Long-Term Outcomes and Predictors of Response to Anti-VEGF Therapy for Diabetic Macular Edema in Clinical Practice

Summary

This protocol details a 24-month retrospective cohort analysis utilizing electronic medical records and inverse probability of treatment weighting (IPTW) to evaluate the real-world effectiveness of class-level anti-VEGF therapy and the prognostic value of baseline optical coherence tomography (OCT) biomarkers in patients with center-involved diabetic macular edema.

Abstract

This protocol provides a standardized methodology for evaluating the 24-month effectiveness and safety of anti-vascular endothelial growth factor (anti-VEGF) therapy in center-involved diabetic macular edema (DME) using real-world clinical data. The procedure involves a structured electronic medical record (EMR) screening algorithm based on ICD-10 codes, followed by manual clinical verification. Patients are categorized into a class-level anti-VEGF treatment group (receiving ranibizumab, aflibercept, or conbercept) or a non-biologic control group (focal/grid laser or observation). To mitigate non-random allocation bias inherent in retrospective data, inverse probability of treatment weighting (IPTW) is employed to balance baseline demographic and tomographic covariates across groups. The methodology specifies standardized visual acuity assessment using ETDRS charts and detailed OCT acquisition parameters to ensure longitudinal data consistency. By analyzing visual and anatomical trajectories alongside structural biomarkers, such as the disorganization of the retinal inner layers (DRIL), this approach provides a reproducible framework for assessing long-term outcomes in routine clinical settings. While acknowledging the limitations of a single-center retrospective design, these methods offer critical evidence to help characterize the efficacy-effectiveness gap and optimize personalized management strategies for DME.

Introduction

Macular edema secondary to diabetes remains a principal driver of severe visual decline within the global working-age demographic¹. The underlying pathology involves a complex cascade of prolonged hyperglycemia that compromises the blood-retinal barrier, prompting the overexpression of vascular endothelial growth factor (VEGF) and subsequent fluid extravasation into the macula^2,3. If left unmanaged, this chronic intraretinal and subretinal fluid accumulation inevitably precipitates irreversible central vision deterioration, severely limiting patient autonomy.

The introduction of intravitreal VEGF inhibitors fundamentally redefined therapeutic protocols for this vision-threatening condition. Extensive randomized controlled trials (RCTs) have unequivocally validated that biological agents, such as ranibizumab, aflibercept, and conbercept, outperform traditional macular grid laser photocoagulation in both anatomical restoration and visual acuity enhancement^4,5. Consequently, current international clinical guidelines universally position these targeted biologic therapies as the premier intervention for center-involved cases with visual impairment^6,7,8.

However, a pronounced disparity frequently emerges between the idealized efficacy demonstrated in RCTs and the actual effectiveness observed in routine clinical environments^9,10. In everyday practice, strict trial protocols are rarely replicated due to practical barriers such as patient non-compliance with rigorous injection schedules, complex systemic comorbidities, and inevitable deviations from standardized pro re nata (PRN) algorithms^7,9. Furthermore, while controlled studies often emphasize the comparative molecular superiority of individual drugs or next-generation bispecific antibodies, clinical reality dictates a different approach. In many tertiary and regional medical centers, the selection among ranibizumab, aflibercept, or conbercept is predominantly driven by pragmatic constraints, including immediate medication availability, dynamic healthcare insurance policies, and patient financial capacity, rather than isolated molecular profiles. This reality necessitates evaluating anti-VEGF interventions as a unified pharmacological class to accurately capture macro-level practice patterns.

To address these complexities, this protocol provides a systematic framework for utilizing electronic medical record (EMR) data to evaluate long-term therapeutic outcomes. This methodological approach is particularly applicable for researchers and clinicians operating in high-volume settings where prospective randomized trials are ethically or financially unfeasible. By applying inverse probability of treatment weighting (IPTW) to balance heterogeneous cohorts, this method allows for a statistically robust evaluation of treatment effectiveness that accounts for real-world switching and availability-driven patterns. The uniqueness of this research lies in its focus on class-level, long-term effectiveness in unselected populations, offering a generalizable benchmark that bridges the established efficacy-effectiveness gap.

We hypothesize that, when deployed as a collective pharmacological class, anti-VEGF therapies deliver sustained structural and functional advantages in real-world populations, and that specific baseline clinical and tomographic predictors strongly dictate these outcomes. Accordingly, the primary objective of this retrospective analysis is to evaluate the 24-month visual trajectories, anatomical fluid resolution, and safety profile of class-level anti-VEGF administration. By analyzing a substantial cohort from a representative high-volume regional hospital, this study authentically mirrors the heterogeneous patient demographics and practical constraints characteristic of everyday ophthalmology. Ultimately, these findings are intended to characterize the efficacy-effectiveness gap and offer evidence-based predictors to optimize individualized long-term management strategies for diabetic macular edema.

Protocol

The study adhered to the tenets of the Declaration of Helsinki and received institutional ethics committee approval (KY2025-133-01). Written informed consent for data use was obtained from all participants. The study followed STROBE guidelines for observational research and incorporated key items from the RECORD checklist (Supplementary File 1).

Database query and patient screening algorithm

An initial digital screening of the hospital’s electronic medical record (EMR) system was performed to identify potential participants treated between January 2021 and December 2023. A query algorithm based on ICD-10 diagnostic codes for Type 1 (E10.3) or Type 2 (E11.3) diabetes mellitus accompanied by macular edema was utilized.

