Retinal Structure and Function in Perinatally HIV-Infected and cART-Treated Children: A Matched Case–Control Study

Abstract

Purpose.: Subtle structural and functional neuroretinal changes have been described in human immunodeficiency virus (HIV)–infected adults without retinitis treated with combination antiretroviral therapy (cART). However, studies on this subject in HIV-infected children are scarce. This study aimed to assess the presence of (neuro)retinal functional and structural differences between a group of perinatally HIV-infected children on cART and age-, sex-, ethnicity-, and socioeconomically matched healthy controls.

Methods.: All participants underwent an extensive ophthalmological examination, including functional tests as well as optical coherence tomography, to measure individual retinal layer thicknesses. Multivariable mixed linear regression models were used to assess possible associations between HIV status (and other HIV-related parameters) and ocular parameters, while accounting for the inclusion of both eyes and several known confounders.

Results.: Thirty-three HIV-infected children (median age 13.7 years [interquartile range (IQR), 12.2–15.8], median CD4⁺ T-cell count 760 cells/mm³, 82% with an undetectable HIV viral load [VL]), and 36 controls (median age 12.1 years [IQR, 11.5–15.8]) were included. Contrast sensitivity (CS) was significantly lower in the HIV-infected group (1.74 vs. 1.76 logCS; P = 0.006). The patients had a significantly thinner foveal thickness (−11.2 μm, P = 0.012), which was associated with a higher peak HIV VL (−10.3 μm per log copy/mL, P = 0.016).

Conclusions.: In this study, we found a decrease in foveal thickness in HIV-infected children, which was associated with a higher peak VL. Longitudinal studies are warranted to confirm our findings and to determine the course and clinical consequences of these foveal changes.

The spectrum of human immunodeficiency virus (HIV)–related retinal disease has changed drastically since the introduction of combination antiretroviral therapy (cART), with a major decline in incidence of opportunistic infections, such as cytomegalovirus (CMV) retinitis, as well as noninfectious ischemic HIV retinopathy. 

However, even in cART-treated individuals with well-suppressed HIV infection and absence of opportunistic infections, functional and structural retinal abnormalities have been reported, such as a subtle loss of color vision and/or contrast sensitivity, visual field deficits, and a thinner retinal nerve fiber layer (RNFL) thickness.^1–10 These changes are thought to be part of a “HIV-associated neuroretinal disorder” (HIV-NRD) and may be mediated by several processes, such as longstanding microvasculopathy,^11–17 direct damage of neural tissue by HIV and/or cART,^18–20 and chronic (para)inflammation.²¹ 

It is unclear whether such retinal changes are also present in cART-treated HIV-infected children (without a history of ocular opportunistic infections). So far, one study group has addressed this question and reported a thinning of the peripapillary RNFL in 19 HIV-infected children as compared to 21 healthy controls,²² as well as foveal thickening and multifocal electroretinographical (mf-ERG) abnormalities in a smaller subgroup of HIV-infected children.²³ 

This study is part of an interdisciplinary observational cross-sectional study, evaluating neurological and neurocognitive disorders, neuroimaging, and ophthalmic alterations in perinatally HIV-1-infected children in The Netherlands.²⁴ In the current study, we assessed the presence of retinal structural and functional differences between perinatally HIV-infected children on cART and a group of age-, sex-, ethnicity-, and socioeconomically matched healthy controls. This is the first study employing the Iowa Reference Algorithm^25,26 on spectral-domain optical coherence tomography (SD-OCT) scans in HIV-infected children, enabling the measurement of the thickness of individual retinal layers. 

Subjects and Methods

The study adhered to the tenets of the Declaration of Helsinki, and approval was obtained from the investigational review board at the Academic Medical Center in Amsterdam. Written informed consent was obtained from all parents and from children aged 12 and above. 

