Absence of peripapillary retinal nerve-fiber–layer thinning in combined antiretroviral therapy-treated, well-sustained aviremic persons living with HIV

Purpose To compare peripapillary retinal nerve-fiber–layer (pRNFL) thickness, total retina macular volume, and ganglion-cell-layer (GCL) macular volume and thickness between persons living with HIV (PLHIVs) with well-controlled infections and good immune recovery, and sex- and age-matched HIV-uninfected controls (HUCs). Methods This prospective cross-sectional study (www.clinicaltrials.gov identifier: NCT02003989) included 56 PLHIVs, infected for ≥10 [median 20.2] years and with sustained plasma HIV-load suppression on combined antiretroviral therapy (cART) for ≥5 years, and 56 matched HUCs. Participants underwent spectral-domain optical coherence tomography (SD-OCT) with thorough ophthalmological examinations and brain magnetic resonance imaging (MRI). Their overall and quadrant pRNFL thicknesses, total macular volumes, and GCL macular volumes and thicknesses were compared. Cerebral small-vessel diseases (CSVD) complied with STRIVE criteria. Results Median [interquartile range, IQR] ages of PLHIVs and HUCs, respectively, were 52 [46–60] and 52 [44–60] years. Median [IQR] PLHIVs’ nadir CD4+ T-cell count and current CD4/CD8 T-cell ratio were 249/μL [158–350] and 0.95 [0.67–1.10], respectively; HIV-seropositivity duration was 20.2 [15.9–24.5] years; cART duration was 16.8 [12.6–18.6] years; and aviremia duration was 11.4 [7.8–13.6] years. No significant between-group pRNFL thickness, total macular volume, macular GCL-volume and -thickness differences were found. MRI-detected CSVD in 21 (38%) PLHIVs and 14 (25%) HUCs was associated with overall thinner pRNFLs, and smaller total retina and GCL macular volumes, independently of HIV status. Conclusions SD-OCT could not detect pRNFL thinning or macular GCL-volume reduction in well-sustained, aviremic, cART-treated PLHIVs who achieved good immune recovery. However, CSVD was associated with thinner pRNFLs and GCLs, independently of HIV status.

72.5±12.8 [45] ( Continued ) We undertook this concurrent cohort study to investigate pRNFL and GCL thicknesses in PLHIVs with well-sustained, cART-controlled, immunovirological parameters and HIV-uninfected controls (HUCs). Because we wanted to examine the role of chronic HIV infection itself, we selected PLHIVs with cART-sustained, immunovirological control for at least 5 years, without hepatitis C virus (HCV) infection, past or ongoing acquired immune deficiency syndrome (AIDS)-defining neurological events (ADNEs), and/or alcohol or illicit drug abuse.

Ethics approval
This study, approved by the CPP Île-de-France VI Ethics Committee, adhered to the tenets of the Declaration of Helsinki. Written informed consent was obtained from all participants.

PLOS ONE
No pRNFL thinning in well-sustained, aviremic HIV+ individuals

Study population
In this cross-sectional study (NCT02003989), we prospectively included PLHIVs followed in two Infectious Diseases Departments in University Hospitals, caring for about 8,000 PLHIVs in the Paris area (France). Inclusion criteria were: (1) HIV seropositivity known for �10 years; sustained CD4+ T-cell count �350/μL and plasma HIV load (plVL) <20 copies/mL for �5 years on cART. plVL was quantified using the Amplicor monitor assay (Cobas 2.0, Roche Diagnostics, Basel, Switzerland), which has a lower detection limit of 20 HIV-1 RNA copies/ mL. Exclusion criteria were: (1) transient low-level viremias �20 but �200 copies/mL (viral blip) once within the previous 5 years; (2) history/concomitant ocular trauma or diseases; (3) family history of glaucoma; (4) prior/current treatment with drugs associated with toxic optic neuropathy or retinopathy; (5) prior/current neurological/psychiatric disorders, including ADNE; (6) prior/current diabetes mellitus; (7) prior/current alcohol or illicit substance abuse (with the exception of occasional cannabis use); (8) HCV infection. Cognitive decay was not sought prior to inclusion and was not an exclusion criterion. Age (±5 years)-and sex-matched (1:1) HUCs were selected on a voluntary basis; exclusion criteria were the same as for PLHIVs. The absence of HIV infection was confirmed by ELISA or rapid HIV test.
All participants underwent the same comprehensive neurological and ophthalmological examinations. All SD-OCT were obtained with a Spectralis OCT, which generates four sectoral pRNFL thicknesses and an overall value that is the average of the four quadrants, using the new Nsite Axonal Analytics software (Heidelberg Engineering GmbH, Heidelberg, Germany). Overall and sectoral pRNFL thicknesses were recorded for each eye. The total retinal macular volume was calculated with the Heidelberg software using the Early Treatment for Diabetic Retinopathy Study (ETDRS) grid. We used the Iowa Reference Algorithm to segment the GCL and calculated the GCL thickness for each of the 9 ETDRS-grid subfields and the total GCL macular volume within the ETDRS grid (S1 Fig) [29][30][31]. Cognition was assessed using the Montreal Cognitive Assessment (MoCA), known to be an adequate clinical and ecological screening tool for PLHIVs [32,33]. 3-Tesla brain MRI and orbit images (Philips Healthcare, Best, The Netherlands) were analyzed by one neuroradiologist (J.S.) blinded to all parameters. CSVD was defined according to the Standards for Reporting Vascular Changes on Neuroimaging (STRIVE) Criteria [9].

