Logistic回归联合ROC曲线分析对不明原因复发性自然流产的血栓前状态的诊断价值

目的:研究血栓前状态(PTS)与不明原因复发性自然流产(URSA)的相关性,探寻检查PTS的适合项目,联合Logistic回归和ROC曲线分析建立统计学诊断模型,计算诊断阈值,进行诊断学评价。方法:应用病例分组对照研究方法,收集2011年7月～2012年6月就诊于佛山市中医院妇科的294例妇女,分为URSA组82例和正常生育组212例。检测两组PT、APTT、TT、Fbg、FDP、D-Dimer、HCY 7个与PTS相关的项目。结果:经方差/非参数检验分析,两组间PT、APTT、HCY结果的差异无统计学意义(P>0.05),TT、Fbg和FDP结果的差异有统计学意义(P<0.01);经二元Logistic回归分析显示,Fbg、TT、FDP与URSA的发生有相关性(χ2=51.723,P<0.01),建立统计学诊断模型,Pre(预测值)=1/[1+e(10.696+1.761×1-0.894×2+0.778×3)],Pre值对URSA患者的预测准确性为53.7%,正常者的预测准确性为89.6%,总体准确性为79.6%;应用ROC曲线分析,FDP、Fbg、TT和Pre诊断URSA患者PTS的曲线下面积AUC和95%CI分别是0.665(0.574～0.756)、0.788(0.701～0.874)、0.799(0.718～0.819)、0.843(0.769～0.917);最大Youden指数分别是0.298、0.546、0.512、0.595,对应的诊断阈值分别是:0.55 g/L、2.74 g/L、16.65 s和0.25;相对应的敏感度分别是68.90%、71.70%、70.80%、71.70%,特异度分别是61.00%、82.90%、80.50%、87.80%。结论:PTS与URSA的发生相关;通过检测FDP、Fbg、TT凝血项目,应用二元Logistic回归和ROC曲线分析建立的统计学诊断模型能有效地诊断URSA患者的血栓前状态。

Diagnosis value of Logistic regression combined with ROC curve analysis on unexplained recurrent spontaneous abortion of the prothrombotic state

Objective:To study the correlation of prethrombotic state(PTS) and unexplained recurrent spontaneous abortion(URSA),search the suitable project to check PTS,combined with Logistic regression and ROC curve analysis to establish statistical diagnosis model,computate diagnostic threshold,and conduct diagnostic evaluation.Methods:Case control study method of grouping was applied,and 294 women cases who were cliniced in Foshan Hospital of Traditional Chinese Medicine from 2011 July to 2012 June were collected and divided into 82 cases of URSA patients and normal fertile group(control group) in 212 cases.Related items of PT,APTT,TT,Fbg,FDP,D-Dimer,HCY 7 and PTS were detected.Results:After Variance/Kruskal-Wallis analysis,there was no significant difference(all P>0.05) between the two groups' results in PT,APTT,HCY.TT,Fbg and FDP results showed significant difference(all P<0.01).By binary logistic analysis,there was the correlation among Fbg,TT,FDP and URSA(χ2=51.723,P<0.01),statistical diagnosis model was builded,Pre(predicted) = 1/[1+e(10.696+1.761x1-0.894x2+0.778x3)].The value of Pre in the patients with URSA forecast accuracy was 53.7%,the prediction accuracy of normal was 89.6%,the overall accuracy was 79.6%.Applied the ROC analysis of FDP,Fbg,TT and Pre in diagnosis of patients with URSA PTS area under the curve of AUC and 95 CI were 0.665(0.574~0.756),0.788(0.701~0.874),0.799(0.718~0.819),0.843(0.769~0.917).The largest Youden exponent was respectively 0.298,0.546,0.512,0.595,the corresponding diagnostic thresholds were: 0.55 g/L,2.74 g/L,16.65 seconds and 0.25;the corresponding sensitivity was respectively 68.9%,71.7%,70.8%,71.7%,the specificity was respectively 61%,82.9%,80.5%,87.8%.Conclusion:PTS and URSA are related to occurrence;by detecting FDP,Fbg,TT coagulation application project,and applying binary logistic regression and ROC curve analysis to establish the statistical diagnosis model can more effectively diagnose the patients with URSA prothrombotic state.