Radiomics for Predicting Anti-VEGF Therapy Response

Age-related macular degeneration (AMD) is a significant cause of vision impairment, particularly in patients older than 50, affecting developed nations profoundly. Neovascular AMD (nAMD) presents a severe form characterized by choroidal neovascularization (CNV), leading to progressive vision loss. Antivascular endothelial growth factor (VEGF) therapy has emerged as a standard treatment, blocking the pathophysiology of nAMD and restoring retinal morphology. Optical coherence tomography (OCT) technology allows noninvasive imaging of retinal cross-sections, providing crucial insights into morphological alterations associated with nAMD and therapeutic response.

The Role of Radiomics

Numerous studies have investigated clinical features associated with visual acuity (VA) outcomes in patients with nAMD, identifying predictive biomarkers such as CNV area, intraretinal fluid (IRF), subretinal fluid (SRF), and pigment epithelial detachments. However, limited research has explored the association between disease progression, treatment effectiveness, and morphological alterations in retinal tissue compartments using computational imaging analysis, known as radiomics.

Radiomics involves extracting biologically relevant features from clinical images and holds promise in predicting therapeutic response and guiding treatment decisions. A recent study assessed radiomic features in OCT scans from a phase 2 trial (OSPREY) and found associations between texture-based features and treatment response in patients with nAMD. However, the small sample size necessitated further validation.

For a study published in Translational Vision Science & Technology, Justis Ehlers, MD, and colleagues tested the hypothesis that a machine learning (ML) model trained on a larger dataset from a phase 3 clinical trial could identify robust features associated with CNV and other disease manifestations, potentially predicting the efficacy of anti-VEGF agents in patients with nAMD. Specifically, their goal was “to evaluate the role of texture-based baseline radiomic features and dynamic radiomics alterations within multiple targeted compartments on [OCT] scans to predic response to anti-[VEGF] therapy in [nAMD],” they wrote.

The phase 3 HAWK trial compared the effectiveness of brolucizumab and aflibercept in treating nAMD. OCT scans were obtained at regular intervals, and complete responders were defined as those achieving and maintaining retinal fluid resolution. Radiomic features were extracted from baseline and post-treatment OCT scans, focusing on fluid, subretinal hyperreflective material (SHRM), and sub-retinal pigment epithelium (sub-RPE) compartments.

Three ML experiments were conducted to evaluate the discriminative ability of radiomic features in predicting treatment response:

Radiomic Features Associated With Anti-VEGF Treatment Response

Among 519 participants of the HAWK trial, there were “280 complete responders and 219 incomplete responders,” wrote Dr. Ehlers and team. “Compartmental analysis of radiomics featured identified the sub-RPE and SHRM compartments as the most distinguishing between the two response groups. The [quadratic discriminant analysis (QDA)] classifier yielded areas under the curve of 0.78, 0.79, and 0.84, respectively, using [radiomic features], [dynamic radiomics alterations], and combined [radiomic features], [dynamic radiomics alterations], and [clinical parameters] on [an independent test set].”

The study’s findings suggest that radiomic features within the sub-RPE compartment and SHRM alterations are associated with treatment response in patients with nAMD. The machine learning model showed promising predictive capabilities, although further prospective validation is warranted. The study underscores the potential of radiomics in guiding personalized treatment approaches for nAMD, providing insights into disease progression and treatment efficacy. “Imaging biomarkers, such as radiomics features identified in this analysis, for predicting treatment response are needed to enhance precision medicine in the management of nAMD,” concluded the study authors.