Ophthalmic pathologies in an older HIV-positive Tanzanian population and their association with neurological morbidity and evaluation of a low-resource screening strategy


 Background

Globally, 43 million people are living with HIV, 90% in developing countries. Increasing life expectancy with combination antiretroviral therapy (cART) results in chronic complications, including HIV-associated neurocognitive disorders (HAND) and eye diseases. HAND screening is currently challenging.
Objectives

To evaluate clinical utility of HIV-retinopathy as a screening measure of HAND in older cART-treated individuals in Tanzania and feasibility of smartphone-based retinal screening in this low-resource setting.
Methods

A cross-sectional systematic sample aged ≥ 50-years attending routine HIV follow-up in Tanzania were comprehensively assessed for HAND by American Academy of Neurology criteria and received ophthalmic assessment including smartphone-based retinal imaging. HAND and ophthalmic assessments were independent and blinded. Diagnostic accuracy was evaluated by AUROC curves.
Results

Of 129 individuals assessed, 69.8% were visually impairment. Six had suspected HIV retinopathy. HAND prevalence was 66.7%. HIV retinopathy was significantly associated with HAND but HIV-disease factors (CD4, viral load) were not. Diagnostic accuracy of HIV-retinopathy for HAND was poor (AUROC 0.545–0.617) but specificity and positive predictive value were high.
Conclusions

Ocular pathology and HAND appear highly prevalent in this low-resource setting. Although retinal screening cannot be used alone identify HAND, prioritization of individuals with abnormal retinal screening is a potential strategy in low-resource settings.


Results
Of 129 individuals assessed, 69.8% were visually impairment. Six had suspected HIV retinopathy. HAND prevalence was 66.7%. HIV retinopathy was signi cantly associated with HAND but HIV-disease factors (CD4, viral load) were not. Diagnostic accuracy of HIV-retinopathy for HAND was poor (AUROC 0.545-0.617) but speci city and positive predictive value were high.

Conclusions
Ocular pathology and HAND appear highly prevalent in this low-resource setting. Although retinal screening cannot be used alone identify HAND, prioritization of individuals with abnormal retinal screening is a potential strategy in low-resource settings.

Background
Almost 43 million people are infected by the human immunode ciency virus (HIV) globally, with 90% of cases in developing countries [1]. In sub-Saharan Africa (SSA), almost 25 million individuals currently live with HIV [2].
Increasing provision of combination antiretroviral therapy (cART) is resulting in both improving life expectancy and emergence of chronic HIV-related complications and chronic diseases as previously seen in high-income countries. HIV-associated neurocognitive disorders (HAND) affect up to 50% of people treated for HIV and are associated with increased morbidity and mortality [3,4]. HAND are classi ed by severity: asymptomatic neurocognitive impairment (ANI); mild neurocognitive disorder (MND) and; HIV-associated dementia (HAD) [5].
Early identi cation of HAND and subsequent prompt initiation of treatment, prior to progression of HAND stage, may reduce mortality and reverse neurocognitive de cits [3]. Screening for, and identi cation of HAND are currently challenging, in SSA and elsewhere. Di culties include shortage of neurology and psychiatry specialist services in SSA, lack of accurate brief screening measures [6] and current consensus diagnostic criteria requiring detailed neuropsychological assessment, impractical in busy clinical settings. Additionally, the use of mobile technologies to help address health and human resource shortages in low-resource settings is a recent and expanding area of research and has previously been evaluated for use in remote grading of ophthalmic disease in SSA [7].
HIV-associated eye diseases occur in up to 70% of HIV positive individuals [8]. Early recognition and treatment can prevent or cure 80% of cases [9]. Features observed in the retina may provide insight about the systemic health of individuals [10], such as in previous studies where ophthalmoscopy identi ed biomarkers of hypertension and its severity [11]. Vascular diseases are also common amongst older people living with HIV, and are often undertreated in SSA [12]. There appears to be a strong association between HIV and vasculopathies such as stroke [13], and renal disease [14] and retinal disease [15]. Retinal vascular changes has been shown to positively correlate with white matter microstructural changes observed using magnetic resonance imaging in non-HIV positive patients [10]. As white matter disease is also seen in both treated and untreated HIV cases and is associated with HAND[16, 17], we hypothesized that retinal imaging could help identify individuals at risk of HAND.
In this study we aimed to determine the clinical utility of HIV-retinopathy as a potential predictive biomarker of HIV-associated neurological complications in older cART-treated HIV positive individuals in Tanzania. A second objective was to determine if retinopathy could be a potential alternative to detailed neuropsychological assessment in a resource poor setting. In addition, we aimed to determine if remote retinal screening using a smartphone based retinal camera was a potentially useful strategy in this context.

