Association Between Anti-vascular Endothelial Growth Factor Therapy Exposure and Visual Outcomes in Diabetic Macular Edema and Retinal Vein Occlusion

Summary

In this retrospective cohort, greater injection frequency, longer treatment duration, and earlier treatment initiation were each independently associated with better visual outcomes. Although non-standard treatment patterns were common and associated with poorer unadjusted outcomes, composite standard-treatment classification did not retain independent prognostic value after accounting for these specific treatment-exposure metrics.

Abstract

Anti-vascular endothelial growth factor (anti-VEGF) therapy is first-line treatment for diabetic macular edema (DME) and retinal vein occlusion (RVO), but real-world practice often differs from clinical trial findings. This single-center retrospective cohort study used hospital electronic medical records to evaluate anti-VEGF treatment patterns and associations with visual outcomes. We included 332 adults who received at least one intravitreal injection of ranibizumab, aflibercept, or conbercept between January 2022 and January 2024 and had at least one post-baseline best-corrected visual acuity (BCVA) measurement. The cohort included 180 patients with DME, 138 with RVO, and 14 with other retinal vascular diagnoses. Standard treatment required three monthly loading doses, a treat-and-extend regimen with intervals of 4–16 weeks, no interval >16 weeks, and treatment duration ≥12 months; all other patterns were classified as non-standard. The primary outcome was change in LogMAR BCVA at the last available follow-up. The standard and non-standard groups included 132 and 200 patients, respectively. Treatment interruption for ≥6 months was common in the non-standard group (68.50%). Unadjusted visual outcomes favored the standard group (median ΔBCVA, −0.23 vs. −0.07 LogMAR; improvement rate, 68.18% vs. 35.00%; P<0.001). In adjusted analyses, baseline BCVA, total number of injections, and treatment duration were positive predictors of improvement, whereas longer time to first injection was a negative predictor; standard/non-standard grouping was not significant (P=0.853). Safety events were infrequent, limiting comparison of rare events. Non-standard treatment patterns were common and associated with poorer unadjusted visual outcomes; optimizing injection exposure and treatment timing may improve real-world outcomes.

Introduction

Retinal vascular diseases, particularly diabetic macular edema (DME) and retinal vein occlusion (RVO), are leading causes of global visual impairment^1,2,3. Their incidence is rising due to increasing diabetes prevalence and population aging⁴, placing a heavy burden on public health systems⁵. The IDF reports ~500 million adults with diabetes, one-third of whom have diabetic retinopathy⁶. As its leading cause of vision loss, DME affects up to 5% of patients⁷. RVO, the second most common retinal vascular disease after diabetic retinopathy, has an annual incidence of ~0.5%, rising significantly with age^8,9. A recent comprehensive systematic review and meta-analysis further confirms the substantial global burden of RVO, highlighting the urgent need for optimized management strategies to mitigate its visual impact¹⁰. Therefore, exploring effective treatment strategies is of both clinical and social significance.

Anti-VEGF drugs specifically bind to the VEGF-A isoform, blocking VEGF binding to its receptors, thereby inhibiting neovascularization, reducing vascular permeability, and alleviating macular edema, ultimately improving and stabilizing vision^11,12. Commonly used anti-VEGF drugs in clinical practice include ranibizumab, aflibercept, conbercept, and the recently approved faricimab. Contemporary meta-analyses and network comparisons have provided a comprehensive comparative landscape of these agents, confirming the overall efficacy of anti-VEGF therapy while also highlighting nuanced differences in treatment burden and dosing regimens that are particularly relevant to real-world practice¹³. Numerous RCTs have shown that anti-VEGF therapy is significantly superior to conventional laser photocoagulation and corticosteroids in improving visual acuity in DME and RVO, making it the recommended first-line treatment per major international guidelines^14,15. However, a significant gap exists between the ideal treatment outcomes reported in RCTs and those observed in real-world clinical practice. RCTs typically have strict inclusion and exclusion criteria, and the study populations are often highly selected, excluding patients with severe systemic diseases, poor adherence, or limited life expectancy¹⁶. At the same time, RCTs employ standardized treatment protocols and rigorous follow-up monitoring to ensure that patients receive standardized care. In contrast, the real-world setting is more complex and variable, with greater patient heterogeneity, more flexible treatment choices, variable adherence to follow-up, and more frequent treatment interruptions and discontinuations. This highlights the need for in-depth analysis of real-world treatment patterns and the factors that influence them to optimize clinical practice and improve patient outcomes.

Treatment pattern is a key factor in anti-VEGF efficacy, with three main regimens: fixed, pro re nata (PRN), and treat-and-extend (T&E)¹⁷. The fixed regimen requires patients to receive regular injections at predetermined intervals. While this ensures continuity of treatment, it may lead to overtreatment, increasing healthcare costs and injection-related risks¹⁸. The PRN regimen uses disease activity indicators, such as changes in central macular thickness on optical coherence tomography, to determine whether to inject. Although theoretically enabling individualized treatment, in practice, it often results in undertreatment and greater fluctuations in vision¹⁹. The T&E regimen combines the advantages of the previous two approaches. After completing the initial loading dose, it dynamically adjusts the follow-up interval based on the patient's treatment response, ensuring both individualized therapy and continuity of care. In recent years, it has gradually become the mainstream treatment model²⁰. However, how the T&E regimen is implemented in the real world and whether it achieves the expected therapeutic outcomes currently lacks sufficient local data support. Anti-VEGF therapy requires long-term, repeated intravitreal injections, and the treatment process is burdensome with substantial financial costs, often leading to suboptimal patient adherence. Poor treatment adherence not only directly affects the control of macular edema and maintenance of vision but may also lead to disease recurrence and irreversible visual impairment. Nevertheless, there is currently no unified definition of treatment adherence, and the evaluation criteria vary considerably across different studies, limiting the comparability of research findings.

