食管癌切除术后并发症多因素Logistic回归分析及风险模型建立

Multivariate logistic regression analysis of risk factors for complications after esophagectomy for esophageal cancer and establishment of a discriminant model

目的:探讨食管癌切除术后并发症的主要危险因索,建立Logistic判别模型.方法:1997-01/2010-12,对食管癌切除术后患者(n=1197)的临床资料进行回顾性分析,用Logistic回归分析研究术后并发症发生的相关因素.结果:食管癌术后并发症与年龄、COPD、糖尿病、心脏病、医师手术例数、吻合口部位、手术时间、术中失血量、侵及周围脏器和TNMⅢ期有关,获得预测模型为:P=1/[1+e(18.629-0.223X1-2.418X2-0.885X3-1.760X4-1.549X5-0.779X6-0.939X7-1.072X8-1.195X9-0.916X10)].回代样本检验显示,此概率模型判断食管癌术后并发症的准确性为74.0%,敏感性为67.9%,特异性为81.8%.结论:年龄、COPD、糖尿病、心脏病、医师手术例数、吻合口部位、手术时间、术中失血量、侵及周围脏器和TNMⅢ期是食管癌术后并发症的独立危险因素,所建立的Logistic回归模型能较好的预测术后并发症发生的概率.

AIM： To investigate the main risk factors for complications after esophagectomy for esophageal cancer and to establish a logistic regression model to predict these complications. METHODS： The clinical data for patients with esophageal cancer patients （n = 1197） who underwent esophagectomy between January 1997 and December 2010 were analyzed retrospective-ly. Logistic regression analysis was performed to investigate the risk factors for postoperative complications in these patients. RESULTS： Postoperative severe complications were associated with age, chronic obstructive pulmonary disease （COPD）, diabetes mellitus, heart diseases, surgeon volume, position of the anastomotic stoma, operative time, blood loss, invasion to the adjacent organ, and TNM stage Ⅲ. The logistic regression equation was P = 1/ [1＋e（18.629-0.223X1-2.418X2-0.885X3-1.760X4-1.549X5-0.779X6-0.939X7-1.072X8- 1.195X9-0.916X10）]. The accuracy, sensitivity and speci- ficity of the logistic model for discriminating test ing samples were 74.0%, 67.9% and 81.8%, respectively. CONCLUTION： Age, COPD, diabetes mellitus, heart diseases, surgeon volume, position of the anastomotic stoma, operative time, blood loss, invasion to the adjacent organ, TNM stage Ⅲ are independent risk factors for complications after esophagectomy for esophageal cancer. The logistic regression model established based on these factors is reliable in predicting these com- plications.