Methods
A cross-sectional study at a single government-funded HIV clinic in Northern Tanzania. Participant selection Participants were recruited as part of a larger cross-sectional study of HAND and underwent detailed neurocognitive and clinical assessment. Criteria for inclusion were: ≥50-years of age, HIV-positive, attending routine follow up and not acutely unwell. A systematic sampling technique was employed with every second or third eligible participant approached for inclusion depending on daily capacity. All of those recruited within the larger study were invited to attend an additional appointment for ophthalmoscopic examination. Informed consent The purpose for the study and their speci c involvement was explained to each potential participant and written informed consent was obtained. Where capacity was in doubt, consent was obtained from close relatives. Ethical approval was obtained from the National Institute for Medical Research (NIMR/HQ/R.8a/Vol. IX/21.36) and Kilimanjaro Christian Medical University College Research Ethics Committee (n.896). This study abides by the tenets of the declaration of Helsinki. Assessment Baseline sociodemographic data of participant, plus HIV disease severity data from standard data sheets, were collected (Appendix 1). A set of functional assessments were used, including Karnofsky Performance Status[18], Rockwood Clinical Frailty Scale [19], and Intervention for Dementia in Elderly Africans (IDEA) low-literacy brief cognitive screen, previously validated for dementia in SSA [20]. Hypertensive and diabetic status were recorded. Visual acuity (Va) was assessed with available correction using the Landolt broken-ring low-literacy "C" chart with notations for testing at 3 meters due to unknown rates of illiteracy amongst the cohort. Visual elds were tested using confrontation (quadrant nger counting) and color vision with Ishihara charts. All participants underwent direct ophthalmoscopic examination following pupil dilatation (1% tropicamide) and retinal images were obtained using the iNview retinal camera (VOLK, Cleveland, USA). Images were independently reviewed by 3 ophthalmologists, 2 of whom were based in Tanzania (JR, LU) and 1 from the UK (JH). HAND assessment and diagnosis HAND were diagnosed by consensus using standard criteria [5] based on previously-obtained local normative values for neuropsychological battery measures, clinical/neurological examination ndings and a collateral history from a close relative or friend. All participants underwent a detailed low-literacy neuropsychological assessment battery, the details of which have previously been published [6]. Domains assessed included working memory, verbal memory (learning, delayed recall and recognition memory), ne motor control, motor speed, visuoconstruction, executive function and comprehension. Ophthalmologist criteria for de ning eye disease The de nitions for visual impairment used followed those given in the National Sensory Impairment Partnership's Framework [21]. Mild-moderate visual impairment was de ned as LogMAR 0.30.8. "glaucoma suspect" was termed if cup-to-disc ratio was >0.6 [22]. A Tanzania-based ophthalmology trainee (JR) determined whether glaucoma could be diagnosed based on ICD-10 criteria [22]. HIV retinopathy was assumed if any of the following features were observed upon ophthalmoscopy: retinal hemorrhages; microaneurysms; areas of capillary non-perfusion; or cotton wool spots. Statistical analysis All statistical analyses were performed using IBM SPSS software V24.0. Normally distributed data were described using mean, standard deviation (SD) and 95% con dence intervals (CI). Comparisons were made using independent t-tests. Nonnormally distributed data were described using median, interquartile range (IQR) and frequency and compared using Mann-Whitney U test. Categorical variables were described using frequency and compared using chisquared test and with the Fisher's exact test correction where appropriate. Logistic regression was used to evaluate potential relationships between outcomes and explanatory covariates. Statistical signi cance was de ned as P ≤ 0.05. The performance of the smartphone HIV retinopathy screen was investigated using Area Under the Receiver Operating Characteristic (AUROC) curve analysis. Diagnostic accuracy was analyzed for HAND and symptomatic HAND (s-HAND, MND/HAD). To determine diagnostic accuracy for HAND, ANI/MND/HAD were coded 1 and all others coded 0. For S-HAND, MND/HAD were coded 1, and all others 0.
Diagnostic accuracy was also analyzed for HIV stage 3 or 4 coded as 1 and all others coded 0 as well as analyzed for detectable viral load coded as 1 and undetectable as 0. When grouped together, the predictive value of HIV retinopathy to determine late HIV stage, detectable viral load as well as HAND and s-HAND was analyzed.