In recent years, with the widespread adoption of electronic medical record systems, real-world studies have received increasing attention and have been applied in ophthalmology. Based on data generated in routine clinical practice, real-world studies can reflect treatment conditions and outcomes in broader patient populations, address the limitations of RCTs in generalizability, and provide more practice-relevant evidence to support clinical decision-making. However, real-world studies on anti-VEGF therapy for retinal vascular diseases in China remain relatively limited, with a particular lack of systematic analyses of treatment pattern regularity, adherence, and their relationship to visual outcomes. Most existing studies have small sample sizes, short follow-up periods, and predominantly employ univariate analyses, failing to adequately control for confounding factors.

Against this background, the present study aims to use single-center electronic medical record data to retrospectively analyze real-world anti-VEGF treatment patterns in a cohort of Chinese patients with retinal vascular diseases, focusing on associations among treatment regularity, drug selection, baseline characteristics, and visual outcomes. The primary novelty of this study lies in its systematic and quantitative evaluation of treatment regularity using a locally defined set of criteria and the identification of specific modifiable treatment behaviors—beyond simple group classification—that are independently associated with visual prognosis in a Chinese clinical setting. Through this study, we hope to identify existing problems and deficiencies in current local clinical practice, determine key factors influencing vision improvement, and provide evidence-based support to optimize anti-VEGF treatment strategies, improve treatment regularity, and enhance long-term visual prognosis in patients. Additionally, the findings may serve as a reference for establishing a quality evaluation system for anti-VEGF therapy tailored to the Chinese context, thereby contributing to the standardization of fundus disease diagnosis and treatment.

Protocol

This study was approved by the Tianjin Medical University Ethics Committee (Approval number: 2024-025). This study adhered to the principles of the Declaration of Helsinki and the Measures for Ethical Review of Biomedical Research Involving Humans. As this was a retrospective analysis using previously de-identified clinical data, the study protocol was reviewed and approved by the hospital ethics committee, and the requirement for informed consent was waived.

Study design
This single-center, retrospective cohort study used real-world data from our hospital’s electronic medical records and ophthalmic picture archiving system to analyze anti-VEGF treatment patterns and visual outcomes in patients with retinal vascular diseases. The study period spanned from January 2022 to January 2024. All clinical data were extracted through the hospital information system and the specialized ophthalmic database, including patient demographic characteristics, diagnostic information, treatment regimens, follow-up records, and visual acuity examination results. The detailed study workflow is shown in Figure 1.

Cohort definition and follow-up rules
Index date: The date of the first intravitreal anti-VEGF injection.

Minimum follow-up: To be eligible for the primary outcome analysis, patients were required to have at least one post-baseline BCVA measurement after the index date.

Primary outcome assessment time point: To balance the inherent variability in real-world follow-up, the primary outcome (change in BCVA from baseline) was assessed at the last available follow-up visit within the study window (up to January 2024). This approach reflects real-world practice but introduces variability in observation time.

Censoring rules: Patients were censored at the time of their last recorded clinic visit, death, or the end of the study period (January 2024), whichever occurred first. Patients who switched to a non-anti-VEGF therapy (e.g., a corticosteroid implant) or underwent vitrectomy during follow-up were censored at the time of the switch or surgery and were not included in the primary outcome analysis.

Standard Treatment Group classification rule: Classification into the Standard Treatment Group required that the patient met all criteria (completion of 3 loading doses, T&E regimen, no interval >16 weeks, duration ≥ 12 months) during the entire observable follow-up period. Patients who had not yet reached 12 months of follow-up by the data cut-off were classified based on their observed pattern; if they met all other criteria but had <12 months of follow-up, they were not classified as Standard.

Justification for the standard treatment definition: The selected cutoffs were based on a combination of clinical trial protocols and real-world practice standards. The 16-week (4-month) maximum interval corresponds to the upper limit of the approved dosing schedule for aflibercept and the treat-and-extend (T&E) extension phase recommended in major guidelines²¹. The 12-month treatment duration was chosen because it represents the minimum period required to complete the initial loading phase (3 monthly injections) and achieve stable disease control under a T&E regimen, as supported by previous real-world studies^22,23,24. These thresholds were predefined in the study protocol and were not derived from outcome-driven data exploration.

Critical methodological note on group definition: The definition of the Standard Treatment Group inherently requires that patients remain under active treatment and observation for at least 12 months and maintain regular injection intervals throughout that period. Consequently, this grouping is partly conditional upon treatment opportunity and follow-up duration. Patients with shorter follow-up, those who discontinued early, or those who experienced interval prolongation prior to completing 12 months were automatically assigned to the non-standard group. This temporal conditioning may introduce a form of immortal-time or conditioning bias in comparisons between groups, as the exposure (standard vs. non-standard) is not purely baseline-defined but is ascertained over the follow-up period. Therefore, group comparisons should be interpreted primarily as descriptive associations, and the independent contribution of the grouping variable beyond its constituent components (e.g., injection number, treatment duration) should be evaluated with caution in the multivariable model.

Inclusion criteria
Patients were eligible for inclusion if they met the diagnostic criteria for a retinal vascular disease, which was confirmed by fluorescein fundus angiography (FFA), optical coherence tomography (OCT), and clinical examination, including diabetic macular edema (DME), retinal vein occlusion (RVO) (either central or branch), or other ischemic retinopathies (e.g., ocular ischemic syndrome) as diagnosed by a treating retinal specialist. All included patients were required to have received at least one intravitreal injection of an anti-VEGF agent, including ranibizumab, aflibercept, or conbercept. The study was limited to patients aged 18 years or older. Furthermore, complete records of best-corrected visual acuity (BCVA) at baseline and at least one follow-up visit were mandatory for study participation.

