Differing spectrum of HIV-associated ophthalmic disease among patients starting antiretroviral therapy in India and South Africa

Differences in the prevalence and spectrum of HIV-associated ophthalmic disease in Africa and Asia are not well documented. We studied two comparable cohorts of patients initiating antiretroviral therapy in Mumbai, India, and Cape Town, South Africa. The prevalence of HIV-associated ophthalmic disease was higher in the Indian population (17.5%) than in the South African population (12.1%). This was largely because of vitreo-retinal opportunistic infections (11.4%vs. 2.6%, respectively), notably cytomegalovirus retinitis. This difference persisted after adjusting for confounding factors (adjusted odds ratio = 11.32, 95% confidence interval: 2.67–48.13), confirming a marked geographical difference in the prevalence of HIV-associated retinal disease.

HIV-associated eye disease is estimated to affect 50-75% of HIV-infected people worldwide at some point during their disease course (Kestelyn & Cunningham 2001). In resource-limited settings, patients typically start antiretroviral therapy (ART) at low CD4 counts and may be at high risk of ocular opportunistic infections (OIs). Retinal OIs such as cytomegalovirus retinitis (CMVR) can lead to chronic visual impairment or blindness if untreated. Additionally, immune reconstitution phenomena such as immune recovery uveitis after ART initiation may increase ocular morbidity (Karavellas et al. 1998;Price et al. 2001). Identification and treatment of ocular disease in HIVinfected individuals is therefore important.
Reports suggest that the prevalence and spectrum of HIV-related eye disease differs geographically. In Asia, the predominant manifestation is CMVR, followed by other retinal OIs. Estimates are highest in south-east Asia, where CMVR affects 27-33% of HIV-infected individuals (Biswas et al. 2000;Ausayakhun et al. 2003;Gharai et al. 2008;Shah et al. 2008;Pathai et al. 2009). In contrast, the reported prevalence of CMVR in sub-Saharan Africa is low (0-16.5%), whereas complications affecting the anterior segment of the eye such as herpes zoster and molluscum contagiosum are more frequently reported (Lewallen et al. 1994;Jaffar et al. 1999;Kestelyn 1999). However, it is difficult to draw firm conclusions about these differences as study populations have differed with regard to levels of immunodeficiency, ART exposure and methods of case ascertainment.
We previously reported on the prevalence of ophthalmic disease among patients enrolling in an ART service in Mumbai, India (Pathai et al. 2009), and have since implemented the same study protocol, using the same study personnel, in a comparable cohort of patients starting treatment in a well-characterised ART service in Cape Town, South Africa (Lawn et al. 2006(Lawn et al. , 2007. The aim of this paper was to directly compare the profile of HIV ocular manifestations in these two populations. Data collection comprised demographic information, standardised ocular symptom screening and a full *Re-use of this article is permitted in accordance with the Terms and Conditions set out at http://wileyonlinelibrary.com/ onlineopen#OnlineOpen_Terms. ophthalmic examination among patients starting ART as previously detailed (Pathai et al. 2009). Prevalence of HIV ocular disease was estimated, and factors associated with disease were compared between sites using the chi-squared test. Multiple logistic regression analysis was used to examine associations between retinal OIs and demographic ⁄ clinical characteristics. Three hundred and six patients were studied (157 in Cape Town and 149 in Mumbai). Patients from the African and Indian study sites had similar demographic and clinical characteristics although the Cape Town cohort had a lower median CD4 cell count (Table 1). The prevalence of overall HIV-related eye disease was higher in the Mumbai (17.5%) than in the Cape Town cohort (12.1%; Table 2). The evidence for a difference was strongest among those with CD4 count <200 cells ⁄ ll and those with advanced WHO clinical stage. When stratifying by specific diagnoses, it was clear that the overall difference was largely because of the higher prevalence of retinal OIs in the Mumbai cohort, specifically CMVR ( Table 2).
The difference in the risk of retinal OI between sites persisted after adjusting for age, gender and other a priori confounders (CD4 count, WHO clinical stage, visual impairment and ocular symptoms), with an adjusted odds ratio = 11.32, 95% confidence interval (CI): 2.67-48.13 (Table 3). In addition, low CD4 cell counts and advanced WHO clinical stages of disease were also independently associated with the prevalence of retinal OIs (Table 1).
Our data confirm that the spectrum of HIV-related eye disease differs between these two study sites in India and Africa. The use of identical study protocols and study personnel, as well as demographically and clinically similar study populations, suggests that these findings can be reliably compared. The low prevalence of retinal OIs within the Cape Town cohort (2.6%) is consistent with other data from the region (Lewallen et al. 1994;Jaffar et al. 1999;Kestelyn 1999). Similarly, our prevalence estimate of CMVR in Mumbai is in agreement with other reports from Asia (Biswas et al. 2000;Ausayakhun et al. 2003;Gharai et al. 2008;Shah et al. 2008).
A previously proposed hypothesis for this difference is that patients with advanced immunodeficiency in the African region have high mortality and do not survive to develop CMVR and other complications (Jaffar et al. 1999). However, this would not explain our findings whereby the geographical location was a strong independent predictor of the prevalence of ophthalmic disease after adjustment for CD4 cell count and WHO stage of disease. Previous studies have shown that CMV seroprevalence is similar in the two populations (Kothari et al. 2002;Cannon et al. 2010;Chakraborty et al. 2010;Rabenau et al. 2010), suggesting lack of exposure to CMV within the Cape Town cohort is an unlikely reason for this difference. The disparities in prevalence are unclear but could potentially be related to genetic differences at the host or virus level (Jabs 1995;Jaffar et al. 1999). There are no reports to date of geographical differences in the epidemiology of non-ocular systemic CMV disease, possibly suggesting that host ophthalmic factors may also be involved. The prevalence of HIV retinopathy was similar in both groups, and these non-infectious retino-vascular pathological processes may occur independently of CMVmediated mechanisms.
In conclusion, the higher prevalence of HIV-associated ophthalmic disease in the Indian study population was largely because of the higher prevalence of CMVR. Further studies are needed to document other geographical variations in ocular HIV epidemiology and investigate the underlying mechanisms. Most cases of CMVR in the Indian population were not detectable with symptom screening (Pathai et al. 2009), and this indicates the need for careful retinal screening in this patient population in