Longitudinal Ellipsoid Zone and Outer Retinal Integrity Dynamics Following Epiretinal Membrane Surgery

STRUCTURED ABSTRACT

Purpose

To quantify ellipsoid zone (EZ) changes in integrity following epiretinal membrane (ERM) surgery, correlate findings to visual acuity (VA), and determine predictors for prognosis.

Methods

A post-hoc analysis of eyes undergoing ERM surgery pooled from the prospective DISCOVER intraoperative optical coherence tomography (iOCT) study and eyes undergoing conventional ERM surgery without iOCT. Quantitative EZ features were extracted using a multi-layer machine learning enabled automated segmentation platform following image analyst review/correction for segmentation accuracy. VA and EZ integrity were quantitatively assessed and correlated before and after ERM surgery. Multiple linear regression was performed to assess pre-operative VA and EZ features as predictors for improvement in VA or EZ integrity.

Results

There were 177 eyes from 177 subjects that underwent ERM surgery from the DISCOVER and conventional arms. Improvement in VA and multiple EZ integrity features was noted after ERM surgery, including EZ partial attenuation and EZ-retinal pigment epithelium (RPE) volume (p<0.05). A reduction in EZ partial attenuation and increase in EZ-RPE central subfield thickness (EZ-RPE CST) was significantly correlated with improved VA after ERM surgery (p<0.05). More robust EZ-RPE CST at baseline predicted VA improvement after ERM peel in regression modeling (β=0.005, p<0.05).

Conclusions

Longitudinal assessment of EZ features demonstrate significant postoperative improvement in multiple EZ integrity metrics following ERM surgery. Improving EZ integrity was correlated to improving VA. EZ integrity and visual acuity were significant predictors in regression modeling and may have value in clinical prognostication.

Introduction:

Epiretinal membranes (ERMs) are common in subjects over the age of 50 after posterior vitreous detachment.^1,2 These ERMs can result in poor central vision and metamorphopsia.^1–3 Vitreous surgery with ERM peel is the gold standard approach to improving vision among patients with severe symptomatic ERMs.^2,4 ERMs may create inner tractional forces which displace normal photoreceptor alignment, detectable on optical coherence tomography (OCT) as disruption of the ellipsoid zone (EZ) layer.³ The integrity of this EZ layer has been associated with baseline visual acuity in several retinal. disorders.^5–8 Additionally, EZ recovery after treatment or disease resolution has been shown to drive visual acuity outcomes in numerous retinal disorders.^5–8

Several investigations have correlated EZ integrity with visual outcomes in ERM.^9–12 However, these studies mostly relied on qualitative assessment of outer retinal perturbations, lacking in quantitative assessment of the EZ due to limitations in panmacular outer retinal segmentation. It remains unclear as to whether quantified baseline EZ features can be used to predict visual acuity improvement or EZ resolution among patients undergoing ERM peel. Additionally, a comprehensive assessment of EZ dynamics following ERM repair is needed to better understand how interventions impact anatomic and functional outcomes. Emerging technology now allows for quantitative, volumetric evaluation of EZ integrity.^13–15 In this study, quantitative EZ features were assessed longitudinally after ERM peel among patients with symptomatic ERMs. These EZ parameters were correlated to visual acuity at baseline and after ERM peel. In addition, baseline EZ features were analyzed as predictors for visual acuity improvement or EZ resolution after ERM peel.

