Abstract 418P
Background
The clinical decision-making regarding carrying out surgery alone (SA) or surgery followed by postoperative adjuvant chemotherapy (SPOCT) in esophageal squamous cell carcinoma (ESCC) remains controversial. This study aimed at proposing a pre-therapy PET/CT image-based deep learning approach to improve the survival benefit and clinical management of patients with ESCC.
Methods
This retrospective multicenter study included 837 patients with ESCC from three institutions. Prognostic biomarkers integrating six networks were developed to build an ESCC prognosis (ESCCPro) model and predict the survival probability of patients with ESCC treated with SA and SPOCT. Patients who did not undergo surgical resection were in a control group. Overall survival (OS), disease-free survival, and progression-free survival were the end-point events. Seven clinicians with varying experience evaluated how ESCCPro performed in assisting clinical decision making.
Results
For SA patients, ESCCPro yielded a C-index of 0.705 for predicting OS in the test dataset. No significant differences in survival were found between SA patients with predicted poor outcomes and the control cohort (P > 0.05). For SPOCT patients, ESCCPro yielded a C-index of 0.695 for predicting OS in the test dataset. The clinical implementation of ESCCPro improved the median OS by 8 months and the 2-year OS rate by 12.0% in SA patients. It also improved the median OS by 15 months and the 3-year OS by 24.0% in SPOCT patients, and significantly improved prognosis accuracy, certainty, and the efficiency of clinical experts.
Conclusions
ESCCPro improved the survival benefit of patients with ESCC and aided in the clinical decision-making among the two treatment approaches.
Legal entity responsible for the study
The authors.
Funding
Has not received any funding.
Disclosure
All authors have declared no conflicts of interest.