Immunological and Neurometabolite Changes associated with switch from Efavirenz to an Integrase Inhibitor

Abstract

Background:

The biological mechanisms by which efavirenz (EFV) causes central nervous system (CNS) effects are unclear. The objective of this pilot study was to elucidate the mechanisms underlying these CNS effects by correlating well-described neuropsychological (NP) changes to neurometabolites and immunologic markers with switch off EFV.

Setting:

Two single arm parallel switch studies among HIV-infected adults in Boston, USA from 2015-2017.

Methods:

20 asymptomatic HIV-infected adults on EFV-containing regimens were switched to an integrase strand transfer inhibitor-based regimen for 8 weeks. NP assessments were conducted pre- and post-switch and correlated with neurometabolite changes measured using magnetic resonance spectroscopy (MRS) and immunological markers. All pre- and post-EFV measures were evaluated using matched-paired analyses.

Results:

NP testing demonstrated improvement in the domains of mood, cognition and sleep off EFV. MRS revealed decreases in neurometabolite glutathione (p=0.03), a marker of oxidative stress following switch. Inhibitory neuronal activity as reflected by GABA levels increased (p=0.03) while excitatory neurotransmitters glutamine+glutamate (Glx) and aspartate decreased (p=0.04, 0.001). Switching off EFV was also associated with changes in inflammatory markers; plasma markers sCD14 (p=0.008) decreased while I-FABP and TNFRI levels increased (p=0.05, 0.03). Cellular markers CD4+ and CD8+ HLA+DR−/CD38+ subsets both increased (p=0.05, 0.02).

Conclusions:

Even asymptomatic participants showed improvements in NP parameters when switched off EFV. These improvements were associated with decreased CNS oxidative stress and excitatory neuronal activity. Changes in immune activation biomarkers suggested overall decreased inflammation. EFV may exert CNS effects through oxidative and inflammatory pathways, providing insight into possible mechanisms of EFV neurotoxicity.

Introduction

Despite being among the most frequently used antiretroviral agents worldwide, efavirenz (EFV) use has declined dramatically in recent years. This shift has been largely driven by the availability of potent integrase strand transfer inhibitors (INSTIs),^1,2 as well as accumulating evidence of EFV’s central nervous system (CNS) side effects³ estimated to occur in 40-60% of persons on EFV.^4-6 While still widely used globally due to its relative affordability, the mechanisms by which EFV induces these CNS effects, particularly with chronic use, remain poorly understood.

EFV use has been associated with psychotic symptoms (e.g. delusions, hallucinations, paranoia) and mood disorders (e.g. anxiety, depression, and rarely suicidality) early after start prompting regimen switches.⁶ Sleep disturbances (e.g. insomnia, vivid dreams) are particularly common and may be chronic. However, there is an increasing awareness that many patients on chronic EFV therapy may have subtler symptoms that can impair quality of life but be unrecognized as EFV-related. These include subclinical mood symptoms, as well as vague manifestations such as dizziness and irritability.^3,6,7 Additionally, cognitive impairment with deficits in concentration, attention and memory are typically insidious but have been implicated with EFV use more than with other antiretrovirals (ARVs).^7,8

While human data is lacking, attempts at ascertaining these mechanisms have been made using non-human and in-vitro models. Animal models have suggested EFV-related neurologic effects are mediated by an increase in pro-inflammatory cytokines, such as interleukin-1ß (IL-1) and tumor necrosis factor-α, (TNF-α), which have independently been associated with depressive symptoms.⁹ In-vitro studies have explored the role of mitochondrial function and neural bioenergetics; in EFV-exposed glial cells and neurons mitochondrial dysfunction resulted in greater oxidative stress exacerbated by neuroinflammation,¹⁰ while in astrocytes EFV caused alterations in the cellular oxidative balance with reduction in glutathione levels.¹¹ Although there is heterogeneity amongst these studies, together they suggest inflammatory pathways, as well as oxidative stress, play key roles in EFV-associated neurologic symptoms.

