Outcome Prediction Using Radiomics for Irinotecan-TACE in CRLM
Overview
This study developed and validated radiomics-based machine learning models to predict overall survival (OS), progression-free survival (PFS), and lesion-level response in patients with colorectal liver metastases (CRLM) treated with irinotecan-loaded drug-eluting microsphere TACE. The findings suggest that integrating imaging-derived features with clinical variables can enhance risk stratification for personalized treatment decisions.
Background
Colorectal liver metastases (CRLM) significantly impact patient prognosis, with many patients ineligible for curative treatments. Transarterial chemoembolization (TACE) using irinotecan has emerged as a potential therapeutic option, yet optimal patient selection remains challenging. Radiomics offers a promising approach to improve treatment personalization by analyzing imaging data to predict outcomes.
Data Highlights
Endpoint
Median Survival (Months)
Overall Survival (OS)
14.5
OS (Salvage Setting)
9.9
PFS (Salvage Setting)
3.8
OS (Post-Inductive Therapy)
19.1
Hepatic PFS (Post-Inductive Therapy)
8.7
PFS (Post-Inductive Therapy)
6.0
Key Findings
152 patients were enrolled across 20 centers in 11 European countries.
Overall survival (OS) was 14.5 months for the entire cohort.
In the salvage setting, OS was 9.9 months and PFS was 3.8 months.
As post-inductive therapy, OS reached 19.1 months, with hepatic PFS of 8.7 months.
Radiomics can enhance risk stratification and treatment personalization in CRLM.
Clinical Implications
The integration of radiomics into clinical practice may improve the selection of patients for irinotecan-TACE, potentially leading to better outcomes. Clinicians should consider utilizing advanced imaging analyses to inform treatment decisions and optimize patient management strategies.
Conclusion
Radiomics-based models show promise in predicting outcomes for patients with CRLM undergoing irinotecan-TACE, highlighting the potential for personalized treatment approaches in this challenging patient population.
by Zuhir Bodalal, Francisco Javier Mendoza Ferradás, Olga Maxouri, Roberto Iezzi, Aleksandar Gjoreski, Stavros Spiliopoulos, Zoltan Bansaghi, Belarmino Gonçalves, Bleranda Zeka, Nathalie Kaufmann, Julien Taieb, Regina Beets-Tan, Philippe L. Pereira, Fernando Gómez Muñoz