A certified ophthalmologist conducted manual verification of the identified clinical charts. The presence of center-involved DME was confirmed, which was strictly defined as a central retinal thickness (CRT) ≥ 300 µm on spectral-domain OCT, accompanied by intraretinal or subretinal fluid.

The following exclusion checkpoints were applied during chart review: (1) macular edema secondary to non-diabetic conditions (e.g., retinal vein occlusion) were identified and excluded; (2) eyes with a history of vitreoretinal surgery or active ocular inflammation were excluded; (3) OCT image quality was verified using a quantitative checkpoint of signal strength ≥ 20 (Heidelberg standard); (4) patients with > 20% missing follow-up data were excluded.

A standardized study eye selection rule was implemented for bilateral cases. The eye with the worse baseline best-corrected visual acuity (BCVA) was selected. If BCVA was identical, the eye with the higher CRT was selected. If both parameters were tied, a pre-programmed random number generator was utilized to assign the study eye.

Group allocation and standardized treatment procedures

Group allocation was documented based on clinical records. Patients were categorized into the anti-VEGF group (n = 166) or the non-biologic control group (n = 137). The control group criteria were clearly defined as patients receiving focal/grid laser or observation due to documented medical contraindications, financial constraints, or patient refusal of injections.

Standardized Anti-VEGF Injection Protocol: The pharmacological agents prepared included ranibizumab (0.5 mg / 0.05 mL), aflibercept (2.0 mg / 0.05 mL), or conbercept (0.5 mg / 0.05 mL). Topical anesthesia was administered using proparacaine hydrochloride 0.5%. Antisepsis of the conjunctival sac and eyelids was performed with 5% povidone-iodine. A transscleral injection was performed 3.5 - 4.0 mm from the limbus using a 30-gauge needle in a dedicated sterile room. A “3 + PRN” regimen was executed. Retreatment was triggered if any of the following checkpoints were met: CRT increase ≥ 50 µm, loss of ≥ 5 ETDRS letters, or persistent fluid on OCT.

Standardized Control Interventions: For laser photocoagulation, a 532 nm solid-state laser was utilized. The spot size was set to 50–100 µm and the duration to 0.1 s. Power was titrated to achieve a barely visible, mild white burn. For observation, systemic glycemic and blood pressure regulation were monitored under a standardized endocrinology department protocol.

Longitudinal assessment and image grading workflow

Follow-up visits were organized at 1, 3, 6, 12, and 24 months. Visit tolerance windows of ± 7 days for monthly visits and ± 14 days for annual visits were implemented to ensure data temporal consistency.

BCVA was measured using a standardized retro-illuminated ETDRS chart at a 4 m distance. Controlled room luminance was ensured, and certified technicians masked to the treatment allocation were utilized.

OCT imaging was executed using the Spectralis OCT device. The acquisition protocol was set to a 20° × 20° macular volume scan with 49 continuous B-scans. Automatic Real-Time (ART) mode was enabled with an averaging setting of 9 frames.

A masked grading workflow was implemented, in which two independent specialists manually graded all images. In case of disagreement regarding fluid presence or structural biomarkers, a third senior specialist was involved as an adjudicator to establish consensus.

Advanced statistical implementation workflow

Normality and Descriptive Analysis: Data distribution was tested using the Shapiro-Wilk test. Continuous data were summarized as mean ± SD.

Propensity Score Weighting Procedure: A logistic regression model was constructed to calculate propensity scores. The included covariates were age, sex, diabetes duration, HbA1c, and baseline OCT features. Inverse Probability of Treatment Weighting (IPTW) was implemented using the Average Treatment Effect (ATE) method. As a verification checkpoint, covariate balance was assessed using Standardized Mean Differences (SMD), with robust balance considered achieved if SMD < 0.10.

Mixed-Effects Model Implementation: Longitudinal changes were analyzed using a Mixed-Effects Model for Repeated Measures (MMRM). An unstructured covariance matrix was specified, and a random intercept for each patient was included. The analysis was performed using R software (version 4.2.2) with the WeightIt and cobalt packages.

Results

Patient screening and diagnostic criteria

Following the standardized EMR query (Step 1.1), 366 patients were initially flagged. After manual verification (Step 1.2), 303 individuals (303 eyes) met the clinical and tomographic criteria for center-involved DME and were included in the final analytical cohort (Figure 1). The baseline demographic and clinical characteristics of these participants are summarized in Table 1. Adherence t...

Discussion

Our retrospective analysis demonstrates that intravitreal anti-VEGF therapy, evaluated as a collective pharmacological class, confers sustained functional and anatomical advantages over non-biologic standard care in a real-world DME cohort¹¹. Over the 24 months, the biological intervention yielded significant visual acuity enhancements and macular fluid resolution while maintaining an established safety profile¹². These results provide critical real-world evidence, characte...

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Aflibercept	Bayer AG/Regeneron		Intravitreal anti-VEGF agent used in the study
Conbercept	Chengdu Kanghong		Intravitreal anti-VEGF agent used in the study
Ranibizumab	Genentech/Novartis		Intravitreal anti-VEGF agent used in the study
Faricimab	Genentech/Roche		Dual Ang-2/VEGF-A inhibitor (Next-generation therapy)
Ranibizumab PDS	Genentech		Sustained drug delivery implant (Refilled every 24 weeks)