Exclusion criteria
Patients were excluded from the study if they presented with other ocular diseases that could severely affect vision, such as age-related macular degeneration (AMD), severe cataract, advanced glaucoma, or optic atrophy. Those who had previously undergone panretinal photocoagulation (PRP) or focal macular laser treatment, where the treatment period overlapped with the follow-up period of this study, were also excluded. A history of intraocular surgery, including cataract extraction, vitrectomy, or ocular trauma that could affect the accuracy of visual acuity assessment, was another reason for exclusion. Patients with active ocular infection or inflammation, such as infectious endophthalmitis or active uveitis, were not included. Pregnant or lactating individuals were excluded, as were those with severe systemic diseases and a life expectancy of less than one year, or those who were unable to cooperate with visual acuity examinations. Finally, patients with severely missing medical record data were excluded from the analysis.

Sample size calculation
Based on the difference in vision improvement rates between the standard treatment group (defined as completion of three loading doses, adherence to a treat-and-extend regimen with intervals ≤16 weeks, and treatment duration ≥12 months) and the non-standard treatment group (defined as any deviation from these criteria) in previous real-world studies (expected improvement rates of 65% and 35%, respectively)²⁵. With α=0.05 and β=0.20 (80% statistical power), the sample size calculation formula for comparing two proportions was used. It was estimated that at least 52 patients were required in each group. Given a 20% data-missing rate, a final planned enrollment of at least 130 patients was set. A total of 332 patients were actually included in this study, meeting the statistical requirements.

Treatment regimens and group definitions

Types of anti-VEGF agents and drug-group assignment
The anti-VEGF agents involved in this study included: Ranibizumab: 0.5 mg/0.05 mL, administered by intravitreal injection; Aflibercept: 2 mg/0.05 mL, administered by intravitreal injection; Conbercept: 0.5 mg/0.05 mL, administered by intravitreal injection. In this cohort, no patient switched anti-VEGF agents during the follow-up period; each patient received a single agent exclusively. Therefore, drug-group assignment for subgroup analysis was based on the agent received throughout the entire treatment course.

Grouping criteria for treatment patterns
Based on treatment adherence and follow-up regularity, patients were divided into two groups. Patients were classified in the Standard Treatment Group only if all of the following criteria were met: completion of the initial three consecutive monthly loading-dose injections; adoption of a T&E regimen with intervals dynamically adjusted based on macular edema/retinal ischemia activity (4–16 weeks); no injection interval > 16 weeks (4 months); and treatment duration ≥ 12 months. Patients were classified in the Non-standard Treatment Group if any of the following criteria were met: treatment interruption (no injection for ≥6 consecutive months, with main reasons including financial difficulties, transportation inconvenience, or subjective perception of disease improvement); incomplete loading dose; insufficient injection frequency (mean injection interval >16 weeks); excessively long follow-up interval (two consecutive follow-up intervals >4 months under the T&E regimen); or patient-initiated self-termination without physician assessment.

Clinical outcome measures

Primary outcome measure
Change in best-corrected visual acuity (ΔBCVA): The change in BCVA (after LogMAR transformation) from baseline at the last available follow-up visit. For the 96 patients with bilateral disease (28.92%), the average LogMAR change across both eyes was used as the patient's visual outcome.

Secondary outcome measures
Vision improvement rate: Proportion of patients with a LogMAR improvement of ≥0.2 (equivalent to a gain of ≥2 lines on the Snellen chart). Vision stability rate: Proportion of patients with an absolute change in LogMAR of <0.1 (visual fluctuation within ±1 line). Vision worsening rate: Proportion of patients with a LogMAR increase of ≥0.1 (vision loss of ≥1 line). Anatomical improvement: Reduction in central subfield thickness (CST) from baseline (applicable only to DME patients). Treatment burden: Total number of injections, mean injection interval. Safety indicators: Incidence of endophthalmitis, incidence of retinal detachment, incidence of elevated intraocular pressure (>21 mmHg or an increase of >5 mmHg from baseline), and incidence of other serious adverse events (SAEs).

Statistical analysis
Continuous variables are presented as mean ± SD; categorical variables are presented as frequencies (%). Group comparisons used the t-test or Mann-Whitney U test for continuous variables and the chi-square test or Fisher's exact test for categorical variables. Multivariate logistic regression identified factors associated with vision improvement (LogMAR improvement ≥0.2). Exploratory stratified analyses by disease type and anti-VEGF agent summarized subgroup outcomes; no formal interaction testing was performed. Two-tailed P<0.05 was considered significant. We acknowledge that several clinically important confounders—including but not limited to ischemic vs. non-ischemic RVO, diabetic retinopathy severity, glycemic control (HbA1c), renal function, OCT biomarkers (e.g., presence of intraretinal fluid, subretinal fluid, hyperreflective foci), lens status/cataract progression, insurance/payment status, travel distance, and physician-directed reasons for extending injection intervals—were not consistently available in this retrospective dataset and could not be included in the model. Given the observational, retrospective design and the potential for residual confounding, all findings are reported as associations rather than causal effects.

