A new retrospective study published in Cancer Imaging compared two PET tracers, 18F-FAPI-04 and 18F-FDG, in radiomics models designed to predict early systemic therapy efficacy in PDAC, and found that the FAPI-based model achieved numerically higher discrimination, although the difference did not reach statistical significance in a paired validation cohort.
The researchers analyzed 155 patients with histologically confirmed PDAC who received systemic therapy. Of these, 118 patients formed the training cohorts; all had undergone pre-treatment contrast-enhanced CT and hepatic MRI, and each had received either 18F-FAPI-04 PET/CT (60 patients) or 18F-FDG PET/CT (58 patients).
Feature selection was performed using the least absolute shrinkage and selection operator (LASSO), after which logistic regression was used to construct three models: a baseline radiomics model, a baseline radiomics-FDG model, and a baseline radiomics-FAPI model. The design deliberately combined information from multiple imaging modalities, so the comparison between tracers was made against a common multimodal background rather than in isolation.
A key strength of the study was its validation strategy. The models were tested in an independent cohort of 37 patients who underwent paired 18F-FAPI-04 and 18F-FDG PET/CT, meaning the same patients were scanned with both tracers. This allowed a genuine head-to-head comparison of the two tracer-based models within the same individuals, reducing confounding from differences in patient characteristics between tracer groups.
Performance was evaluated with receiver operating characteristic (ROC) curve analysis, calibration curves, and decision curve analysis (DCA). The baseline radiomics-FAPI model achieved an AUC of 0.863 (95% CI: 0.762–0.946) in the training cohort and 0.829 (95% CI: 0.682–0.952) in the validation cohort. These values were numerically higher than those of the baseline radiomics model, which reached AUCs of 0.791 in training and 0.755 in validation, and the baseline radiomics-FDG model, which reached 0.848 in training and 0.764 in validation.
Despite the numerical advantage for the FAPI-based model, the authors emphasize that the differences were not statistically significant. In the paired validation cohort, DeLong tests comparing the two tracer-based models yielded a ΔAUC of 0.065 (95% CI: −0.064 to 0.202) with P = 0.344. In other words, the apparent superiority of 18F-FAPI-04 in this dataset is exploratory and cannot be considered a demonstrated improvement over 18F-FDG.
The FAPI-based model also showed stable calibration in validation, with a Brier score of 0.171, and decision curve analysis indicated a higher estimated net benefit across most threshold probabilities. Calibration and net benefit matter because a model that discriminates well but is poorly calibrated can mislead clinical decisions; here, the reported metrics suggest the FAPI model’s predicted probabilities tracked observed outcomes reasonably well in the validation cohort.
The study’s conclusion is deliberately measured: the 18F-FAPI-04-based clinical-radiomics model predicts systemic therapy efficacy in PDAC, showing numerically higher, though not statistically significantly different, discrimination compared with the FDG-based model in the paired validation cohort. The authors state that these findings are exploratory and warrant prospective multicenter validation before any clinical application.
Several limitations follow from the study design. The work is retrospective, the training cohorts were not randomized between tracers, and the validation cohort, while valuable for its paired design, comprised only 37 patients, which limits the precision of the performance estimates, as reflected in the wide confidence intervals. Until prospective multicenter studies confirm these results, the comparison between FAPI-04 and FDG radiomics for predicting early therapy response in PDAC remains an open question that this study helps frame rather than settle.
Subject of Research: Comparing 18F-FAPI-04 and 18F-FDG PET/CT radiomics models for predicting early systemic therapy response in pancreatic ductal adenocarcinoma.
Article Title: Head-to-head comparison of 18F-FAPI-04 and 18F-FDG PET/CT in multimodal radiomics for early prediction of systemic therapy response in pancreatic ductal adenocarcinoma
Article References: Chen, T., Li, X., Lin, L., Chen, Y., Xue, X., Xiao, W., Zhang, X., Su, X., & Bai, X. (2026). Head-to-head comparison of 18F-FAPI-04 and 18F-FDG PET/CT in multimodal radiomics for early prediction of systemic therapy response in pancreatic ductal adenocarcinoma. Cancer Imaging. https://doi.org/10.1186/s40644-026-01144-3
Image Credits: AI Generated
DOI: 10.1186/s40644-026-01144-3
Keywords: pancreatic ductal adenocarcinoma, 18F-FAPI-04, 18F-FDG, PET/CT, radiomics, systemic therapy response, prediction model, logistic regression, ROC analysis, decision curve analysis, retrospective study, Cancer Imaging
