Logistics regression analysis and construction of prediction model for lymphatic leakage after radical resection of gynecological malignant tumors
Objective To explore the risk factors of lymphatic leakage after radical resection of gynecological malignant tumors and construct the prediction model so as to provide reference for the prediction of lymphatic leakage.Methods A total of 700 patients with gynecological malignant tumors undergoing radical resection in Xuchang Central Hospital were enrolled between February 2017 and July 2022,and they were randomly divided into modeling group(n=500)and verification group(n=200).According to presence or absence of postoperativelymphatic leakage,patients in modeling group were divided into leakage group(n=43)and non-leakage(n=457).The risk fac-tors of lymphatic leakage after radical resection of gynecological malignant tumors were screened out by logistic re-gression analysis.The prediction model was constructed based on regression coefficients,and the model differentia-tion was evaluated by receiver operating characteristics(ROC)curves.Results Logistic regression analysis showed that overweight,much intraoperative blood loss,pelvic+para-aortic lymph node dissection,much lymph nodes dissection,preoperative anemia and hypoproteinemia were independent risk factors of lymph leakage after radical resection of gynecological malignant tumors.The regression equation to predict lymphatic leakage was as follow:P=1/[1+e-(-6.315+0.944 × BMI+0.881 × intraoperative blood loss+1.496 × lymph node dissection range+0.794 × number of lymph nodes dissection+0.772 ×anemia+0.740 × hypoproteinemia)]the goodness of fit of the regression equation was detected by Hosmer-Lemeshow(P=0.890).The internal verification of data was conducted in modeling group,and its area under ROC curve(AUC),sensitivity,specificity and the cut-off value based on the maximum Youden index were 0.827[95％CI(0.761-0.892)],81.40％,75.71％and 0.094,respectively.The external verification of data was conducted in verification group.In verification group,there were 15 cases with postoperative lymphatic leakage.Taking the above cut-off point as the cut-off value,AUC,sensi-tivity and specificity were 0.817[95％CI(0.710-0.924)],86.67％and 76.76％,respectively.Conclusion The constructed prediction model has good discriminate validity,and which can be applied to identify the high-risk group with lymphatic leakage after radical discriminate of gynecological malignant tumors.
Gynecological malignant tumorRadical resectionLymphatic leakageRisk factorsPrediction model
NETL
NSTL
万方数据
