Predictors of Visual Response after Treatment Lapse in Patients with Neovascular Age-related Macular Degeneration

Abstract

Purpose : The standard of care treatment for neovascular macular degeneration (nAMD) is anti-vascular endothelial growth factor (VEGF) agents. However, for those who undergo lapses in treatment, there is little known about how their vision will respond when restarting. This study aims to determine baseline predictors of visual response in patients with nAMD following treatment lapses.

Methods : This retrospective case-control study consisted of 163 patients with nAMD diagnoses and history of anti-VEGF treatment. A lapse was defined as a minimum of 3 months loss to follow-up despite provider recommendation. Vision loss was defined as at least 10 early treatment of diabetic retinopathy study (ETDRS) letters between last pre-lapse and first post-lapse visits; stable vision was defined as less than 10 letter loss post-lapse. Patients were divided into stable vision (n=125) and vision loss (n=38) groups. Stepwise backward logistic regression was used to analyze baseline medical and ophthalmic factors to develop an algorithm predicting patients most likely to have vision loss.

Results : Both groups were similar (p>0.05) with respect to age (84.5±6.99 vs. 83.8±8.94 years), sex (62.4% vs. 68.4% female), and race (93.6% vs. 92.1% white); stable vision values reported first. There was an observed trend for average lapse length which was greater in the vision loss group (6.13±5.79 vs. 4.57±3.19 months, p=0.12); average total number of injections and starting ETDRS vision in the stable vision group was greater (11.5±8.69 vs. 9.37±6.91, p=0.13 and 62.7±12.5 vs. 58.2±20.2, p=0.2). Hypertension and hyperlipidemia rates were greater in the vision loss group (25% vs. 18%, p=0.35 and 28% vs. 19%, p=0.15). Baseline vision, baseline cube volume, length of lapse, and smoking status were included in the simplified regression model to avoid overfitting. No smoking history reduced odds of vision loss while increased cube volume and lapse length raised odds of vision loss (Table 1). The final model had predictive accuracy of 78.1% and 58% area under the curve (AUC) (Table 2). Further modeling is ongoing to determine if addition of other factors increases AUC.

Conclusions : It may be possible to utilize baseline medical and ophthalmic characteristics to predict visual loss in nAMD patients with treatment lapse; however, further analyses with larger sample sizes and additional baseline variables may be needed.

This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.