Establishment and validation of a nomogram for predicting perioperative complications of retroperitoneal laparoscopic adrenalectomy.

Abstract

While laparoscopic adrenalectomy (LA) represents a gold standard for treating most adrenal lesions, no effective visual model for the prediction of perioperative complications of retroperitoneal laparoscopic adrenalectomy (RLA) exists. A retrospective study was conducted involving all consecutive patients underwent unilateral RLA for adrenal disease from January 2012 to December 2021. The entire cohort was randomly divided into 2 subsets (70% of the data for training, 30% for validation). Subsequently, a Least Absolute Shrinkage Selection Operator (LASSO) regression was performed to select the predictor variables, which were further consolidated via random forest (RF) and Boruta algorithm. Then the nomogram was established using the bivariate logistic regression analysis. Eventually, the receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA) were employed to evaluate discrimination, calibration and clinical usefulness of the model, respectively. A total of 610 patients underwent unilateral RLA for adrenal diseases were enrolled. After machine learning analyses, a weighted nomogram was established with 7 factors associated with complications including operative time, lesion laterality, intraoperative blood loss, pheochromocytoma, body mass index (BMI) and 2 preoperative comorbidities [respiratory diseases, cardiovascular diseases (CVD)]. The model displayed a fine calibration curve for perioperative complications evaluation in both the training dataset (P=0.847) and validation dataset (P=0.248). ROC with area under the curve (AUC) revealed excellent discrimination in the training dataset (0.817, 95% CI: 0.758-0.875) and validation dataset (0.794, 95% CI: 0.686-0.901). DCA curves showed that using this nomogram provided a more net benefit where threshold probabilities lay in the range of 0.1 to 0.9. An effective nomogram that incorporating 7 predictors was established in this study to identify patients at high risk of perioperative complications for RLA. It would contribute to the improvement of perioperative strategy due to its accuracy and convenience.