In this pilot study, our goal was to extend beyond the insights of prior non-human studies by using advanced neuroimaging and immunological markers to better investigate the biological mechanisms driving subclinical CNS effects of EFV. We enrolled participants without CNS signs or symptoms who were on long-term EFV and switched them to a guideline preferred INSTI-containing regimen, while correlating neuropsychological (NP) and cognitive symptoms with inflammatory/cell activating blood and CNS biomarkers. We hypothesized that despite tolerating long-term EFV treatment many participants would have demonstrable subclinical NP abnormalities corresponding with both systemic and CNS biological changes. The objective of this study is to understand the mechanisms by which EFV causes CNS symptoms, as well as provide a novel framework for using advanced imaging, systemic blood markers and NP testing to evaluate future ARVs for potential CNS effects.

Methods

Study Participants

HIV-infected individuals virally suppressed for ≥6 months on efavirenz/tenofovir disoproxil fumarate/emtricitabine (EFV/TDF/FTC) were enrolled. Eligible patients had to be free of known or reported CNS/NP symptoms to ensure they were without expectation of symptomatic improvement from a regimen switch that could bias perceived outcomes. Individuals who were co-infected with hepatitis B and/or hepatitis C, older than 65 years of age, non-English speaking/reading and lacked right-handedness were excluded, as were those with severe psychiatric (e.g. diagnosis of psychiatric disorder requiring medical treatment, history of suicidal ideation) or neurological disorders (e.g. epilepsy, structural brain lesions, neurodegenerative disorders), history of CNS infection, substance abuse/dependence or heavy alcohol use (i.e. >12 oz./week) within the last year, severe vascular disease, on medications that could affect cerebral flow or had any contraindications to magnetic resonance imaging (MRI).

Study Design

Participants were enrolled into one of two parallel studies that included a switch to INSTI-based regimens ( NCT01978743, NCT01929759) without change in the nucleoside reverse transcriptase backbone of TDF/FTC. The study protocol and all amendments were approved by the Partners Hospital Institutional Review Board. Written informed consent was obtained from each participant before study procedures were performed. A control group was not included, partly limited by high cost of repeated neuroimaging, and largely due to a specific focus on understanding the biological mechanisms underlying EFV-induced neurotoxicity using each participant as their own internal, longitudinal control. Participants were switched to either 1) raltegravir (RAL) plus TDF/FTC or 2) elvitegravir (EVG)/cobicistat (COBI) /TDF/FTC, with the following assessments performed at baseline and repeated 8 weeks post-switch; this duration was chosen based on pharmacokinetic data showing that EFV and its metabolites are cleared by then:¹²

1. Laboratory Parameters: A complete blood count, including T-cell subsets, HIV quantitative RNA viral load, fasting lipid profile, renal and hepatic function tests were obtained. EFV levels as well as those of its metabolites, 7-hydroxy-efavirenz (7-OH-EFV) and 8-hydroxy-efavirenz (8-OH-EFV) were measured using high-performance liquid chromatography fractionation. These plasma levels have been previously correlated with CNS toxicity.⁵

2. Neuroimaging: MRS was conducted in two brain locations implicated as important nodes in many different psychiatric and neurobiological pathways: the posterior cingulate gyrus (PCG)^13-15 and the anterior cingulate gyrus (ACG).^16,17 N-acetyl aspartate (NAA), choline (Cho), creatine (Cr), and glutathione (GSH) and glutamate+glutamine (Glx), and myoinositol (mI) which were measured using point resolved spectroscopy (PRESS) with an echo time (TE) of 30ms (repetition time (TR)=2 sec, 64 averages, 2×2×2 cm³). Glx is used due to the significant overlap between the glutamate and glutamine resonances such that despite the predominating glutamate signal, the contribution of glutamine cannot be ignored. Spectra were pre-processed using OpenMRSLab¹⁸ in order to frequency- and phase-correct spectra before quantification using LCModel.¹⁹ Aspartate (Asp) and gamma-amino butyric acid (GABA) were measured using 2D correlated spectroscopy (COSY) using starting TE of 30ms (TR=1.5 sec) and 64 increments of 0.8ms acquiring a vector size of 1024 points and acquisition time of 512ms to provide a spectral width of 2000 Hz in F1 and 1250 Hz in F2. GABA was quantified using FelixNMR as previously described.²⁰

