Longitudinal Assessment of Ellipsoid Zone Mapping Parameters in Retinal Venous Occlusive Disease With Associated Macular Edema

Abstract

Purpose:

This work evaluated the longitudinal dynamics of ellipsoid zone (EZ) integrity in retinal vein occlusion (RVO) with macular edema and their relation to outcomes.

Methods:

Clinical characteristics and optical coherence tomography data of patients with RVO and associated macular edema were collected at baseline and at 3 and 12 months. Macular cube scans were exported into EZ and retinal-layer analysis software. Longitudinal EZ parameters and visual acuity (VA) outcomes were regressed and correlated.

Results:

The study included 108 eyes of 108 patients; all eyes were treated with antivascular endothelial growth factor therapy at the baseline visit. VA improved from 20/97 at baseline to 20/52 at 3 months and 12 months (P < .001), correlating with EZ integrity at each time point (P < .001). At 12 months following initiation of antivascular endothelial growth factor therapy, EZ partial attenuation and EZ total attenuation improved over 12 months from 16.4% to 8.5% (P < .001) and from 12.3% to 5.9% (P < .001), respectively. VA improvement from baseline to 12 months correlated with improvement of EZ partial and total attenuation (P < .001). Baseline EZ characteristics did not predict VA outcomes, but at 3 months, EZ parameters did predict improvement in visual outcomes by 12 months (P < .01).

Conclusions:

EZ and outer retinal integrity are correlated with functional outcomes in RVO. Following treatment, EZ integrity improves and is associated with functional improvement. In RVO baseline, EZ features were not associated with 1-year VA outcomes, but evaluation of EZ integrity at 3 months was linked to 1-year outcomes.

Introduction

Retinal vein occlusions (RVOs) are the second most common retinal vascular disease and include central RVOs (CRVOs), branch RVOs (BRVOs), and hemiretinal RVOs (HRVOs). Secondary complications of RVOs include macular edema (ME) and less commonly neovascularization of the retina or anterior segment. ^{1 -3} Vascular endothelial growth factor (VEGF) has been demonstrated to be significantly elevated in RVO and is a major driver of ME and neovascularization. Treatment with intravitreal VEGF inhibitors has become the gold-standard treatment of RVO management. ^2,4 However, there is a spectrum of improvement in visual acuity (VA) that remains difficult to predict. ^5,6

Optical coherence tomography (OCT) has revolutionized the ability to noninvasively diagnose, monitor, and evaluate treatment in macular-involving retinal pathologies. ⁷ OCT has been used extensively in the investigation of RVO, including evaluating changes in central subfield thickness (CST), intraretinal and subretinal fluid (SRF), photoreceptor damage, ellipsoid zone (EZ) integrity, and disorganization of inner retinal layers (DRIL). ^{8 -11} The EZ is 1 of 4 outer retinal hyperreflective bands. It is the second-most inner band just posterior to the external limiting membrane. ¹² Research has suggested that the EZ on OCT corresponds to a high density of mitochondria in the photoreceptor inner segment ellipsoid. ¹²

EZ integrity has been strongly linked to VA in a number of retinal diseases; however, quantitative assessment of EZ integrity has been challenging. One previous study found potential prognostic value of baseline EZ using qualitative assessment. ¹³ Our group has previously demonstrated the feasibility of EZ mapping and quantitative assessment of EZ integrity in RVO that also confirmed the association of EZ integrity metrics with presenting VA. ¹⁴ This methodology enables a more robust characterization of EZ integrity using panmacular assessment rather than reliance on a single or small number of OCT B-scans for evaluation.

The purpose of this study is to evaluate longitudinal quantitative EZ integrity dynamics using a novel EZ mapping analysis platform and to correlate the impact of EZ dynamics on functional outcomes in eyes with ME secondary to RVO treated with anti-VEGF therapy. In addition, this study evaluates whether baseline or 3-month OCT EZ parameters can prognosticate long-term (ie, 12-month) visual outcomes in this same population.