Methods:

Study Population:

This study was a pooled analysis including post-hoc analysis of the prospective DISCOVER intraoperative OCT study and a retrospective assessment of eyes undergoing conventional ERM surgery with EZ features and visual acuities assessed before ERM surgery and at 6–12 months (+/− 3 months) following surgery. DISCOVER is a single-center, multi-surgeon, IRB-approved prospective study assessing the impact of intraoperative OCT technology on surgical decision making, including extent of ERM removal.¹⁶ All eyes underwent standard 3-port small gauge (23, 25, or 27) vitrectomy with ICG staining. All eyes pooled from the DISCOVER study underwent ICG staining and ERM peel with optional removal of the internal limiting membrane (ILM) during ERM surgery.¹⁶ Intraoperative OCT was performed after initial peel to confirm complete ERM removal but not complete ILM removal. Additional peeling or staining was optional if residual membranes were identified on iOCT. The conventional ERM group included all patients that underwent standard of care ERM surgery with compulsory ILM removal without intraoperative OCT. In this group, complete ILM removal was confirmed after initial peel with repeat ICG staining. The pooled analysis was performed after Institutional Review Board (IRB) approval was obtained and research pertaining to this study adhered to the tenets of the Declaration of Helsinki.¹⁷ Diagnosis of ERM was determined by fundoscopic and spectral domain-OCT (SD-OCT) findings. The best corrected VA available (Snellen, pinhole, or refraction) was used for each patient before and after ERM surgery. “Good visual outcomes” were defined as eyes that gained 10 or more letters and/or had a final visual acuity of 20/25 or better; “poor visual outcomes” were defined as eyes with 0 letter gain or loss and a final visual acuity of 20/40 or worse. Combination surgery was defined as a study eye that underwent phacoemulsification in addition to ERM surgery.

Inclusion criteria were all eyes from DISCOVER and conventional groups undergoing ERM surgery between 2013–2017 at the Cleveland Clinic Cole Eye institute (n = 282). Exclusion criteria were subjects with unavailable pre-operative or 6–12 month (+ 3 months) postoperative OCT data of sufficient quality (n = 73), poorly controlled ocular co-morbidities (n=29), and data collection limitations (n=3). Co-morbidities included subjects with fungal endophthalmitis with subretinal abscess, diabetic retinopathy, retinal detachment, central corneal abrasion, branch retinal vein occlusion, age-related macular degeneration, choroidal neovascular membrane, non-resolving vision changes due to Irvine gas, and poorly controlled extensive macular edema resulting in significant retinal structural disturbance including outer retinal disruption and shadowing as well as macular edema secondary to concurrent macular diseases such as diabetic macular edema and neovascular age-related macular degeneration. The resulting population included 177 eyes from 177 subjects meeting inclusion and exclusion criteria with 103 eyes pooled from the DISCOVER intraoperative OCT study and 74 eyes pooled from the conventional ERM surgery group.

Image Acquisition and Software Analysis:

All subjects underwent retinal imaging with the Cirrus 5000 SD-OCT system (Carl Zeiss Meditec, Dublin, CA, USA) before ERM peel procedure and at a six or twelve-month follow-up visit (± 3 months), heretofore referred to as, follow up after ERM peel. Pre-ERM peel OCTs were done prior to surgery with 88.7% of them performed within 60 days before surgery. At the time of this study, 6.0mm x 6.00mm (width, length) images from the iOCT and conventional groups were exported into a novel multi-layer machine learning enabled automated segmentation platform developed at the Cleveland Clinic that has been previously described (OCT Viewer) which provides linear, area and volumetric EZ measurements based on relative proximity to the RPE line, serving as a surrogate to photoreceptor outer segment length.^14,15 This platform has demonstrated EZ response to pharmacologic and surgical intervention and association with VA in several retinal pathologies including retinal venous occlusive disease, macular telangiectasia type 2, macular hole repair and neovascular age-related macular degeneration.^{14,15,18–20} Eyes with completed automatic segmentation were reviewed by trained image analysts for segmentation accuracy. Manual correction was performed in areas of segmentation error. After initial review by the image analyst, a sequential review by a senior project lead was performed for final quality review. All image analysts underwent standardized training for approach to EZ segmentation and performed segmentation with standardized lighting, monitors, computers, and input systems.