3. Surveys: Affective symptoms, cognitive function and sleep quality were assessed by the Hamilton Depression Scale (HAMD), Spielberger State-Trait Anxiety Inventory (STAI), Frontal Systems Behavior Scale (FRSBE), Pittsburgh Sleep Quality Index (PSQI), Depression Anxiety Stress Scales-21 (DASS-21), Trail making Test Parts A and B (TMT), and Wechsler Adult Intelligence Scale Revised Digit Symbol Substitution Test (WAISR-DSST). This focused battery of well-validated NP tests was chosen to quantitate changes in the NP domains already previously associated with long-term EFV use; specifically, domains of executive function, anxiety, depression, and sleep quality. NP tests used in prior EFV studies, such as DASS-21,^21,22 PSQI,²³ as well as WAISR-DSST and TMT,^7,24 were combined with additional measures (e.g. HAMD,^25,26 STAI,²⁷ FRSBE²⁸) in order to detect and quantify with better resolution the subclinical changes in NP symptoms in participants, who were otherwise tolerating EFV without frank symptoms. Participants were also asked about ARV preference at the end of the study.

4. Immunologic markers: Cellular markers (CD4/CD8/CD38/HLA−DR/PD-1/TIGIT/TIM-3/Lag3) were measured via flow cytometry using fluorescently conjugated antibodies (BD biosciences), run on an LSR Fortessa (BD), and analyzed using Flowjo software v9.9.6 (Treestar). Soluble markers (sCD14, IP-10, sCD163, MCP-1, IL-6, intestinal fatty acid binding protein (I-FABP)) were measured via an enzyme-linked immunosorbent assay (ELISA (R&D Systems).

Statistical Analysis

Two-tailed paired t-test was used to compare pre- and post-EFV levels in MRS neuro-metabolites and changes in NP testing scores, specifically; timed TMT scores were log transformed prior to analysis. Wilcoxon matched-paired test was used to assess changes in inflammatory markers. We also used mediation analysis in an exploratory manner (given small sample size) to evaluate the potential impact of each inflammatory and neurological biomarker as a direct mediator of the relationship between EFV and specific NP outcomes. This method assesses the impact of particular biomarkers on a causal pathway. A mediator is a variable on the causal pathway between an exposure (i.e. EFV) and an outcome (i.e. change in NP measures) and explains at least part of the observed effect of the exposure on the outcome. In this study we estimated the degree to which each biomarker mediated the impact of EFV on an individual’s NP outcomes. The specific biomarkers used for mediation analysis were limited to those that were first identified to be significantly associated with EFV usage, using paired t-tests and Wilcoxon tests, as appropriate. For NP endpoints, we identified those that were significantly associated with EFV by generalized linear regression models fit with generalized estimating equations (GEE) to account for the repeated measures on each individual (See Supplementary Digital Content 1 for further details).

Results

A total of 20 HIV-infected participants were enrolled; 10 in each INSTI arm. There were no reported serious adverse events and all participants completed the study. Data from the two groups were combined to increase power due to lack of differences in baseline characteristics (Table 1), pre-switch serum levels of EFV, 7-OH EFV and 8-OH EFV, or post-switch levels (all undetectable after 8 weeks off EFV).

Table 1:

Baseline Characteristics

	Switch from EFV+FTC/TDF to:
	EVG/c/FTC/TDF (n=10)	RAL+FTC/TDF (n=10)	p-value	All participants (n=20)
No. of participants	10.0	10.0	-	20.0
Age (mean)	46.5 (9.0)	47.9 (9.9)	0.7	47.2 (9.2)
Male gender (%)	90	60	-	75
CD4 T cell (mean)	657 (187)	709 (280.5)	0.6	683 (234)
HIV-1 RNA <20 copies/ml (%)	80.0	80.0	-	80.0
On ART (%)	100	100	-	100
Creatinine (mean)	0.9 (0.2)	0.8 (0.2)	0.4	0.9 (0.2)
Cholesterol (mean)	178.8 (24.2)	200.9 (37.0)	0.1	190.0 (32.5)
% on statin	10	30	-	20
ALT (mean)	26.2 (15.2)	28.6 (15.1)	0.6	27.4 (14.8)
EFV plasma conc (mcg/ml)*	2.3 (0.9)	3.9 (2.3)	0.1	3.1 (1.9)
8-OH EFV conc	0.6 (0.4)	0.5 (0.4)	0.8	0.5 (0.4)
7-OH EFV conc	0 (0)	0.01 (0.01)	0.1	0.01 (0.01)
Post- EFV plasma conc (mcg/ml)	0.02 (0.05)	0 (0)	0.4	0.4 (0.7)
Post 8-OH EFV conc	0 (0)	0 (0)	0.3	0 (0)
Post 7-OH EFV conc	0 (0)	0.01 (0.01)	0.05	0 (0)