Study Participants

All HIV-infected children between 8 and 18 years of age attending the pediatric HIV outpatient clinic of the Academic Medical Center were approached for study participation between December 2012 and January 2014. Healthy controls were recruited through parental evenings at schools, sports clubs, and churches situated in areas in Amsterdam, aiming to capture an ethnically diverse population with a lower socioeconomic status (SES) than the general Dutch population, as similar as possible to patients.²⁷ Exclusion criteria were chronic (non–HIV-associated) neurological diseases like epilepsy, (history of) intracerebral neoplasms, and psychiatric disorders. Intelligence quotient (IQ) was measured using the WISC III and WAIS III for children older than 16 years of age.^28,29 Frequency matching for age, sex, ethnicity, and SES was performed. Socioeconomic status was determined using parental education and occupational status. Parental education was scored according to the International Standard Classification of Education (ISCED). Occupational status was defined as no, one, or two caregivers with a paid job. The remaining sociodemographic data were obtained using standardized questionnaires. 

HIV- and cART-Related Characteristics

Historical HIV viral load (VL) and CD4⁺ T-cell counts, Centers for Disease Control (CDC) clinical stage category, and cART treatment history were derived from the Dutch HIV Monitoring Foundation database.³⁰ Clinical, immunological, and virological data prior to migration to The Netherlands were collected for the immigrant children and registered as “missing” when not traceable. Outpatient visits in our center occur every 3 months, and all clinical, immunological, and virological data are registered by the HIV Monitoring Foundation. The time of HIV diagnosis was defined as the first known documented positive HIV test, also using test data from the country of origin of children who were born outside The Netherlands. Viral load results were based on different assays used between 1995 and 2013 with decreasing detection limits (<1000 copies/mL in 1995 to <40 copies/mL in 2013). An undetectable HIV VL was defined as a VL below the detection limit of the assay used at that time. The HIV VL during study participation was determined by the Abbott (Chicago, IL, USA) Real Time HIV-1 assay. The peak HIV VL was defined as the highest VL prior to cART initiation, or the highest HIV VL due to interruption of cART or virological failure. The proportion of life spent with a detectable VL was calculated by adding all days between two detectable HIV VL and half of the days between the last detectable and the next undetectable HIV VL, and dividing the cumulative number of days with a detectable VL by the participant's age at inclusion into this study. 

To account for physiological age-related differences in CD4+ T-cell counts, all registered CD4⁺ T-cell counts were transformed into z scores by subtracting the reference value for the age at the time of the CD4⁺ T-cell measurement and dividing this by the age-related SD. A z score of 0 represents the age-appropriate mean.³⁰ The nadir CD4⁺ T-cell z score was defined as the lowest z score prior to cART initiation or to a maximum of 3 months after the start of cART. At inclusion, absolute CD4⁺ T-cell counts were analyzed instead of z scores, as all children were >8 years of age. Lastly, we calculated the cumulative time with a CD4⁺ T-cell count below 50, 100, 200, 350, and 500 cells/mm³ by adding all days between two CD4⁺ T-cell measurements of <500 cells/mm³ to half of the days between two CD4⁺ T-cell measurements with one <500 cells/mm³ and one >500 cells/mm³. 

Ophthalmic Examination

Ophthalmic exclusion criteria were high refractive errors (spherical equivalent [SE] > +5.5 or > −8.5 diopters [D]), visual acuity below 0.1 logMAR, intraocular pressure (IOP) higher than 21 mm Hg, significant media opacities, and a history of ocular surgery or ocular disease. One patient with a history of CMV retinitis in both eyes and one control with a refractive error > +5 D were excluded for not meeting the inclusion criteria. In addition, two left eyes of two HIV-infected children were excluded from analysis due to the presence of uveitis and congenital toxoplasmosis lesions, respectively. 

