverse effects on CNS function.
To our knowledge, this is the first study to assess this question in a controlled manner by investigating the effects of a switch away from EFV on neurocognitive performance, cerebral metabolites, brain activity, and sleep. Despite the fact that a very high proportion of subjects self-reported CNS symptoms at baseline (including 81% reporting memory problems), we observed no objective changes in neurocognitive performance across multiple domains, in cerebral metabolites, or in brain activity. There was, however, an increase in self-reported sleep quality (PSQI).
This was a pilot study, and so our ability to detect small effect sizes will have been limited by the sample size. Notwithstanding this, for almost all parameters studied, the effect sizes observed were very close to zero, and there were no consistent trends in directions of change across related domains (for example in cerebral metabolites or cognitive performance). We therefore conclude that we are unlikely to have missed any clinically important effects.
Given that previous studies have suggested adverse effects of EFV on neurocognitive function, what are the possible explanations for our findings? As previous studies associating cognitive impairment and EFV were observational and cross-sectional in nature, they were liable to being confounded by differences between groups exposed and unexposed to EFV [5]. In the one randomized controlled trial (RCT) that has compared EFV and a protease inhibitor, the study was performed in the setting of treatment-na€ ıve subjects starting HAART [20]. Thus, the predominant effect was changes in cerebral metabolites associated with suppression of viral replication and immune reconstitution, and differences observed between drugs may not have been wholly independent of their effects on those parameters. Although most of our subjects did report at least some CNS symptoms at baseline, those patients who experience severe early neuropsychiatric side effects of EFV are likely to switch away from the drug within the first few months and therefore would not be represented in our study. Furthermore, previous EFV switch studies have documented improvement in CNS symptomatology, but have not included objective measures of CNS performance [21,22]. In theory, a switch off EFV might have prevented a decline in cognitive function that would otherwise have occurred, but this seems very unlikely given the short study period. Alternatively, if LPV/r were slightly less effective at suppressing viral replication in the CNS, this could potentially offset any benefits from EFV withdrawal. We did note some VL blips and reported missed doses after treatment switch. Again, it seems unlikely that this would affect CNS function within the timeframe studied, but this should be explored in future studies by correlation with cerebrospinal fluid (CSF) HIV VL. A final possibility is that long-term EFV therapy could cause a mild but permanent deficit in CNS function (e.g. as a result of neuronal loss) which would not be reversible following therapy switch. We did observe around half of subjects with mild cognitive impairment at baseline, but this study cannot determine whether this was attributable to EFV or other factors. This question could only be explored by a very long follow-up RCT.
In summary, our study did not find support for an acutely reversible effect of EFV on either CNS metabolites or cognitive function in otherwise stable patients. These findings should be confirmed by further randomized studies.