This retrospective study included 165 eyes (114 patients) with resolved DME in the Department of Ophthalmology at Guangdong Provincial People's Hospital between January 2017 and May 2020. Initially, each eye received best-corrected visual acuity (BCVA) measurement with a decimal chart and clinical examination, such as spectral-domain OCT (SD-OCT) scanning (Spectralis; Heidelberg Engineering, Heidelberg, Germany), intraocular pressure (IOP) measurement, and slit-lamp biomicroscope. At 1, 3, and 6 months after the loading treatment, all included eyes underwent BCVA measurement and SD-OCT scanning. The study was approved by the Institutional Review Board of Guangdong Provincial People's Hospital and was conducted according to the Declaration of Helsinki. Since the study is about the retrospective analysis of outcomes of a standard DME treatment and no individual patient could be identified from the data, formal informed consent was waived.

Eyes with DME affecting the central fovea at baseline and resolved DME with complete integrity of EZ at 6 months were included. We included eyes that were untreated or received previous anti-vascular endothelial growth factor (VEGF) or pan-retinal photocoagulation (PRP) at least 6 months ago and eyes with the central foveal thickness (CFT) over 275 μm and BCVA between 0.3 and 1.0 logarithm of the minimal angle of resolution (LogMAR) (20/200–20/40) at baseline (12, 19). Eyes with macular edema secondary to other causes, such as retinal artery/vein occlusion, age-related macular degeneration, and polypoidal choroidal vasculopathy, eyes with macular ischemia, glaucoma or IOP > 21 mmHg, a history of macular grid photocoagulation or vitrectomy, refractive error over 6 diopters (D), severe cataracts, or previously treated with PRP or intravitreal or periocular injection <6 months were excluded (12, 17). We also excluded eyes with unsatisfactory SD-OCT images resulting from poor patient cooperation or media opacity.

All included eyes received a loading dose of 3 monthly consecutive 0.5-mg intravitreal injections of ranibizumab (IVR) treatment. After the loading treatment, patients were followed on monthly basis and received 1 IVR injection if they met any of the following criteria: (A) CFT increases by ≥ 100 μm; (B) BCVA decreases by ≥ 0.1 LogMAR; or (C) the decrease of BCVA attributed to newly formed SRD or intraretinal cyst, based on the surgeons; or (D) the decrease of BCVA due to enlargement of previous SRD or intraretinal cyst, based on the surgeons (9, 12, 17). IVR injection was suspended if either of the following criteria was met: (A) BCVA ≤ 0.0 LogMAR (20/20) observed at the two last consecutive follow-ups; or (B) stable BCVA over three consecutive follow-ups that include the current follow-up evaluation, specifically no improvement of BCVA due to IVR injections at the two last consecutive follow-ups (9, 12, 17).

A set of high-speed SD-OCT scans (Spectralis; Heidelberg Engineering, Heidelberg, Germany) was obtained using a custom 20° × 20° volume acquisition protocol. With this protocol, we can obtain 25 horizontal and central vertical cross-sectional B-scan images, each composed of 512 A-scans (12, 17). According to the morphologic characteristics on OCT examination, DME was classified into three types (i.e., SRD, DRT, and CME). SRD type was defined as an optically clear space between the retinal pigment epithelium (RPE) and retina and a shallow elevation of the retina (9, 12, 14). DRT type was defined as sponge-like retinal swelling of the macula with reduced intraretinal reflectivity (9, 12, 14). CME type was defined as highly reflective septa separating cystoid-like cavities and low reflective intraretinal cystoid spaces in the macular area (9, 12, 14). Three DME types were further classified into two categories: SRD and non-SRD (i.e., DRT and CME) according to their different pathogenesis (17). Two ophthalmologists (QW and BL) exported the horizontal SD-OCT images through the fovea of all the included eyes for independent reading and measurement of CFT, outer nuclear layer (ONL) thickness, height of SRD (SRDH) and non-SRD (NSRDH), and inner segment (IS) and outer segment (OS) length (Figures 1, 2) (12). If there were discordance between the two ophthalmologists, arbitration was performed by a retinal specialist (HY) to generate the final decision.

Based on the existence of FB at 6 months after the loading treatment, the included eyes were divided into the FB (+) group and the FB (–) group. As a qualitative OCT parameter, FB was defined as a dome-shaped structure of the EZ where the length of OS located at the central fovea is ≥10 μm longer than the mean OS length located at 250 μm nasal and temporal from the central fovea (Figure 1) (10, 12). SD-OCT images of typical cases of the two groups are shown in Figures 3, 4.

The two ophthalmologists manually measured CFT, ONL thickness, SRDH, NSRDH, IS length, and OS length at the central fovea in a masked manner (Figures 1, 2). Mean values of OCT measurements of the two ophthalmologists were used for statistical analysis. CFT was measured as the vertical distance between the outer border of RPE and the surface of the internal limiting membrane (ILM) (10, 12, 17). SRDH was measured as the vertical distance at the fovea between the anterior boundary of the RPE-choriocapillaris region and the top of subretinal fluid, while NSRDH was measured as the vertical distance at the fovea between the top of subretinal fluid and the surface of ILM (12, 17). ONL thickness was measured as the vertical distance between the outer border of ELM and the outer border of ILM (10, 12). IS length was measured as the vertical distance between the outer border of EZ and the outer border of ELM (10, 12). OS length was measured as the vertical distance between the inner border of RPE and the outer border of EZ (10, 12).