Visual acuity (VA) was measured using a modified Early Treatment Diabetic Retinopathy Study (ETDRS) chart with Sloan letters (Lighthouse, New York, NY, USA) at 4 m. Visual acuity was recorded in logMAR units. Intraocular pressure was measured by air-puff tonometry (computerized tonometer, CT80; Topcon Medical Systems, Inc., Oakland, NJ, USA). All subjects underwent pupil dilation (0.5% tropicamide and 5% phenylephrine) and a standard ophthalmic examination, including slit-lamp biomicroscopy with a handheld lens, as well as fundus photography. 

Mars Contrast Sensitivity (CS) Charts

All children were tested with the Mars Letter Contrast Sensitivity Test (Mars Perceptrix, Chappaqua, NY, USA), a portable chart measuring 23 × 36 cm, consisting of 48 letters arranged in eight rows of six Sloan letters each. The Mars test letters subtend 2° (at 50 cm), the change in contrast between successive letters is 0.04 log units, and the contrast range is from 0.04 to 1.92 log units. The Mars test has test–retest reliability equal to or better than the Pelli Robson test and has proven to be a useful and practical alternative to the Pelli Robson contrast sensitivity chart.^23,31–34 

Lanthony D-15 Color Vision Panel Test

Color vision was determined using the Lanthony Desaturated 15-hue (D-15) color vision test.³⁵ This test is more sensitive to subtle color discrimination deficiencies and is easier to perform and score than the Farnsworth-Munsell 100-hue test (FM-100).³⁶ Testing was performed under standard illuminant conditions and repeated once when errors were made. Color confusion index (CCI), as described by Vingrys and King-Smith,³⁷ was determined for each eye. Errorless performance is scored with a CCI of 1.0, and higher values indicate a worse test result. The best outcome per eye was used for analysis. 

Rarebit Perimetry and Rarebit Fovea Test

The Rarebit Perimetry (RBP) and Rarebit Fovea Test (RFT) are visual function tests developed to detect subtle central visual field damage and have been described extensively elsewhere.^38,39 The RBP (inner and outer tests) evaluates the central 30° visual field, while the RFT evaluates the central 4° visual field. The results of the Rarebit test are presented as mean hit rate (MHR), a percentage of the stimuli seen by the subject. 

SD-OCT and Retinal Layer Segmentation

Optical coherence tomography images of the subjects were obtained with SD-OCT (Topcon 3D OCT-1000; Topcon, Inc., Paramus, NJ, USA) using the 3D macular and disc volume scan protocols. Only high-quality images with a Topcon image quality factor (QF) > 60 were used. From each 3D macular volume, individual retinal layers were segmented automatically by the Iowa Reference Algorithm,^25,26 which uses an extensively validated, robust, fully three-dimensional graph search approach (Fig. 1A). The Iowa Reference Algorithm^25,26 allows for calculation of the thickness of all individual retinal layers for each of the nine ETDRS grid defined regions (Figs. 1B–E). 

Figure 1

 

(A) Macular SD-OCT B-scan with intraretinal surfaces as indicated by the colored lines and segmented using the Iowa Reference Algorithm.^25,26 In this study, the highly reflective layer between inner and outer segments and the outer segments up to the retinal pigment layer were taken together as one layer, the outer segment layer (OSL), ignoring the line ascribed to the cone outer segments.⁶³ Corresponding retinal layers: 1, retinal nerve fiber layer, 2, ganglion cell layer, 3, inner plexiform layer, 4, inner nuclear layer, 5, outer plexiform layer, 6, outer nuclear layer + inner segments (photoreceptors), 7, outer segments (photoreceptors), 8, retinal pigment epithelium. (B–E) ETDRS grid. Nine subfields of the nine ETDRS regions in each eye. (B) Right eye. (C) Left eye. For each retinal layer, three areas were defined using this ETDRS grid: the fovea, the central circle with a diameter of 1 mm (depicted as 1 in [B, C]); the pericentral ring, a donut-shaped ring centered on the fovea with an inner diameter of 1 mm and an outer diameter of 3 mm (D); and the peripheral ring, with an inner diameter of 3 mm and outer diameter of 6 mm (E). Thickness measurements of the pericentral and peripheral rings were estimated by averaging the thickness measurements of the four corresponding quadrant areas (segments 2–5 for the pericentral ring and segments 6–9 for the peripheral ring).