Population characteristics
Of the 762 individuals aged ≥50 years attending follow-up during the study period, 145 were systematically sampled and consented to inclusion. Of these, 129 completed the ophthalmologic assessment and had complete data (Figure 1). There was no signi cant difference in age, CD4 count or viral load between those who had complete data and those who did not.
Thirty-eight (29.5%) of patients were hypertensive in clinic (blood pressure (BP) >140/90 millimeter of mercury (mmHg). The mean BP in clinic was 134.5/81.2 mmHg with a maximum systolic of 219 mmHg. Seven (5.4%) patients reported being diabetic, 5 (71.4%) of whom were on treatment. HIV Retinopathy and its association with HAND HIV retinopathy features were signi cantly associated with both HAND diagnosis and HAND severity (Table 1).
HIV retinopathy was associated with patient-reported vision loss during the previous 3 months (χ 2 (1, n=129): 5.111; p=0.037). Using logistic regression, HAND severity and self-reported visual change were assessed for signi cance against HIV retinopathy. The signi cance intervals, odds ratios and con dence intervals are shown in Table 2. This shows HAND severity is associated with HIV retinopathy, adjusting for co-variates.  Although the relationship between HIV retinopathy and HAND is unclear, there are commonalities in the pathogeneses of both HIV-related complications which may explain their relationship. HIV retinopathy most likely occurs secondary to microvasculopathy from either immune complex deposition, increased plasma viscosity or vascular endothelium invasion by HIV, opportunistic infections or malignancy [23]. HAND occurs through a number of mechanisms including opportunistic central nervous system (CNS) infections, direct neurotoxic effect of the HIV virus and neurotoxic effect of cART [4]. Contributory factors include HIV disease severity, accelerated vascular disease and frequent co-morbidities [4].
The World Health Organization (WHO) advocates that screening for mental disorders be integrated into chronic disease monitoring to address the shortage of specialist personnel in low-resource settings [24]. Eye disease is currently a major cause of morbidity and disability in both HIV and non-HIV populations and disproportionately affects older people [25]. Cataracts, glaucoma, uncorrected refractive error and trauma are the major reported causes of visual impairment [25]. Identi cation and screening for ophthalmic disease in low-resource settings is currently challenged by similar issues to those noted for chronic complications of HIV. These include lack of specialist personnel, particularly in rural and remote areas, a focus on acute intervention rather than primary care and prevention, lack of integrated referral systems and access to affordable necessary equipment[26,27].
Using mobile technologies to help address health and human resource shortages in low-resource settings is a recent and expanding area of research and is increasingly feasible given the rapid increase in smartphone use across SSA [28]. Low-cost smartphone applications have been evaluated for use in remote grading of ophthalmic disease in SSA [7]. Non-clinical imagers were able to capture images at a standard that enabled remote grading at the level of a desktop retinal camera [7]. In some regions of SSA there is only one ophthalmologist per 2.5 million people, so remote retinal imaging could ameliorate the impact of understa ng [29].
In this study, we demonstrate a high prevalence of visual impairment in our cohort of HIV-positive patients, with 41.1% having mild-moderate visual impairment and 28.7% severe visual impairment. A previously published study reported that 11.2% of HIV-positive individuals aged 18 years or greater had a visual acuity of 0.2 or more in at least one eye [30]. One previous study in HIV-negative individuals aged 50 or greater report visual impairment of 13.6% in SSA [31]. Our prevalence was much higher. This difference may relate to our cohort being 50 years or older whereas other studies investigated younger cohorts. In addition, other explanations may include a risk of false positives from unidenti ed refractive causes of acuity loss, a lower threshold for diagnosis of visual impairment or our small sample size [30]. Factors relating to the willingness of patients to present with ophthalmic features may also contribute, perhaps due to lack of symptom knowledge, cost or impaired cognition. Since there is a high prevalence of eye disease in this cohort, it is prudent to screen for it.
We suspect the high prevalence of visual impairment may be seen also in the background local population, due to lack of routine eye tests and low patient presentation, even when high quality, affordable services are available [32]. Rates of visual impairment in the remote and rural areas may in fact be worse than those regularly attending hospital due to cost and availability of transport. In addition, low-cost retinal imaging may enable the effective monitoring of patients with HIV-associated ophthalmic diseases and could address challenges in healthcare provision relating to the limited availability of healthcare professionals in these regions [29].

