Early dynamic response on spectral-domain OCT predicts long-term visual outcome in diabetic macular edema treated with aflibercept Authors: Aiping Deng, Yuhe Gan, Junyan Da, Qiuyan Ning, Run Tian Published: 2026 ABSTRACT AbstractObjectiveA retrospective evaluation of the long-term visual recovery prediction in patients with diabetic macular edema (DME) following aflibercept treatment, based on spectral-domain optical coherence tomography (SD-OCT) parameters at one and four months post-treatment.MethodsThis study involved 112 eyes of 112 DME patients which received at least one injection of aflibercept. SD-OCT parameters at 1 and 4 months post-aflibercept treatment-including central subfield thickness (CST), the continuity of external limiting membrane (ELM)/ellipsoid zone (EZ) and cone outer segment tip (COST), as well as the counts of retinal hyperreflective retinal foci (HRF), along with best-corrected visual acuity (BCVA) improvement at 12 months post-treatment-were analyzed. The correlation between SD-OCT parameters and BCVA improvement was evaluated using the chi-square or Wilcoxon test, and the predicted value of SD-OCT parameters was assessed via receiver operating characteristic (ROC) curve and binary logistic regression model.ResultsTwelve months after aflibercept treatment, BCVA was improved from 0.66 ± 0.46 to 0.48 ± 0.42 logMAR (p = 0.002). The SD-OCT parameters were significantly improved: CST was decreased from 397.05 ± 199.60 to 289.08 ± 119.39 μm; the affected eyes with damaged ELM/EZ and COST decreased from 46 (41.1%) and 71 (63.4%) to 31 (27.7%) and 43 (38.4%), respectively; the number of HRF decreased from 11.00 ± 11.38 to 8.03 ± 9.58. The AUC values for predicting BCVA gain of ≥ 5 letters at 12 months based on SD-OCT parameters at 1 and 4 months post-aflibercept treatment and their combination were 0.812, 0.824, and 0.873, respectively. Multivariate logistic analysis revealed that CST, ELM/EZ, COST, and HRF at 1 and 4 months were all independent predictors of long-term visual improvement outcomes.ConclusionIntravitreal injection of aflibercept can effectively improve the vision and fundus structure of DME patients. Improvements in short- and medium-term CST, ELM/EZ, COST integrity, and HRF values are key physiological indicators for visual recovery.Keywords: Aflibercept, Diabetic macular edema, Spectral-domain optical coherence tomography, Best-corrected visual acuity, Retinal microstructures Introduction Diabetic macular edema (DME) is the most common cause of visual impairment resulting from diabetic retinopathy. Its pathological features include disruption of the retinal barrier and leakage of retinal vessels, leading to fluid accumulation in the macular area [1, 2]. The anti-vascular endothelial growth factor (VEGF) drugs, such as aflibercept, have become the first-line treatment strategy for DME, effectively reducing vascular leakage and improving vision [3]. Aflibercept is a recombinant fusion protein that can bind to VEGF-A, VEGF-B and placental growth factor and inhibit their functions, thereby blocking angiogenesis [4]. The integrity of the retinal microstructure, particularly the integrity of the neurosensory retinal layer, outer retinal layer, the external limiting membrane (ELM), and the ellipsoid zone (EZ), is crucial for maintaining normal visual function [5]. Once these structures are damaged, it will lead to irreversible vision loss if not treated promptly. Previous studies have confirmed that aflibercept can effectively alleviate DME. However, most of these studies have focused on changes in central subfield thickness (CST) and vision [6–8], and there is relatively little research on the specific improvement of the retinal microstructure, especially the repair of ELM and EZ. The visual prognosis of DME patients shows significant individual differences. In recent years, spectral-domain optical coherence tomography (SD-OCT) has been able to display the microscopic structures of the retina with high resolution, providing a new perspective for understanding the pathophysiology and treatment response of DME. Hyperreflective retinal foci (HRF) and the disruption of ELM/EZ and cone outer segment tip (COST) integrity has been proven to be related to the initial visual acuity and prognosis of DME patients [9–12]. However, more evidence is still needed regarding