In addition, peripapillary RNFL thickness measurements were acquired from the 3D optic nerve head OCTs using the same Iowa Reference Algorithm.^25,26 The peripapillary ring was centered manually if needed, with the center of the ring coinciding with the center of the optic disc. 

Statistical Analysis

Demographic characteristics were compared between groups using the unpaired t-test, the Mann-Whitney U test, or the χ² test. Univariable and multivariable linear regression models with mixed effects were used to explore associations between HIV status and ocular variables in all study participants while accounting for the inclusion of both eyes and potential confounders (age at study visit,^40–44 sex,^41–43 IQ [corrected for in the function test analyses], OCT QF,^40,44–46 and SE^44,47 [corrected for in the OCT analyses]). Covariates with a P value < 0.2 in univariable analysis were incorporated in the multivariable models. In the multivariable models the level of significance was set at a P value < 0.05. 

The outcome variables that were significantly different between HIV-infected and healthy children in the models described above were further investigated in the HIV-infected group only, again using multivariable mixed linear regression models. Associations between the specified parameters and (1) disease-related factors (HIV VL at time of study visit, peak HIV VL, the proportion of life spent with a detectable VL, nadir CD4⁺ T-cell z score, duration of CD4⁺ T-cell counts < 500 cells/mm³, CDC clinical category) and (2) cART-related factors (age at cART initiation, duration of cART use, current cART use, and duration of exposure to didanosine and/or stavudine (which can cause a toxic retinopathy^48–50) were explored. 

Furthermore, possible correlations between visual function test results and retinal layer thickness were assessed, in particular, focusing on the significant parameters. 

Data entry and management was performed using OpenClinica open source software (Waltham, MA, USA). Statistical analyses were carried out using Stata Statistical Software, release 12 (StataCorp LP, College Station, TX, USA). 

Results

Demographic and Clinical Characteristics

Table 1 shows the demographic and clinical characteristics of all study participants. In total, 33 HIV-infected children (median age 12.1; interquartile range [IQR], 11.5–15.8) and 36 healthy controls were included (median age 13.7; IQR, 12.2–15.8). Most children were of black (HIV: 79%; healthy: 75%) or mixed black (HIV: 12%; healthy: 17%) ethnicity. Although mean SE, IOP, and VA differed significantly between the two groups, all values were within a normal range. Among the HIV-infected children, 32 (97%) had ever received cART, and 28 (85%) were using cART at time of the study assessment; among these, 27 (96%) had an undetectable plasma HIV VL. The median CD4⁺ T-cell count was 760 cells/mm³ at the time of assessment. 

Table 1

 

Participant Characteristics

 

Visual Function Tests

No significant differences in color vision and central visual field were detected between the two groups (Table 2). Contrast sensitivity was significantly lower in the HIV-infected children, although the difference was only half a letter (1.74 vs. 1.76 logCS, P value = 0.006). 

Table 2

 

Visual Function Test Results in HIV Patients and Controls

 

Thickness of Individual Retinal Layers

Multivariable mixed linear regression analyses (adjusted for age, sex, OCT QF, and SE) were performed to detect differences between the HIV-infected and healthy groups. Mean retinal layer thicknesses (individual retinal layers, peripapillary RNFL, and total retinal thickness) for patients and controls are shown in Table 3. Human IV-infected children had a significantly thinner total foveal thickness (−11.2 μm, P value = 0.012), predominantly due to a thinner foveal outer nuclear layer and inner segments (ONL-IS; −6.2 μm, P value = 0.011). 

Table 3

 

OCT Individual Retinal Layer Thicknesses in HIV Patients and Controls

 

Multivariable Analyses Within the HIV-Infected Group

Visual function and OCT parameters that differed significantly between HIV-infected and healthy children were further investigated in the HIV-infected group to identify potential associations between these parameters and HIV- and/or cART-related variables. 

