Predictors of visual outcome in post-fever retinitis: a retrospective analysis

To identify the predictors of final visual outcome in cases with post-fever retinitis (PFR). This is a retrospective study of cases with diagnosis of post-fever retinitis. Colour fundus photograph and optical coherence tomography (OCT) parameters at presentation and final visit were analysed. Various factors at presentation [age, systemic illness, best-corrected visual acuity (BCVA), area of retinitis and hard exudates, OCT parameters], at final visit (OCT parameters) and the treatment modalities used were correlated with BCVA at final visit. Twenty-four eyes of 16 patients with PFR were included in the study. Median BCVA at presentation was 6/60 and at final visit was 6/9. By multiple linear regression after adjusting for other variables, for every 1 unit increase in height of subretinal fluid (SRF) at fovea at presentation, the value of final BCVA decreased by 0.001 unit. For every 1 unit increase in extent of ellipsoid zone (EZ) loss and subfoveal deposit height, the value of final BCVA decreased by 0.0001 unit and 0.004 unit, respectively. The baseline OCT parameters that had negative correlation with final BCVA included central macular thickness (r: − 0.5182, p: 0.02), maximum SRF height (r: − 0.5539, p < 0.01) and SRF height at fovea (r: − 0.582, p < 0.01). The OCT parameters at final visit which had a negative correlation with final BCVA included disorganisation of retinal inner layers (DRIL) within 1000 microns from centre of fovea (r: − 0.6494, p < 0.01), height of subfoveal deposit (r: − 0.7627, p < 0.01), horizontal extent of subfoveal deposit (r: − 0.6695, p < 0.01) and extent of EZ loss (r: − 0.8216, p < 0.01). Height of SRF at presentation, extent of EZ loss and subfoveal deposit height at final visit were associated with poor final BCVA in PFR.


Introduction
Post-fever retinitis (PFR) is a term used to describe the posterior segment manifestations following a febrile illness [1,2]. It is a relatively newer entity, described in the literature by various names like epidemic retinitis, post-febrile retinitis, acute multifocal retinitis and post-febrile uveitis [1][2][3]. Patients with PFR typically develop unilateral or bilateral retinitis after 2-4 weeks of febrile illness [4]. It is characterised by the presence of unifocal or multifocal retinitis, predominantly involving the posterior pole and peripapillary region. It is associated with macular oedema, neurosensory detachment and minimal vitritis. Vascular involvement is characterised by the presence of perivascular haemorrhages and perivascular yellow plaques [3]. These manifestations are hypothesised to be either immune mediated or due to the direct invasion of causative agent [3]. The treatment protocol is not established. The natural course of disease is believed to be self-limiting. However, few clinicians prefer to treat with systemic steroids and antibiotics [1,3]. The manifestations and course of PFR are described in the literature. Irrespective of the aetiology of fever and treatment modalities, the disease is reported to follow a natural course. So the final visual outcome is probably not related to the aetiology or the treatment. The studies evaluating the factors influencing the final visual outcome are lacking. In this retrospective study, the factors predicting the visual outcome in PFR were analysed.

Materials and methods
This was a retrospective study conducted in a tertiary eye care centre in south India. The study was approved by the Institute ethics committee on 4 December 2020 and adhered to the tenets of the declaration of Helsinki.
Patients diagnosed with post-fever retinitis between August 2018 and May 2020 were included. The diagnosis of post-fever retinitis was considered in patients presenting with acute onset unifocal or multifocal retinitis with or without vascular involvement following an episode of febrile illness within the last 6 weeks. Diagnosis of PFR was based on clinical examination and was aided by confirmation of macular thickening and subretinal fluid on optical coherence tomography (OCT). Age, systemic illness, the interval between the onset of visual symptoms and febrile illness were recorded from the medical records. Best-corrected visual acuity (BCVA) at presentation and final visit measured with Snellen's chart were recorded. Fundus photograph and OCT images captured with Topcon 3D OCT 2000 (Topcon Medical Systems, Japan) at presentation and at complete resolution were reviewed. '3D macula' and '6 mm radial 16 mm overlap' OCT protocols were used for imaging. Patients with regular follow-up with goodquality colour fundus photograph (45 degree) and OCT images centred on fovea at each visit up to complete resolution of retinitis were included. All the images were captured in a single machine with fixed protocol. 'Resolution' was defined as the absence of retinal whitening, intraretinal fluid and subretinal fluid. Patients lost to follow up and those with poor quality fundus or OCT images were excluded from the study. Details related to the treatment during the course of retinitis were recorded. Patients with retinitis involving the papillomacular bundle (PMB) and perifoveal area were treated initially with intravenous methyl prednisolone (IVMP) 1 g/day for 3 days followed by oral steroids. In rest of the patients, oral steroids were used as first-line therapy. Patients were treated with oral prednisolone 1 mg/kg and were tapered over 6 to 8 weeks. In cases with poor response to oral steroids, IVMP was considered as rescue therapy at 1-2-week follow-up. In cases with poor response to both oral and intravenous steroids, intravitreal triamcinolone acetonide was considered. Oral doxycycline 100 mg twice a day was administered for 2 weeks in few cases, based on treating ophthalmologist's preference.
