基于logistic回归的外阴上皮内非瘤样病变和外阴上皮内瘤变2及以上病变风险预测模型的构建

Construction predictive model of non-neoplastic epithelial disorders of the vulva and VIN 2+risk based on logistic regression

目的探讨外阴上皮内非瘤样病变(non-neoplastic epithelial disorders of the vulva,NNEDV)及外阴上皮内瘤变(vulvar intraepithelial neoplasia,VIN)2及以上病变(VIN2+)的危险因素,应用临床资料构建NNEDV及VIN 2+风险预测模型。方法以122例NNEDV及48例VIN 2+患者的临床资料为模型样本,同期68例正常外阴组织者为对照组。采用SPSS 26.0软件进行单因素分析、logistic多因素回归分析,并构建logistic回归模型。采用受试者工作特征(ROC)曲线下面积(AUC)评估模型的预测能力,Hosmer-Lemeshow拟合优度检验评估其校准度,混淆矩阵回代分析验证模型的正确率。结果logistic回归分析结果显示,绝经(OR=2.699,95%CI:1.184~6.151,P<0.05)、临床主要表现为外阴瘙痒(OR=3.758,95%CI:1.537~9.190,P<0.01)与外阴上皮内非瘤样病变患病风险存在显著关联,通过显著性检验,风险预测模型有统计学意义(P<0.001),预测模型ROC曲线下面积AUC为0.796(95%CI:0.727~0.864),Hosmer-Lemeshow检验具有显著性(P>0.05),模型总体预测正确率为88.5%(108/122)。HPV 16阳性(OR=7.027,95%CI:2.128~23.201,P<0.01)、临床主要表现为外阴皮损(OR=8.273,95%CI:1.810~37.819,P<0.01)与VIN2+患病风险存在显著关联,通过显著性检验,风险预测模型有统计学意义(P<0.001),预测模型ROC曲线下面积AUC为0.826(95%CI:0.747~0.904),Hosmer-Lemeshow检验具有显著性(P>0.05),该模型总体预测正确率为75.9%(88/116)。结论绝经、临床表现为外阴瘙痒与NNEDV患病风险显著相关,HPV16阳性、临床表现为外阴皮损与VIN2+患病风险显著相关。

Objective To investigate the risk factors for non-neoplastic epithelial disorders of the vulva(NNEDV)and VIN 2+and to construct a risk prediction model for NNEDV and VIN2+by applying clinical data.Methods The clinical data of 122 patients with NNEDV and 48 patients with VIN2+were used as model samples,and 68 patients with normal vulvar tissues were used as control group.SPSS 26.0 software was used for the Chi-square Test.Fisher Test and logistic regression analysis,and the logistic regression model were constructed.Using pathological diagnosis results as the gold standard,the value of the prediction model for NNEDV and VIN2+was assessed through the analysis of the ROC curve.The Hosmer-Lemeshow goodness of fit test was used to assess its calibration,and confusion matrix back generation analysis was used to verify the correctness of the model.Results Menopause(OR=2.699,95%CI:1.184~6.151,P<0.05),pruritus of the vulva(OR=3.758,95%CI:1.537~9.190,P<0.01)were significantly associated with the risk of NNEDV,and the risk prediction model was statistically significant(P<0.001).The areas under the curve(AUC)of the prediction model was 0.796(95%CI:0.727~0.864).Hosmer-Lemneshow test was significant(P>0.05),and the overall prediction correctness of the model was 88.5%(108/122).HPV 16 positive(OR=7.027,95%CI:2.128~23.201,P=0.001),vulvar skin lesion(OR=8.273,95%CI:1.810~37.819,P<0.01)were significantly associated with the risk of VIN2+,and the risk prediction model was statistically significant(P<0.001).The areas under the curve(AUC)of the prediction model was 0.826(95%CI:0.747~0.904).Hosmer-Lemneshow test was significant(P>0.05)and the overall prediction correctness of the model was 75.9%(88/116).Conclusions Menopause and the clinical manifestation of vulvar skin lesions are significantly associated with the risk of NNEDV.HPV 16 positive and the clinical manifestation of vulvar mass are significantly associated with the risk of VIN2+.