how aflibercept treatment specifically affects these microscopic structures, and the causal relationship between these structural changes and visual improvement. This study aims to evaluate the impact of aflibercept treatment on SD-OCT parameters in patients with DME, and to identify which improvements in microscopic structures are influence factors of visual improvement. Methods Research design and patient inclusion This study was a single-center retrospective study. A total of 112 patients diagnosed with DME and treated with aflibercept in Affiliated Hospital of Yunnan University were included. Inclusion criteria: (1) Diagnosed with type 2 diabetes and DME; (2) Newly diagnosed or more than 3 months since the last laser treatment and no prior other anti-VEGF therapy. Exclusion criteria: (1) Complicated with other eye diseases affecting vision; (2) Underwent vitreoretinal surgery during the follow-up period; (3) Had severely turbid refractive medium affecting the quality of SD-OCT imaging. A total of 112 eyes were finally included. For both affected eyes of the same patient, we randomly selected one eye for analysis. Treatment plan and follow-up All patients initially received an injection of 2 mg aflibercept (Regeneron Pharmaceuticals, Inc., USA) once a month for a total of three consecutive times. Monthly SD-OCT and fundus examinations were conducted. If the patient’s best-corrected visual acuity (BCVA) improved by ≥ 5 letters and CST decreased by ≥ 20%, the treatment was suspended. If CST increased by more than 10% compared to the lowest value, or macular edema recured, supplementary injections were given. BCVA examination BCVA was measured using the ETDRS visual acuity chart at a distance of 4 m, and the visual acuity was converted to logMAR for statistical analysis. logMAR refers to the logarithmic value of the minimum resolvable angle with base 10. The minimum resolvable angle refers to the smallest angle between two points that the eye can distinguish. In scientific research and clinical trials, logMAR can detect subtle yet clinically significant changes in visual acuity. The lower the logMAR score, the better the vision. Conversely, the higher the score, the poorer the vision. SD-OCT data collection The macular area centered on the fovea was scanned using SD-OCT (Heidelberg Engineering, Germany) and macular imaging of the 20 × 20 scanning area (49 grating B-scan lines, 1024 A-scans per B-scan lines) was obtained. The images were analyzed by three experienced ophthalmologists who independently measured the following parameters without any knowledge of other clinical information, and the average value was taken: CST: This refers to macular central thickness, which is the average thickness within a 1 mm radius of the macula center. ELM/EZ and COST continuity: ELM, EZ and COST were defined as the first three hyper-reflective bands in the outermost layer of the SD-OCT image. ELM and EZ were evaluated as a whole. The assessment of ELM/EZ and COST integrity was conducted by two independent, experienced retinal specialists who were unaware of the patients’ clinical information and group assignments. Any discernible interruption or omission is deemed a damage. For the ELM/EZ integrity assessment, the agreement rate between two observers was 90.0%, with a Cohen’s kappa value of 0.83 (95% CI: 0.77–0.96). For the COST integrity assessment, the agreement rate was 88.9% with a kappa value of 0.80 (95% CI: 0.67–0.89), indicating high agreement. The number of HRF: HRF was the hyperreflective foci, small bright spots with clear boundaries, no artifacts, a diameter of ≤ 30 μm, and a reflectivity of ≥ retinal pigment epithelium (RPE) located between the internal limiting membrane and RPE. Statistical analysis Statistical analysis was conducted using SPSS 26.0 software. Continuous variables were expressed as mean ± standard deviation. The Wilcoxon test was used to compare the changes in continuous variables. The chi-square test was used to compare the differences between groups of categorical variables. Receiver operating characteristic (ROC) curve and binary logistic regression analysis was used to evaluate the influencing factors of BCVA improvement of ≥ 5 letters [13–15]. A p value < 0.05 was considered statistically significant. Results Baseline characteristics A total of 112 eyes were included in