Multivariable mixed regression analysis showed an inverse association between total foveal retinal thickness and peak HIV VL (−10.7 μm per log copy/mL, P value = 0.016, Fig. 2); a similar relationship was observed between the foveal ONL-IS and peak HIV VL (−7.1 μm per log copy/mL, P value = 0.013). No other associations were found between visual function, OCT parameters, and HIV- or cART-related parameters. 

Figure 2

 

Correlation between total foveal thickness and peak HIV viral load.

Structure–Function Relationships

Finally, we explored potential correlations between visual function test results (color vision, CS, and central visual field) and OCT retinal layer thickness in HIV-infected children. No associations were observed (data not shown). 

Discussion

This study aimed to assess retinal structure and visual function in a group of perinatally HIV-infected children, compared to a group of matched healthy controls. Subtle structural retinal changes and visual dysfunction, termed HIV-associated neuroretinal disorder, have been described in HIV-infected adults on cART without infectious retinitis^{1–10,14,16,17,51–59}; however, data on this subject in HIV-infected children are limited and derived from one study group.^22,23 This is the first study assessing individual retinal layer thicknesses and exploring associations between various HIV-and cART-related factors and ocular parameters in HIV-infected children. 

Our findings indicate that HIV-infected children have a thinner foveal retinal thickness than healthy controls while having a comparable peripapillary RNFL thickness and visual function outcomes. 

We found no significant differences in peripapillary RNFL thickness between HIV-infected children and age-, sex-, ethnicity- and socioeconomically matched controls, which is not in line with the previous pediatric study by Moschos et al.²² They reported a thinner peripapillary RNFL thickness in a group of 19 HIV-infected children compared to a group of 21 age-matched healthy children.²² Of note, however: In our study, we used spectral-domain OCT, which is more accurate in measuring retinal and RNFL thickness than time-domain OCT.⁶⁰ Furthermore, in our assessment of peripapillary RNFL thickness, we adjusted for known confounders such as age,^40–44 sex,^41–43 SE,^40,44,47 and OCT QF.^40,44,46 In addition, we applied multilevel mixed linear modeling to take correlations between right and left eyes into account. 

In their second study, Moschos et al.²³ found an increase in foveal thickness as well as multifocal ERG abnormalities in a subgroup, consisting of 10 eyes (the number of children was not stated), in their cohort. We observed the opposite in our study, with a significantly lower total foveal thickness in the group of HIV-infected children, predominantly due to a thinner ONL-IS of the photoreceptors as well as thinner inner retinal layers. Since we are the first group measuring individual OCT retinal layer thicknesses in HIV-infected children, comparison of our findings to other studies in children is not yet attainable. In HIV-infected adults, however, it is thought that damage (caused by HIV and/or other factors) to the optic nerve leads to thinning of the peripapillary RNFL.^1,7,9 Multiple adult studies indeed reported a decrease in peripapillary RNFL thickness, particularly in patients with (a history of) low (<100 cells/mm³) CD4⁺ T-cell counts.^4,52,57–59 Since the axons of the ganglion cell layer make up the optic nerve in large part, a decrease in peripapillary RNFL implies that a decrease in ganglion cell layer thickness (and possibly other inner retinal layers) would also be expected, but no study has reported on this. If we extrapolate this hypothesis of HIV-associated neuroretinal degeneration to HIV-infected children, a decrease of peripapillary RNFL thickness and inner retinal layers (especially in the pericentral retinal area, where the ganglion cell layer is thickest) would be expected in this group. Surprisingly, we detected a thinner foveal thickness and no other OCT differences in the HIV-infected children as compared to the controls. 

