多变量Logistic回归分析在渗出性胸水病因鉴别诊断中的作用

Differential diagnosis of pleural fluid by multivariate logistic analysis

目的利用多变量Logistic回归,综合分析多个临床常规检查结果,探索诊断渗出性胸水良、恶性质的简便方法。方法收集首次就诊时发现渗出性胸水的236例病例资料,对可能有助于诊断的性别、年龄、有无发热、有无异常包块、有无5年以上吸烟史以及胸水中CEA水平等6个指标分别进行单变量分析和多变量Logistic回归分析。结果单变量分析显示年龄、异常包块、吸烟史和胸水CEA具有统计学差别,但诊断价值有限。多变量Logistic回归分析显示性别、年龄、吸烟史、异常包块和胸水CEA水平是恶性胸水的预测指标。异常包块的预测能力最强(OR=10.5,95%可信区间7.8-12.4)。对没有异常包块的患者,性别、年龄、吸烟史和胸水CEA水平仍然是有意义的预测指标。结论利用多变量Logistic回归综合分析性别、年龄、异常包块、吸烟史及胸水CEA等常规临床指标,有助于提高诊断水平。

Objective The aim of this study was to analyze multiple clinical radices with multivariate logistic analysis, to explore a simple and easy method to determine the nature of pleural fluid. Methods We collected data of 236 patients with plemal exsudative effusion, analyzed 6 indices possible helpful for clinical diagnosis as univariate or multivariate logisbc analysis, including age,gender, fever, mass, smoking more than 5 years and level of pleuml CEA. Results Univariate analysis showed age, srnoking, mass and pleural CEA were predictive marker with limited diagnosis value. In multivariate logistic analysis, age, gender, smoking, mass and pleural CEA were able to predict the nature of pleural fluid.Mass was the most predictive index （OR = 10.5,95%CI = 7.8 - 12.4）. Age, gender, smoking and CEA remained helpful for the diagnosis in those without mass. Conclusion The multivariate logistic analysis was helpful to combine multiple clinical indices to improve the diagnosis. Age,gender,smoking,mass and pleural CEA were predictive to the nature of pleural fluid.