Methods

Participants

Inclusion criteria were patients with a diagnosis of CRVO, BRVO, or HRVO, presence of ME, and 3- and 12-month follow-up visits. Exclusion criteria included age younger than 18 years, presence of other macular disease aside from RVO, lack of macular cube OCT scans at any time point (baseline, 3-month, or 12-month visit), or OCT imaging of limited quality or signal strength that impaired software analysis.

Clinical Treatment and Data

Patients were treated at the initial encounter with an intravitreal anti-VEGF injection. For all follow-up visits, the treatment decision was at the discretion of the retina specialist based on standard of care. Patient demographics, clinical characteristics, and spectral-domain OCT data were collected including age, sex, VA, CST measurement, and type and number of anti-VEGF injections.

Panmacular Quantitative Assessment

All eyes underwent macular spectral-domain OCT (Cirrus, Zeiss) at baseline and each evaluated follow-up interval (3-month and 12-month) for this analysis. Macular cube scans (6 mm × 6 mm) were exported into a retinal layer–segmentation and EZ-mapping software platform. Following automated analysis, each frame was reviewed for segmentation accuracy by a trained reader and corrected as needed. ^14,15 After verification of optimal segmentation, multiple EZ parameters were exported for analysis. EZ integrity metrics included percentage of macular cube en face area with EZ–retinal pigment epithelium (RPE) partial attenuation (EZ-RPE thickness ≤ 20 µm), percentage of macula area with EZ-RPE total attenuation (EZ-RPE thickness = 0 µm), EZ-RPE macular volume, and EZ-RPE CST were evaluated (Figure 1). The percentage of partial attenuation and total attenuation represents the macular cube A-scans that had a measurement for the EZ-RPE thickness that met criteria for a given attenuation. For example, if there was 5% total attenuation measured, then 5% of all A-scans represented a 0-µm EZ-RPE thickness within the macular cube. These EZ mapping parameters were analyzed for correlation with functional outcomes based on VA.

Figure 1.

Ellipsoid zone (EZ)–to–retinal pigment epithelium (RPE) thickness maps. The color scales adjacent to the thickness maps correspond to the distance between the EZ and RPE at any given point, with pink representing atrophy (EZ-RPE distance = 0 μm) and purple representing partial attenuation (EZ-RPE distance ≤ 20 μm). A normal retina is represented (A) alongside 2 maps from a retinal vein occlusion patient at (B) baseline and (C) follow-up, demonstrating longitudinal improvement in EZ integrity.

Statistical Analysis

VA was converted to logMAR for statistical analysis. Variables with continuous measures were evaluated using Pearson correlations, whereas Wilcoxon rank sum and Kruskal-Wallis tests were used for categorical factors. Analysis was performed using SAS software (version 9.4, SAS Institute) and JMP Pro 14 (SAS Institute). Multiple variable linear regression analyses evaluated the effect of baseline and 3-month OCT parameters on change in visual outcome between baseline and the final 12-month time point. To account for variation in extent of ME, particularly between the 0- to 3-month periods, CST at baseline and change in CST from baseline to 3 months were included in all models of analysis. In addition, baseline VA was included in all models given its known significant contribution to changes in outcome. Multicollinearity, correlation between parameters that can lead to model overfitting, between EZ variables was evaluated using variation inflation factor testing, and no models were run with variation inflation factor scores exceeding 4 for any predictor.

Results

Patient Characteristics

A total of 108 eyes of 108 patients with RVO were included in the study: There were 49 (45.4%) cases of CRVO, 48 (44.4%) cases of BRVO, and 11 (10.2%) cases of HRVO. Mean age was 70 years (range, 26-93 years). In total, 61 (56.5%) were female. The mean follow-up time between baseline and 3-month visits was 99 ± 16 days, and 371 ± 39 days for the 12-month visit. All eyes were treated with anti-VEGF therapy at presentation (bevacizumab 77.8%, ranibizumab 19.4%, and aflibercept 2.8%). Across 12 months of therapy, the mean number of injections was 6 (range, 1-12). Within the study duration, 1 patient developed neovascularization requiring panretinal photocoagulation, and one other patient received macular laser. Demographic and baseline characteristics are presented in Table 1.

