The numbers of patients recruited and examined at 34th Regiment Military Hospital were 82 EVD survivors (161 eyes; 2 missing retina images and 1 prosthetic eye) and 105 never-infected controls (208 eyes; 2 missing retinal images). Male-to-female ratio was 1:1.48 of EVD survivors and 1:0.64 of controls. Median age at time of ophthalmic examination was 28 years (interquartile range [IQR] 22–38 years) for EVD survivors and 41 years (IQR 30–48 years) for controls. Median time from Ebola treatment unit discharge to ophthalmic examination for survivors was 411 days (n = 70) (IQR 368–470 days). Ophthalmic examination findings were summarized for survivors and controls (Table).

Figure 1

Figure 1. Prevalence of retinal scar lesion types in a case–control study of ocular signs in Ebola virus disease survivors, Sierra Leone, 2016. Type 1, uniform pigmented lesion; type 2, uniform pigmented lesion...

We subclassified pigmented and nonpigmented retinal lesions into 10 discrete groups (Technical Appendix 1) and noted frequency of each lesion type (Figure 1). We found no occurrences of the retinal lesion documented in a previous case report (9) in this EVD survivor cohort. Only the type 6 subcategory of retinal lesion was observed exclusively in EVD survivors, occurring in 12/82 (14.6% [97.5% CI 7.1%–25.6%]) EVD survivors and 0/105 controls (0% [97.5% CI 0%–4.1%]) (p<0.01). In 50% of EVD survivors, this type of lesion was observed bilaterally.

Figure 2

Figure 2. Composite scanning laser ophthalmoscope retinal images showing type 6 Ebola peripapillary and peripheral lesions, observed following the anatomic distribution of the ganglion cell axons (retinal nerve fiber layer), in a case–control...

Figure 3

Figure 3. Characteristic features of lesions observed in a case–control study of ocular signs in Ebola virus disease survivors, Sierra Leone, 2016. A) Composite scanning laser ophthalmoscope retinal image, left eye. Arrow indicates...

Two fundal distributions of type 6 lesions were evident: isolated or multifocal lesions in the peripheral retina or peripapillary lesions observed emanating from the optic disc (Figure 2). Each lesion shape was variable but often exhibited characteristic sharp angulations, resembling a diamond or wedge (Figure 3). Surrounding these lesions was a well-demarcated area of darkened retina in comparison with the adjacent retina. Presence of any retinal lesions of types 1–10, excluding type 6, were observed in 21/82 (25.6% [97.5% CI 15.5%–38%]) EVD survivors and 25/105 (23.8% [97.5% CI 15.1%–34.4%]) controls.

The aqueous humor of 2 EVD survivors with white cataract and no anterior chamber inflammation was negative for EBOV RNA on qRT-PCR assay. Postprocedure conjunctival swabs also were negative. The aqueous humor sampling procedure was uncomplicated and well-tolerated. No complications were reported on follow up.

This case–control study identified a novel retinal sign that appears to be specific to EVD survivors. This sign occurred among a local population with a high rate of background chorioretinal disease. Uveitis after EVD has been reported (3,8), and a recent case report included a published fundus image from a survivor with a chorioretinal lesion attributed to EVD (9). That patient went on to have panuveitis.

The retinal lesions specific to EVD survivors were located either adjacent to the optic disc or in the fundus periphery. In the 8 cases in which lesions appear adjacent to the optic disc, their curvilinear projections from the disc margin appear to align with the anatomic pathways of the retinal ganglion cell axons that constitute the optic nerve. This distribution suggests a neurotrophic spread into the eye from the optic nerve and along the retinal ganglion cell axons. The other possible mode of entry into the eye is hematologic. Although the retinal ganglion cell axons often have parallel curvatures around the retinal arcade vessels, the lesions clearly follow the nerve fiber distribution in the absence of major vessels (Figure 1, panels A and C). Furthermore, we have not found any signs suggestive of associated vascular involvement, such as vasculitis, vascular occlusions, retinal ischemia, or secondary neovascularization, to support a hematologic spread. Neurotrophic properties are increasingly being recognized in EBOV (18). West Nile virus disease, caused by a known neurotropic virus, is associated with retinal lesions that follow a similar pattern of distribution to the pattern we have observed in our study (19).

Each Ebola lesion shape is variable, but a characteristic angulated appearance often resembling a diamond or wedge shape appears unique (Figure 2). As far as we are aware, the appearance of these lesions is not characteristic of any other retinal disease. The reason for the sharp angulated appearance of these lesions might be explained by the tight triangular packing of the retinal cone mosaic (Figure 3, panel D) (17,20). The regular pattern of the photoreceptor triangular mosaic is disrupted by larger blue cones (17) and diminishes with eccentricity (20), which might explain the variability in shape. Optical coherence tomography indicates that these lesions are limited to the retina (Figure 3, panel B), and the resemblance of the lesion shape to the photoreceptor mosaic suggests that the ganglion cell axons act merely as a means of transportation to the photoreceptor end target.

Despite the proximity of the lesions to the optic nerve head, we observed no optic nerve head swelling or pallor in our study. This fact is in contrast to the 10% of optic nerve swelling reported in 1 abstract (5), although the time from acute infection to ophthalmic examination in that case was not stated. This difference might be attributable to varying durations since acute infection, allowing for any potential disc swelling to resolve in our cohort, for whom the median time since discharge was 411 days. Further optic nerve functional assessment, such as visual field analysis or color vision testing, has yet to be conducted.