Colour fundus photograph at presentation was analysed for the area of retinitis and area of hard exudates within the macula by Image J. Area of retinitis (retinal whitening) was marked on colour photograph by the inbuilt marker of Image J software and was calculated. Area of the circular 45-degree colour photograph was measured with Image J. As the image resolution can affect the measurements in Image J, the area was recorded as ratio of area of retinitis to the area of colour photograph to maintain uniformity (Fig. 1). Similarly, the area of hard exudates was measured as ratio. The presence or absence of perivascular plaque was documented.
OCT parameters at presentation were analysed using both 3D macula and 6 mm radial 16 mm overlap scan centred on fovea. The OCT parameters recorded at presentation and final visit are described in Tables 1 and 2, respectively. Figures 2 and 3 depict the measurement of OCT parameters.
Statistical analysis: Statistical analysis was carried out using the SPSS 25.0 version (SPSS, Chicago, IL, USA) software for windows. The data were analysed descriptively by mean, median and standard deviation. The OCT parameters with continuous variables at baseline and final follow-up were correlated with final BCVA by Pearson correlation. All the factors including age, systemic illness, interval between fever and retinitis, BCVA at presentation, area of retinitis and hard exudates at presentation, the OCT parameters at presentation and final visit and the mode of treatment were correlated with BCVA at final visit using univariate and multiple linear regression.

Results
The medical records revealed 28 patients with postfever retinitis during the study duration. Twenty-four eyes of 16 patients satisfied the inclusion criteria and were included in the study. Nine (56.25%) patients were men. The mean age at presentation was 44.38 ± 17.28 years (range 19 to 74 years). Mean interval between fever and onset of symptoms was 4.13 ± 2.13 weeks. Mean interval between onset of symptoms to presentation was 20 ± 18.91 days (range 5 to 90 days). Eight (50%) patients had bilateral involvement. Two patients had systemic hypertension and 2 patients had both systemic hypertension and diabetes mellitus. The aetiology of fever was evaluated during the febrile illness in 9 patients. Two patients were diagnosed with dengue fever, one with typhoid fever and aetiology of fever was not established in rest of the cases. Considering the less number of patients with established aetiology, statistical analysis of its significance on outcome was not possible.
At presentation, the median BCVA was 6/60 (range 1/60 to 6/6, mean 0.29 ± 0.35 in decimal notation). BCVA was B 6/60 in 13 eyes (54.16%). On fundus photograph, the mean area of retinitis within the macula was 0.20 ± 0.19 (range 0 to 0.63) and that of hard exudates was 0.015 ± 0.02 (range 0 to 0.10). Perivascular plaques were seen in 6 (25%) eyes. Five cases had periarterial plaques and 1 case had perivenous plaque. In all the cases, perivascular plaques were located within or adjacent to area of retinitis. None of the cases had optic disc oedema.
Eight patients were treated with oral steroids alone (prednisolone 1gm/kg/day tapered over 6-8 weeks); 8 were treated with IVMP for 3 days in addition to oral steroid therapy. One eye of a patient with bilateral PFR was treated with intravitreal triamcinolone in view of poor response to oral and intravenous steroids. Four patients with bilateral disease and 1 patient with unilateral disease received oral doxycycline 100 mg BD for 2 weeks in addition to systemic corticosteroids.
The median follow-up duration was 17 weeks (range 4 to 220 weeks). Fourteen eyes (58.3%) showed progressive resolution of retinitis with treatment. Initial progression of retinitis for 1-2 weeks followed by resolution was noted in 8 (33.3%) eyes and recurrence of retinitis after completion of steroid therapy was seen in 2 (8.3%) eyes. The median final BCVA was 6/9 (range 1/60 to 6/6, mean 0.59 ± 0.40 Fig. 1 a Measurement of area of retinitis from colour photograph using Image J marked with yellow outline. b Measurement of area of colour photograph using Image J marked with yellow outline in decimal notation). BCVA was B 6/60 in 6 (25%) eyes and better than 6/12 in 16 (66.6%) eyes.