Table 1.

Baseline Demographic and Ocular Characteristics.

Age ± SD, y	70.3 ± 11.9
Visual acuity, logMAR ± SD (Snellen equivalent)	0.68 ± 0.45 (20/100)
CST from standard OCT, µm ± SD	469 ± 166
Sex, No. (%)
Male	47 (43.5)
Female	61 (56.5)
Eye, No. (%)
Left	56 (51.9)
Right	52 (48.1)
Type of RVO, No. (%)
CRVO	49 (45.4)
BRVO	48 (44.4)
HRVO	11 (10.2)
Baseline injection, No. (%)
Bevacizumab	84 (77.8)
Ranibizumab	21 (19.4)
Aflibercept	3 (2.8)

Abbreviations: BRVO, branch retinal vein occlusion; CRVO, central retinal vein occlusion; CST, central subfield thickness; HRVO, hemiretinal vein occlusion; OCT, optical coherence tomography; RVO, retinal vein occlusion.

Traditional Clinical Outcomes

VA improved from 20/97 at baseline (range, 20/20 to 20/4000) to 20/52 (range, 20/20 to 20/2666) at 3 months (P < .001) and remained at 20/52 (range, 20/20 to 20/1000) at 12 months (P < .001). Mean CST obtained by standard OCT was 469 µm at baseline and significantly decreased both at the 3-month follow-up (354 µm, P < .001) and 12-month follow-up (323 µm, P < .001).

Longitudinal EZ Integrity

Following initiation of anti-VEGF therapy, EZ partial attenuation and EZ total attenuation across the macular surface area improved from 16.4% to 8.5% (baseline to month 12, P < .001) and from 12.3% to 5.9% (baseline to month 12, P < .001), respectively (Figure 2). Over the 12-month follow-up, mean EZ-RPE CST increased from 16.3 µm to 27.1 µm (P < .001) and EZ-RPE volume increased from 1.044 mm³ to 1.134 mm³ (P < .001). In 19 of the 108 patients, VA decreased from baseline to 12 months (P < .05). In this subgroup of anti-VEGF nonresponders, EZ metrics including partial attenuation, total attenuation, EZ-RPE volume, and EZ-RPE central foveal thickness (CFT) were not statistically different at 12 months from baseline (P > .05).

Figure 2.

Optical coherence tomography B-scans and representative posttherapeutic ellipsoid zone (EZ)–to–retinal pigment epithelium (RPE) thickness maps at (top row) baseline and (lower row) posttreatment showing EZ improvement (A) without fluid resolution and (B) with resolving fluid with comparatively minimal EZ change. On the B-scans, the internal limiting membrane is segmented in dark blue, the EZ in yellow, and the RPE in light blue. The orange lines demonstrate 2 parameters, point thicknesses from the internal limiting membrane to the RPE and from the EZ to the RPE.

Correlation and Regression Analysis

At all given time points, VA was correlated with the associated time point’s EZ integrity, reflected by logMAR VA’s correlation with EZ volume (baseline r = –0.61, P < .001; 3-month r = –0.28, P = .003; 12-month r = –0.52, P < .001), EZ partial attenuation (baseline r = 0.70, P < .001; 3-month r = 0.46, P < .001; 12-month r = 0.62, P < .001), EZ total attenuation (baseline r = 0.63, P < .001; 3-month r = 0.55, P < .001; 12- month r = 0.62, P < .001), and EZ central foveal area (baseline r = –0.61, P < .001; 3-month r = –0.43, P < .001; 12-month r = –0.63, P < .001). VA improvement from baseline to 12 months was significantly correlated with improvement of EZ volume (r = 0.38, P < .001), EZ partial attenuation (r = 0.47, P < .001), total attenuation (r = 0.49, P < .001), as well as an increase in central foveal area (r = 0.39, P < .001). Change in mean EZ-RPE CST was also significantly associated with VA improvement (r = 0.43, P < .001). A comparison of EZ characteristics is shown in Table 2. Baseline VA (r = 0.46) and 3-month VA (r = 0.67) both correlated with 12-month VA (P < .001).