Utilizing the image processing and analysis features, quantitative values of several EZ features were generated after final review. EZ-retinal pigment epithelium (EZ-RPE) volume is defined as the retinal volume contained within the macular cube bounded by the EZ and RPE lines (mm³). EZ partial attenuation is defined as the percent macular map area with partial EZ disruption (EZ-RPE height 20μm or less) (Figure 1). This threshold for partial EZ attenuation has been demonstrated in prior studies and is based on a very small percentage of the normal macula that demonstrates partial attenuation as well as association of partial EZ attenuation, as defined, with retinal diseases such as hydroxychloroquine toxicity, macular holes, age related macular degeneration, and ocriplasmin toxicity when compared to normal.^18–23 EZ total attenuation is defined as the percent map area with total EZ disruption (EZ-RPE height = 0 μm, i.e., no EZ is visible). EZ-RPE central subfield mean thickness (EZ-RPE CST, μm) is the mean thickness of the central 1mm. The EZ features for each OCT image were generated before ERM removal and at follow-up after ERM peel.

Fig. 1. Selected Ellipsoid Zone (EZ) Partial and Total Attenuation Example.

(A) Horizontal foveal B-scan before epiretinal membrane (ERM) peel corresponding to horizontal crosshair in (C). Retinal pigment epithelium (b, cyan) and EZ band (a, yellow) lines are automatically segmented at each B-scan throughout the macular cube. Orange vertical lines identify center placement of fovea (middle) and 1.0 mm radius demarcating the parafoveal region (left and right lines). Number figures next to vertical lines identify retinal thickness between segmentation lines. After automatic segmentation, segmentation is carefully reviewed by trained technicians and quality specialists with manual corrections as needed. (B) Horizontal foveal B-scan 12 months after ERM peel corresponding to horizontal crosshair in (D). (C) Macular cube EZ-retinal pigment epithelium (EZ-RPE) distance map before ERM peel. Partial attenuation includes percent area of the macula cube where the EZ-RPE height is 20 microns or less, while total attenuation is percent area of cube equal to 0 microns distance for EZ-RPE height (e.g., no EZ present). Partial and total attenuation are 1.37 and 1.16 percent respectively. (D) Macular cube EZ-RPE distance map 12 months after ERM peel showing a reduction in EZ partial and total attenuation to 0.04 and 0.00 percent respectively. (E) & (F) EZ-RPE thickness maps with color transition to enhance attenuation visualization before (E) and after (F) ERM surgery.

Clinical Data Review and Statistical Analysis:

Records from both the conventional and iOCT treatment cohorts were reviewed at the time of this study for visual acuities and pre-operative and post-operative characteristics/co-morbidities. Age, sex, phakic status, and combination surgery were readily available existing data from iOCT DISCOVER study. Data were compared using either student’s t-test (age), Wilcoxon rank-sum or Wilcoxon signed-rank (VA, EZ and follow-up time), Spearman’s correlation, McNemar’s test (categorical visual outcome groups), Fisher’s exact method (all categorical variables), or multiple linear regression. Multiple linear regression analysis was used to assess effect of pre-surgical variables and intraoperative variables (combination surgery) on change in BCVA and change in EZ integrity from pre- to post-ERM surgery. Pre-ERM variables were obtained from the same day as pre-operative OCT date. Post-ERM variables were obtained from the same day as one of two follow-up OCT dates at 6 or 12 months (+ 3 months), selected based on availability. Multiple linear regression models tested whether each EZ feature at baseline significantly predicted change in VA (LogMAR) from baseline to after ERM peel while controlling for other variables. Multiple linear regression analysis was also used to test if each EZ parameter at baseline predicted recovery in said EZ feature from baseline to after ERM peel while controlling for other variables. Each EZ feature was evaluated in separate multiple linear regression models to account for multicolinearity. Correlation analysis was performed between follow-up time and VA and EZ variables to determine need for incorporation into regression models. VA means, standard deviations, and standard error of means (SEM) are reported as Snellen, LogMAR or ETDRS letters converted from Snellen measurement.²⁴ LogMAR was utilized in statistical tests to determine significance.²⁴ All statistical analyses utilized JMP Pro 15.1 Software (SAS Institute, Inc., Cary, NC, USA) or Microsoft Excel (Microsoft, Redmond, WA, USA) while graphs were created utilizing GraphPad Prism (version 8.4.2). Photoshop (version 22.3.1) was used to enhance visualization of EZ attenuation in figure 1 as specified in figure.