The Ebola retinal lesions did not affect visual acuity. Overall, no difference was observed in uncorrected visual acuity between EVD survivors and controls. The most common cause of visual impairment in EVD survivors was white cataract (7.3%), which was accompanied by hypotony (low intraocular pressure) in 80% of EVD survivors. Hypotony suggests inadequate aqueous humor production and can limit the visual potential of an eye through complications such as retinal folds at the macula (i.e., hypotensive maculopathy).

Concern exists about the safety of cataract surgery in EVD survivors in Sierra Leone because of the unknown duration of EBOV ocular persistence. A sample size of 2 negative aqueous humor samples in this study is too small to make any definitive conclusions but shows that EBOV does not necessarily persist in aqueous humor in those with cataract but no ongoing intraocular inflammation. This finding suggests that cataract surgery can be conducted safely, providing an opportunity to restore vision and remove the stigma of EVD survivor status associated with having a visible white cataract. At present, we would recommend that anterior chamber sampling with EBOV PCR and a negative result should precede cataract surgery. However, cataract surgery might be challenging and visual outcomes disappointing in cases of secondary hypotony, which occurred in 80% of EVD survivors.

Before this study, only 1 aqueous humor sample had been obtained in an EVD survivor (9), enabling the detection of viable EBOV in aqueous humor during acute uveitis 9 weeks after discharge from hospital (9). Virus persistence in aqueous humor has also been observed in uveitis after Marburg virus infection (21), becoming negative on being repeated at 10 weeks (22). In EVD and Marburg virus–associated uveities, intraocular pressure was markedly elevated (9,21). Although Ebola-related acute uveitis has been reported to be associated with high intraocular pressure, we did not find any evidence of persistently high intraocular pressure in survivors with Ebola retinal lesions.

Uveitis accounts for 24% of blindness in Sierra Leone and is second only to cataracts as the leading cause (23). A proportion of those cataracts might be a consequence of intraocular inflammation, especially in younger patients. Given the high endemic rates of parasitic, viral, and fungal disease in the region, infectious uveitis is likely to have a higher prevalence than in Western populations (24). Nevertheless, the proportion of controls with chorioretinal lesions and retinal vasculitis was surprising. Pigmented and atrophic chorioretinal scars not in keeping with the Ebola retinal lesions were no more common in EVD survivors than controls, and it is important not to attribute these findings to EBOV infection in survivors documented in case series (7,12).

The leading cause of uveitis in Sierra Leone is onchocerciasis, but this disease is in decline because of the systematic distribution of ivermectin to affected areas (25,26). The rate of other uveitis-associated blindness appears to be increasing in Sierra Leone (23). This study was conducted in Freetown, where the incidence of onchocerciasis is lower than in rural regions, and other causes are probably responsible. Toxoplasmosis accounted for 43% of symptomatic cases of posterior uveitis in 1 study (27), and it was probably a common cause among the patients in our study, although no serologic testing for toxoplasmosis was available. HIV prevalence in persons >15 years of age in Sierra Leone was estimated to be 1.25% in 2015 (28). The Ebola outbreak disrupted the fragile health system, including HIV reporting mechanisms and AIDS response (29). This HIV rate is still relatively low compared with many other African nations. Further diagnostic investigation is required to attempt to attribute causation to the various chorioretinal lesions observed in this study. Geographic areas of retinal whitening (white without pressure) are thought to be normal variants (30,31). Areas of retinal darkening (dark without pressure) have previously been attributed to sickle cell disease (32).

Our study is subject to 1 limitation with regard to the control group, who were selected opportunistically with unmatched cases and controls, and differences in age and sex ratios between the groups. This fact reflects the difficulties and limitations of conducting research in the post-Ebola setting in Freetown in 2016. The study was conducted in a military hospital, which housed the Ebola treatment unit and the continuing EVD survivors clinic. The hospital also serves the local civilian community and a military barracks community. The use of a non-EVD control group, even without matching, allowed a comparison in the fundus findings between post-EVD and control groups. We found a higher prevalence of retinal disease in the symptomatic clinic-attending control group than in the asymptomatic population control group; both groups included some military members of staff and families. This comparison allows us to be more positive about the specificity of the Ebola retinal lesion. Given our aim to compare EVD with non-EVD fundus findings, an age- and sex-matched population control group probably would not change the study conclusions.

EVD survivors were identified by the possession of an Ebola treatment center discharge certificate. Forgery of these certificates has been known given the free access to healthcare it confers. IgG confirmation of previous EBOV infection is planned for ongoing follow-up studies. Our study provides information on the medium-term ocular sequelae of EVD survivors with a median time of 411 days since hospital discharge. Our study does not provide data on acute uveitis and ocular disease in the immediate aftermath of EVD as reported elsewhere (2,4,6).

Although we can reasonably conclude the retinal lesions described in our study are sequelae of EVD, no pre-EVD retinal imaging was available to conclusively identify the timing of acquisition of the lesions. Our control group demonstrates the common retinal signs and pathologies that are present in the population before Ebola exposure.

We have documented a novel retinal abnormality in EVD survivors that appears to be specific to EVD, although the proportion in the cohort with the condition is small. The background prevalence of chorioretinal abnormalities, including scarring with pigmentation, in the population is high and should not be attributed to EVD. Although further studies with larger sample sizes are required, EBOV does not necessarily persist in the aqueous humor of those with cataracts and no ongoing intraocular inflammation. These initial results raise the possibility of safe cataract surgery for EVD survivors with no signs of ongoing intraocular inflammation.