Results

Baseline characteristics of patients
A total of 332 patients were included. Among them, 180 (54.22%) had diabetic macular edema (DME), 138 (41.57%) had retinal vein occlusion (RVO; comprising 72 with central RVO and 66 with branch RVO), and 14 (4.22%) carried other diagnoses, including retinal vasculitis and ischemic retinopathy. Measured baseline characteristics were not statistically different between the standard and non-standard treatment groups, including age, sex, affected side, disease type distribution, duration of diabetes, baseline BCVA, baseline CST, comorbidities, and anti-VEGF drug selection (all P>0.05). Standardized mean differences (SMDs) for all baseline variables were below 0.20, suggesting acceptable balance on observed covariates. However, residual confounding from unmeasured or imprecisely measured factors (e.g., ischemic status, glycemic control, OCT biomarkers, socioeconomic factors) cannot be excluded. See Table 1.

Analysis of treatment patterns

Distribution of non-standard treatment types
Table 2 presents the distribution of various non-standard treatment behaviors in the non-standard treatment group. Among the 200 patients with non-standard treatment, treatment interruption for ≥6 months was the most common (137 cases, 68.50%), with the main reasons being financial difficulties (39.00%), transportation inconvenience (16.00%), and patients subjectively perceiving disease improvement and subsequently stopping treatment on their own (13.50%). In addition, incomplete loading dose (26.50%), insufficient injection frequency (19.50%), and excessively long follow-up interval (12.50%) also accounted for considerable proportions. Moreover, 15.50% of patients exhibited multiple non-standard behaviors simultaneously.

Comparison of treatment burden
Table 3 summarizes treatment burden indicators. By design, the standard treatment group had, on average, more injections (median 8.00 vs. 5.00, P<0.001), longer observed treatment duration (19.00 vs. 11.00 months, P<0.001), shorter injection intervals (10.00 vs. 18.00 weeks, P<0.001), and a shorter time from diagnosis to first injection (15.50 vs. 26.00 days, P<0.001) compared to the non-standard group. These differences largely reflect the definitional criteria used to assign patients to each group rather than representing independent evidence of therapeutic superiority.

Analysis of visual outcomes
Table 4 summarizes the primary and secondary visual outcome measures for the two groups. The standard treatment group showed greater median improvement in BCVA than the non-standard treatment group (−0.23 vs. −0.07 LogMAR, P<0.001). For secondary outcomes, the standard treatment group had a higher rate of vision improvement (68.18% vs. 35.00%) and a lower rate of vision worsening (8.33% vs. 27.00%) than the non-standard treatment group (both P<0.001). The vision stability rate was lower in the standard treatment group (23.48% vs. 38.00%, P=0.006), consistent with more patients meeting the improvement threshold. The proportion of patients achieving a final decimal BCVA of 0.3 or better was higher in the standard treatment group (59.09%) than in the non-standard treatment group (41.00%, P=0.001).

Subgroup analyses

Subgroup analysis by disease type
Table 5 presents exploratory subgroup analyses stratified by disease type. These analyses were not pre-specified to include formal interaction testing; therefore, the findings should be interpreted as hypothesis-generating. Due to limited sample sizes within subgroups, particularly for central retinal vein occlusion (CRVO), the absence of a statistically significant difference in ΔBCVA should not be overinterpreted as evidence of a lack of effect. No formal interaction tests were conducted to assess whether the association between treatment regularity and outcomes differed by disease type or drug agent. Among patients with DME, the standard treatment group had a median ΔBCVA of -0.24 LogMAR and a vision improvement rate of 70.27%, both significantly better than those in the non-standard treatment group (-0.06 LogMAR, 33.02%; both P<0.001). Similar trends were observed in patients with RVO, where the standard treatment group had a median ΔBCVA of -0.21 LogMAR and an improvement rate of 65.38%, significantly better than those in the non-standard treatment group (-0.08 LogMAR, 37.21%; P<0.001). After further stratifying RVO into CRVO and branch retinal vein occlusion (BRVO) subgroups, although the difference in ΔBCVA between the standard and non-standard treatment groups in the CRVO subgroup did not reach statistical significance (P=0.133), the vision improvement rate remained significantly better with standard treatment (60.71% vs. 31.82%, P=0.016). In the BRVO subgroup, standard treatment demonstrated significant advantages in both ΔBCVA and improvement rate (both P<0.05). Formal interaction tests for treatment effect by disease type were not performed due to limited sample sizes in the subgroups.

Subgroup analysis by different anti-VEGF agents
Table 6 presents exploratory subgroup analyses by anti-VEGF agent. Among patients receiving ranibizumab, aflibercept, or conbercept, the standard treatment group consistently showed greater median ΔBCVA and higher rates of vision improvement compared to the non-standard treatment group (all P<0.05 for improvement rates). These subgroup findings are descriptive and exploratory; no formal interaction testing was conducted to assess whether the association between treatment regularity and outcomes differed by drug type.

Anatomical outcomes (in DME patients)
Table 7 compares anatomical outcomes among DME patients in the standard and non-standard treatment groups. The standard treatment group had greater median CST reduction (−135.00 vs. −92.00 µm, P<0.001), a higher rate of CST <300 µm (56.76% vs. 28.30%, P<0.001), and a higher rate of CST reduction ≥100 µm (70.27% vs. 39.62%, P<0.001).

Analysis of influencing factors

Univariate analysis
Table 8 presents the results of univariate logistic regression analysis for factors influencing vision improvement (LogMAR improvement ≥ 0.2). Higher baseline LogMAR BCVA values (indicating worse baseline visual acuity), standard treatment, higher total number of injections, and longer treatment duration were significantly associated with greater odds of vision improvement (all P<0.001). In contrast, longer mean injection interval and longer time from diagnosis to first injection were significantly associated with lower odds of vision improvement (both P<0.001). Age, sex, disease type, baseline CST, and drug choice showed no significant associations in this univariate analysis.