Best-corrected visual acuity with a decimal chart was converted to the LogMAR for statistical analyses. All values are presented as mean ± SD. We used the SPSS 20.0 (SPSS. Inc, Chicago, IL, USA) to perform all statistical analyses and calculated intra-class coefficient (ICC) values to assess the reproducibility of OCT measurements between the two ophthalmologists (QW and BL). The mean values and frequency of parameters between the two groups were compared using the unpaired Mann-Whitney test and the chi-square Test, respectively, after confirming the data normality.

Univariate and multinomial logistic regression analyses were performed to identify possible baseline predictors for the presence of FB and to evaluate the correlations between the presence of the FB and post-treatment factors. In the univariate analysis, the association between the detection of the FB and each variable was solely examined. Subsequently, stepwise backward multivariate regression analysis was performed with the following parameters: age, baseline BCVA, baseline CFT, SRDH, detection of SRD and disrupted ELM at baseline, post-treatment BCVA, and post-treatment OS length. The variables significantly associated with the detection of the FB in the univariate analysis or showing statistically significant differences between the two groups in the unpaired Mann-Whitney test and the chi-square Test were introduced as independent variables. The multinomial logistic regression analysis generates an odds ratio (OR) for each category of the dependent variable relative to the reference category. The OR value includes the 95% CI allowing estimating the degree of accuracy. All continuous variables were categorized as the quartile-categorical variables for univariate and multinomial logistic regression analyses. For all the tests and analyses, a value of P < 0.05 was considered statistically significant.

This study included 165 eyes with complete integrity of EZ at the central fovea and completely resolved DME. At 6 months, FB was present in 65 (39.39%) eyes. In the FB (+) group, there were 47 eyes (72.31%) received previous anti-VEGF or PRP at least 6 months ago, compared to 79 eyes (79.00%) in the FB (–) group (P = 0.323). There were no significant differences in other baseline characteristics between the FB (+) group and the FB (–) group (Table 1). Moreover, the proportion of eyes from the FB (+) group was not significantly different between eyes with and without prior treatment (37.30% vs. 46.15%, P = 0.323). The interobserver ICC for the measurement of ONL thickness was 0.914, for IS length was 0.855, and for OS length was 0.872, indicating that the OCT measurements between the two ophthalmologists (QW and BL) with good reproducibility.

At baseline, BCVA was significantly better, and CFT, SRDH, incidence of SRD, and disrupted ELM were significantly lower in the FB (+) group than the FB (–) group. NSRDH has no significant difference between the two groups (Table 2). BCVA of the FB (+) group was 0.38 ± 0.18 LogMAR, and BCVA of the FB (–) group was 0.59 ± 0.19 LogMAR (P < 0.001). CFT of the FB (+) group was 442.65 ± 155.87 μm, and CFT of the FB (–) group was 548.08 ± 169.83 μm (P < 0.001). SRDH of the FB (+) group was 94.54 ± 127.82 μm, and SRDH of the FB (–) group was 161.28 ± 186.71 μm (P = 0.034).

At 6 months, BCVA was significantly better, and OS length was significantly longer in the FB (+) group than the FB (–) group. CFT, ONL thickness, and IS length at 6 months were not significantly different between the two groups (Table 2). BCVA of the FB (+) group was 0.10 ± 0.16 LogMAR, and BCVA of the FB (–) group was 0.37 ± 0.16 LogMAR (P < 0.001). There were 40 eyes with a BCVA ≤ 0.0 LogMAR, and 33 (82.50%) eyes with FB, compared to 7 (17.50%) eyes without FB (P < 0.001). At 6 months, the OS length of the FB (+) group was 44.09 ± 5.62 μm, and the OS length of the FB (–) group was 32.96 ± 2.99 μm (P < 0.001). At 6 months, FB was present in 25 (31.25%) eyes with SRD at baseline, compared to 40 (47.06%) eyes without SRD at baseline (P = 0.038).

The multivariate regression analysis indicated better baseline BCVA, lower baseline CFT, and younger age as significant predictors of the existence of FB at 6 months (Table 3). Better baseline BCVA has the most important contribution to the existence of FB, with an OR of 0.278 (P < 0.001) for the FB (+) group vs. the FB (–) group. Moreover, patients with lower baseline CFT (OR = 0.616, P = 0.007) and younger age (OR = 0.655, P = 0.020) were also significantly associated with the existence of FB at 6 months. The correlations between the existence of FB and post-treatment factors after the resolution of DME were determined by univariate and multivariate logistic regression analysis. The existence of FB was negatively correlated with the post-treatment BCVA (OR = 0.453, P = 0.005) and positively correlated with post-treatment OS length (OR = 11.314, P < 0.001) in the multivariate regression analysis (Table 3).