Results:

Subject Demographics:

There was a total of 177 eyes from 177 subjects included in the study with a mean preoperative VA of 20/51 and combined follow-up period of 10.8 (range; 3.7 – 15.2). The mean age was 69.9 ± 9.3 years with 94/177 (53%) being female. Subject characteristics included 103 in the iOCT-guided ERM peel group and 74 eyes in the conventional ERM surgery group with no significant differences in follow-up period between groups (table 1).

Table 1:

Baseline Subject Demographics

	OCT-Guided ERM Peel (n =103)	Conventional Technique (n=74)	All (n=177)	p-value
Age; mean (SD)	69.9 (9.0)	69.8 (9.8)	69.9 (9.3)	0.93
VA; mean (SD in Letters)	20/53 (10.6)	20/49 (9.9)	20/51 (10.3)	0.56
Gender (female); number (percent)	51 (49.5%)	43 (58.1%)	94 (46.9%)	0.29
Phakic; number (percent)	70 (68.0%)	52 (70.3%)	122 (69.0%)	0.87
Combination Surgery (PPV and ERM Peel + Phaco)	66 (64.1%)	52 (70.3%)	118 (66.7%)	0.42
Post-ERM surgery Follow-up (mean, months)	10.8 [4.2 – 15.2]	10.9 [3.7 – 14.6]	10.8 [3.7 – 15.2]	0.96

Abbreviations: OCT, optical coherence tomography; VA, visual acuity; PPV, pars plana vitrectomy; ERM, epiretinal membrane; Phaco, phacoemulsification

Values reported are number (%) and median [range]

Statistical analyses were student’s t-test (age), Wilcoxon rank-sum (VA & post-ERM OCT interval), and Fisher’s exact (all others)

Statistical significance defined as p<0.05

Ellipsoid Zone Integrity and Visual Acuity Outcomes:

Mean VA improved 13.2 letters from 20/51 (64.5 ± 0.8 letters; mean + SEM) to 20/28 (77.7 ± 0.6) after ERM peel (p <.0001) (figure 2) with no difference in VA at baseline between surgical groups (p = 0.56). At baseline, 39.6% of subjects had 20/40 vision or better compared to 89.8% after ERM peel (p < 0.0001).

Fig 2. Mean Visual Acuity (ETDRS Letters) after Epiretinal membrane (ERM) Peel.

Mean Visual acuity (VA) increases 13.19 letters from 64.51 ± 0.78 (SEM) to 77.70 ± 0.60 (SEM) after ERM peel. ****p <.0001.

Among all eyes, EZ partial attenuation reduced from 2.89% at baseline to 0.80% (p = 0.003) (figure 3). Similarly, EZ total attenuation reduced from 2.71% at baseline to 0.63% (p = 0.0001) (table 2). EZ-RPE volume increased 2.29% from 1.31 ± 0.02 mm³ to 1.34 ± 0.02 mm³ after ERM peel (p = 0.006) (figure 3). Mean EZ-RPE CST increased 2.68% from 41.41 ± 0.69 μm to 42.52 ± 0.57 μm after ERM peel, although this increase was not significant (p = 0.230) (figure 3).

Fig 3. Mean Ellipsoid zone (EZ) Feature after Epiretinal Membrane (ERM) peel.