Multivariate analysis
Table 9 presents the independent factors influencing vision improvement identified by multivariate logistic regression analysis. After controlling for other variables in the model, higher baseline LogMAR BCVA values (OR=1.057, 95% CI: 1.043–1.074, P<0.001), total number of injections (OR=1.331, 95% CI: 1.179–1.514, P<0.001), and treatment duration (OR=1.104, 95% CI: 1.042–1.173, P=0.001) remained positive predictors of vision improvement, while time from diagnosis to first injection (OR=0.957, 95% CI: 0.924–0.991, P=0.015) was a negative predictor. The grouping variable of standard treatment did not reach statistical significance in the multivariate model (P=0.853). The Hosmer-Lemeshow goodness-of-fit test yielded χ²=6.452 (P=0.597), indicating that the model had acceptable calibration.

Safety analysis
Table 10 compares patient-level safety indicators during anti-VEGF therapy. In the standard versus non-standard treatment groups, endophthalmitis incidence was 0.76% (1 case) vs. 0.50% (1 case) (P=1.000); no retinal detachment or persistent ocular hypertension occurred. Vitreous hemorrhage occurred only in the non-standard group (1.00%, 2 cases, P=0.520). Cataract progression was similar (2.27% vs. 2.50%, P=1.000). Overall serious adverse event rates were 0.76% vs. 1.50% (P=0.926). Safety events were infrequent, and comparisons of rare adverse events should be interpreted cautiously.

Data Availability Statement:
The de-identified datasets generated and analyzed during the current study are available as supplementary material accompanying this article.

Figure 1. Research process. Flow diagram summarizing patient screening, exclusions, eligibility, grouping, and analysis. Abbreviations: AMD = Age-related macular degeneration; PRP = Panretinal photocoagulation; CST = Central subfield thickness; BCVA = Best-corrected visual acuity; VEGF = vascular endothelial growth factor. Please click here to view a larger version of this figure.

Variable	Total (n=332)	Standard Treatment Group (n=132)	Non-standard Treatment Group (n=200)	Statistics	P	SMD
Demographic Characteristics
Age (years), mean±SD	62.89±11.21	61.80±10.84	63.62±11.42	t=-1.450	0.148	0.164
Sex, n (%)				χ2=0.645	0.422	0.090
Male	190 (57.23)	72 (54.55)	118 (59.00)
Female	142 (42.77)	60 (45.45)	82 (41.00)
Affected Side, n (%)				χ2=0.205	0.651	0.051
Unilateral	236 (71.08)	92 (69.70)	144 (72.00)
Bilateral	96 (28.92)	40 (30.30)	56 (28.00)
Disease Characteristics
DME, n (%)	180 (54.22)	74 (56.06)	106 (53.00)	χ2=0.300	0.584	0.061
RVO, n (%)	138 (41.57)	52 (39.39)	86 (43.00)	χ2=0.426	0.514	0.073
CRVO, n (%)	72 (21.69)	28 (21.21)	44 (22.00)	χ2=0.029	0.865	0.019
BRVO, n (%)	66 (19.88)	24 (18.18)	42 (21.00)	χ2=0.397	0.529	0.071
Other, n (%)	14 (4.22)	6 (4.55)	8 (4.00)	χ2=0.059	0.809	0.027
Diabetes Duration (years), M (Q1, Q3)*	12.00 (9.00, 16.00)	12.00 (10.00, 14.00)	12.00 (9.00, 17.00)	Z=-0.460	0.645	0.101
Baseline BCVA (LogMAR), M (Q1, Q3)	0.71 (0.52, 0.88)	0.70 (0.50, 0.86)	0.72 (0.54, 0.90)	Z=-1.209	0.227	0.134
Baseline CST (μm), mean±SD	422.14±102.99	412.55±98.34	428.47±105.71	t=-1.381	0.168	0.156
Comorbidities
Hypertension, n (%)	213 (64.16)	77 (58.33)	136 (68.00)	χ2=3.231	0.072	0.199
Diabetic Kidney Disease, n (%)*	92 (51.11)	34 (45.95)	58 (54.72)	χ2=1.342	0.247	0.176
Cardiovascular Disease, n (%)	76 (22.89)	28 (21.21)	48 (24.00)	χ2=0.350	0.554	0.067
Drug Selection
Ranibizumab, n (%)	118 (35.54)	48 (36.36)	70 (35.00)	χ2=0.065	0.799	0.028
Aflibercept, n (%)	100 (30.12)	38 (28.79)	62 (31.00)	χ2=0.185	0.667	0.048
Conbercept, n (%)	114 (34.34)	46 (34.85)	68 (34.00)	χ2=0.025	0.873	0.018

Table 1: Comparison of baseline characteristics. Baseline demographic and clinical variables are shown for the total cohort and by treatment group. Note: *Only in DME patients

Non-standard Types	n	Constituent ratio (%)
Treatment interruption ≥6 months	137	68.50
Financial difficulties	78	39.00
Transportation inconvenience	32	16.00
Subjective perception of disease improvement	27	13.50
Loading dose incomplete	53	26.50
Insufficient injection frequency	39	19.50
Excessive follow-up interval	25	12.50
Coexistence of multiple non-standard behaviors	31	15.50

Table 2: Distribution of non-standard treatment types in the non-standard treatment group (n = 200). Counts and constituent ratios are shown for treatment interruption and other non-standard treatment behaviors.

Variable	Standard Treatment Group (n=132)	Non-standard Treatment Group (n=200)	Statistics	P
Total number of injections (times)	8.00 (6.00, 10.00)	5.00 (3.00, 6.00)	Z=9.712	<0.001
Treatment duration (months)	19.00 (16.00, 21.25)	11.00 (8.00, 16.25)	Z=9.364	<0.001
Mean injection interval (weeks)	10.00 (8.75, 12.00)	18.00 (13.00, 23.00)	Z=-12.102	<0.001
Time from diagnosis to first injection (days)	15.50 (11.00, 20.00)	26.00 (19.00, 33.00)	Z=-9.966	<0.001

Table 3: Comparison of treatment burden indicators, M (Q1, Q3). Injection number, treatment duration, injection interval, and time to first injection are compared between groups.

