Analysing early changes of photoreceptor layer thickness following surgery in eyes with epiretinal membranes

Abstract

Background/Objectives

To analyse short-term changes of mean photoreceptor thickness (PRT) on the ETDRS-grid after vitrectomy and membrane peeling in patients with epiretinal membrane (ERM).

Subjects/Methods

Forty-eight patients with idiopathic ERM were included in this prospective study. Study examinations comprised best-corrected visual acuity (BCVA) and optical coherence tomography (OCT) before surgery, 1 week (W1), 1 month (M1) and 3 months (M3) after surgery. Mean PRT was assessed using an automated algorithm and correlated with BCVA and central retinal thickness (CRT).

Results

Regarding PRT changes of the study eye in comparison to baseline values, a significant decrease at W1 in the 1 mm, 3 mm and 6 mm area (all p-values < 0.001), at M1 (p = 0.009) and M3 (p = 0.019) in the central 1 mm area, a significant increase at M3 in the 6 mm area (p < 0.001), but no significant change at M1 in the 3 mm and 6 mm area and M3 in the 3 mm area (all p-values > 0.05) were observed. BCVA increased significantly from baseline to M3 (0.3LogMAR-0.15LogMAR, Snellen equivalent = 20/40-20/28 respectively; p < 0.001). There was no correlation between baseline PRT and BCVA at any visit after surgery, nor between PRT and BCVA at any visit (all p-values > 0.05). Decrease in PRT in the 1 mm (p < 0.001), 3 mm (p = 0.013) and 6 mm (p = 0.034) area after one week correlated with the increase in CRT (449.9 µm–462.2 µm).

Conclusions

Although the photoreceptor layer is morphologically affected by ERMs and after their surgical removal, it is not correlated to BCVA. Thus, patients with photoreceptor layer alterations due to ERM may still benefit from surgery and achieve good functional rehabilitation thereafter.

Introduction

An idiopathic epiretinal membrane (ERM) represents a fibrocellular layer that develops on the vitreoretinal interface and leads to metamorphopsia and vision loss. The prevalence ranges between approximately 1% to 29%, differing widely between ethnicities. Risk factors associated with idiopathic ERM include old age and posterior vitreous detachment among other factors, such as cataract surgery and diabetes [1]. Pars plana vitrectomy (PPV) combined with membrane peeling is the current standard for the surgical treatment of symptomatic ERM. Nevertheless, the indications for the surgical procedure have not been standardized yet and postoperative visual acuity varies. Thus, identifying predictive factors for postoperative functional outcome is vital to determine whether and at which point in time to perform surgery in order to achieve the best clinical outcome possible [2].

Through continuous advancement in image resolution and acquisition speed, spectral-domain optical coherence tomography (OCT) has become the gold standard in the assessment of ERM ahead of surgery and thereafter [3]. Recently, several predictive factors for postoperative visual acuity assessed via OCT were proposed. Govetto et al. introduced an OCT-based grading scheme that allows classifying the severity of ERMs according to changes of the inner retinal structures, such as the presence of the foveal pit, the definition of retinal layers and the presence of ectopic inner foveal layers. They were able to show that the preoperative presence of ectopic inner foveal layers was significantly associated with lower postoperative best-corrected visual acuity (BCVA). Furthermore, the thickness of these inner foveal layers correlated negatively with pre- and postoperative BCVA [4]. Inner retinal structures seem to be significantly altered by the tractional stress exerted by the membrane. In a previous study, we measured changes in the papillofoveal distance in order to assess foveal movement after successful surgery in patients with idiopathic ERM. We observed that movement of the fovea towards the optic disc started one day after surgery and continued up to three months. The extent of foveal movement correlated with higher central retinal thickness (CRT) and lower BCVA values, thus we suggested that more extensive postoperative foveal movement occurs in patients with a more advanced disease stage [5]. Besides changes of the inner retinal structures, alterations of the outer retinal layers, especially the photoreceptor layer (PRL), may also predict postoperative visual outcomes. Several studies showed that postoperative visual acuity is significantly worse if preoperative disruptions or irregularity of the inner segment/outer segment junction were detected [6–10]. Defects of the ellipsoid- and interdigitation zone and the photoreceptor outer segment (PROS) length seem to be predictors of poor visual acuity after surgery as well [11, 12]. However, there is only sparse literature concerning changes in the PRL, which can provide an insight into the postoperative functional rehabilitation [13].