^*mean (stdv)

Neuropsychological measures improved following switch off EFV

Although none of the participants had overt signs of cognitive impairment or NP symptoms as assessed by referring clinicians or self-report, we detected longitudinal changes in sleep and NP tests after switching off EFV (Table 2). Significant improvements in several NP domains were measured, in particular fewer depressive symptoms on the HAMD (score difference −2.30, p=0.003) and decreased anxiety as evaluated by STAI (score difference −3.40, p<0.001). Switching off EFV also resulted in improved executive function and sleep quality as assessed by the WAISR-DSST (score difference +4.75, p<0.001) and PSQI (score difference −1.60, p<0.001), respectively; these findings echo those of a recent randomized controlled trial⁸ as well as other studies among asymptomatic patients.^7,29

Table 2:

Neuropsychological measures and changes

Description of neuropsychological measures				Neuropsychological changes with EFV switch
	Measures	Domains	Scale Range	Pre-Switch (SD)	Post- Switch (SD)	Change after EFV switch	P- value	Clinical interpretation
WAISR-DSST	Weschler Adult Intelligence Scale	Cognitive test to assess executive function	0-100 Lower score = more cognitive dysfunction	48.7 (11.8)	53.5 (12.0)	+4.75	<0.001	Improved executive function
Trail Making	Trail Making test Parts A and B	Tests speed of information processing, attention and task switch to detect cognitive impairment	Time (log transformed, sec) Slower speed = greater cognitive impairment	3.98	3.95	−0.036	0.19	Part A: faster information processing
				4.52	4.67	0.146	0.31	Part B: unchanged executive function
HAMD	Hamilton Rating Scale for Depression	Standard clinical measure of severity of depression	0-7 is normal, >20 = moderate-severe depression	5.1 (3.6)	2.8 (2.6)	−2.30	0.003	less depressive symptoms
DASS-21	Depression Anxiety Stress Scale	Self-reported symptoms over the past week for depression, anxiety and stress	0-63 Higher score = increased depressive symptoms	Depression				less reported depression
				7.0 (9.0)	3.4 (4.2)	−3.60	0.074
				Anxiety				less self-reported anxiety
				4.3 (5.3)	2.5 (2.2)	−1.80	0.067
				Stress				less self-reported stress level
				6.9 (4.6)	4.4 (4.2)	−2.50	0.013
FRSBE	Frontal Systems Behavioral Scale	Self-reported assessment for 3 frontal system: apathy, disinhibition and executive dysfunction	37-186 Higher score = greater behavioral impairment	78.4 (11.4)	71.4 (11.3)	−7.00	0.004	less self-reported behavioral symptoms
STAI	Spielberger state trait anxiety inventory	Assess specifically for presence and severity of anxiety	20-80 Higher score = more anxiety	29.1 (5.5)	25.7 (5.2)	−3.40	0.003	decreased anxiety
PSQI	Pittsburgh Sleep Quality Index	Subjective sleep quality to assess duration, sleep efficiency and disturbances	0-21 Higher score = greater sleep disturbances	5.1 (2.6)	3.5 (1.8)	−1.60	<0.001	improved sleep quality

Decreased oxidative stress and neuroexcitatory signals on MRS

Pre- and post-switch changes in CNS neuro-metabolite levels were measured by MRS in the ACG and PCG (Figure 1a). These regions were selected as they can achieve higher spectral data quality allowing for greater sensitivity to metabolites with lower concentrations. The PCG has been shown to be highly sensitive to early metabolic changes in HIV,^13,14,30 especially after ARV therapy¹⁵ and has been implicated in depression.^31,32 The ACG shows similar changes although is more closely associated with mood disorders, as well as being sensitive to neurocognitive changes.^16,17,33 Unlike another MRS study utilizing alternative areas,²³ our use of these regions facilitated detection of subtle neurometabolite changes in our HIV-infected asymptomatic participants.

