3686 - Development of an Lqrg-TCP Model Incorporating DCE-MRI Derived Ktrans to Predict Heterogeneous Tumor Response In Hypo-Fractionated Radiotherapy for Locally Advanced Non-Small Cell Lung Cancer
Presenter(s)
H. Zhang1, D. Wang2, and H. Liu2; 1Sun Yat-sen University Cancer Center, Guangzhou, China, 2State Key Laboratory of Oncology in South China, Guangdong Provincial Clinical Research Center for Cancer, Department of Radiation Oncology, Sun Yat-sen University Cancer Center, Guangzhou, Guangdong, China
Purpose/Objective(s): To develop and validate a tumor control probability (TCP) model that incorporates dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI)-derived vascular permeability (Ktrans) to predict local control in locally advanced non-small cell lung cancer (LA-NSCLC) treated with hypo-fractionated radiotherapy (hypo-RT).
Materials/Methods: The unresectable LA-NSCLC patients from three independent cohorts (NCT03006575, NCT03900117 and NCT04212052) were enrolled retrospectively to construct and validate an extended LQRG-TCP model. All patients received hypo-RT and pretreatment DCE-MRI was performed to calculate the volume transfer constant (Ktrans). A training cohort was used for model development, and a separate, external cohort was used for validation. The primary outcome was local progression-free survival (LPFS). A Ktrans-based LQRG-TCP model was developed. Model performance was assessed using the standard deviation (SD), mean absolute error (MAE), and mean relative error (MRE), and compared to a conventional LQRG-TCP model without Ktrans stratification.
Results: A total of 339 patients (training cohort, n=226; validation cohort, n=113) were included. The DCE-MRI driven model stratified tumors into high- and low-Ktrans groups with distinct fitted a/ß ratios (high-Ktrans: 10.54; low-Ktrans: 9.29). In the training cohort, the Ktrans-based LQRG-TCP model significantly improved LPFS prediction over the conventional LQRG-TCP model (MAE: 2.24% vs. 12.70%, MRE: 2.87% vs 21.35% for high-Ktrans group, P < .001; MAE: 6.47% vs. 12.70%, MRE: 11.18% vs 21.35% for low-Ktrans group, P < .001). This superior performance was confirmed in the external validation cohort (MAE: 4.29% vs. 12.70%, MRE: 5.18% vs 21.35% for high-Ktrans group, P < .001; MAE: 8.67% vs. 12.70%, MRE: 13.10% vs 21.35% for low-Ktrans group, P < .001).
Conclusion: In this study, a novel LQRG-TCP model incorporating DCE-MRI-derived Ktrans significantly improved the prediction of local tumor control after hypo-RT for LA-NSCLC by identifying distinct radiobiological phenotypes. These findings suggest that imaging biomarkers can personalize radiobiological modeling, potentially aiding in clinical prognosis and therapeutic strategy.