In this prospective study, our aim was to evaluate the photoreceptor thickness (PRT) as quantified by automated artificial intelligence (AI) in patients with different ERM stages who underwent surgery. Additionally, we assessed the relationship between PRT and functional as well as morphological outcomes.

Materials and methods

Forty-eight eyes of 48 patients (24 females, 24 male) with idiopathic ERM that underwent vitrectomy for combined ERM and inner limiting membrane (ILM) peeling were included in this prospective, non-randomized clinical study after giving written informed consent (see Fig. 1). Exclusion criteria comprised any additional diseases causing a decrease in vision apart from cataract, i.e. glaucoma, age-related macular degeneration, any vascular disease and diabetic retinopathy. Patients with opacities of the optical media causing a decrease in OCT image quality as well as patients receiving endotamponade during the surgery were not included in the study. Eighteen healthy fellow eyes served as intraindividual comparison for baseline PRT. The study was approved by the institutional review board of the Medical University of Vienna and adhered to the tenets of the Declaration of Helsinki.

Fig. 1. Overview of the study design from patient inclusion to OCT-analysis.

ERM stage was assessed according to Govetto et al. [4] IRB institutional review board, ERM epiretinal membrane, ILM inner limiting membrane, BCVA best corrected visual acuity, SD-OCT spectral-domain optical coherence tomography, PRL photoreceptor layer, CRT central retinal thickness.

Surgical procedure

In all cases, 23-gauge 3 port pars plana vitrectomy with combined ERM and ILM peeling was performed. View-ILM® (n = 32; Alchimia, Italy), Membrane Blue™ (n = 14; Dutch Ophthalmic, USA) or Brilliant Peel DualDye (n = 2; Fluoron GmbH, Ulm, Germany) were used for membrane staining and removed after 1 min under balanced salt solution. A microforceps (Vitreq®) and a superview HTC® contact lens (Insight instruments) were used for the removal of the ERM and ILM. If indicated, a combined cataract surgery was performed. In all cases, no endotamponade was necessary.

Measurements

Examinations were performed before surgery (baseline) as well as 1 week, 1 month and 3 months after surgery. At each visit, patients underwent a BCVA test and a complete ophthalmologic examination including funduscopy and spectral-domain OCT (Spectralis HRA + OCT; Heidelberg Engineering, Heidelberg, Germany). OCT images were acquired using a 6x6mm grid with a resolution of 1024 × 49 (A-scans × B-scans) centred on the fovea. The follow-up function provided by the systems software was used to obtain images at week 1, month 1 and month 3 after surgery. CRT and ERM stage according to the scheme of Govetto [4] were assessed for each visit.

Volume scans were exported for a fully automated thickness segmentation of the PRL using a deep-learning-based image algorithm published previously [14, 15]. The algorithm identified the PRL as the region between the inner limit of the ellipsoid zone and the inner interface of the retinal pigment epithelium, corresponding to the ellipsoid section of the photoreceptors, which is densely packed with mitochondria, and the contact cylinder between the RPE and cones [15–17]. The thickness of the PRL was computed in alignment with an overlying Early Treatment Diabetic Retinopathy Study (ETDRS)-grid with a diameter of 6mm, mean thickness values were obtained for the 1 mm, 3 mm and 6 mm area. To evaluate the precision of the applied algorithm, the segmentation was checked systematically and, if necessary, manually corrected using the in-house developed annotation tool Optimus (version 1.8.6.4; Medical University Vienna, Vienna, Austria), which has previously been used by our research group [18].