this study, and the overall baseline BCVA was 0.66 ± 0.46 logMAR. The baseline characteristics of the patients were shown in Table 1. The patients were divided into two groups based on the mean BCVA values at 12 months following aflibercept treatment. In the poorer vision group (BCVA > 0.48), a thicker CST was observed, the integrity of ELM/EZ and COST was more frequently impaired, and the number of HRF was greater (Table 1), suggesting that the changes in these microstructures may be the key to visual impairment and recovery in patients with DME. Table 1The baseline characteristics of DME patientsCharacteristicsTotal(n = 112 eyes)BCVA ≤ 0.48 logMAR(n = 64 eyes)BCVA > 0.48 logMAR(n = 48 eyes)*pAge (years)59.38 ± 10.5658.00 ± 9.2261.21 ± 11.090.114Gender (female/male)47/6531/3316/320.109DM duration (years)12.05 ± 5.9511.23 ± 5.3913.15 ± 6.460.094Laser treatment history (yes/no)67/4535/2932/160.201Aflibercept treatment (times)4.51 ± 2.904.83 ± 2.804.08 ± 2.980.182BCVA (logMAR)0.66 ± 0.460.51 ± 0.380.87 ± 0.48< 0.001HbA1c (%)7.50 ± 1.527.30 ± 1.537.78 ± 1.470.101CST (µm)397.05 ± 199.60384.41 ± 184.88413.92 ± 216.530.443ELM/EZ damage (yes/no)46/6617/4729/19< 0.001COST damage (yes/no)71/4133/3138/100.003HRF (number)11.00 ± 11.388.69 ± 10.7414.08 ± 11.490.013DME, diabetic macular edema. BCVA, best-corrected visual acuity. DM, diabetes mellitus. HbA1c, glycated hemoglobin. CST, central subfield thickness. ELM, external limiting membrane. EZ, ellipsoid zone. COST, cone outer segment tip. HRF, hyperreflective retinal foci*p, Wilcoxon test or chi-square test The baseline characteristics of DME patients DME, diabetic macular edema. BCVA, best-corrected visual acuity. DM, diabetes mellitus. HbA1c, glycated hemoglobin. CST, central subfield thickness. ELM, external limiting membrane. EZ, ellipsoid zone. COST, cone outer segment tip. HRF, hyperreflective retinal foci *p, Wilcoxon test or chi-square test The effectiveness of aflibercept treatment Prior to treatment, patients with DME had a baseline BCVA of 0.66 ± 0.46 logMAR, CST thickness of 397.05 ± 199.60 μm, ELM/EZ and COST damage of 46 (41.1%) and 71 (63.4%), respectively, and an HRF count of 11.00 ± 11.38 (Table 2). One month and four months after aflibercept treatment: BCVA was improved to 0.61 ± 0.39 and 0.55 ± 0.41 logMAR, respectively; CST thickness decreased to 324.45 ± 130.41 and 296.70 ± 115.29 μm, respectively; ELM/EZ damage decreased to 44 (39.3%) and 45 (40.2%); COST damage decreased to 65 (58.0%) and 50 (44.6%); and the number of HRFs decreased to 6.28 ± 8.02 and 7.63 ± 9.66 (Table 2). Twelve months after aflibercept treatment, BCVA improved to 0.48 ± 0.42 logMAR (an improvement of 9 letters, p = 0.002), CST decreased to 289.08 ± 119.3 μm (p < 0.001), ELM/EZ and COST lesions reduced to 31 (27.7%, p = 0.035) and 43 (38.4%, p < 0.001), respectively, and HRF decreased to 8.03 ± 9.58 (p = 0.036). Table 2Changes in SD-OCT parameters before and after aflibercept treatmentParametersBaseline1 month4 months12 months*pBCVA (logMAR)0.66 ± 0.460.61 ± 0.390.55 ± 0.410.48 ± 0.420.002CST (µm)397.05 ± 199.60324.45 ± 130.41296.70 ± 115.29289.08 ± 119.39< 0.001ELM/EZ damage (yes/no)46/6644/6845/6731/810.035COST damage (yes/no)71/4165/4750/6243/69< 0.001HRF (n)11.00 ± 11.386.28 ± 8.027.63 ± 9.668.03 ± 9.580.036BCVA, best-corrected visual acuity. CST, central subfield thickness. ELM, external limiting membrane. EZ, ellipsoid zone. COST, cone outer segment tip. HRF, hyperreflective retinal foci*p, Wilcoxon test or chi-square test (baseline vs. 12 months) Changes in SD-OCT parameters before and after aflibercept treatment BCVA, best-corrected visual acuity. CST, central subfield thickness. ELM, external limiting membrane. EZ, ellipsoid zone. COST, cone outer segment tip. HRF, hyperreflective retinal foci *p, Wilcoxon test or chi-square test (baseline vs. 12 months) The SD-OCT images showed that, after aflibercept treatment, pathological cystoid spaces disappeared, and the continuity of the ELM/EZ and COST was significantly improved (Fig. 1A). The retinal detachment caused by subretinal fluid resolved, and the number of HRF lesions decreased (Fig. 1B). These indications suggest that aflibercept’s effects on these retinal microstructures may be related to the vision improvement of patients with DME. SD-OCT images show the changes in fundus microstructure before