There was no statistically significant correlation between age, gender, systemic illness, interval between fever and retinitis, area of retinitis, area of hard exudates, initial worsening of retinitis, different treatment modalities and final BCVA. The OCT parameters at presentation and its correlation with final BCVA are summarised in Table 3. The OCT parameters at final visit and its correlation with final BCVA are summarised in Table 4.
By Pearson correlation, the baseline OCT parameters that had negative correlation with final BCVA included central macular thickness (CMT) (r: -0.5182, p: 0.02), maximum SRF height (r: -0.5539, p \ 0.01) and SRF height at fovea (r: -0.582, p \ 0.01). The OCT parameters at final visit which correlated negatively with final BCVA included disorganisation of retinal inner layers (DRIL) within 1000 microns from centre of fovea (r: -0.6494, p \ 0.01), height of subfoveal deposit (r: -0.7627, p \ 0.01), horizontal extent of subfoveal deposit (r: 3. Horizontal extent of SRF was measured in all the 12 radial scans and the mean value was recorded. If the extent of SRF was beyond the scan length, it was considered as 6000 micron 4. Reflectivity of SRF was graded as  Table 3. By multiple linear regression, height of SRF at fovea had a statistically significant effect on final BCVA. After adjusting for other variables, for every 1 unit increase in height of SRF at fovea, the value of final BCVA decreased by 0.001 unit. By univariate linear regression, DRIL within 1000 micron of centre of fovea, presence of subfoveal deposit, height and length of subfoveal deposit, extent of ellipsoid zone (EZ) loss at fovea and presence of vitreoretinal (VR) interface abnormality had a statistically significant effect on final BCVA. Unadjusted coefficients with p value of univariate linear regression are reported in Table 4. By multiple linear regression, subfoveal deposit height and extent of EZ loss at fovea had a statistically significant effect on Description of the methods used to record the parameter Central macular thickness (CMT) As mentioned in Table 1 Disorganisation of retinal inner layers within 1000 microns from centre of fovea (DRIL-1000) As mentioned in Table 1 Horizontal extent and height of subfoveal deposits Subfoveal deposits were identified as hyperreflective material between neurosensory layer and RPE. Horizontal extent and height of subfoveal deposits was measured in a single well centred radial scan Of the 8 patients with bilateral disease, significant difference in the disease severity between the two eyes was noted in two patients. Fourfold difference in area of retinitis between the 2 eyes was noted in one case and threefold difference in other case. Correlation between the two eyes was significant in terms of presenting BCVA (r: 0.82, p = 0.01), area of retinitis (r: 0.86, p \ 0.01) and final EZ loss (r: 0.88, p \ 0.01), but the correlation was not significant in terms of height of SRF at fovea (r: 0.68, p = 0.06) and final BCVA (r: 0.49, p = 0.21).

Discussion
Post-fever retinitis predominantly involves the inner retina. Subretinal fluid and intraretinal oedema are characteristic features of PFR, which can be due to vasculitis and break down of inner blood retinal barrier (iBRB) by infective or inflammatory retinitis. Inflammatory mediators including vascular endothelial growth factor (VEGF) would play an important role in breakdown of iBRB. So the retinal damage in postfever retinitis might be due to neuro-retinal loss secondary to retinitis and loss of photoreceptors due to subretinal fluid. In this study, the parameters which assess the neuro-retinal involvement and exudation were analysed. Studies on diabetic retinopathy have noted that the accumulation of SRF and hard exudates are due to vascular leak with protein and lipid exudation secondary to breakdown of iBRB [5]. So OCT parameters such as altered foveal contour, CMT, intraretinal cystic spaces, hyperreflective foci secondary to hard exudates, height of subretinal fluid and its reflectivity at presentation are indicators of breakdown of iBRB. Parameters such as area of retinitis, IRH within 500 and 1000 micron from centre of fovea, PMB hyperreflectivity, DRIL within 1000 microns from centre of fovea at presentation are indicators of neuro-retinal involvement. In our study, we found a negative correlation between CMT at presentation and final BCVA. Few studies have analysed the relation between macular parameters and outcomes in intermediate uveitis. A study by Matas J et al. showed that the final visual acuity was negatively affected by the higher base line macular thickness [6]. Similarly, study by Markomichelakis N N et al. showed negative correlation between visual acuity and macular thickness [7].
In this study, multiple linear regression showed that the increase in SRF height was associated with  [8][9][10]. Hyperreflectivity of SRF at presentation was associated with poor visual outcome. The hyperreflectivity of SRF would indicate the presence of protein and lipid which can lead to development of subretinal deposits at resolution (discussed later). Hyperreflectivity of SRF would be an indicator of severity of disruption of iBRB. IRH within 500 microns of fovea and papillomacular bundle had statistically significant effect on final visual acuity. But the area of retinitis had a weak correlation with final BCVA. This would imply that retinitis closer to the fovea and papillomacular bundle has more detrimental effect than the overall extent of retinitis.
