Visual and anatomical failure of anti-VEGF therapy for retinal vascular diseases: a survival analysis of real-world data.

Importance: Predicting undesirable outcomes following anti-VEGF initiation in macular oedema is critical for effective clinical decision-making and optimised care.

Objective: To estimate the time to undesirable events in diabetic macular oedema (DMO), central and branch vein occlusions (CRVO and BRVO) after appropriate loading doses with either ranibizumab or aflibercept and identified baseline predictors of negative outcome.

Design, Setting, Participants: A retrospective cohort study of 3277 patients (N = 2107 in DMO, N = 413 in CRVO and N = 757 in BRVO) collected over a 10-year period, in a large UK tertiary centre. Only one eye was included per patient. Inclusion criteria pre-specified a minimum of two clinic visits with one being at least 6 months post treatment.

Main Outcome And Measures: The main outcome measure was absence of visual acuity (VA) improvement due to macular oedema failure of anti-VEGF therapy (defined as VA gain <5 ETDRS letters and CST increase of 50 µm or CST > 325 µm) modelled using time-event analyzes of appropriately loaded patients. Secondary outcomes included survival curves by individual condition (DMO, CRVO, BRVO) and factors associated with negative outcomes.

Results: After starting anti-VEGF, there was a 50% chance of undesirable outcomes at 2.3, 5.24 and 6.16 years for DMO, CRVO and BRVO, respectively. Cox proportional hazards modelling identified presenting age, intraretinal (IRF) volume, presence of DMO and VA as predictors of negative outcomes, whilst South East Asian ethnicity conferred an independent protective effect.

Conclusion: Real-world data suggest that undesirable events following anti-VEGF injections is likely to in 50% of patients by the third year of treatment in spite of appropriate loading. The definition of undesirable treatment events captured nearly all patients who were escalated to another therapy, but this proportion represented a small percentage of our definition of failed response.