Figure 1a:

Comparison of pre- and post-EFV switch neurometabolite levels measured by MRS in the anterior cingulate gyrus (ACG) or the posterior cingulate gyrus (PCG). Statistical comparison done by two-tailed paired t-test. A) GSH = glutathione, B) GABA/Cr = gamma-aminobutyric acid creatine ratio, C) Glx = glutamate + glutamine, D) Asp = aspartate.

A decrease in GSH, a metabolite involved in cellular oxidative stress, was observed (p = 0.028) in the ACG. Off EFV there was attenuation of the excitatory neurosignaling. Increases in the inhibitory neurotransmitter GABA (p = 0.026) were seen off EFV in the PCG, with corresponding decreases in excitatory neurotransmitters Glx (p=0.038) and Asp (p=0.001). We did not observe significant changes in markers associated with more severe neurodegenerative diseases, choline, mI and NAA, off EFV.

Reductions in systemic inflammation off EFV

Plasma sCD14, a marker of monocyte activation associated with neurocognitive impairment,^34-36decreased off EFV (Figure 1b; mean decrease 482.9 ng/ml; p=0.008). I-FABP, a marker of enterocyte damage and TNFRI, one of the major receptors for TNF-α involved in regulating inflammation³⁷, both increased (227.3 pg/ml, p=0.05 and 51.4 pg/ml, p=0.035 respectively). Increases in TNFRI level may downregulate TNF-α activity, as has been observed in other disease conditions.^38,39 With the switch to INSTI-containing regimens, increases in HLA−DR−/CD38+ were observed among both CD4+ and CD8+ T-cell subsets (p=0.05, 0.02, Table 3). By contrast, no significant changes in the degree of T-cell activation (HLA−DR+/CD38+) or exhaustion were observed after discontinuation of EFV (Table 3).

Figure 1b:

Changes in level of soluble inflammatory makers pre- and post-EFV switch measured by ELISA. sCD14, IP-10, sCD163, MCP-1, Il-6, I-FABP, LPS, TNFRI were measured. Only those with significant changes are presented: A) sCD14 with mean decrease of 482.9 ng/ml, B) I-FABP with meanS increase of 227.3 pg/ml, and C) TNFR1 with mean increase of 51.4 pg/ml.

Table 3:

Changes in cellular markers pre- and post-switch

	CD4+ T cells		CD8+ T cells
	mean difference*	paired t-test	mean difference*	paired t-test
HLA−DR−/CD38+	1.94	0.05	3.88	0.02
HLA−DR+/CD38+	−0.21	0.61	−0.14	0.89
Lag3+	−0.21	0.07	−0.25	0.03
PD-1+	0.65	0.25	−0.04	0.89
TIGIT+	0.53	0.26	0.57	0.62
TIM-3+	−0.26	0.66	0.22	0.79

^*Difference of post- and pre levels

Oxidative stress and inflammation may mediate EFV effects

Mediation analysis (See Supplementary Digital Content 2) was performed as an exploratory analysis to determine the contribution of biomarkers in driving the EFV-specific effects on particular NP measures and identified GSH as a possible mediator. Specifically, 18.7% (95% CI; −28.8 - 59.7%) of the EFV effect on cognition assessed by WAISR-DSST was mediated through GSH (See Supplementary Digital Content 3). Similarly, I-FABP was identified as possible mediator of the EFV effect on anxiety measures (per DASS and STAI). Combinations of I-FABP and GABA mediated 63.6% (95% CI; =1.07 – 17.4) of EFV’s effect on anxiety outcomes.

Laboratory and HIV parameters remained stable post-switch

Laboratory parameters overall did not change post-switch, supporting the well-known safety and tolerability profiles of INSTIs (Figure 2). Total cholesterol and LDL values decreased and creatinine declined slightly post-switch (median difference in both INSTI groups: serum creatinine 0.025 mg/dL, p<0.002; eGFR 7 ml/min/1.73 m³, p<0.002); the latter is often seen among patients switching to INSTIs.⁴⁰ Total bilirubin increased while aspartate aminotransferase and alkaline phosphatase decreased off EFV. HIV parameters including CD4+ T-cell counts, other T-cell subsets viral suppression remained stable, consistent with other switch studies to INSTI-based regimens, and further confirming adherence to study medications.²¹

Figure 2:

Clinical Laboratory changes off EFV. Shown are mean ratio of post-pre levels of each clinical laboratory value with 95% Confidence Interval (paired t-test). No changes were observed in HIV laboratory parameters.