Statistical analysis

All statistical analyses were computed using R (version 3.6.1) and SPSS Statistics 27. Descriptive statistics are given as mean ± SD or median (IQR), whichever was appropriate. A paired t-test was conducted to determine differences in PRT of study eyes and healthy fellow eyes at baseline. To quantify the change of PRT and BCVA, which was our primary outcome, linear mixed models were computed. Mean PRT was defined as the dependent variable and patient age, visit, an indicator variable for study eye/fellow eye as well as an interaction term between visit and eye as the explanatory variables. A random effect for the patient was included. The estimated change in PRT from baseline was computed for each visit and each study eye separately. P-values were computed to investigate whether PRT of the study eye differed significantly from the PRT baseline value. To investigate changes in BCVA over the 4 visits we fitted a linear mixed model with BCVA as the dependent and the categorical variable visit as the explanatory variable. Again, we included a random effect for the patient. We performed an ANOVA and post-hoc comparisons for each contrast between visits and computed Bonferroni-Holm adjusted p-values. To analyse whether baseline PRT may predict postoperative BCVA, correlation coefficients between baseline PRT and postoperative BCVA were computed using Spearman’s Rho, as were correlation coefficients between PRT and BCVA, CRT and ERM stages at each visit. The correlation coefficients between the respective variables were computed for each visit separately. Additionally, we computed correlation coefficients between the differences from the PRT and CRT baseline to week 1 after surgery using Spearman’s Rho. Fischer’s exact test was applied to the differences of PRT and CRT from baseline to week 1 to investigate whether an increase in CRT resulted in a higher than average decrease in PRT. To classify the data, a threshold value of 6.6 µm decrease in PRT was chosen by calculating the mean decrease plus two standard deviations of PRT between baseline and week 1 of all patients that showed a decrease in CRT from baseline to week 1. Regarding PRT, the statistical tests were computed using the values of each macular area, respectively (1 mm, 3 mm and 6 mm) and due to the exploratory character of this study, correction for multiple testing was not applied. Linear mixed models were computed using the lmerTest package [19]. Contrasts were calculated with the multcomp package [20].

Results

Forty-eight patients (24 females, 24 male) with a mean age of 70 years (SD 6.7, range 57–84 years) were included into the study. Fifteen patients were pseudophakic before surgery, in 23 patients, concomitant cataract surgery was performed, and 10 phakic patients underwent membrane peeling without additional cataract surgery (see Table 1).

Table 1.

Demographic and surgical characteristics of the study patients.

	Mean (SD; Range)
Age in years	70 (6.7; 57–84)
	N (%)
Sex
Male	24 (50)
Female	24 (50)
Pars Plana Vitrectomy with combined ERM and ILM-peeling	48 (100)
Dye used for membrane staining
View-ILM® (Alchimia, Italy)	32 (67)
Membrane Blue™ (Dutch Ophthalmic, USA)	14 (29)
Brilliant Peel DualDye (Fluoron GmbH, Ulm, Germany)	2 (4)
Lens status
Phakic at baseline, no concomitant cataract surgery	10 (21)
Phakic at baseline, concomitant cataract surgery	23 (48)
Pseudophakic at baseline	15 (31)

SD standard deviation, N number, % percentage.

Course of photoreceptor thickness

At baseline, the PRT in the study eye was 38.23 µm ± 6.88 µm, 34.52 µm ± 4.09 µm and 32.40 µm ± 3.69 µm the 1 mm, 3 mm and 6 mm area, respectively (Table 2 and Fig. 2). PRT in the healthy fellow eye was 40.82 µm ± 5.62 µm, 36.29 µm ± 4.61 µm and 33.46 µm ± 4.68 µm in the 1 mm, 3 mm and 6 mm area, respectively. There was no significant difference between PRT in the study eye and the fellow eye in the 1 mm, 3 mm and 6 mm area at baseline (p = 0.261, p = 0.055 and p = 0.098, respectively).

Table 2.

Course of PRT, BCVA, CRT and ERM stage of the study eye.