A EZ partial attenuation decreased to 0.80% ± 0.19% (SEM) from 2.89% ± 0.70% (SEM) after ERM peel (p = 0.003). B EZ-RPE volume increased 2.29% from 1.31 mm³ (0.02, SEM) to 1.34 mm³ (0.02, SEM) after ERM peel (p = 0.006). Baseline EZ-RPE volume was greater in the OCT group than conventional group (p = 0.0004). **p < 0.01

Table 2:

Ellipsoid Zone Integrity Dynamics Following Epiretinal Membrane Peel

	Total Study Cohort (n=177)
	Baseline	After Peel	p value
Percentage 20 μm EZ or less Partial Attenuation	2.89 ±.70	0.80±.19	0.003
Percentage 0 μm EZ Total Attenuation	2.71±.68	0.63±.17	0.0001
EZ-RPE Central Subfield Mean Thickness (μm)	41.41±.69	42.53±.57	0.230
EZ-RPE Volume (mm3)	1.31±.02*	1.34±.02	0.006

Abbreviations: OCT, optical coherence tomography; EZ, ellipsoid zone; RPE, retinal pigment epithelium;

Statistical analysis before and after peel conducted with Wilcoxon signed-rank test

^*Statistical significance difference between iOCT and conventional surgical groups at baseline (p = 0.0004) demonstrated with Wilcoxon rank-sum test

Statistical significance defined as p<0.05, highlighted in bold

Correlation of Ellipsoid Zone Integrity and Visual Acuity

Higher EZ-RPE CST was correlated with better VA at baseline (r=0.29, p = 0.0001) and after ERM peel (r=0.21, p = 0.0043). An increase in EZ-RPE CST was correlated with improving VA after ERM peel (r=0.18, p = 0.014). Less EZ partial attenuation was correlated with better VA at baseline (r=−0.23, p = 0.0017) and after ERM peel (r=−0.27, p = 0.0003). A reduction in EZ partial attenuation was correlated with improving VA after ERM peel (r=−0.16, p = 0.0305). Similarly, less EZ total attenuation was correlated with better VA at baseline (r=−0.22, p = 0.003) and after ERM peel (r=−0.25, p = 0.0009) while a reduction in EZ total attenuation was correlated with improving VA after ERM peel (r=−0.21, p = 0.005). Greater EZ-RPE volume was correlated with superior VA after ERM peel (r=0.15, p = 0.046) but not at baseline (r=0.10, p = 0.191). Changes in EZ-RPE panmacular volume were not correlated to changes in VA after ERM peel (r=0.05, p = 0.4839). Follow-up duration was not correlated with final EZ or VA measurements as well as improvement in EZ or VA (all p > 0.05).

Predictors of Improved Visual Acuity and Ellipsoid Zone Outcomes

Multiple linear regression modeled change in VA after ERM peel based on baseline EZ-RPE CST, baseline VA, cystoid macular edema (CME), combination surgery and age. This model showed that baseline EZ-RPE CST and baseline VA were associated with postoperative visual acuity (R² = 0.54, F (5,171) = 39.99, p <.0001). Specifically, higher baseline EZ-RPE CST (β = 0.005, p = 0.0004) and poorer baseline VA (β = 0.8035, p <.0001) predicted more improvement in VA after ERM peel, while preoperative CME, combination surgery and age did not significantly predict change in VA. Other EZ features were not significant predictors of VA improvement after ERM peel when exchanged into the same model individually.

In a separate regression model, we assessed the change in each selected EZ feature after ERM peel based on each corresponding EZ feature at baseline, CME and age. For EZ-RPE volume, this model showed that two predictors, baseline EZ-RPE volume and CME explained 23% of the variance in the model (R² = 0.23, F (3,173) = 16.98, p <.0001). Two predictors, baseline EZ-RPE volume (β = 0.250, p < 0.0001) and CME (β = 0.023, p = 0.0197), predicted change in EZ-RPE volume after ERM peel, while age did not. Meaning, a lower baseline EZ-RPE volume and presence of baseline CME predicted greater EZ-RPE volume restoration. Higher EZ partial (p < 0.0001) or total attenuation (p < 0.0001) and lower EZ-RPE CST (p < 0.0001) were significant predictors of EZ improvement in their models as well while presence of CME and younger age were predictors for improvement in the model for change in EZ-RPE CST (p = 0.049).

