放射敏感性指数(RSI):一个用于预测宫颈癌放疗抵抗的患者来源类器官平台

Purpose Radiotherapy is one of the major treatment modalities for cervical cancer; however, treatment responses vary among patients, and no reliable method currently exists to predict individual radiosensitivity. We aim to establish a radiosensitivity index (RSI) model using cervical cancer patient-derived organoids (PDOs) and to evaluate its clinical performance in relation to actual patient outcomes. Patients and Methods Fresh cervical cancer tissues were obtained from 14 patients at diagnosis, and clinical information including stage, radiation history, histology, recurrence, and progression-free survival was retrieved from their medical records. Organoids were established from patient-derived cervical cancer tissues, and their histopathological and genomic features were compared with the original tumors to confirm reproducibility and model fidelity. Each organoid underwent irradiation assays to evaluate radiosensitivity. Radiation-response metrics including area under the survival curve (AUC-survival), growth rate-adjusted slope (GR-slope), and ΔG2/M% as well as clinical stage were standardized using z-score normalization and integrated into a composite RSI. Predictive performance of the RSI was assessed using receiver operating characteristic (ROC) curve analysis, and AUC, sensitivity, specificity, PPV, and NPV were calculated based on Youden's index. Kaplan-Meier analysis was performed to evaluate differences in survival outcomes according to RSI-predicted radiosensitivity. Results All 14 PDOs successfully underwent irradiation assays, and multiparametric integration of AUC-survival, GR-slope, ΔG2/M%, and clinical stage yielded an individualized RSI value. The RSI effectively predicted recurrence after radiotherapy (AUC = 0.844, p = 0.039; sensitivity 80.0%, specificity 88.9%). In addition, the predicted recurrence probability was significantly higher in recurrent patients compared with non-recurrent patients. Survival analysis demonstrated a trend toward improved progression-free survival in the RSI-predicted radiosensitive group compared with the radioresistant group (median 42.3 vs. 15.5 months, p=0.257), suggesting concordance between RSI-based predictions and actual clinical outcomes. Conclusions We successfully developed a personalized preclinical platform based on a cervical cancer PDO-derived RSI model. This is the first PDO-based approach that predicts individual radiosensitivity and validates these predictions against actual clinical outcomes. The RSI model may serve as a preclinical decision-support tool to personalize treatment decisions and guide radiotherapy optimization for cervical cancer patients.