Prediction of resection after preoperative FOLFIRINOX in patients with localized pancreatic adenocarcinoma: a Trans-Atlantic Pancreatic Surgery (TAPS) Consortium study

BACKGROUND: Neoadjuvant chemotherapy is increasingly used for patients with localized pancreatic ductal adenocarcinoma (PDAC), although the likelihood of subsequent resection remains uncertain. The aim of this study was to develop a prediction model using baseline characteristics to estimate the probability of meeting prespecified criteria for resection after induction FOLFIRINOX. METHODS: This retrospective study included patients with localized PDAC who received (m)FOLFIRINOX as initial treatment in five referral centers in the United States and the Netherlands (2012-2019). Multivariable logistic regression identified independent predictors of resection and was used to develop a prediction model. RESULTS: Among 1,835 patients, 18.9% were classified as potentially resectable (PR), 28.9% as borderline resectable (BR), and 52.2% as locally advanced (LA). Observed resection rates were 70.5% for PR, 53.1% for BR, and 17.8% for LA PDAC. Unfavourable independent factors for resection were advanced anatomical stage (BR, OR 0.51; 95% CI 0.37 to 0.69 and LA, OR 0.11; 95% CI 0.08 to 0.15), baseline CA19-9 > 500U/mL (OR 0.60; 95% CI 0.46 to 0.77), a WHO performance status of ≥ 1 (OR 0.41; 95% CI 0.33 to 0.52), and tumor size on baseline imaging >40mm (OR 0.62; 95% CI 0.48 to 0.80). Depending on these factors, the predicted probability of resection ranged from 6.7% to 81.7%. CONCLUSION: At diagnosis, the probability of resection after induction (m)FOLFIRINOX for localized PDAC ranged from 7% to 82%, depending on anatomical stage, CA19-9 level, performance status, and tumor size at baseline. This prediction model may help communicating realistic expectations for the probability of resection. TABLE OF CONTENTS: Resection probabilities after induction (m)FOLFIRINOX for localized pancreatic ductal adenocarcinoma range from 7% to 82% at diagnosis, based on baseline characteristics. The prediction model developed in this study can guide shared decision-making by providing individualized, evidence-based estimates of resectability.