and after aflibercept treatment. The effect of aflibercept in treating cystic macular edema (A) and serous retinal detachment (B) Factors influencing the outcome of vision According to the BCVA gain in the treated eye after 12 months of aflibercept therapy, they were divided into the slightly poor visual recovery group (BCVA gain < 5 letters) and the good visual recovery group (BCVA gain ≥ 5 letters). The chi-square test or Wilcoxon test indicated that changes in CST, ELM/EZ damage, and COST damage at 1 and 4 months post-treatment may be potentially correlated with the final visual recovery in patients with DME (Tables 3 and 4). Although no statistically significant difference in HRF was observed one month later, a trend toward a decrease in HRF number was noted alongside improvements in BCVA (Table 3). ROC analysis revealed that the AUCs for predicting long-term visual recovery based on SD-OCT parameters (CST, ELM/EZ, COST, HRF) at one month and four months post-treatment were 0.812 (sensitivity = 71.9%, specificity = 83.3%) and 0.824 (sensitivity = 79.9%, specificity = 75.0%), respectively (Fig. 2). The AUC for the combined prediction of visual recovery using SD-OCT parameters at these two time points was 0.873 (sensitivity = 78.1%, specificity = 85.4%), demonstrating more reliable predictive value. Further multivariate logistic analysis revealed that CST, ELM/EZ, COST, and HRF at 1 and 4 months post-treatment were all independent predictors of visual recovery at 12 months after adjusting for age, gender, diabetes mellitus (DM) duration, laser treatment history, number of aflibercept treatments, and HbA1c levels (Tables 5 and 6). Table 3The relationship between the parameters at one-month-node of aflibercept treatment and the vision improvement after 12 monthsParametersBCVA gain < 5 letters(n = 64 eyes)BCVA gain ≥ 5 letters(n = 48)*pAge (years)59.13 ± 11.5759.71 ± 9.030.775Gender (female/male)28/3619/290.658DM duration (years)11.61 ± 5.6612.65 ± 6.260.366Laser treatment history (yes/no)35/2932/160.201Aflibercept treatment (times)4.92 ± 2.813.96 ± 2.930.084HbA1c (%)7.39 ± 1.467.66 ± 1.590.356CST (µm)352.66 ± 152.75286.83 ± 78.100.008ELM/EZ damage (yes/no)37/277/410.001COST damage (yes/no)49/1516/320.001HRF (number)6.45 ± 8.246.04 ± 7.710.791BCVA, best-corrected visual acuity. DM, diabetes mellitus. HbA1c, glycated hemoglobin. CST, central subfield thickness. ELM, external limiting membrane. EZ, ellipsoid zone. COST, cone outer segment tip. HRF, hyperreflective retinal foci*p, Wilcoxon test or chi-square test The relationship between the parameters at one-month-node of aflibercept treatment and the vision improvement after 12 months BCVA, best-corrected visual acuity. DM, diabetes mellitus. HbA1c, glycated hemoglobin. CST, central subfield thickness. ELM, external limiting membrane. EZ, ellipsoid zone. COST, cone outer segment tip. HRF, hyperreflective retinal foci *p, Wilcoxon test or chi-square test Table 4The relationship between the parameters at 4-month-node of aflibercept treatment and the vision improvement after 12 monthsParametersBCVA gain < 5 letters(n = 64 eyes)BCVA gain ≥ 5 letters(n = 48)*pAge (years)59.13 ± 11.5759.71 ± 9.030.775Gender (female/male)28/3619/290.658DM duration (years)11.61 ± 5.6612.65 ± 6.260.366Laser treatment history (yes/no)35/2932/160.201Aflibercept treatment (times)4.92 ± 2.813.96 ± 2.930.084HbA1c (%)7.39 ± 1.467.66 ± 1.590.356CST (µm)330.00 ± 131.48252.29 ± 67.170.003ELM/EZ damage (yes/no)35/2910/38< 0.001COST damage (yes/no)38/2612/36< 0.001HRF (number)7.80 ± 10.517.40 ± 8.380.001BCVA, best-corrected visual acuity. DM, diabetes mellitus. HbA1c, glycated hemoglobin. CST, central subfield thickness. ELM, external limiting membrane. EZ, ellipsoid zone. COST, cone outer segment tip. HRF, hyperreflective retinal foci*p, Wilcoxon test or chi-square test The relationship between the parameters at 4-month-node of aflibercept treatment and the vision improvement after 12 months BCVA, best-corrected visual acuity. DM, diabetes mellitus. HbA1c, glycated hemoglobin. CST, central subfield thickness. ELM, external limiting membrane. EZ, ellipsoid zone. COST, cone outer segment tip. HRF, hyperreflective retinal foci *p, Wilcoxon test or chi-square test ROC curves for predicting long-term BCVA improvement based on SD-OCT parameters at 1 and 4 months post-aflibercept