[Prognostic factors of postoperative delayed gastric emptying after pancreaticoduodenectomy: a predictive model]

Objective: To study the prognostic factors of delayed gastric emptying(DGE) after pancreaticoduodenectomy(PD) and construct a prognostic predictive model for clinical application. Methods: Clinic data of 401 consecutive patients who underwent PD between January 2012 and July 2016 in the First Affiliated Hospital of Harbin Medical University were retrospectively collected and analyzed. The patients were randomly selected to modeling group(n=299) and validation group(n=102) at a ratio of 3∶1. The data of modeling group were subjected to univariate and multivariate analysis for prognostic factors and to construct a prognostic predictive model of DGE after PD. The data of validation group were applied to test the prognostic predictive model. Results: DGE after PD occurred in 35 of 299 patients(11.7%) in the modeling group. The multivariate analysis of the modeling group showed that upper abdominal operation history(χ(2)=6.533, P=0.011), diabetes mellitus(χ(2)=17.872, P=0.000), preoperative hemoglobin <90 g/L(χ(2)=14.608, P=0.000) and pylorus-preserving pancreaticoduodenectomy(PPPD)(χ(2)=8.811, P=0.003) were associated with DGE after PD independently. A prognostic predictive model of DGE after PD was constructed based on these factors and successfully tested. The area under the receiver operating characteristic(ROC) curve was 0.761(95%CI: 0.666-0.856) of the modeling group and 0.750(95% CI: 0.577-0.923) of the validation group. Conclusions: Upper abdominal operation history, diabetes mellitus, preoperative hemoglobin<90 g/L and PPPD are associated with DGE after PD independently. The preoperative assessment of a patient's prognostic for DGE after PD is feasible. The model is a valid tool to take precautions against DGE after PD.