Variable	Standard Treatment Group (n=132)	Non-standard Treatment Group (n=200)	Statistics	P
Primary Outcome
ΔBCVA (LogMAR)	-0.23 (-0.31, -0.07)	-0.07 (-0.23, 0.11)	Z=-5.253	<0.001
Secondary Outcomes
Vision improvement rate	90 (68.18)	70 (35.00)	χ2=35.067	<0.001
Vision stability rate	31 (23.48)	76 (38.00)	χ2=7.670	0.006
Vision worsening rate	11 (8.33)	54 (27.00)	χ2=17.597	<0.001
Final BCVA ≥ 0.3	78 (59.09)	82 (41.00)	χ2=10.424	0.001

Table 4: Comparison of visual outcome indicators. Primary and secondary BCVA outcomes are compared between groups. Abbreviation: BCVA = Best-corrected visual acuity.

Disease Type	Group	n	ΔBCVA (LogMAR)	Statistics	Visual improvement rate, n (%)	Statistics
DME	Standard Treatment Group	74	-0.24 (-0.31, -0.08)	Z=-4.633; P<0.001	52 (70.27)	χ²=24.215; P<0.001
	Non-standard Treatment Group	106	-0.06 (-0.23, 0.11)		35 (33.02)
RVO	Standard Treatment Group	52	-0.21 (-0.28, -0.04)	Z=-2.393; P=0.017	34 (65.38)	χ²=10.310; P<0.001
	Non-standard Treatment Group	86	-0.08 (-0.23, 0.09)		32 (37.21)
CRVO	Standard Treatment Group	28	-0.21 (-0.23, -0.04)	Z=-1.503; P=0.133	17 (60.71)	χ²=5.827; P=0.016
	Non-standard Treatment Group	44	-0.07 (-0.22, 0.09)		14 (31.82)
BRVO	Standard Treatment Group	24	-0.25 (-0.32, -0.04)	Z=-2.176; P=0.030	17 (70.83)	χ²=4.799; P=0.028
	Non-standard Treatment Group	42	-0.08 (-0.26, 0.08)		18 (42.86)

Table 5: Comparison of visual outcomes in disease-specific subgroups. Exploratory subgroup outcomes are shown for DME, RVO, CRVO, and BRVO. Abbreviations: BRVO = Branch retinal vein occlusion; CRVO = Central retinal vein occlusion; DME = Diabetic macular edema; RVO = Retinal vein occlusion.

Drug type	Group	n	ΔBCVA (LogMAR)	Statistics	Visual improvement rate, n (%)	Statistics
Ranibizumab	Standard Treatment Group	48	-0.23 (-0.31, -0.09)	Z=-4.134; P<0.001	34 (70.83)	χ²=15.218; P<0.001
	Non-standard Treatment Group	70	0.00 (-0.24, 0.11)		24 (34.29)
Aflibercept	Standard Treatment Group	38	-0.23 (-0.28, -0.06)	Z=-2.285; P=0.022	26 (68.42)	χ²=10.240; P<0.001
	Non-standard Treatment Group	62	-0.07 (-0.23, 0.09)		22 (35.48)
Conbercept	Standard Treatment Group	46	-0.21 (-0.31, 0.05)	Z=-2.521; P=0.012	30 (65.22)	χ²=9.855; P=0.002
	Non-standard Treatment Group	68	-0.08 (-0.24, 0.09)		24 (35.29)

Table 6: Comparison of visual outcomes in drug-specific subgroups. Exploratory subgroup outcomes are shown by anti-VEGF agent. Abbreviation: VEGF = vascular endothelial growth factor.

Variable	Standard Treatment Group (DME patients, n=74)	Non-standard Treatment Group (DME patients, n=106)	Statistics	P
ΔCST (μm)	-135.00 (-168.00, -92.75)	-92.00 (-116.00, -50.50)	Z=-4.847	<0.001
CST <300 μm	42 (56.76)	30 (28.30)	χ2=14.702	<0.001
CST reduction ≥100 μm	52 (70.27)	42 (39.62)	χ2=16.405	<0.001

Table 7: Comparison of anatomical outcomes in DME patients. CST reduction and CST response thresholds are compared between treatment groups among DME patients. Abbreviations: CST = Central subfield thickness; DME = Diabetic macular edema.

Variable	β	S.E	OR(95%CI)	P
Age	-0.012	0.01	0.988 (0.969, 1.007)	0.215
Sex (Male vs. Female)	-0.077	0.222	0.926 (0.599, 1.431)	0.728
Disease Type (DME vs. Other)	0.256	0.221	1.291 (0.838, 1.995)	0.247
Baseline BCVA (LogMAR)	0.04	0.005	1.041 (1.030, 1.053)	<0.001
Baseline CST (μm)	-0.001	0.001	0.999 (0.997, 1.001)	0.395
Aflibercept vs. Ranibizumab	-0.228	0.272	0.796 (0.466, 1.357)	0.402
Conbercept vs. Ranibizumab	-0.208	0.263	0.812 (0.484, 1.359)	0.428
Standard Treatment	1.381	0.238	3.980 (2.508, 6.398)	<0.001
Total Number of Injections (times)	0.286	0.044	1.331 (1.226, 1.455)	<0.001
Treatment Duration (months)	0.144	0.022	1.155 (1.108, 1.208)	<0.001
Mean Injection Interval (weeks)	-0.071	0.018	0.932 (0.899, 0.964)	<0.001
Time from Diagnosis to First Injection (days)	-0.059	0.012	0.943 (0.920, 0.964)	<0.001

Table 8: Univariate logistic regression analysis of factors influencing vision improvement. Odds ratios are shown for baseline variables, drug selection, treatment pattern, and treatment-exposure metrics.

