2154 - Noninvasive Radiomic Biomarkers for Predicting Pseudoprogression and Hyperprogression in Non-Small Cell Lung Cancer Patients Treated with Immune Checkpoint Inhibitors

Y. Li¹, P. Wang², J. Xu¹, X. Shi¹, T. Yin^1,3, J. Yu⁴, and F. Teng¹; ¹Department of Radiation Oncology, Shandong Cancer Hospital and Institute, Shandong First Medical University and Shandong Academy of Medical Sciences, Jinan, Shandong, China, ²Shandong Cancer Hospital & Institute, Cheeloo college of medicine, Shandong University, Jinan, China, ³Cancer Center, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, Hubei, China, ⁴Shandong Cancer Hospital and Institute, Shandong First Medical University and Shandong Academy of Medical Sciences, Jinan, Shandong, China

Purpose/Objective(s): The occurrence of atypical responses, such as pseudoprogression (PP) and hyperprogression (HPD) induced by immune checkpoint inhibitors(ICIs), can significantly impact treatment strategies. However, the identification of reliable biomarkers to predict these responses remains elusive. This study aims to develop and validate a radiomics model that utilizes computed tomography (CT) imaging for predicting PP and HPD in patients with non-small cell lung cancer (NSCLC) treated with ICIs. Materials/

Methods: Between 2019 and 2022, A total of 105 NSCLC patients treated with ICIs were enrolled in this retrospective study, Based on the RECIST V1.1 criteria, the cohort was divided into three groups: PP (n=16), HPD (n=44), and progressive disease (PD, n=45). The target lesions were delineated using ITK-SNAP software, from which a total of 6008 radiomic features capturing both intra- and peritumoral texture patterns were extracted. T-tests and LASSO regression analysis were employed to identify the most significant radiomic features. Logistic Regression (LR) and Support Vector Machine (SVM) models were developed and compared for classification efficacy. The models performances were evaluated using Receiver Operating Characteristic (ROC) curves and the Area Under the Curve (AUC). Additionally, Kaplan–Meier survival analysis and log-rank tests were conducted to determine the prognostic value of the models.
Results: No significant clinical differences were observed across the three groups. Four intra- and three peritumoral radiomic features demonstrated strong capability in distinguishing PP from true progression (PD and HPD) using the Logistic Regression (LR) model, achieving an AUC of 0.951 (95% CI: 0.911-0.991) in the training set and 0.875 (95% CI: 0.657-1) in the testing set. Additionally, the Support Vector Machine (SVM) model, utilizing selected radiomic features, exhibited superior performance in differentiating PD from HPD, with AUCs of 0.969 (95% CI: 0.931-1) in the training set and 0.868 (95% CI: 0.723-1) in the testing set. Moreover, our model exhibited exceptional proficiency in accurately predicting patient prognosis. The KM survival curves for the high-risk and low-risk groups, as distinguished by the model, show good concordance with the survival curves of the PP and PD patient populations. This finding aligns with our empirical observations that patients with PP generally have better OS outcomes than those with true progression.
Conclusion: Our study demonstrates that radiomics features extracted from baseline CT scans are effective in predicting PP and HPD patients with NSCLC treated with ICIs.