treatment Table 5Multivariate logistic regression analysis of factors predicting improved vision after 12 months based on one-month-parametersParametersOR95% CI p lowhighAge0.9890.9261.0560.734Gender2.4070.8376.9230.103DM duration0.9640.8421.1050.600Laser treatment history1.4120.3585.5770.622Aflibercept treatment1.0420.8681.2510.660HbA1c0.7970.5571.1400.214CST1.0091.0021.0160.016ELM/EZ9.2401.85046.1380.007COST5.3521.51018.9700.009HRF0.8800.8050.9630.005OR, odds ratio. CI, confidence interval. DM, diabetes mellitus. HbA1c, glycated hemoglobin. CST, central subfield thickness. ELM, external limiting membrane. EZ, ellipsoid zone. COST, cone outer segment tip. HRF, hyperreflective retinal foci Multivariate logistic regression analysis of factors predicting improved vision after 12 months based on one-month-parameters OR, odds ratio. CI, confidence interval. DM, diabetes mellitus. HbA1c, glycated hemoglobin. CST, central subfield thickness. ELM, external limiting membrane. EZ, ellipsoid zone. COST, cone outer segment tip. HRF, hyperreflective retinal foci Table 6Multivariate logistic regression analysis of factors predicting improved vision after 12 months based on 4-month-parametersParametersOR95% CI p lowhighAge0.9870.9331.0440.643Gender1.7960.6624.8750.251DM duration0.9710.8551.1030.655Laser treatment history0.7300.1812.9420.658Aflibercept treatment1.1470.9731.3530.103HbA1c0.7730.5461.0940.146CST1.0091.0031.0150.004ELM/EZ4.9101.48216.2640.009COST5.5361.76217.3900.003HRF0.9350.8790.9930.030OR, odds ratio. CI, confidence interval. DM, diabetes mellitus. HbA1c, glycated hemoglobin. CST, central subfield thickness. ELM, external limiting membrane. EZ, ellipsoid zone. COST, cone outer segment tip. HRF, hyperreflective retinal foci Multivariate logistic regression analysis of factors predicting improved vision after 12 months based on 4-month-parameters OR, odds ratio. CI, confidence interval. DM, diabetes mellitus. HbA1c, glycated hemoglobin. CST, central subfield thickness. ELM, external limiting membrane. EZ, ellipsoid zone. COST, cone outer segment tip. HRF, hyperreflective retinal foci Discussion This study aims to investigate the key factors influencing long-term visual prognosis in patients with DME treated with aflibercept, as well as their predictive value, through analysis of clinical and imaging data. Although previous studies have confirmed the efficacy of aflibercept in treating DME and preliminarily explored the association between certain SD-OCT parameters and visual prognosis [16–19], this study builds upon existing literature to further clarify the independent predictive value of multiple SD-OCT microstructural parameters. By integrating dynamic assessments (at 1 and 4 months post-treatment), a more timely and comprehensively predictive model was developed, offering a novel perspective for early and precise evaluation of individual treatment responses in clinical settings. The most significant finding of this study is that combined assessment of retinal microstructural changes at two critical time points (1 month and 4 months) during treatment provides a more reliable prediction of long-term visual recovery. We confirm that apixaban therapy significantly improves visual acuity and anatomical structure, consistent with findings from previous studies. However, this study goes beyond describing baseline or final outcomes at a single time point, systematically analyzing the association between early treatment dynamics (at 1 month and 4 months) and visual acuity outcomes at 12 months. After adjusting for multiple potential confounding factors including age, DM duration, prior laser treatment history, injection frequency, and HbA1c, multivariate logistic regression analysis indicated that CST at 1 and 4 months post-treatment, ELM/EZ integrity, COST integrity, and HRF count were all independent predictors of visual recovery at 12 months. This finding underscores the importance of evaluating early structural response rather than relying solely on baseline characteristics. Although multiple studies have reported the relationship between individual parameters such as CST, ELM/EZ, COST or HRF and prognosis [20–24], the ROC analysis in this study demonstrated that integrating information from these key parameters at two early treatment time points to construct a combined predictive model achieved