Analysis between Subjects with Good and Poor Visual Outcomes:

Among all eyes, there were 146 eyes (82.5%) in the good or poor outcome groups. Between these groups there were 132 eyes (90.4%) with good outcomes and 14 eyes (9.6%) with poor outcomes. Subjects with good outcomes had a median follow-up of 11.0 (range; 3.7 – 15.2) compared to 6.2 (range; 3.8 – 13.6) months in the poor outcomes group (p = 0.0033). Among the good outcome group, follow-up period was not correlated with final EZ or VA as well as change in EZ or VA (p > 0.05). Among the poor outcome group, longer follow-up period was correlated with worsening of VA (r =−0.56, p = 0.0365) as well as higher final EZ total attenuation (r=0.54, p = 0.0476). In the good outcome group, mean VA improved 16.65 letters from 20/52 (64.27 ± 0.95 letters; mean + SEM) to 20/24 (80.92 ± 0.42; mean + SEM) after ERM peel (p <.0001). Mean VA decreased 1.59 letters from 20/53 (63.83 ± 2.50 letters; mean + SEM) at baseline to 20/57 (62.24 ± 2.50; mean + SEM) after ERM peel in the poor outcome group (p =.0313).

Among subjects with good VA outcomes, EZ partial attenuation, EZ total attenuation, EZ-RPE CST, and EZ-RPE volume were 2.95%, 2.80%, 41.90 + 0.81 μm, and 1.32 + 0.02 mm³ at baseline compared to 0.65% (p < 0.0001), 0.51% (p < 0.0001), 43.25 + 0.60 μm (p = 0.29), and 1.35 + 0.02 mm³ (p = 0.02) after ERM peel. Among subjects with poor VA outcomes, EZ partial attenuation, EZ total attenuation, EZ-RPE CST, and EZ-RPE volume were 1.50%, 1.40%, 37.82 + 2.53 μm, and 1.24 + 0.04 mm³ at baseline compared to 1.37% (p = 0.84), 1.19% (p = 0.55), 35.94 + 3.17 μm (p = 0.67), and 1.26 + 0.05 mm³ (p = 0.27) after ERM peel. Baseline EZ features did not differ between the two outcome groups (p > 0.05). However, the change in EZ total attenuation from baseline to after ERM peel between good and poor outcome groups was significantly different (p = 0.03), whereas the other EZ features were not (p > 0.05).

Discussion:

This study assessed EZ quantification in ERM surgery for evaluating longitudinal EZ progression and predicting functional outcomes at baseline. VA and quantified EZ features significantly improved after ERM peel. Healthier EZ integrity features were correlated with better visual acuity at baseline and after ERM peel. Specifically, higher EZ-PRE CST and less EZ partial or total attenuation were correlated with better VA at baseline and after ERM peel while higher EZ-RPE volume was correlated with better VA after ERM peel only. An increase in EZ-RPE CST and reduction in EZ partial or total attenuation from pre- to post-ERM peel were significantly correlated with improvement in visual acuity. Multiple regression analyses showed that a poorer baseline VA and higher baseline EZ-RPE CST predict greater improvement in VA after ERM peel. Modeling also suggested that lower baseline EZ-RPE volume and presence of CME may be associated with improvement of EZ-RPE volume after ERM peel. Other baseline EZ measures, including higher EZ partial or total attenuation and lower EZ-RPE CST, also predicted improvement in EZ integrity in each feature’s separate model. While presence of CME and younger age were also independent predictors in the model for improvement in EZ-RPE CST. Improvement in EZ total attenuation after ERM peel was greater among patients with good visual outcomes compared to patients with poor visual outcomes although follow-up between groups was significantly different.

