Efficacy and prognostic factors of anti-VEGF treatment for neovascular age-related macular degeneration: An OCTA imaging-based deep learning analysis.

BACKGROUND

Anti-VEGF therapies improve visual acuity and reduce central macular thickness (CMT) in neovascular age-related macular degeneration(nAMD) patients. However, In our study, 35.76 % patients do not respond adequately to the anti-VEGF treatment.. Therefore, this study aimed to investigate imaging biomarkers and forecast the influencing factors of anti-VEGF treatment response in nAMD, aiming to enhance clinical evaluation.

METHODS

We retrospectively analyzed data from patients treated with anti-VEGF at Tianjin Medical University General Hospital from August 2018 to August 2023, including 165 patients with exudative AMD. We investigated their sex, age, best-corrected visual acuity (BCVA)(LogMAR), number of anti-VEGF treatments, optical coherence tomography angiography (OCTA) results. A deep learning model, the improved LUNet model was used to analyze OCTA images, focusing on retinal features including foveal avascular zone (FAZ), vessel density (VD), vessel diameter index (VDI), vessel dispersion (Vdisp) of retinal layers(deep vascular complex and deep vascular complex) to assess their relationship with treatment response.

RESULTS

After 6 months of anti-VEGF treatment, the patients' BCVA improved from 0.84 ± 0.40 to 0.70 ± 0.40 (t = 2.85, P = 0.005) (corresponds to an improvement from 20/140 to 20/100 on the Snellen scale.), CMT decreased from 300.16 ± 118.92 μm to 249.53 ± 84.39 μm (t = 4.46, P < 0.001). FAZ perimeter increased from 2.68 mm (2.29, 3.50) to 3.04 mm (2.57, 3.59) (Z=-2.05, P = 0.040), VDI decreased from 6.09 ± 0.83 to 5.90 ± 0.71 (t = 2.33, P = 0.021), central VD decreased (t = 2.21, P = 0.028). Logistic regression showed the Vdisp of macular neovascularization (Vdisp-MNV) (P = 0.007, OR 1.41), Vdisp of deep vascular complex(DVC) (P = 0.006, OR 0.45), neovascular surface area (SA) (P = 0.002, OR 0.26), and pigment epithelial detachment(PED) (P = 0.009, OR 0.29) significantly affected the response to anti-VEGF treatment. There were statistical differences in the association between the base line Vdisp of DVC (b = 0.07, t = 2.50, P = 0.014), base line intraretinal fluid (IRF) (b = 0.17, t = 2.02, P = 0.045) and base line BCVA, shown with multiple linear regression analysis.

CONCLUSION

This study analyzed factors affecting treatment response. Vdisp in MNV showed positive correlation with anti-VEGF treatment response; SA and PED showed negative correlation with anti-VEGF treatment response. It allowed for making personalized treatment strategy to tailor interventions based on individual risk profiles and vascular characteristics which can enhance patient outcomes.