The preoperative thickness of each retinal layer, including the inner retinal layer and ONL, is a significant factor in the postoperative visual prognosis of ERM patients. Thus, accurate measurement of each thickness is crucial. However, studies on the reliability of each retinal layer thickness measurement in ERM patients have been insufficient so far. Furthermore, to the best of our knowledge, there has been no report on how low the reliability of automatic measurement is in ERM patients with EIFLs, compared to those without EIFLs. We found that the measurement of each retinal layer thickness in the foveal area showed low short-term reproducibility in ERM patients with EIFLs. In patients without EIFLs, inner retinal layer thickness measurements had good short-term reproducibility compared to the measurements in patients with EIFLs, but outer retinal layer thicknesses, including OPL and ONL, were less reproducible. Additionally, the short-term reproducibility of the inner retinal layers was significantly correlated with the average CMT in patients with EIFLs, and the difference between the two measurements of the inner retinal layer was significantly associated with the average CMT.

Previous studies have reported good repeatability and reproducibility of CMT measurements using SD-OCT in ERM patients. Lee et al. [18] reported that the ICC and CV of CMT in ERM patients were 0.995 and 0.8%, which was relatively good reproducibility. Pinilla et al. [14] also reported that macular thickness measurements showed a mean CV of 2.95% with a higher ICC than 0.919 using SD-OCT in ERM patients. In our study, the CV and ICC were 1.49% and 0.964 in Group 1, and 1.70% and 0.962 in Group 2, respectively, consistent with previous reports. CMT measurements using SD-OCT in ERM patients are considered highly reliable, regardless of the presence of EIFLs.

On the other hand, most of the retinal layers had low reproducibility in patients with EIFLs, with CVs > 11.00% and ICCs < 0.550. Additionally, the inner retinal layer, including NFL, GCL, and IPL, of patients with EIFLs were more scattered on Bland-Altman plots with larger reproducibility limits compared to patients without EIFLs. Although a previous study reported an ICC of 0.881 for the ganglion cell-inner plexiform layer (GCIPL) in ERM patients, suggesting good reproducibility, this could be attributable to the enrollment of patients without EIFLs, which also showed relatively good reproducibility in our study [18]. In patients with EIFLs, the boundaries of the retinal layers are often tortuous and unclear, and could be blurred and indistinguishable on OCT images, which can cause segmentation errors. Thick EIFLs can be an indication of surgery, but inaccurate measurements by these segmentation errors may interfere with treatment decisions and can be a confounding factor in predicting postoperative prognosis. Therefore, physicians should not rely on automatic retinal layer thickness measurements without first checking the segmentation accuracy of OCT images, and manual adjustments should be performed when obvious segmentation errors are found especially in patients with EIFLs.

Here, the short-term reproducibility of the inner retinal layer thickness measurements in patients without EIFLs was better than in patients with EIFLs. The relatively high CV values of the NFL may be the result of a thinner NFL compared to other retinal layers around the fovea because the CV was calculated as a standard deviation/mean value. Unlike the inner retinal layers’ measurements, which showed relatively good reproducibility, OPL and ONL measurements showed low reproducibility in patients without EIFLs. In patients with advanced ERM without EIFLs, an abnormally wide attachment to the ILM of the ONL and bulging ONL were identified in many OCT images. In these cases, it tends to frequently fail in segmenting the bulged parts well, which could result in unreliable measurements. Apart from the inner retinal layer, the outer retinal layer thickness is also significantly associated with visual function in ERM patients. Arichika et al. [19] reported that outer retinal thickening in the fovea, parafovea, and perifovea was significantly correlated with visual acuity in patients with idiopathic ERM. Cacciamani et al. [20] also found a significant correlation between retinal sensitivity impairment and ONL thickness. Therefore, accurate measurement of the outer retinal layer thickness, including OPL and ONL, is important in ERM patients, and dependence on automatic measurements of outer retinal layer thickness can lead to inaccurate visual prognosis analysis.

Lee et al. [21] reported that the repeatability of GC-IPL thickness measurements was lower in patients with macular edema caused by various retinal diseases including age-related macular degeneration, retinal vein occlusion, central serous chorioretinopathy, and diabetic macular edema. Another study reported low repeatability of GC-IPL measurement in ERM patients with a CMT greater than 450 μm. The proposed explanation was based on frequent auto segmentation errors following the more distorted configuration of the macula, compared with patients with a CMT less than 450 μm [18]. Our study also demonstrated a significant negative correlation between CMT and the short-term reproducibility of inner retinal layer thickness measurements, including NFL, GCL, and IPL, in patients with EIFLs. This was supported by the significant association of the difference between the two measurements of inner retinal layer thickness with the average CMT. However, no significant correlation between the average CMT and reproducibility of GCL and IPL was seen in ERM patients without EIFLs, and the difference between the two measurements was not significantly associated with the average CMT. Therefore, in cases with thick CMT, the importance of manual measurement in patients with EIFLs is more emphasized than in those without EIFL for accurate visual prognosis analysis and adequate explanation for patients because of the low reliability of automatic measurement.

This study has several limitations. First, the retrospective nature of the work inevitably introduced some selection bias. Second, the measurements were acquired with only the Spectralis OCT device; there could be subtle differences among the OCT devices currently in use. Additionally, there are several methods of measuring retinal layer thickness in Spectralis OCT such as a 20° × 20° capture mode with 25 fast scans or the posterior pole algorithm, which can result in varying reproducibility. Fourth, this study demonstrated the short-term reproducibility of OCT in ERM patients with examination intervals of less than one week. However, subtle changes in ERM may occur in a week, and the intervals between the examinations were different among the patients, which could cause some bias. The main strength of this study was that we evaluated the short-term reproducibility of each retinal layer thickness measurement, including the outer retina, in ERM patients, which has rarely been reported. This was also the first study to statistically analyze the short-term reproducibility of each retinal layer thickness measurement in patients with EIFLs and to compare it with that of measurement in patients without EIFLs.