Impact of surgical approach according to PD-ROBOSCORE in pancreaticoduodenectomy: A single-center external validation and comparative study.

Abstract

Pancreaticoduodenectomy remains a technically demanding procedure associated with substantial postoperative morbidity. The PD-ROBOSCORE was developed to quantify procedural difficulty using preoperative variables, but external validation and its clinical relevance across different surgical approaches remain limited. This study aimed to validate PD-ROBOSCORE and assess its prognostic performance in predicting major postoperative complications. We conducted a retrospective single-center study including consecutive patients undergoing open or robotic pancreaticoduodenectomy between 2020 and 2025. PD-ROBOSCORE was calculated using the original weighted formula and analyzed as a continuous variable, by quartiles, and using the predefined high-difficulty threshold (≥9). The primary end point was major postoperative complications (Clavien-Dindo ≥3). Associations were evaluated using logistic regression. Discrimination was assessed using receiver operating characteristic analysis. Incremental predictive value was examined by comparing a baseline clinical model with and without PD-ROBOSCORE. A total of 381 patients were included; 36.5% developed major complications. Higher PD-ROBOSCORE values were independently associated with increased odds of major morbidity (odds ratio, 1.12; 95% confidence interval, 1.01-1.15; P = .02), with a significant trend across quartiles (P = .011). The score demonstrated good discrimination (area under the curve, 0.727), with similar performance in open and robotic pancreaticoduodenectomy (P = .91). Among high-difficulty cases (n = 98), no statistically significant difference in major postoperative complications was observed between robotic and open pancreaticoduodenectomy. Addition of PD-ROBOSCORE improved predictive accuracy compared with the clinical model alone (area under the curve, 0.748 vs 0.673; P = .003). PD-ROBOSCORE is an effective predictor of major postoperative morbidity after pancreaticoduodenectomy and provides incremental prognostic value beyond patient-related factors. It may support preoperative risk stratification and surgical decision-making, particularly in technically complex cases.