Deep Learning-Based LS-SCLC Patient Selection for the New First-Line Treatment Using High-Dose Hyperfractionated SIB Radiotherapy
Abstract
Purpose
To develop risk-stratification and prognosis-prediction models for limited-stage small cell lung cancer (LS-SCLC) patients, suggesting potential candidates that may benefit from the new treatment protocol using high-dose hyperfractionated simultaneous integrated boost (SIB) radiotherapy to improve overall survival (OS) and avoid over-treatment.
Methods
A total of 182 SCLC patients from an open-label, randomized phase III trial (NCT03214003) were analysed after propensity score matching between 45 Gy and 54 Gy SIB groups. CT deep learning radiomics signature (DLRS) and important clinical factors were incorporated into Logistic regression models to predict prognostic outcomes. Risk stratifications based on DLRS and prophylactic cranial irradiation (PCI) were used to evaluate the impact of 54 Gy SIB across groups.
Results
The models based on DLRS and clinical factors achieved mean time-dependent area under curve (AUCs) of 0.797, 0.627, 0.789, and 0.633 for predicting OS, progression-free survival (PFS), local progression-free survival (LPFS), and metastatic-free survival (MFS), respectively. Using a cutoff value of -0.55, patients classified as having a high DLRS exhibited significantly worse OS, PFS, LPFS, and MFS (all P < 0.05). Notably, patients in the high-DLRS group derived significant benefits from 54 Gy SIB, with improvements observed in OS, PFS, and LPFS (all P < 0.05). Further analyses showed that this group also experienced a significant benefit from PCI, and regardless of whether PCI was administered, 54 Gy hyperfractionated SIB radiotherapy consistently provided a significant OS advantage.
Conclusion
High-DLRS patients are more recommended to receive 54 Gy SIB. For low-DLRS patients, 54 Gy SIB may be unnecessary, otherwise more frequent follow-up using magnetic resonance imaging should be considered. By stratifying patient-specific risk, this work proposed predictive models to select LS-SCLC patients that may potentially benefit from the new 54 Gy hyperfractionated SIB radiotherapy and PCI to avoid unnecessary over-treatment, follow-up examination resource waste, and radiation toxicity.