Related Experiment Video
Updated: Jul 25, 2026
Therapy Testing in a Spheroid-based 3D Cell Culture Model for Head and Neck Squamous Cell Carcinoma
Published on: April 20, 2018
To develop a prognostic model for neoadjuvant immunochemotherapy efficacy in esophageal squamous cell carcinoma by
Zhou Yehan1, Qin Sheng1, Yang Hong1
1Department of Pathology, Sichuan Cancer Hospital & Institute, Sichuan Cancer Center, School of Medicine, University of Electronic Science and Technology of China, Chengdu, China.
Objective:
The choice of neoadjuvant therapy for esophageal squamous cell carcinoma (ESCC) is controversial. This study aims to provide a basis for clinical treatment selection by establishing a predictive model for the efficacy of neoadjuvant immunochemotherapy (NICT).
Methods:
A retrospective analysis of 30 patients was conducted, divided into Response and Non-response groups based on whether they achieved major pathological remission (MPR). Differences in genes and immune microenvironment between the two groups were analyzed through next-generation sequencing (NGS) and multiplex immunofluorescence (mIF). Variables most closely related to therapeutic efficacy were selected through LASSO regression and ROC curves to establish a predictive model. An additional 48 patients were prospectively collected as a validation set to verify the model's effectiveness.
Results:
NGS suggested seven differential genes (ATM, ATR, BIVM-ERCC5, MAP3K1, PRG, RBM10, and TSHR) between the two groups (P < 0.05). mIF indicated significant differences in the quantity and location of CD3+, PD-L1+, CD3+PD-L1+, CD4+PD-1+, CD4+LAG-3+, CD8+LAG-3+, LAG-3+ between the two groups before treatment (P < 0.05). Dynamic mIF analysis also indicated that CD3+, CD8+, and CD20+ all increased after treatment in both groups, with a more significant increase in CD8+ and CD20+ in the Response group (P < 0.05), and a more significant decrease in PD-L1+ (P < 0.05). The three variables most closely related to therapeutic efficacy were selected through LASSO regression and ROC curves: Tumor area PD-L1+ (AUC= 0.881), CD3+PD-L1+ (AUC= 0.833), and CD3+ (AUC= 0.826), and a predictive model was established. The model showed high performance in both the training set (AUC= 0.938) and the validation set (AUC= 0.832). Compared to the traditional CPS scoring criteria, the model showed significant improvements in accuracy (83.3% vs 70.8%), sensitivity (0.625 vs 0.312), and specificity (0.937 vs 0.906).
Conclusion:
NICT treatment may exert anti-tumor effects by enriching immune cells and activating exhausted T cells. Tumor area CD3+, PD-L1+, and CD3+PD-L1+ are closely related to therapeutic efficacy. The model containing these three variables can accurately predict treatment outcomes, providing a reliable basis for the selection of neoadjuvant treatment plans.