		Baseline	Week 1	Month 1	Month 3
PRT Mean (SD)	1 mm	38.23 (6.88)	32.13 (5.61)†	35.87 (6.04)*	35.41 (4.25)*
	3 mm	34.52 (4.09)	29.78 (4.24)†	34.01 (4.00)	34.39 (3.03)
	6 mm	32.40 (3.69)	28.17 (3.72)†	32.28 (3.96)	32.99 (2.60)†
BCVA Mean (SD)		0.3 (0.24)	0.25 (0.17)	0.24 (0.17)	0.15 (0.16)†
CRT Mean (SD)		449.9 (78.97)	462.20 (52.53)	430.16 (49.45)	415.44 (56.04)
ERM Stage Median (IQR)		3 (2–3)	2 (2–2.5)	2 (2–2)	2 (2–2)

Comparison of the mean photoreceptor thickness (PRT), mean best corrected visual acuity (BCVA), mean central retinal thickness (CRT) and median epiretinal membrane (ERM) stage of the study eye (SE) between baseline and follow-up data. To quantify the change of PRT, linear mixed models were computed. P-values were computed to investigate whether PRT of the study eye differed significantly from the PRT baseline value. To investigate changes in BCVA over the 4 visits we fitted a linear mixed model. We performed an ANOVA and post-hoc comparisons for each contrast between visits and computed Bonferroni-Holm adjusted p-values. Statistically significant changes of the PRT in the SE in comparison to the baseline thickness values and of the BCVA in comparison to the baseline BCVA are marked with asterisks and daggers.

*p = <0.05; ^†p = <0.001.

Fig. 2. Photoreceptors of patients with different ERM stages before surgery, arranged in descending order (Stage 1 – upmost, Stage 2 – upper middle, Stage 3 – lower middle, Stage 4 – lowest).

The images depicted are central B-scans of fovea-centred spectral-domain OCT scans with overlying segmentation of the PRL, comprising the ellipsoid zone (IS/OS), outer segment of photoreceptors and interdigitation zone (IZ). The segmentation of the OCT volume scans (6 × 6 mm, 1024 a-scans × 49 b-scans) was performed using an automated algorithm. To evaluate the precision of the applied algorithm, the segmentation was checked systematically and, if necessary, manually correction applied.

At week 1, the PRT in the study eye was 32.13 µm ± 5.61 µm, 29.78 µm ± 4.24 µm and 28.17 µm ± 3.72 µm in the 1 mm, 3 mm and 6 mm area, respectively. A significant decrease at week 1 in the PRT in the study eye in comparison to baseline was observed in each disc size (p < 0.001 for each disc size, respectively).

At month 1, PRT in the study eye was 35.87 µm ± 6.04 µm, 34.01 µm ± 4 µm and 32.28 µm ± 3.96 µm in the 1 mm, 3 mm and 6 mm area, respectively. PRT in the 1 mm area was significantly reduced compared to baseline values (p = 0.009), whilst there was no significant difference in the 3 mm and 6 mm area between baseline and month 1 (p = 0.628 and p = 0.643, respectively).

At month 3, PRT in the study eye was 35.41 µm ± 4.25 µm, 34.39 µm ± 3.03 µm and 32.99 ± 2.6 µm. In comparison to baseline, PRT in the 1 mm area was significantly reduced (p = 0.019), whilst it was significantly increased in the 6 mm area (p < 0.001). PRT in the 3 mm area did not differ from baseline values (p = 0.326). The changes of PRT in the study eye are illustrated in Table 2.

Relation between photoreceptor thickness and BCVA, CRT and ERM stage

BCVA was 0.30 ± 0.24 logMAR (Snellen equivalent approximately 20/40) at baseline and improved to 0.25 ± 0.17 logMAR (Snellen equivalent approximately 20/36; p = 0.151) at week 1, 0.24 ± 0.17 logMAR (Snellen equivalent approximately 20/35; p = 0.070) at month 1 and 0.15 ± 0.16 logMAR (Snellen equivalent approximately 20/28; p < 0.001) at month 3. There was no significant correlation between PRT in the 1 mm, 3 mm or 6 mm area and BCVA at any of the visits (in all cases p > 0.05). Baseline PRT in the 1 mm, 3 mm or 6 mm did also not correlate with BCVA at week 1, month 1 or month 3 (in all cases p > 0.05). CRT increased from 449.90 µm ± 78.97 µm to 462.20 µm ± 52.53 µm at week 1 and decreased to 430.16 µm ± 49.45 µm at month 1 and 415.44 µm ± 56.04 µm at month 3. A significant positive correlation between PRT in the 1 mm area and CRT was observed at week 1, month 1 and month 3 (rs = 0.464, p = 0.003; rs = 0.366, p = 0.017 and rs = 0.324, p = 0.044 respectively). The median ERM stage was 3 (2–3) at baseline, 2 (2–2.5) at week 1, 2 (2–2) at month 1 and 2 (2–2) at month 3. There was no significant correlation between the PRT and the ERM stages at any visit (in all cases p > 0.05). All values regarding BCVA, CRT and ERM stage can be found in Table 2.

