Beyond fluid resolution: Long-term real-world retinal biomarker evolution following faricimab for treatment-naïve diabetic macular oedema

Purpose: To characterise longitudinal changes in retinal biomarkers following faricimab treatment in treatment-naïve diabetic macular oedema (DMO), evaluate associated visual, anatomical and treatment-durability outcomes, and explore baseline predictors of treatment response.

Methods: This retrospective single-centre study included treatment-naïve eyes with centre-involving DMO receiving ≥6 intravitreal faricimab injections between January 2024 and January 2026. Visual acuity, central retinal thickness (CRT), treatment intervals and retinal imaging biomarkers were assessed at baseline, after four initial 4-weekly injections (post-loading) and final follow-up. OCT biomarkers comprised intraretinal fluid (IRF), neurosensory detachment (NSD), hyperreflective foci (HRF), disorganisation of the retinal inner layers (DRIL) and epiretinal membrane; fundoscopic biomarkers comprised dot/blot haemorrhages (DBH), hard exudates (HE) and cotton-wool spots (CWS). Images were independently graded by two reviewers using predefined criteria. Exploratory baseline-adjusted regression models with patient-clustered robust standard errors evaluated predictors of visual and anatomical response, with false discovery rate (FDR) correction applied within outcome families.

Results: Forty-five eyes from 34 patients were included, with mean follow-up of 15.1±4.8 months. Retinal biomarkers demonstrated distinct temporal patterns of resolution. NSD showed the most rapid response, decreasing from 26.7% at baseline to 4.4% post-loading and resolving completely by final follow-up. IRF decreased from 100% to 73.3% post-loading and 57.8% finally, while HRF showed slower resolution from 93.3% to 82.2% and 64.4%, respectively. DRIL prevalence was 17.8%, 4.4% and 4.4%, respectively. Fundoscopic biomarkers also improved heterogeneously: DBH decreased from 93.3% to 57.8%, HE from 57.8% to 42.2%, and CWS from 33.3% to 11.1% by final follow-up. Reductions in IRF, NSD, HRF, DBH, HE and CWS were significant (all P≤0.016), whereas DRIL showed no significant overall change (P=0.453).

These structural changes occurred alongside a mean ETDRS improvement of 3.7±12.1 letters (P=0.047) and CRT reduction of 171.2±96.6 µm (P<0.001). Most visual improvement occurred during loading, whereas CRT continued to decrease thereafter. CRT reduction did not correlate with ETDRS gain (r=0.06, P=0.70). Mean final treatment interval was 14.9±5.6 weeks, with 46.7% achieving ≥16-week intervals. Following FDR correction, baseline qualitative retinal biomarkers did not independently predict visual, anatomical or durability outcomes; higher baseline CRT was the only independent predictor of treatment response, predicting greater CRT reduction.

Conclusions: Faricimab produced substantial anatomical improvement in treatment-naïve DMO, but retinal biomarkers demonstrated markedly different trajectories of resolution. Fluid biomarkers, particularly NSD, resolved rapidly, whereas HRF and DRIL persisted despite substantial retinal thinning, and anatomical improvement was not correlated with visual gain. Baseline qualitative biomarkers did not independently predict treatment response after correction for multiple testing. These findings suggest that retinal recovery following faricimab extends beyond fluid resolution and that longitudinal assessment of biomarkers reflecting retinal integrity may provide greater insight into functional recovery than baseline fluid status alone.