Zimmerer J, Haubitz I, Mainos D, Hadass H, Tittor W
Stoffwechselklinik der LVAW, Bad Mergentheim.
Z Gastroenterol. 1996 Jul;34(7):421-7.
A survival analysis was carried out based on the data of 190 male patients with alcoholic liver cirrhosis (Child A: 82.2%; Child B: 17.8%). Patients (mean age: 49.6 +/- 7.1 years) were examined during the period 1983-1990. Censoring in May 1993 was based on the recordings of the "Rentenversicherungsanstalten". There were no "drop-outs". During follow-up (mean: 4.2 years) 64 (33.7%) of the patients died. 13 potential prognostic variables were examined individually by drawing Kaplan-Meier curves and performing log-rank tests. Portal pressure, determined during hepatic vein catheterization as hepatic vein pressure gradient HVPG (P), size of esophageal varices, serum bilirubin, serum albumin, prothrombin time (Quick), thromboplastin time (PTT), cholinesterase (ChE) and Child scores were correlated to survival (p < 0.05), whereas age, gamma GT, IgA, drinking habits and additional diagnoses were not. A multivariate Cox regression analysis stepwise eliminated all but three variables: ChE, albumin and variceal size were included in the prognostic index PI of the final model. The usefulness of the model was tested by a cross validation method. No significant difference was found between estimated and observed survivorship functions. To compare the PI of the Cox model with Child's scores, ROC curves of sensitivity and specificity of predicting death within one, three and five years were constructed. Better prognostic efficiency was indicated for PI. Because ChE, albumin and the size of varices are determined as a routine in our clinic, we consider the construction of PI an advisable alternative to Child's classification.
基于190例男性酒精性肝硬化患者的数据进行了生存分析(Child A:82.2%;Child B:17.8%)。患者(平均年龄:49.6±7.1岁)于1983年至1990年期间接受检查。1993年5月的截尾数据基于“养老保险机构”的记录。没有“失访”情况。在随访期间(平均:4.2年),64例(33.7%)患者死亡。通过绘制Kaplan-Meier曲线并进行对数秩检验,对13个潜在的预后变量进行了单独检查。通过肝静脉导管插入术测定的门静脉压力作为肝静脉压力梯度HVPG(P)、食管静脉曲张大小、血清胆红素、血清白蛋白、凝血酶原时间(Quick)、凝血活酶时间(PTT)、胆碱酯酶(ChE)和Child评分与生存相关(p<0.05),而年龄、γ-谷氨酰转移酶、IgA、饮酒习惯和其他诊断则无相关性。多变量Cox回归分析逐步排除了除三个变量之外的所有变量:ChE、白蛋白和静脉曲张大小被纳入最终模型的预后指数PI。通过交叉验证方法对模型的有效性进行了测试。估计的和观察到的生存函数之间未发现显著差异。为了将Cox模型的PI与Child评分进行比较,构建了预测1年、3年和5年内死亡的敏感性和特异性的ROC曲线。结果表明PI具有更好的预后效率。由于ChE、白蛋白和静脉曲张大小在我们诊所是常规检测项目,我们认为构建PI是Child分类的一个可取替代方法。