Strengths and Limitations of the study
To our knowledge, this is the only study on HIV retinopathy in Tanzania, thus providing unique data on visual impairment and ocular disease prevalence. Previous studies looking at disease amongst HIV-infected individuals in SSA, excluded patients with confusion or altered mental state [30]. Since this study was able to assess the link between HIV ophthalmic diseases and HAND, this was an additional strength.
There are several limitations to this study. Images obtained from VOLK iNview retinal camera were of low-tomoderate quality so subtle changes or more complex pathologies may have been missed. Therefore, there is a substantial reliance on external referrals to ensure a correct diagnosis is made.
There are non-HIV causes of the fundal signs used to de ne suspected HIV retinopathy, such as hypertensive and diabetic retinopathy. A large proportion of our cohort had a blood pressure over 140mmHg, however 'whitecoat effect' is relatively common in SSA [33]. Seven of the 13 (53.8%) suspected HIV retinopathy patients had hypertension in clinic; we cannot exclude hypertensive retinopathy in these individuals. Despite individuals not self-reporting as diabetic and being under regular clinic follow-up, we found that 9 patients had glycosuria, a clinical sign of undiagnosed or untreated diabetes. 5 of these 9 had retinal signs thought to be consistent with HIV retinopathy but without measuring HbA1c, we cannot exclude diabetic retinopathy in these individuals.
This research only relates to HIV positive adults aged 50-years or older. This may be a strength in that studies of this older cART-treated population are few in SSA, but limits generalizability.

Conclusions And Recommendations For Future Work
Ocular pathology and HAND were both common in this patient population. We demonstrate that HIV retinopathy is associated with the presence of HAND. This has implications for clinical practice as retinal imaging may have 'rule in' value in predicting which individuals with HIV are likely to be at risk of cognitive impairment. Substantial clinical need for low-resource screening tools for HAND has been identi ed and although retinal screening cannot be used alone to determine who might need assessment for HAND, it may offer healthcare professionals a tool to enable patient prioritization.
As early diagnosis of ocular pathology is crucial to allow for timely intervention, portable and affordable tools which image the retina and enable remote diagnosis and monitoring of ophthalmic conditions in HIV-positive individuals. As technology develops, such assessments will become more accurate and useful.  Of the 762 individuals aged ≥50 years attending follow-up during the study period, 145 were systematically sampled and consented to inclusion. Of these, 129 completed the ophthalmologic assessment and had complete data (Figure 1).