The parameters which would assess the severity of neuro-retinal and photoreceptor damage at resolution were also analysed. Parameters such as DRIL with-in1000 microns of centre of fovea and papillomacular bundle at resolution are indicators of the severity of neuro-retinal damage. Parameters such as presence of subfoveal deposits, its height and length, extent of EZ loss at fovea are indicators of severity of photoreceptor loss. This study showed that these parameters had statistically significant effect on final BCVA.
Multiple linear regression showed that final BCVA is poor in cases with higher subfoveal deposit height at resolution. On resolution of SRF, the protein and lipid may condense to form subfoveal deposits. In studies on diabetic macular oedema (DMO) and central serous chorioretinopathy (CSC), the presence of subfoveal deposits was associated with poor visual acuity [11,12]. They hypothesised that hindrance of interaction between the neurosensory retina and retinal pigment epithelium (RPE) due to the presence of the subfoveal deposits resulted in poor visual outcome. But the effect of subfoveal deposits in PFR is not described in the literature. Increase in length of subfoveal EZ loss showed poorer final visual acuity. Ellipsoid zone integrity reflects the presence of healthy photoreceptors. Shen et al. reported that the EZ integrity is more strongly correlated with BCVA [13]. The presence of intact EZ is a positive predictive factor for visual acuity in other retinal disorders like DMO, vein occlusion and CSC [14]. The disruption of EZ is considered to represent the cellular damage of photoreceptors. The final CMT did not correlate with the final BCVA. In PFR, the final CMT depends on two factors: the presence of subfoveal deposits which leads to increase in CMT, loss of retinal layers which leads to decrease in CMT. Both the factors result in poor visual outcome. This could explain the lack of correlation between final CMT and BCVA.
Progression of retinitis is considered to be a part of natural course of disease and can occur despite initiation of steroid therapy [3]. Initial progression was seen in about one third of cases in the present study, but it did not correlate with final BCVA. This could be due to small sample size. There was no difference in the final outcome between the cases treated with steroid monotherapy and combination of steroid and doxycycline. But considering the small sample size, it would not be possible to draw such conclusion. Similar results are reported in previous studies in the literature. A Study by Vishwanath S et al. showed that the patients treated with oral steroids showed improvement in visual acuity irrespective of aetiology [4]. Similarly, study by Kawali A et al. showed that there was no statistically significant difference in gain in visual acuity in patients treated with antivirals ? doxycycline ? steroids, doxycycline ? steroids, steroids alone, anti-VEGF ? steroids [2]. Route of steroid administration did not have an effect on final BCVA.
Half of the cases were unilateral, which indicates that PFR has differential effect on two eyes. But in cases with bilateral disease, the disease was symmetrical between the eyes, except for 2 cases. But, due to the small sample size, it would be difficult to arrive at a conclusion. This study gives a better understanding of the various factors predicting visual outcome in post-fever retinitis. Neurosensory detachment, hyperreflectivity of SRF, inner retinal hyperreflectivity, papillomacular bundle hyperreflectivity, DRIL close to fovea at initial presentation had poor visual outcome. DRIL close to fovea, presence of subfoveal deposits, extent of EZ loss, presence of VR interface abnormality at resolution had poor visual outcome. By multiple linear regression, height of SRF at fovea was the most important factor predicting the final BCVA. So, we hypothesise that treatment aimed at rapid resolution of SRF such as anti-VEGF may be considered in cases with more SRF [2]. But further studies would be required to test this hypothesis. The limitations of this study are that it is a retrospective study with small sample size and variable follow-up duration. The interval between the onset of symptoms and presentation was variable which might influence the baseline characteristics. The aetiology of fever was not established in all the cases and the effect of aetiology on the final outcome could not be established. Macular ischaemia can be one of the predictors of poor visual  Fig. 4b showing ellipsoid zone loss (arrow head) away from the centre of fovea outcome, but it was not evaluated in the study. It was not possible to study the course of improvement in terms of time required for resolution of retinitis or subretinal fluid, as the follow-up schedule varied from patient to patient. collection of data. Ravishankar HN conceptualised the study and involved in manuscript preparation. P Mahesh Shanmugam involved in critical analysis of results and manuscript preparation. Aanal Shah and Pradeep Tekade involved in collection of data. Vidya Mooss involved in manuscript preparation. All authors critically revised the paper.