Logistic回归模型对不明位置妊娠患者早期异位妊娠的预测价值

目的:初步建立预测不明位置妊娠患者(PUL)早期异位妊娠(EP)的Logistic回归模型,评价该模型的应用价值。方法:对妊娠试验阳性的受检者常规进行经阴道能量多普勒超声检查(TV-PDU),对子宫内膜形态学及内膜下血流动力学参数进行测定,联合临床、生化资料行单变量非条件Logistic回归分析,筛选出有意义的变量,在此基础上行多变量Logistic回归,对有意义的研究因素采用逐步法,经优化组合后建立预测EP的Logistic回归模型并进行模型稳定性验证。利用ROC曲线确定诊断阈值和评估各单项指标及其不同指标组合的诊断价值。结果:最终197例(28.8%)为PUL患者,184例资料完整,62例(33.7%)确诊为EP,90例(48.9%)为宫内早孕,32例(17.4%)为宫内早孕流产。多变量Logistic回归分析显示只有4种超声参数对于鉴别EP有统计学意义(P<0.001),以此而建立了三个Logistic回归模型,Youden指数分别为81.2、95.2和97.6,以内膜形态 内膜厚度 对称性 内膜下动脉血流(M3)这一指标组合最优,联合评价诊断价值最高,模型稳定性较好,其ROC曲线下面积(AUC)最大,灵敏度98.4%,特异度99.2%,阳性预测值98.4%,阴性预测值99.2%。结论:Logistic回归分析为研究子宫内膜形态学、血流动力学与EP之间的关系提供了有效的方法,TV-PDU检查早期PUL患者子宫内膜的多变量Logistic回归模型可更早期地估测EP,在临床上无创性预测早期妊娠结局的危险性是可行的。

The prediction of early ectopic pregnancy in the women with pregnancy of unknown location with logistic regression model

Objective:To establish a logistic regression model of predicting early ectopic pregnancy in the women with pregnancy of unknown location(PUL)and to estimate its diagnostic value.Methods:Women with a positive pregnancy test and suspected early pregnancy complications were referred for transvaginal power Doppler ultrasonography(TV-PDU)assessment.Women were first examined by TV-PDU to detect morphology of endometrium and subendometrial hemodynamics.All women with pregnancies that could not be located on the scan had a blood sample taken to quantify the serum human chorionic gonadotropin(hCG)and the clinical data was recorded.All parameters were tested by univariate analysis and then analyzed in multivariate analysis and a stepwise procedure to form logistic regression models for predicting ectopic pregnancy and their stability were tested.Receiver operating characteristics(ROC)was constructed and the Cut-off value of endometrial sonographic findings and clinical data to diagnose ectopic pregnancy were studied,and the accuracy of every univariate and multivariate were evaluated.Results:A total of 683 women were included in the study.In 197 cases(28.8%),the location of pregnancy was unknown.Complete data sets were obtained in 184 cases,including 62(33.7%)ectopic pregnancies,90(48.9%)normal intrauterine pregnancies,32(17.4%)intrauterine pregnancies miscarriages.Multivariate Logistic analysis showed that four ultrasonographic diagnostic parameters contributed significantly to the predictive power of ectopic pregnancy.Three logistic model were constructed and Youden index were 81.2,95.2 and 97.6,respectively.With the use of this model 3 the thickness of the endometrium,the endometria symmetry,endometrial pattern type and blood flow display type,could predict the EP well at the initial visit with the sensitivity of 98.4%,specificity of 99.2%,the positive predictive value of 98.4%,and negative predictive value of 99.2%.The area under the ROC curves of model 3 was the larger.The stability was better.Conclusion:The study among morphology of endometrium,subendometrial hemodynamics and EP are offered by Logistic regression analysis.It is feasible to use logistic regression model for predict-ing ectopic pregnancy earlier and predicting the risk of pregnancy outcome in practice.