Participant Preference

The survey employed at the end of the study to assess regimen preference showed no participant preferred the EFV-containing regimen; 80% preferred the INSTI-containing regimen while the remaining 20% had no preference.

Discussion

In this pilot study we identified potential biological mechanisms underlying the insidious NP effects of chronic EFV use, using neuroimaging techniques and measures of inflammation. We found that in neuroasymptomatic participants, there were detectable improvements in mood and cognition measured by NP testing when they were switched from long-term EFV to an INSTI-based regimen; observations all consistent with earlier clinical studies.^21,41 Moreover, through the simultaneous use of MRS imaging techniques and measurement of plasma biomarkers, we were able to isolate important inflammatory and oxidative pathways as possible mechanisms for the CNS effects of EFV. These findings fill an important knowledge gap with regard to the mechanisms of ARV-associated CNS effects which have pertinent diagnostic and therapeutic implications for the care of HIV-infected patients.

EFV-associated CNS effects are a major barrier to its use even though many symptoms wane with continued use or are tolerated. However, chronic subclinical symptoms have been reported in post-switch setting,⁸ as well as in cross-sectional comparative studies.^41,42 With these well-characterized symptoms ascribed to EFV use, our study focused on testing these major domains in asymptomatic patients; specifically, mood, cognition and sleep, through in-depth and focused NP testing. Our findings were as expected; improvement in these domains with EFV cessation, albeit of perhaps lower magnitude compared to earlier studies with symptomatic patients⁴³ and in line with a recently reported EFV randomized controlled trial that was also conducted among asymptomatic patients.⁸

Using MRS to characterize the CNS effects of ARVs amongst otherwise asymptomatic HIV-infected persons we were able to provide possible mechanistic insights for EFV CNS effects. We observed reductions in CNS oxidative stress, measured by decreased GSH levels off EFV. Higher GSH in the ACG has been previously associated with mild cognitive impairment,⁴⁴ and as GSH levels serve as a putative measure of CNS oxidative stress, the observed decline after switching off EFV implicates oxidative stress as a potential mechanism.

MRS identified several alterations in neurosignaling associated with EFV use. With mood improvement measured by NP testing off EFV we observed a concurrent rise in neuroinhibitory transmitter GABA. This finding corroborates earlier murine models which found EFV-induced NP symptoms were linked to decreased striatal levels of GABA.⁴⁵ This attenuated inhibitory neurosignaling has been similarly observed in patients with psychiatric disorders who demonstrated lower levels of GABA inhibitory signaling but with subsequent rise in GABA as NP symptoms remitted.⁴⁶ Similarly, we found decreases in Glx and Asp, excitatory neurotransmitters with important roles in mood and cognition, once EFV was removed. Dysregulated Glx levels have been linked with mood disorders in HIV-uninfected populations, specifically elevated in bipolar mania states and decreased in depressive disorders.⁴⁷ Elevated Asp levels in post-mortem neuroimaging in the PCG of patients with Alzheimer’s Disease,⁴⁸ implicates a role in neurocognition as well. Together, these findings suggest EFV-induced mood symptoms may result from an imbalance of excitatory and inhibitory neurosignaling.

Two neurometabolites, NAA and mI, typically associated with more advanced neurodegenerative disease did not change in our asymptomatic cohort following cessation of EFV. NAA levels have been correlated with cognitive impairment in patients with severe dementia, with HIV-associated as well as Alzheimer’s alone.^48,49 Given its prior association with neuronal cell death in severe neurological diseases,⁵⁰ the absence of significant change in NAA in our study suggests that milder neurocognitive symptoms associated with EFV are mediated through more reversible neuronal mechanisms rather than neuronal cell death. Similarly, we did not detect or expect changes in mI, a marker of glial activation that has been observed to be increased in individuals with advanced Alzheimer’s disease⁵¹ and AIDS dementia,⁴⁹ compared to their asymptomatic counterparts. The consistency of our MRS findings are highly compatible with those seen in other neuropsychological conditions in addition to being upheld by the clinical improvements measured by NP testing, and together, highlight the potential utility of MRS in studying HIV-associated neurocognitive symptoms and screening therapeutic agents for potential neurotoxicity.

