Sci Rep. 2025 Dec 5;15(1):43222. doi: 10.1038/s41598-025-27350-w.
ABSTRACT
Diabetic macular edema (DME) is a major cause of visual impairment, and treatment response can vary based on retinal structural changes. This post-hoc analysis of the Phase III VISTA trial evaluated ellipsoid zone (EZ) integrity and fluid dynamics over 100 weeks using a machine learning-enabled OCT segmentation and feature extraction platform. Among 443 eyes randomized to intravitreal aflibercept injections (IAI) or laser photocoagulation, IAI treatment provided greater reductions in intraretinal fluid volume (- 0.933 ± 1.344 mm2 vs. - 0.386 ± 1.080 mm2; p < 0.0001) and better preservation of EZ integrity (- 24.6% vs. + 125.6%; p < 0.0001) compared to laser. The IAI every 4 weeks (2q4) regimen achieved the most sustained improvement in fluid control, while increased fluid volatility in the IAI every 8 weeks (2q8) group was associated with poorer visual outcomes. In contrast, laser treatment was associated with early reduction in EZ integrity, consistent with known effects of photocoagulation, which this study quantified in vivo over time. At week 100, EZ integrity strongly correlated with best-corrected visual acuity (r = 0.61, p < 0.05). No significant differences in treatment response were observed across racial/ethnic groups. These findings highlight the importance of consistent fluid control and demonstrate the utility of machine learning-derived EZ metrics and fluid stability as imaging biomarkers to guide personalized treatment strategies and optimize long-term visual outcomes.Registry: ClinicalTrials.gov, TRN: NCT01363440, Registration date: 27 May 2011.
PMID:41350590 | DOI:10.1038/s41598-025-27350-w