However, the consequences of chronic HIV infection in perinatally infected children are likely to be different from those in adults because their infection occurs during rather than after (neuronal) development. This may result in different findings between adults and children. Of all foveal layers, the outer nuclear layer shows the most distinct increase during foveal maturation (from infancy to 16 years of age), creating a bulge.⁶¹ Therefore, if we speculate that HIV infection may interrupt this maturation, this would be mostly reflected in the outer nuclear layer. This may explain the significantly thinner fovea—in particular, the ONL-IS—that we detected in the HIV-infected children in our study. Supporting this hypothesis was the significant association we observed between a higher peak HIV VL and thinner foveal ONL-IS. 

A recent study in 22 perinatally HIV-infected children (median age 16.6 years) on cART, with nadir CD4 counts > 200 cells/mm³, reported an absence of vision-threatening disease but a high prevalence (18%) of strabismus,⁶² which is thought to be a developmental disorder, again suggesting that HIV infection may hamper the development of the visual system. 

We did not detect any correlations between retinal structure and visual function in HIV-infected children. This was to be expected considering the high interindividual variability in retinal layer thickness and functional test outcomes and the small changes observed. 

Strengths of our study are the inclusion of an age-, sex-, ethnicity-, and socioeconomically matched control group and the adjustment for relevant potentially confounding factors in our statistical analyses. Of note, we found IQ to be an important confounder when analyzing the functional test outcomes, with a strong positive association between test results and total IQ. Nonetheless, there are some limitations. Even though this is the largest ophthalmological study in clinically stable HIV-infected children without retinitis at present, the relatively small sample size may have hampered the detection of some potential associations. Secondly, a cross-sectional study of parameters known to have a high interindividual variability is less able to detect small changes in retinal structure and function than a longitudinal study. 

Although we aimed to capture a control population as similar as possible to the patients by matching for age, sex, and SES, there is a possibility that some of the variability in the results we detected was caused by non–HIV-related factors. 

The visual function tests we used are psychophysical techniques and involve an element of subjectivity; however, we used standardized protocols (i.e., lightning conditions, testing distance) to minimize this problem. A more objective method for evaluating retinal function is mf-ERG, but it requires adequate cooperation and can therefore be challenging to perform in children. Lastly, we did not measure peripapillary RNFL thickness in quadrants; therefore, it is possible that there were quadrantal changes in RNFL thickness in our patients while the average peripapillary RNFL thickness was not different between the two groups. 

In summary, our findings indicate that HIV-infected children have a thinner foveal retinal thickness compared to matched controls while having comparable peripapillary RNFL thickness and visual function outcomes. Our results do not confirm the results of Moschos et al.^22,23; however, they are in accordance with some recent studies in HIV-infected adults observing similar average peripapillary RNFL thickness between patients and controls,^1,9 and little difference in CS.^7,9 A novel finding in our study is the decrease in foveal thickness in the group of HIV-infected children; the clinical significance of this is yet unclear, since both VA and visual function were adequate. We postulate that HIV infection may disturb foveal maturation, leading to a thinner fovea. Although our results do not support the presence of a HIV-neuroretinal disorder, it is also possible that retinal (neuro)degeneration will occur at a later time in patients' lives. The long-term effects of HIV infection on the retina are unknown, and as life expectancy of HIV-infected patients is increasing with the global rollout of cART, vision loss might become more prevalent and symptomatic with time. Therefore, longitudinal studies are warranted to investigate the effect of chronic HIV infection and long-term cART on the retinal structure and visual function of both HIV-infected adults and children. 

Acknowledgments

Supported by Global Ophthalmology Awards Program 2012, Bayer HealthCare and the Emma Foundation (Grant 11.001). 

Disclosure: N. Demirkaya, None; S. Cohen, None; F.W.N.M. Wit, None; M.D. Abramoff, IDx LLC (S), P; R.O. Schlingemann, Novartis (S); T.W. Kuijpers, None; P. Reiss, Gilead Sciences (S), Janssen Pharmaceutica (S); D. Pajkrt, None; F.D. Verbraak, Bayer HealthCare (F), Bayer NL (S), Novartis NL (S)