Multivariate logistic regression analysis of postoperative severe complications and discriminant model establishment in gastric cancer post gastrectomy

Abstract

To investigate the main risk factors for postoperative severe complications, and establish Logistic regression model to predict severe complications in gastric cancer following gastrectomy. The data of 1728 gastric cancer patients underwent gastrectomy between June 2001 and June 2007 were analyzed retrospectively. Logistic regression analysis was used to investigate the risk factors for postoperative severe complications in those patients. Postoperative severe complications were associated with extent of lymph node dissection (D(2)(+)-D(3)), chronic obstructive pulmonary disease (COPD), invasion to the adjacent organ, combined organ resection, extent of lymph node dissection (D(2)), diabetes mellitus (DM), TNM staging IV, heart diseases, malnutrition, surgeon's operative volume, operative time, blood loss and age. The Logistic regression model was P = 1/[1+e((14.806-2.523X1-1.792X2-1.558X3-1.551X4-1.270X5-1.150X6-1.101X7-0.981X8-0.817X9-0.657X10-0.578X11-0.542X12-0.309X13))]. A testing sample showed that the accuracy, sensitivity and specificity of the Logistic model were 72.5%, 70.0% and 75.0%, respectively. The extent of nodal dissection (D(2)(+)-D(3)), COPD, invasion to the adjacent organ, combined organ resection, extent of nodal dissection (D(2)), diabetes mellitus, TNM staging IV, heart diseases, malnutrition, surgeon's operative volume, operative time, blood loss and age are the independent risk factors associated with severe complications in gastric cancer post gastrectomy. The Logistic regression model based on these factors is reliable in predicting the severe complications.