Relation between the course of photoreceptor thickness from baseline to week 1 and CRT

Interestingly, the decrease of the PRT in the 1 mm, 3 mm and 6 mm area correlated with the increase in CRT (rs = −0.604, p < 0.001; rs = −0.399, p = 0.013 and rs = −0.344, p = 0.034, respectively) from baseline to week 1 after surgery. Unexpectedly, Fischer’s exact test furthermore revealed that a decrease of more than 6.6 µm in PRT in the 1 mm, 3 mm and 6 mm area significantly differed in CRT values (p < 0.001, p = 0.003 and p = 0.017, respectively; Fig. 3).

Fig. 3. Correlation of the difference of mean photoreceptor thickness (PRT) in the 1 mm, 3 mm and 6 mm area and the difference of central retinal thickness (CRT) from baseline (BSL) to week 1 (W1).

The difference of PRT thickness in µm in the 1 mm, 3 mm and 6 mm area disc from BSL to W1 is displayed on the x-axis. The y-axis denotes the difference in CRT in µm from BSL to W1.

Discussion

The aim of our study was to detect changes in PRT using automated AI after surgery for epiretinal membrane. The applied algorithm showed great accuracy: The median difference between its results and those obtained after manual correction ranged between 0.5% in the 6 mm area and 3% in the 1 mm area. Thus, the algorithm offers a fast and simple tool to analyse the PRL, as it was able to reproduce the results of retinal expert grading.

Our most important discovery was a significant decrease in the PRT one week after vitrectomy in the central millimetre, 3 mm and 6 mm area in the macula. Moreover, the decrease in PRT was associated with an increase in CRT one week after surgery.

There are multiple hypotheses that may explain our results. The changes in PRT at week 1 might stem from the traction release after membrane peeling. The ERM exerts a tractional force on the inner retinal surface that is transmitted to the entire retina through the Müller cell network and the ELM and eventually stretches the photoreceptors [21, 22]. The surgical removal of the ERM drastically reduces the traction on the Müller cells, which, as an immediate response, might result in an initial contraction of the photoreceptors. We observed a prolonged initial decrease in PRT in the foveal centre, which was termed “Müller cell cone” by Gass and is mostly composed of intertwining cone cells and Müller cells [23, 24]. This effect seems plausible when taking into account that the traction exerted onto the photoreceptors by the Müller cells is strongest in the foveal centre. Furthermore, photoreceptors in this area react especially sensitive to mechanical stress [25–27]. Nevertheless, regeneration of the PRL seems to become more homogenous over time, which is reflected in the observed increase of PRT in the 3 mm and 6 mm area after week 1.

The increase of CRT that is correlated with a PRT decrease of more than 6.6 µm from baseline to week 1 (Fig. 3) could be an indication of a more pronounced trauma response. Foveal Müller cells seem to respond inversely to the traction release, which might be caused by the fact that Müller cells are mechanosensitive and respond to retinal stretch with changes in intracellular calcium levels and protein expression [26, 28]. Bringmann et al. also proposed that stiffening and displacement of the Müller cells accounts for the absence of regeneration of the normal foveal structure [26]. Similar to our results, Ruberto et al. found out that CRT decreased only in patients with intact IS/OS junction integrity 3 months after surgery and Miyamoto et al. observed an increase of CRT and ellipsoid zone disruptions 3 months after membrane peeling in a patient whose BCVA worsened after surgery [29, 30].