Table 2.

Visual Acuity and Optical Coherence Tomography Parameters at Baseline, 3 Months, and 12 Months. P Values Were Calculated By Comparing Follow-up Time Point Parameters to Baseline.

Parameter	Baseline	3-month	P	12-month	P
Visual acuity (range)	20/92 (20/20-20/4000)	20/52 (20/20-20/2666)	< .001	20/52 (20/20-20/1000)	< .001
CST from standard OCT, µm ± SD	469 ± 166	354 ± 152	< .001	323 ± 117	< .001
Partial attenuation (EZ-RPE thickness ≤ 20 µm), % ± SD	16.4 ± 17.3	7.4 ± 10.8	< .001	8.5 ± 16.6	< .001
Total attenuation (EZ-RPE thickness = 0 µm), % ± SD	12.3 ± 16.3	4.6 ± 7.0	< .001	5.9 ± 13.4	< .001
Central foveal thickness, µm ± SD	24.7 ± 8.2	30.3 ± 7.2	< .001	29.4 ± 7.4	< .001
Mean EZ-RPE CST, µm ± SD	16.3 ± 15.4	27.2 ± 12.9	< .001	27.1 ± 12.5	< .001
EZ-RPE volume, mm3 ± SD	1.038 ± 0.207	1.140 ± 0.149	< .001	1.102 ± 0.207	.02

Abbreviations: CST, central subfield thickness; EZ, ellipsoid zone; OCT, optical coherence tomography; RPE, retinal pigment epithelium.

Regression models, controlling for age, baseline VA, and intraretinal fluid, evaluated individual EZ metrics separately because of multicollinearity to predict change in VA from baseline and at 3 months and 12 months. Baseline VA was controlled for in all models given its strong correlation with VA improvements and reflection of baseline disease (P < .001). No baseline EZ parameter predicted change in VA at the 12-month time point. However, several EZ parameters at 3 months including EZ volume (R ² = 0.50, P = .02), EZ attenuation (R ² = 0.52, P < .001), EZ atrophy (R ² = 0.51, P = .006), and EZ-RPE CFT (R ² = 0.52, P = .003) did predict improvement in visual outcomes by 12 months when controlling for age, baseline VA, and intraretinal fluid. EZ-RPE mean CST trended toward significance for predicting visual improvement (P = .07).

Conclusions

This study reported the longitudinal dynamics of EZ integrity following initiation of anti-VEGF therapy in RVO, specifically the efficacy of EZ parameters in both prognosticating functional outcomes in RVOs and their time point–specific correlations. Final EZ and outer retinal integrity were directly correlated with final functional outcomes in RVO-associated ME undergoing longitudinal anti-VEGF therapy. The most significant predictor for final visual outcome was baseline visual function, which previous research had highlighted as a significant contributor. ^6,13,16 Following anti-VEGF therapy, VA improvement significantly correlated with improvement of EZ partial attenuation, EZ total attenuation, and improved central foveal EZ-RPE area and volume. In regression analysis, no baseline EZ characteristic predicted functional visual improvements at 12 months. However, EZ parameters measured at month 3, including EZ attenuation and EZ atrophy, did predict 1-year improvements in best-corrected VA from baseline. These results affirm the utility of EZ integrity as a biomarker for functional improvement at the 3-month time point in RVOs.

The cause and progression of EZ disruption in RVOs has yet to be comprehensively understood, although acute interphotoceptor edema may play a role, while chronic degeneration may arise from continuing ischemia and/or edema. ^6,10 At baseline this study showed substantial EZ attenuation and atrophy (see Table 2). Longitudinally, the EZ on average recovered significantly after initiation of anti-VEGF injections, with EZ attenuation and atrophy both dropping by more than 50% at 3 months posttherapy and remaining stable until 12 months (see Table 2). These improvements were assessed across the macula including in the central subfield, where EZ-RPE thickness increased by more than 50% by 3-month follow-up.

