We thank Lee et al1 for their letter and interest in our article.2
The authors commented on missing a test of repeatability of the different optical coherence tomography (OCT) devices in our article. Yet, the focus of our study was not the repeatability of the respective devices, as the question of repeatability of the devices in the context of various ophthalmic diseases has already been investigated in detail by several former studies, and showed satisfactory intradevice reproducibility.3–7 Besides that, Lee et al emphasized that low visual acuity might result in an inability of our patients to gaze at the fixation target during the examination period. All the eligible patients were able to focus on the fixation target during the imaging procedure, and we included only the scans centered on the fovea.
In addition, they report to have shown that the ganglion cell inner plexiform layer (GCIPL) thickness of the group of patients with macular edema was thinner compared with the control group in their latest study, and that they have attributed this incidence to a high frequency of segmentation errors.8 We did not include a control group in our series. Therefore, a comparison of GCIPL thicknesses between healthy control eyes and patients with diabetic macular edema (DME) was not addressed in our study, and we did not discuss or suggest a pathogenic reduction of GCIPL thickness in these patients with DME.
Lee et al8 also emphasized that they and Keane et al9 have shown previously that OCT measurements in retinal diseases involving the macular area may be misjudged as a result of segmentation errors due to structural damage in the presence of intraretinal cysts, exudates, or subretinal fluid. In response, as reported in our methods section, we excluded images with visible eye motion and blinking artifacts or those with poor image quality, defined as signal strength <6 of 10 in Cirrus HD-OCT or a signal-to-noise ratio <25 of 40 dB in Spectralis OCT. As explained in our discussion, these thresholds were chosen, as the manufacturer claims that the analysis algorithms of their devices work accurately, given that the proposed signal quality is reached, and the purpose of our study was to compare the performance of these individual software algorithms for GCIPL analysis regardless of possible segmentation errors, that we additionally discussed and outlined in Table 1. We excluded these lower-quality images to reduce other confounding factors that may then be accountable for the low performance of the individual algorithms. Regarding the goal of our study to compare the GCIPL thickness results of two different devices, we intentionally did not exclude those with visible segmentation errors. However, we would like to emphasize that segmentation errors might contribute to deviant thickness measurements. The true impact of segmentation errors still needs to be quantified in separate analyses.
The authors suggest to additionally perform a measurement of the GCIPL with the two OCT devices in healthy eyes3,10 and to compare the results with GCIPL thicknesses of our patients with DME. However, we emphasize again that the aim of this study was to assess the interdevice agreement of GCIPL thickness values in patients with DME, as DME represents a frequent pathology in patients with diabetes in clinical practice. An accurate measurement of GCIPL would be particularly crucial in patients with diabetes. Reduced GCIPL thickness might be a potential biomarker providing insight into the development of diabetic neurodegeneration.