However, the surgical procedure itself may also account for our observations. Datlinger et al. observed a postoperative movement of the fovea already one day after surgery and proposed that the iatrogenic forces applied during surgery may constitute for the process of foveal movement [5]. Shiono et al. observed a decrease in PROS length one month after surgery that they attributed to perioperative traumatic mechanisms, for instance postoperative inflammation or intraoperative injury, which reflects the amount of reversible detriment that can be inflicted on the PRL [12]. We hypothesize that the decrease in PRT detected in our study might occur due to the same reasons suggested by Datlinger et al. and Shiono et al. Consequently, perioperative traumatic mechanisms may also account for the increase in CRT, which was correlated to the reduction of the PRT. This might also explain our highly interesting finding that the decrease in PRT from baseline to week 1 was greater if CRT increased, whilst the decrease was lesser if CRT decreased as well (Fig. 3). The PRL is perhaps more susceptible to perioperative trauma than all retinal layers in the macular centre combined, since its thickness decreased before we detected an increase in CRT. Thus, the extent of early changes in the thickness of the PRL might serve as an indicator for the severity of perioperative trauma.

Comparable to Shiono et al., who registered a decrease in PROS length one month after surgery, we observed a significant decrease in PRT in the central millimetre one month after surgery compared to baseline. PRT in the 3 mm and 6 mm area reapproached baseline values and did not significantly differ from those. Hence, it might very well be possible that the recovery of the PRL begins as early as one month after vitrectomy in the parafoveal and perifoveal region of the macula.

We detected a significant decrease in PRT three months after surgery compared to baseline in the 1 mm area. Our results do not align with those of Kuriyan et al. and Shiono et al., who observed no change in photoreceptor thickness and PROS length in the inner 1.5 mm and central subfield of the macula, respectively, around three months after surgery compared to baseline [12, 13]. Nevertheless, we observed a significant increase in PRT in the 6 mm area at month 3. We hypothesize that these changes might be due to a more rapid recovery of the perifoveal retinal region compared to the parafoveal and central foveal retinal region, of which the latter seems to still be affected three months after surgery. Treumer et al. as well as Zou et al. also proposed that the perifoveal retinal region recovers faster than the parafoveal and foveal region [31, 32]. In contrast to our results, Treumer et al. found no change of the PRL in particular and Zou et al. even observed a postoperative decline in perifoveal PRL thickness. These inconsistent results regarding the PRT may stem from differences in study design and follow-up period.

We did not detect any correlation between PRT and ERM stage or PRT and BCVA at any visit. Furthermore, baseline PRT in the 1 mm, 3 mm and 6 mm area did not correlate with BCVA at any visit after surgery. These findings further animate the debate whether changes in the PRL are correlated with BCVA in patients with ERM. Whilst some research supports this association [8, 12], our results are in line with multiple studies in which no significant correlation between these two parameters was ascertainable [9, 33, 34]. The missing correlation between PRT, ERM stage and BCVA leads us to agree with Colakoglu et al., who proposed that although metamorphopsia and loss of visual acuity are induced by pathologies in both the inner and outer retina, ERMs affect the inner retina and Müller cells to a higher degree than the outer retina and furthermore, changes in the outer retina are induced by those in the inner retina [35]. Nevertheless, our results also imply that patients with poor PRL status due to ERM shall not be discouraged from surgery, as they may still profit from it and achieve good functional rehabilitation thereafter. Lastly, a positive correlation between PRT in the 1 mm area and CRT was discovered at week 1, month 1 and month 3. PRT in the 3 mm and 6 mm area did not correlate with CRT. These observations might be explained by the fact that the PRT is, as described above, a main component of the CRT, and an increase of PRT in the central millimetre also leads to an increase in CRT, even if the thickness of the other retinal layers remains unchanged.

