Abstract
BACKGROUND: Many imaging scoring models have been developed for tumor surgery to provide critical guidance for the selection of surgical methods. However, little research has been aimed at developing scoring models for adrenal tumors and retroperitoneal laparoscopic adrenal surgery (RLAS), which has become the primary technique for treating adrenal tumors. The study set out to establish a computed tomography (CT)-based adrenal tumor scoring model for predicting perioperative outcomes in patients with adrenal tumors who have undergone RLAS. METHODS: The retrospective analysis included 306 patients with adrenal tumors diagnosed by preoperative unenhanced or enhanced CT from January 2014 to August 2018 in the First Affiliated Hospital of Fujian Medical University. CT images were used to quantify the tumor location and size; the relationships of the tumors with the surrounding organs and tissues, the large abdominal blood vessels, and the upper poles of the kidneys and renal hila; the adhesion of periadrenal fat (PF); and the tumor CT enhancement value. We conducted multivariate ordinal logistic regression analysis to screen variables and performed principal component analysis to construct a novel scoring model for RLAS. The perioperative outcomes of RLAS were evaluated according to postoperative length of stay, operative time (OT), intraoperative blood loss (IBL), and postoperative complications. RESULTS: The final scoring model included tumor size; the relationships of the tumors with the surrounding organs and tissues, the large abdominal blood vessels, and the upper poles of the kidneys and renal hila; the tumor CT enhancement value; the adhesion of the PF; and the functional status of adrenal tumors. The total score had positive correlations with the OT (r(s)=0.431), IBL (r(s)=0.446), and postoperative length (r(s)=0.180) (all P values <0.001). Compared to any single metric, the total score provided better prediction of OT and IBL. The grading system for RLAS based on the scoring model also performed well in predicting the complexity and difficulty of RLAS. The coincidence rate for these factors was good (all P values <0.001). CONCLUSIONS: The developed model is feasible and repeatable in the prediction of the perioperative outcomes, complexity, and difficulty of RLAS.