Propose a CT-based noninvasive method to assess tumor fibrosis, guiding precision chemotherapy for unresectable pancreatic cancer

1. Research Background

Pancreatic ductal adenocarcinoma (PDAC) is one of the most aggressive malignancies of the digestive system, with a 5-year survival rate of only approximately 13%, earning it the title of the "king of cancers". Chemotherapy is the mainstay of treatment for unresectable PDAC, with gemcitabine/nab-paclitaxel (AG), FOLFIRINOX, and SOXIRI as first-line regimens. However, the efficacy of the same regimen varies significantly among patients, and there is a lack of effective stratification biomarkers in clinical practice.

Tumor fibrosis is a core pathological feature of PDAC. Dense stromal fibrosis not only affects tumor progression but also is closely related to the delivery and efficacy of chemotherapeutic drugs. Traditional fibrosis assessment relies on invasive biopsy and histological staining, which not only has sampling bias but also fails to reflect the spatial heterogeneity of tumors, making it difficult to meet the needs of clinical personalized treatment. Therefore, the development of a noninvasive and accurate method for assessing tumor fibrosis has become the key to breaking through the precision of pancreatic cancer chemotherapy. Furthermore, transcriptome analysis confirmed that pathways such as collagen metabolism and cell-matrix interaction were significantly enriched in highly fibrotic tumors, validating the biological rationale for quantifying fibrosis.

2. Research Progress

This multicenter study, jointly completed by teams from Sun Yat-sen University Cancer Center, Shenzhen University, and Xiangya Hospital of Central South University, was published in Research 2025. Through multi-cohort and multi-dimensional analysis, the study achieved noninvasive quantification and clinical efficacy prediction of tumor fibrosis in pancreatic cancer, with the core progress as follows:

(1) Establish a quantitative analysis method for WSI fiberization: Based on deep learning technology, tissue segmentation was performed on hematoxylin and eosin (H&E)-stained whole-slide images (WSI) of 361 patients with resectable PDAC, defining fibrosis as stromal proportion and completing quantification. The results showed that patients with high fibrosis had significantly longer overall survival (OS) in the TCGA, SYSUCC, and XYCSU cohorts, verifying the reliability of fibrosis as a prognostic biomarker.

(2) Developing a CT-based non-invasive fibrosis prediction model: Using preoperative contrast-enhanced CT images, 15 fibrosis-related radiomic features were extracted from venous-phase images to construct and validate a radiomics model. The model achieved an area under the curve (AUC) of 0.718 in the external validation cohort, which can noninvasively predict the level of tumor fibrosis, and the prediction results are highly consistent with the WSI quantification results.

(3) Validation of the model's clinical treatment guiding value: Among 295 patients with unresectable PDAC receiving chemotherapy, those with CT-predicted high fibrosis who received AG therapy had their progression-free survival (PFS) extended from 4.70 months to 6.23 months, and overall survival (OS) from 7.73 months to 13.37 months; while in patients receiving FOLFIRINOX or SOXIRI regimens, there was no significant correlation between fibrosis level and efficacy. This finding for the first time confirms that CT-quantified tumor fibrosis can serve as a specific predictive biomarker for the efficacy of AG therapy.

3. Future Prospects

Rapid clinical translation and application: The CT-based fibrosis assessment model can be directly integrated into hospital imaging systems, and preoperative routine CT examinations can quickly determine whether patients are suitable for AG therapy without additional invasive operations. It is expected to become a routine tool for stratification of pancreatic cancer chemotherapy, reducing ineffective treatment and medical costs.

Expanding research on multimodal therapy: Based on fibrosis characteristics, further explore the synergy with targeted therapy and immunotherapy. For example, for high-fibrosis tumors, matrix-targeted drugs can be combined to improve the drug delivery efficiency of AG chemotherapy and further enhance efficacy.

Technology iteration and cross-tumor application: Optimize the model by combining multimodal imaging (MRI, PET-CT) and artificial intelligence technology to improve the accuracy of fibrosis prediction; at the same time, extend this noninvasive assessment strategy to solid tumors rich in stroma such as breast cancer and colorectal cancer, providing a reference for the precision treatment of more tumors.

Sources: https://spj.science.org/doi/10.34133/research.0937