Variable	β	S.E	OR(95%CI)	P
Intercept	-4.637	1.396		0.001
Age	-0.019	0.014	0.981 (0.954, 1.009)	0.191
Sex (Male vs. Female)	-0.033	0.305	0.967 (0.531, 1.760)	0.913
Disease Type (DME vs. Other)	0.336	0.317	1.399 (0.754, 2.622)	0.289
Baseline BCVA (LogMAR)	0.056	0.007	1.057 (1.043, 1.074)	<0.001
Baseline CST (μm)	-0.001	0.002	1.000 (0.997, 1.002)	0.745
Aflibercept vs. Ranibizumab	-0.579	0.372	0.560 (0.267, 1.155)	0.119
Conbercept vs. Ranibizumab	-0.369	0.370	0.692 (0.332, 1.422)	0.318
Standard Treatment	0.091	0.491	1.095 (0.415, 2.861)	0.853
Total Number of Injections (times)	0.285	0.064	1.331 (1.179, 1.514)	<0.001
Treatment Duration (months)	0.099	0.030	1.104 (1.042, 1.173)	0.001
Mean Injection Interval (weeks)	-0.009	0.031	0.991 (0.932, 1.051)	0.760
Time from Diagnosis to First Injection (days)	-0.044	0.018	0.957 (0.924, 0.991)	0.015

Table 9: Multivariate logistic regression analysis of factors influencing vision improvement. Adjusted odds ratios are shown for variables included in the final logistic regression model.

Adverse Events	Standard Treatment Group (n=132)	Non-standard Treatment Group (n=200)	Statistics	P
Endophthalmitis	1 (0.76)	1 (0.50)	-	1.000
Retinal Detachment	0 (0.00)	0 (0.00)	-	-
Persistent Ocular Hypertension	0 (0.00)	0 (0.00)	-	-
Vitreous Hemorrhage	0 (0.00)	2 (1.00)	-	0.520
Cataract Progression	3 (2.27)	5 (2.50)	χ2=0.000	1.000
Any Serious Adverse Event	1 (0.76)	3 (1.50)	χ2=0.009	0.926

Table 10: Comparison of safety indicators [n (%)]. Patient-level adverse event counts and percentages are shown by treatment group.

Discussion

In this single-center, real-world cohort, we systematically characterized anti-VEGF treatment patterns in patients with retinal vascular diseases and explored their associations with visual outcomes. The main findings are threefold. First, non-standard therapy was common, affecting 60.24% of the cohort, with treatment interruption for ≥6 months being the most frequent manifestation. Second, patients in the standard-treatment group had significantly better unadjusted visual outcomes than those in the non-standard group. Third, in multivariable analysis, the composite grouping variable itself was not independently associated with vision improvement; rather, the specific treatment-exposure metrics—total number of injections, treatment duration, and earlier treatment initiation—were the key independent predictors. These results highlight that the apparent benefit of being classified as “standard” is largely mediated by modifiable treatment behaviors, rather than by the label itself.

The management of diabetic retinopathy and its complications, including DME, has evolved into a complex therapeutic landscape where treatment patterns significantly impact long-term visual outcomes²⁶. The high frequency of non-standard therapy observed here (60.24%) is comparable to rates reported in other real-world studies from different healthcare settings. For instance, Licht et al. found a 52.8% loss-to-follow-up rate among patients receiving intravitreal injections in Germany, with subjective perception of improvement and transportation difficulties as leading reasons²⁵. Similarly, Abualhasan et al. reported a 59.8% non-adherence rate in a Palestinian cohort, driven by financial constraints and mobility issues²⁷. The concordance across these geographically and economically diverse populations suggests that treatment interruption and irregular follow-up may represent a broader real-world challenge in anti-VEGF delivery, although direct cross-study comparisons are limited by differences in definitions and healthcare systems.

The negative association between treatment irregularity and visual outcomes was evident in our unadjusted comparisons. The non-standard group had a median ΔBCVA of only −0.07 LogMAR and a 27.00% worsening rate, versus −0.23 LogMAR and 8.33% in the standard group. These figures are in line with prior work, such as that by Shellvarajah et al., who found a 17-letter difference between adherent and non-adherent nAMD patients at 12 months²⁸. However, it is critical to interpret such group-level differences with caution. Because the standard-group definition inherently required patients to remain on treatment for at least 12 months with regular intervals, this classification is partly ascertained after the start of follow-up. Consequently, the unadjusted comparison between groups is susceptible to conditioning or immortal-time bias, and should not be taken as evidence that the composite “standard” label directly causes better outcomes.

The multivariable analysis offers a more refined picture. After adjusting for baseline BCVA, injection number, treatment duration, and time-to-first injection, the standard/non-standard grouping variable was no longer significant (P=0.853). In contrast, each additional injection (OR=1.331), longer treatment duration (OR=1.104), and shorter diagnostic-to-treatment delay (OR=0.957 per day) remained independent predictors of vision improvement. From a mechanistic standpoint, more frequent injections sustain VEGF suppression, thereby stabilizing macular structure and preventing recurrent edema²⁹; longer treatment duration ensures durable disease control and reduces the risk of irreversible retinal damage³⁰; and earlier intervention limits the duration of macular edema, which is known to cause photoreceptor injury^31,32. Our findings align with broader therapeutic strategies for macular edema, where consistent VEGF suppression is crucial for achieving both functional and anatomical improvements³³. These findings suggest that clinical efforts should focus on optimizing these specific, actionable treatment parameters rather than on achieving a composite categorization per se.