Main outcome measure and estimated number of participants
The main outcome measure was the overall pRNFL thickness. Previous studies found a standard deviation of~9 μm for it with the OCT machine used herein [34]. With that standard deviation, a unilateral 5% alpha-risk and 95% power, we estimated that 50 participants in each group would be sufficient to detect �7-μm pRNFL thinning in PLHIVs compared to HUCs.

Statistical analyses
Only one eye of each participant was randomly selected for analysis. The non-parametric Mann-Whitney U-test and chi 2 or Fisher's exact test were used as appropriate to compare groups. Analysis of variance (ANOVA) of repeated measures was used to detect differences among the different pRNFL quadrants and among the different GCL ETDRS-grid subfields.
Linear-regression models were used to test associations between OCT measurements (outcome variables: overall and temporal pRNFL thicknesses, total retinal macular volumes, GCL macular volumes) and visual function measurements, axial length and age (predictor variables). HIV status was included in all these linear-regression models to test a possible HIV effect on an association.
For the PLHIV group, associations between overall and temporal pRNFL thicknesses, total retinal macular volumes, GCL macular volumes (outcome variables) and duration of HIV infection (predictor variable) were tested with a multivariate model including age as a covariate. Univariate regression analyses were used to identify associations between PLHIVs' overall and temporal pRNFL thicknesses, total retinal macular volumes or GCL macular volumes (outcome variables) and CD4+ T-cell nadirs (predictor variable).
Significance was defined as p<0.05. All analyses were computed with Statistica software (Statsoft, Inc, Maison Alfort, France) and no statistical correction was made for multiple comparisons.
pRNFL thicknesses, total retinal macular volumes and GCL macular volumes and thicknesses (Table 3) did not differ between PLHIVs and HUCs. These analyses were repeated using the other eye when both eyes were assessable and the results were comparable (S1 Table). Potential confounding factors (axial length, spherical equivalent and Optical Quality Analyzing System (OQAS)-assessed media opacity) were comparable for PLHIVs and HUCs.
Among the functional parameters (Table 3), high contrast VA, low contrast VA and color vision were comparable for PLHIVs and HUCs. However, visual field mean deviations (MDs) and intraocular pressure (IOP) differed significantly, being slightly lower for PLHIVs than HUCs but still within normal limits.
Associations between structural measures and other variables are reported in Table 4. Age was associated with overall pRNFL thickness, GCL macular volume and total retinal macular volume, with HIV status having no significant effect (Fig 1). Among the PLHIVs, no significant association between structural measures and HIV-infection duration or CD4+ T-cell nadir was found (Table 4, Fig 2).
MRIs did not reveal a lesion that could cause optic neuropathy or trans-synaptic retrograde degeneration within the optic nerves. MRI detected CSVD in 21 (38%) PLHIVs and 14 (25%) HUCs (p = 0.15); it mainly reflected WMHs of presumed vascular origin (21 PLHIVs versus 12 HUCs, p = 0.06). Documented CSVD was associated with overall thinner pRNFLs, smaller whole retinal macular volumes and smaller GCL macular volumes for all participants, with HIV status having no significant effect (Fig 3, S1 Table).

Discussion
Our results showed that overall and 4-quadrant (localized) pRNFLs, total retinal macular volumes, GCL macular volumes and EDTRS-grid-defined GCL regional thicknesses were not smaller in long-term-sustained, immunovirologically controlled PLHIVs compared to ageand sex-matched HUCs.
The higher frequency of CVSD-surrogate WMHs in aviremic, cART-treated PLHIVs compared to HUCs, not related to any ART classes, was recently reported [10,47]. We found only a trend toward significance for WMHs of presumed vascular origin between PLHIVs and HUC (p = 0.06). That failure to reach significance is probably due to a lack of statistical power of our study. These vascular abnormalities were associated with thinner overall pRNFLs, smaller total retina macular volumes and smaller CGL macular volumes, independently of HIV status. A

PLOS ONE
No pRNFL thinning in well-sustained, aviremic HIV+ individuals pRNFL thinning or defect was reported previously for the arteriosclerotic CSVD form [26] and cerebral autosomal-dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL) [48], a genetic form of CSVD. To the best of our knowledge, that association has not been reported previously in adult, cART-treated PLHIVs with CSVD. CSVD is