Prior literature has consistently identified EZ integrity as predictive of VA at a specific time point across numerous diseases; our results replicated these findings at all time points (ie, baseline EZ integrity correlated with baseline VA). ^12,14,15 However, in the case of RVOs, this study found that EZ integrity at baseline cannot prognosticate outcomes after initiation of anti-VEGF therapy. Such limited baseline prognostic ability is notable given the correlation of several baseline central subfield parameters (central foveal EZ to RPE, EZ-RPE CST, and EZ-RPE CFT) with the presence of SRF in a prior analysis of the eyes studied in this cohort. ¹⁴ This follow-up analysis on these eyes suggests that central SRF at baseline in patients with RVO does not affect long-term VA. In contrast, EZ integrity at 3-month follow-up and related improvement in EZ integrity from baseline to 3 months does predict final VA improvement.

Concordantly, VA improved from 20/97 at baseline to 20/52 at months 3 and 12. These functional and anatomic improvements likely reflected eyes that had the strongest response to anti-VEGF therapy, permitting more optimal EZ recovery. Evidence supports this with an analysis of CRUISE (the CRVO Study) showing that patients with complete anti-VEGF response, assessed via CST changes, by 3 months had improved 1-year outcomes. ¹⁷ This study further suggests that functional and anatomic improvements in RVOs are both generally attained by 3 months post initiation of anti-VEGF therapy and remain stable from that point onward with continued, standard therapy.

There have been conflicting reports regarding imaging biomarkers that predict long-term functional outcomes in RVO after anti-VEGF therapy. In a categorical analysis splitting patients into EZ intact vs not groups, Fujihara-Mino et al ¹³ found that EZ integrity at baseline did correlate with final vision, although this effect was not retained in adjusted analysis. Chan and colleagues ⁶ and Mimouni et al ¹⁸ investigated the role of both DRIL and EZ integrity in longitudinal RVO studies in contrast to this work that focused on EZ characteristics. Both identified DRIL as a significant predictor of final outcomes, but Chan et al, ⁶ in agreement with this analysis, reported that 3-month EZ changes correlated with 12-month VA. As Chan et al ⁶ described, lack of adjusting for baseline VA in the study by Mimouni and colleagues ¹⁸ may have driven the difference in results. One technical limitation of these studies assessing EZ integrity is reliance on a small number of B-scans that assessed anatomic perturbations by evaluating linear disruptions. ⁶ Our work greatly adds to these findings given the highly quantitative volumetric methodology used to determine EZ metrics at a more advanced level through multilayer retinal segmentation with panmacular EZ mapping. ^19,20

It is important to acknowledge the limitations of this study. First, fluid segmentation was not performed in this study, resulting in the use of proxies for evaluating ME changes in modeling analysis.

Second, macular ischemia and DRIL were not quantified in this analysis. Whether including such parameters as predictors in regression modeling would have reduced the prognostic utility of the EZ parameters remains unclear. At the baseline visits in particular, intact EZ may have been obscured by shadowing from hemorrhage. The EZ disruption at baseline may have been affected by variability in symptom duration before patients presented for evaluation. In addition, this analysis evaluated the OCT signal that corresponds to the anatomic correlate. In that regard other factors such as shadowing from fluid or hemorrhage may affect EZ appearance and contribute to “apparent” attenuation. Future work to correct for shadowing may help to mitigate this issue, but eyes that have underlying preserved EZ but extensive shadowing may exhibit more robust “apparent” EZ recovery and associated visual recovery. In addition, the analysis was retrospective with no specific protocol for anti-VEGF therapy, which may have resulted in undertreatment or overtreatment. Similarly, there was no protocol for refracting at all visits; this limited the gathering of best-corrected VA at each time point.

EZ mapping provides an additional metric for evaluating RVO impact on retinal anatomy and VA prognosis during the follow up-period but not at baseline. Such mapping will not replace decision-making regarding further treatment in RVO cases but rather will enable further refining of a patient’s expectations in an individualized fashion. Further research is needed to better elucidate the predictive value of these biomarkers on overall outcomes and the role for assessment in disease management.