the highest predictive performance (AUC = 0.873). This model (sensitivity of 78.1%, specificity of 85.4%) is statistically superior to prediction models based on a single time point. Its clinical significance lies in providing ophthalmologists with a relatively straightforward early assessment framework based on routine OCT examinations. Within the first four months after aflibercept treatment, more accurate individualized predictions of a patient’s long-term treatment response potential and visual recovery trajectory can be made by observing the degree of macular edema (CST) resolution, early signs of outer retinal structure (ELM/EZ, COST) repair, and reduction in inflammatory/degenerative activity markers (HRF). This facilitates early identification of patients who may respond poorly to standard therapy, enabling more timely decision-making regarding intensified treatment, combination therapy, or management strategy adjustments for more proactive, personalized disease management. Our findings support and refine the current understanding of DME treatment response. ELM/EZ and COST, as key structural markers of photoreceptor cell health, their early repair is associated with a favorable visual prognosis [25–27], confirming that the integrity of photoreceptor cells is the cornerstone of ultimate visual function recovery. HRF, as a marker representing glial cell activation or rigid exudation [28–31], may reflect the therapeutic suppression of intra-retinal inflammatory or leakage activity through its dynamic quantitative changes. This study confirmed the independent contributions of these parameters through multivariate analysis and highlighted the predictive significance of their dynamic changes (rather than merely baseline or final states), offering a valuable addition to existing knowledge. It suggests that treatment response is a dynamic process, with early positive anatomical improvements, particularly structural repair, portending a more favorable long-term biological course. However, the limitations of this study are as follows: This study employed a retrospective design with a limited sample size. Although the statistical model adjusted for key confounding factors, the influence of all unmeasured factors could not be entirely ruled out. Our proposed joint prediction model requires external validation in larger-scale, multicenter prospective cohorts to confirm its generalizability. Furthermore, precise quantitative analysis of parameters such as HRF (e.g., distribution, morphology) may provide deeper insights in the future. Conclusion In summary, this study not only validated the efficacy of aflibercept in treating DME but also constructed and preliminarily validated a vision prognosis assessment model with high predictive efficacy. This was achieved by emphasizing and quantifying the dynamic changes in multidimensional OCT microstructural parameters during the early treatment phase (1 month and 4 months). This model integrates key information from CST, ELM/EZ, COST, and HRF, emphasizing the importance of early comprehensive assessment. It provides a novel tool with potential clinical utility for achieving personalized treatment and precise prognostic evaluation in DME patients. Future prospective studies will focus on optimizing and validating this predictive strategy. Acknowledgements Not applicable. Author contributions All authors designed this study. AP D, YH G, JY D, QY N and R T conducted the experiment and analyzed the data. AP D wrote the manuscript. R T revised the manuscript. All authors reviewed and approved for publication. Funding This study was funded by Medical Research Fund of Yunnan University (Key projects YDYXJJ2024- 0010). Data availability All data generated or analyzed during this study are included in this article. Declarations Ethics approval and consent to participate The study protocol was approved by The Ethics Committee of Affiliated Hospital of Yunnan University. All procedures performed in studies involving human participants were in accordance with the 1964 Helsinki Declaration and later versions. In addition, for investigations involving human subjects, informed consent has been obtained from the participants involved. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement All data generated or analyzed during this study are included in this article.