Prior studies have linked soluble inflammatory markers to mood disturbances and neurocognitive symptoms in HIV-infected persons. We observed a decrease in sCD14 levels similar to findings from a study among women switching from EFV to an INSTI.⁵² In our study, this corresponded with improved monocyte activation profiles reflected by increases in CD4+ and CD8+ naïve T-cell populations. The clinical implications of the observed increase in TNFR1, which may downregulate TNF-α, however, remain unclear. In one study, rats exposed to EFV experienced depressive-like symptoms associated with an increase in TNF-α.⁹ Nevertheless, it should be noted that the responses of TNFRI to changes in TNF-α level longitudinally are not well understood.³⁷ In fact, a similar increase in TNF receptor levels was observed in another EFV-to-INSTI switch study, yet the initial increase became non-significant with additional follow up time at 48-weeks, suggesting the possibility of a transient effect immediately post-switch that wanes over time.⁵² Our observed increase in I-FABP after 8 weeks post-switch seems to suggest increased enterocyte turnover but has not been observed in prior studies.^52,53 On the other hand, we did not detect a corresponding increase in lipopolysaccharide-binding protein, an acute phase reactant which binds to bacterial lipopolysaccharide of intestinal bacteria, to suggest this finding was mediated through an increase in gut translocation. It is yet to be determined whether I-FABP is a useful biomarker in patients on long-term ART suppression as in our cohort; normal variations among treatment-experienced patients have not been characterized.⁵³

While used here in a novel application, the prohibitively high cost of MRS limited sample size. We were able to, nonetheless, combine data from two parallel INSTI switch studies to increase power and demonstrate that any changes in the measured parameters were due to class (INSTI) effects and not to one particular agent. Because the CNS effects of EFV has been well-described in many earlier cohorts,^8,29,41 we focused on characterizing the biological underpinnings of these phenomena through measuring intra-patient changes longitudinally rather than with a control group. Although follow-up time was short, we were assured by complete clearance of EFV and its metabolites by the time of post-switch assessment. Practice effect is a potential limitation with any study using repeated cognitive tasks in serial assessment, however, a recent randomized controlled study on EFV switch (without use of MRS) showed similar cognitive improvements, providing further support that the NP improvements we observed were likely true clinical change rather than practice effect.⁸ Furthermore, tests have varying sensitivity to practice effects and are often augmented with high frequency testing, not done in our study. While the influence of practice effects on WAISR-DSST and TMT, the most susceptible tests in our panel, cannot be excluded we only observed significant improvement in the former and not both as would be expected if practice effect was prominent. We purposely selected a focused panel of NP tests based on the recognized CNS effects of EFV use because the main objective was to correlate changes in these domains with biological markers. Since it was beyond the scope of the study to perform a comprehensive NP battery we may have been limited in detecting potential changes in unmeasured domains. Lastly, as in any study with self-report questionnaires there is the potential for expectation bias; however, our participants were completely asymptomatic when enrolled without expectation of any change post-switch. We used mediation analyses here in an exploratory manner to describe potential relationships between neuropsychological measures and biomarkers, however, its interpretability is limited given the small sample size. In addition, while ARV neurologic effects are of particular importance in patients with mental health and substance use comorbidities, our exclusion of these individuals limits the generalizability of this data to patients with these particular comorbidities, however was critical to isolate the NP and biomarker changes specific to EFV without the potential confounding influence of other conditions. Ultimately, larger scale studies will be needed to address these limitations, further validate and extend the findings from this work.

To our knowledge, this is the first study correlating EFV-associated NP symptoms with both blood and neurometabolite markers. We identified specific oxidative and inflammatory pathways as potential mechanisms by which EFV may exert CNS effects with the possibility of GSH, I-FABP, GABA being mediators of EFV-associated neurocognitive symptoms. Ultimately, these data may be applied to help clinicians tailor ARV regimens, particularly for patients with comorbid mental health and CNS disorders associated with increased inflammation and oxidative stress. These clinical concerns are also particularly pertinent now with aging of the HIV-infected population where avoiding ARV regimens that have even minor adverse CNS effects may be appropriate. Finally, this methodology can serve as an innovative model for the use of MRS imaging to correlate CNS effects when evaluating newer ARVs for potential neurotoxicity.