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

To develop and validate a pre-treatment clinical-radiomics model based on 18 F-FAPI–04 PET/CT for predicting early systemic therapy efficacy in pancreatic ductal adenocarcinoma (PDAC), and to compare its predictive utility with an ¹⁸F-FDG PET/CT model. A total of 155 patients with histologically confirmed PDAC receiving systemic therapy were retrospectively analyzed. Of these, 118 patients underwent pre-treatment contrast-enhanced CT, hepatic MRI, and either 18 F-FAPI-04 PET/CT ( n = 60) or 18 F-FDG PET/CT ( n = 58) were included as the training cohorts. Radiomics features were extracted from the tumor regions on multimodal images. After feature selection by the least absolute shrinkage and selection operator, logistic regression constructed baseline radiomics, baseline radiomics-FDG, and baseline radiomics-FAPI models. Crucially, the models were validated in an independent cohort of 37 patients who underwent paired 18 F-FAPI-04 and 18 F-FDG PET/CT, allowing for a head-to-head comparison of the two models based on different tracers. Model performance was evaluated using receiver operating characteristic (ROC) curve analysis, calibration curves, and decision curve analysis (DCA). The baseline radiomics-FAPI model achieved AUCs of 0.863 (95% CI: 0.762–0.946) and 0.829 (95% CI: 0.682–0.952) in the training and validation cohorts, respectively, numerically higher than the baseline radiomics model (training AUC: 0.791; validation AUC: 0.755) and the baseline radiomics-FDG model (training AUC: 0.848; validation AUC: 0.764). Notably, DeLong tests in the paired validation cohort showed that these differences were not statistically significant (baseline radiomics-FAPI vs. baseline radiomics-FDG: ΔAUC = 0.065, 95% CI: −0.064 to 0.202, P = 0.344). The baseline radiomics-FAPI model demonstrated stable calibration (validation Brier score: 0.171), and DCA indicated a higher estimated net benefit across most threshold probabilities. 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 a paired validation cohort. These findings are exploratory and warrant prospective multicenter validation before any clinical application.