Clinicoradiomic model incorporating orbital adipose tissue radiomic features for predicting response to immunotherapy in patients with metastatic non-small cell lung cancer

BACKGROUND. Host-related biomarkers may improve prediction of response to immune checkpoint inhibitors. However, the predictive value of radiomic features derived from orbital adipose tissue remains unknown.OBJECTIVE. To evaluate the prognostic significance of orbital adipose tissue radiomic features and develop a combined clinicoradiomic model for predicting objective response to immunotherapy.MATERIALS AND METHODS. This study included 86 patients with stage IV non-small cell lung cancer (NSCLC) treated with immunotherapy. Baseline head computed tomography scans were used for segmentation of bilateral orbital adipose tissue. Radiomic features (Variance, Entropy, and Size Zone Non-Uniformity) were extracted using PyRadiomics. Independent predictors were identified by multivariable logistic regression, and predictive performance was evaluated using receiver operating characteristic analysis.RESULTS. Objective response was significantly associated with lower radiological density (≥-77 HU; p=0.001), lower Entropy (p<0.001), and lower Size Zone Non-Uniformity (p=0.011). Radiological density (p=0.0001), Variance (p=0.002), and Entropy (p=0.029) were also associated with disease control. Multivariable analysis identified first-line immunotherapy, PD-L1 expression, radiological density, and Entropy as independent predictors of objective response. The combined clinicoradiomic model demonstrated the highest predictive accuracy (area under the curve 0.896).CONCLUSIONS. Radiomic features of orbital adipose tissue are independent predictors of immunotherapy response in metastatic NSCLC.