Exploratory disease-specific subgroup analyses showed associations between standard treatment and better visual outcomes in DME and RVO, though the effect appeared less pronounced in the CRVO subgroup, where the ΔBCVA difference did not reach statistical significance. This may reflect the more severe ischemia and inflammatory component in CRVO, which can attenuate the response to anti-VEGF monotherapy^34,35, or simply limited statistical power due to small sample sizes. Similarly, drug-specific subgroup analyses showed greater visual improvement in the standard-treatment group within each anti-VEGF agent subgroup. These results should not be interpreted as evidence of uniform treatment effects or therapeutic equivalence because no interaction tests or head-to-head comparisons were performed. The anatomical data in DME patients supported the functional results: the standard-treatment group had greater CST reduction (−135 vs. −92 µm) and a higher rate of achieving CST <300 µm (56.76% vs. 28.30%). Safety events were infrequent and similar between groups; however, the low event rates and patient-level reporting limit any meaningful comparison of rare complications.

Several limitations must be acknowledged. First, this was a single-center retrospective study, and the findings may not be generalizable to other populations, healthcare systems, or geographic regions. Second, despite adjustment for measured covariates, residual confounding from unmeasured or imprecisely measured factors—including ischemic status, glycemic control (HbA1c), renal function, OCT biomarkers (e.g., intraretinal fluid, subretinal fluid, hyperreflective foci), lens status, socioeconomic variables (insurance type, travel distance), and physician-directed reasons for extending intervals—cannot be excluded. It is worth noting that the handling of bilateral disease warrants methodological consideration. For the 96 patients (28.92%) with bilateral involvement, we used the average of the LogMAR change values from both eyes as the patient-level outcome. While this approach is commonly employed in ophthalmic real-world studies to generate a single summary measure per patient, it does not account for within-patient inter-eye correlations and may obscure clinically meaningful inter-eye asymmetry.

Third, the definition of the standard treatment group was conditional on treatment duration and follow-up regularity; patients who discontinued early or had prolonged intervals before completing 12 months were automatically assigned to the non-standard group. This introduces a form of conditioning or immortal-time bias, such that group comparisons are partly confounded by the opportunity to remain in the study. Fourth, there is inherent circularity between the grouping definition and the treatment-burden variables (e.g., injection number, treatment duration) used in the multivariable model, which may over-adjust and obscure the effect of the grouping variable; conversely, the non-significance of the group variable should not be interpreted as proof that treatment regularity is irrelevant, but rather that its effect is mediated through these specific exposure metrics. Fifth, follow-up duration was variable, and the primary outcome was assessed at the last available visit rather than a fixed time point, introducing heterogeneity in observation windows. Sixth, no patient switched anti-VEGF agents in this cohort, so the handling of drug switching was not applicable; however, this also limits the generalizability to clinical settings where switches occur. Seventh, the safety comparisons are underpowered due to rare events, and no inferential statistics should be overinterpreted. Finally, the study cannot establish causal effects; all findings are associations, and any causal language has been explicitly avoided.

In conclusion, in this single-center cohort, non-standard treatment patterns were frequent, and lower injection frequency, shorter treatment duration, and delayed treatment initiation were each independently associated with poorer visual outcomes. The composite classification of “standard treatment” did not show independent prognostic value after accounting for these specific exposure metrics. These findings underscore that optimizing modifiable treatment behaviors—particularly ensuring adequate injection frequency, maintaining treatment duration, and minimizing time to first injection—may be more clinically relevant than striving for a categorical label. The results do not establish that standard treatment per se leads to better vision, nor do they directly evaluate the effectiveness of health policy interventions. Future research should employ prospective, multicenter designs with standardized definitions and longer follow-up to validate these associations. Additionally, investigations into the effectiveness of targeted adherence-support programs, patient education, and financial assistance schemes—as well as the real-world impact of newer agents with extended dosing intervals—are warranted to translate these observational findings into improvements in practice.

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Aflibercept (Eylea)	Bayer	61755-005-02	Anti-VEGF agent administered by intravitreal injection.
Alternative SD-OCT System	Carl Zeiss Meditec AG	Cirrus HD-OCT 5000	Alternative spectral-domain OCT system used for retinal thickness analysis and retinal imaging.
Best-Corrected Visual Acuity (BCVA) Assessment	Precision Vision, La Salle, IL, USA	Revised 2000 ETDRS Charts	ETDRS charts used for standardized BCVA assessment.
Conbercept (Langmu)	Chengdu Kanghong Biotech	S20130002	Anti-VEGF agent administered by intravitreal injection.
Fundus Fluorescein Angiography (FFA) System	Heidelberg Engineering GmbH	222611 / FDA 510(k): K101223	System used for fluorescein angiography and retinal imaging.
Ranibizumab (Lucentis)	Novartis	50242-080-03	Anti-VEGF agent administered by intravitreal injection.
Slit-Lamp Biomicroscope	Haag-Streit AG, Bern, Switzerland	BQ 900	Slit-lamp biomicroscope used for anterior and posterior segment examination.
Spectral-Domain Optical Coherence Tomography (SD-OCT) System	Heidelberg Engineering, Germany	222611	Used to measure central subfield thickness (CST) and assess macular edema morphology.
SPSS Statistics Software	IBM Corp.	Version 26.0	Used for statistical analyses, including t-tests, Mann-Whitney U tests, chi-square tests, Fisher's exact tests, and logistic regression.