PLOS ONE
No pRNFL thinning in well-sustained, aviremic HIV+ individuals characterized by thickening of the walls of the small perforating arteries in the brain, resulting in low cerebral blood flow [49]. A causative role of cerebral hypoperfusion and decreased perfusion of the inner retinal layers has been advanced to explain the association of CSVD with pRNFL and GCL thinning, as shown for CADASIL [50]. Our results agree with an emerging change of the NeuroHIV paradigm, highlighting the potential contribution of vascular brain damage in aging PLHIVs [16]. Despite being the first combined OCT-MRI study in middle-aged, cART-treated PLHIVs with well-sustained immune restoration, our study might suffer from a lack statistical power. Although our study-sample size might be considered relatively small, it is in accordance with previous similar publications (Table 1) and is sufficient to highlight pRNFL differences in these subjects. Although PLHIV and HUC pRNFL-thickness sameness cannot be established, our study's statistical power to detect thinning of -5 μm was 90%. Interestingly, the test-retest variability of SD-OCT-measured pRNFL thickness was~5 μm [51]. Because the primary endpoint and the power calculation were based on the overall pRNFL thickness, interpretation of the lack of a smaller GCL volume and localized pRNFL thinning in our study requires prudence. Localized pRNFL thinning could result from methodological biases secondary to the multiplicity of statistical comparisons without control for inflation of the type-1 error [52]. That bias was also seen in macula studies, in which multiple quadrants and multiple retinal

PLOS ONE
No pRNFL thinning in well-sustained, aviremic HIV+ individuals

PLOS ONE
No pRNFL thinning in well-sustained, aviremic HIV+ individuals

Fig 2. No association between PLHIVs' overall peripapillary retinal nerve-fiber-layer (pRNFL) thickness (top) or ganglion-cell-layer (GCL) volume (bottom) and CD4+ T-cell count nadirs of the persons living with human immunodeficiency virus (PLHIVs).
The vertical line represents the CD4+ T-cell count nadir of 100 cells/μL, because previous studies found that only PLHIVs with nadirs <100 cells/μL were more likely to have thinner pRNFLs. In our study only nine PLHIVs had a nadir <100 cells/μL but their pRNFL thicknesses and macular GCL volumes did not differ from those of the other PLHIVs.  Table).
layer thicknesses generated high numbers of p values. Such methodological biases were previously highlighted by Demirkaya et al. [5]. Although our series is not representative of the entire HIV+ population, it is representative of PLHIVs receiving care in northern Europe, where >90% are successfully treated [53]. However, the 2019 UNAIDS world epidemiological data showed that 79% of PLHIVs are aware of their seropositivity, 78% of PLHIVs knowing their HIV status are cART-treated and 86% of those cART-treated PLHIVs have a plVL below the detection threshold (unaids.org). Moreover, it was recently demonstrated that low plVLs of 51-200 copies/mL were strongly associated with virological failure [54]. Thus, HIV-induced brain damage may be more a legacy effect resulting from prior incomplete virus control [54,55]. Those findings provide support for the European definition of virological failure as persistent plVLs of >50 copies/mL (eacsociety.org). Our findings should not be compared or held as a contradiction to historical studies and may better apply to future PLHIV cohorts, for whom the therapeutic guidelines recommend that cART must be initiated for a CD4+ T-cell count threshold �500 cells/μL [56]. Indeed, the CASCADE study showed that PLHIVs are mostly diagnosed and treated with a CD4+ T-cell count nadir >200 cells/μL [57]. Recent findings showed that full viral suppression may preserve long-term PLHIVs' brain health [58]. Hence, continuing to report results concerning virologically uncontrolled, cART-treated PLHIVs is not really suitable [54,55,58].
In addition to the OCT evaluation of visual pathway structures, most visual evaluation and ocular parameters were similar for PLHIVs and HUCs, with the exception of visual field MDs and IOP. Decreased visual field MD is a classic sign of HIV-NRD but in our study it cannot be explained by pRNFL thinning or decreased GCL macular volume. That decline could however be explained by functional changes without structural damage of the retinal ganglion cells or other cells implicated in vision. Indeed, decreased cone-photoreceptor density in HIV+ participants was found using an adaptive optics camera [59] and SD-OCT visualized possible changes in retinal layers other than GCL and RNFL [5,60,61]. Lower IOP in HIV+ participants was previously described and its causes are more likely multifactorial [61]. Our results showed that this diminished IOP persists in PLHIVs with long-term, well-sustained immune control.
Our results showed that cART-treated PLHIVs who successfully achieved sustained immunovirological control, even with HIV infection lasting 20 years, do not have pRNFL thinning. That is an optimistic finding in terms of PLHIVs aging. However, in line with recent results from a study on HIV-infected children [28], as in the general population, CSVD in our middle-aged PLHIVs was associated with thinner pRNFLs and GCLs.