Several limitations to this study have to be acknowledged. Patients were not examined earlier than one week after surgery to avoid any misinterpretation of our results by minimizing the effect of other surgically induced factors such as corneal astigmatism and inflammation in the anterior chamber [36]. The follow-up time of three months was chosen so that our results would not interfere with cataract formation that might occur after vitrectomy in phakic patients that did not receive concomitant cataract surgery [37]. Our results show that three months only suffice to show partial regeneration of the PRL. Furthermore, some patients received concomitant cataract surgery, which induces inflammatory processes that might affect the posterior segment [38, 39]. However, it is still unclear if such responses outweigh those induced by vitreomacular surgery [40, 41]. Future studies should investigate long-term changes of the complete photoreceptor layer, whether its rehabilitation in the foveal centre occurs at a later time point and if additional cataract surgery affects the PRL. Although we were able to show that AI offers the capability to extensively evaluate the PRT in the macula, the resolution of SD-OCT machines used in the current clinical routine is not high enough to depict changes of the cellular structures of the individual photoreceptor and Müller cells. Moreover, it is uncertain which components of the photoreceptor cells are actually depicted on OCT in the ellipsoid zone and interdigitation zone termed by the consensus nomenclature [16, 17, 42–44]. Thus, further studies with higher resolution imaging techniques, such as adaptive-optics OCT, are urgently needed to gain insight into the changes of the individual cellular constituents of the PRL and Müller cells in more detail and to confirm the former named hypotheses in their entirety.

In conclusion, PRT was significantly reduced in the 1 mm, 3 mm and 6 mm area one week after surgery compared to baseline. PRT reapproached its baseline values in the 3 mm and 6 mm area one month after surgery, but remained reduced in the central millimetre until three months after surgery. In contrast, PRT in the 6 mm area increased three months after surgery. The difference of PRT in the 1 mm, 3 mm and 6 mm area between baseline and week 1 correlated negatively with the difference of CRT between the aforementioned points in time. A decrease of more than 6.6 µm in PRT in the 1 mm, 3 mm and 6 mm area lead to significantly different CRT values. PRT in the 1 mm area correlated with CRT 1 week, 1 month and 3 months after surgery. There was neither correlation between PRT in the 3 mm and 6 mm area and CRT, nor between PRT in the 1 mm, 3 mm and 6 mm area and BCVA or ERM stage at any visit. Baseline PRT did also not correlate with BCVA at any visit after surgery. Although the PRL is morphologically affected by ERMs and after their surgical removal, it is not correlated to visual acuity. Thus, patients with alterations in the PRL may still profit from vitrectomy and membrane peeling and achieve good functional rehabilitation thereafter.

Summary

What was known before

• Spectral domain-optical coherence tomography has become the gold standard in the assessment of epiretinal membrane ahead of surgery and thereafter. Inner and outer retinal structures of the macula are affected by epiretinal membranes and pars plana vitrectomy with combined membrane peeling.

What this study adds

• The thickness of the photoreceptor layer in the macula is significantly decreased one week after vitrectomy and membrane peeling. Whilst the foveal centre is affected until 3 months afterwards, rehabilitation of the parafoveal and perifoveal region already starts one month surgery. AI helps analysing the photoreceptor layer in the macula in a manner that proves elusive for manual annotation.

Author contributions

MH acquired and analysed patient data and wrote the original manuscript. MG provided input for the study design, performed surgery and interpreted patient data. JI acquired and analysed patient data. JB performed statistical analyses and contributed to writing the article. OL analysed and interpreted patient data. HB analysed patient data and provided his experience and knowledge for patient data interpretation. USE provided their input for the study design, reviewed and edited the manuscript. SS, the corresponding author of this article, designed the study, performed surgery, overviewed the conduction of the study, interpreted patient data, reviewed and edited the manuscript.

Data availability

Data are available from the authors upon reasonable request.

Competing interests

HB received research grants from Heidelberg Engineering and Apellis and speaker fees from Bayer, Roche, and Apellis. Stefan Sacu received funds from Bayer, Roche, Novartis, Iveric and Ionis. USE is a scientific consultant for Apellis, received grant support from Kodiak, Novartis, Apellis and RetInSight and received Patents/Royalty from RetInSight. JI, MH, MG, OL and JB report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.