Early qualitative studies, assessing EZ disruption as a categorical variable based on macular B-scan visual assessment, identified that poor baseline EZ integrity prognosticated poor visual acuity outcomes.^10,11 Advances in image analysis platforms has enabled quantification of the outer retinal layers, including the EZ. Recent studies evaluating ERM have focused on morphological and/or qualitative classification of ERM and retinal. features.^25,26 The quantitative findings, herein, support prior findings of qualitative EZ resolution after ERM peel and associations of qualitative EZ disruption with poorer visual acuity. Notably, this quantitative analysis aligns with qualitative studies showing that a healthier EZ, identified in this study by a higher EZ-RPE CST, may be used to predict improved VA response as well as EZ response to ERM peel.^12,27,28 Significant prediction of VA improvement by EZ-RPE CST compared to other EZ features in this study is likely due to its central foveal location which is strongly linked to VA as compared to the other measures which are panmacular representations of the EZ. Prior work in retinal venous occlusive disease emphasized EZ-RPE CST as possibly predicting improvement in visual acuity (trending towards significance) relative to other measurements.²⁹

Modeling showed that poorer baseline VA predicted greater improvement in VA after ERM surgery. This finding has been described in the prior literature and likely highlights the greater improvement potential for subjects with poorer baseline VA as well as relatively easy reversal of tractional forces exerted by ERMs.^30,31 The data in this study also shows that combination surgery with phacoemulsification was not a significant predictor of improvement in VA following ERM surgery, agreeing with prior literature showing that there is no difference in visual outcomes between patients that undergo combined surgery compared to sequential.³² This data is of clinical import as providers can potentially prognosticate a specific subject’s outcomes at a potentially higher resolution with quantitative data. Moreover, as automated OCT segmentation platforms are developed and optimized, quantified EZ integrity can be extracted more rapidly and efficiently.^33,34 The results of this study support the potential application of these novel technologies in ERM characterization by highlighting the utility of quantifying EZ metrics longitudinally.

There are important limitations of this study. This study was performed at a single tertiary care center and a single surgeon performed ERM peel in the conventional surgery group, possibly limiting the generalizability of these findings. Adding to limitations in generalizability, prospective data from the DISCOVER clinical trial was pooled with data from retrospective collection which may have led to differences in data collection between the groups. To best address this limitation, it is important to note that the primary measures including EZ features and segmentation, clinical VA data and co-morbidities were collected together during this retrospective study. The ranges for follow up visit after ERM peel are broad, complicating the takeaways regarding time course of EZ restoration. Conclusions regarding differences between poor and good visual outcome groups must be interpreted with caution given that follow-up varied significantly between them. When assessing all subject outcomes in regression modeling, we choose not to control for follow-up period to avoid overfitting, given that follow-up did not correlate with VA and EZ improvement in the overall cohort. Baseline EZ-RPE volume differed between the two surgical groups which may limit generalization; however, baseline EZ-RPE was adjusted for in the model for change in EZ-RPE Volume. Additionally, running several multiple linear regression models for the same outcome variable (i.e. change in VA) varying one selected parameter (EZ feature) increases the risk for type 1 error. However, the analyzed EZ features were highly multicolinear which would introduce over fitting into a model if modeled together rather than separately. Finally, although baseline characteristics were similar between iOCT and conventional surgery groups, subjects were not randomized into either group, leaving room for confounding by unknown factors.

This study provides an assessment of a novel quantitative technique for evaluating EZ integrity following ERM surgery. Quantitative measurements of EZ health may afford greater accuracy in prognosis and management in subjects with symptomatic ERMs. Notably, this analysis reproduced findings with panmacular quantitative assessment that EZ integrity is correlated with visual acuity and that a healthier EZ before ERM peel predicts larger improvement in VA. In the future, quantifiable features such as these may optimize management and prognostication for ERM subjects.

Summary Statement.

Quantitative analysis of ellipsoid zone (EZ) integrity after pars plana vitrectomy with epiretinal membrane peel. Visual acuity and EZ integrity improve after epiretinal membrane peel while improvement in EZ integrity is correlated with improvement in visual acuity. Preoperative visual acuity and baseline EZ integrity may predict clinical outcomes following surgery.