基于Logistic回归的外阴上皮内瘤变危险度评价模型的构建

Construction of risk assessment model for vulvar intraepithelial neoplasia based on logistic regression

摘要:

目的 应用临床资料和流行病学资料建立外阴上皮内瘤变危险度评价模型。方法 模型样本为101例外阴上皮内瘤变患者、134例外阴良性病变者及150例外阴完全正常者。对其临床资料和流行病学资料进行单因素分析，通过受试者工作特征（ROC）曲线获取影响因素的截断值，构建Logistic回归模型，采用混淆矩阵回代分析验证模型的正确率。结果 经Logistic回归分析，年龄、吸烟、人类乳头瘤病毒（HPV）感染、外阴瘙痒、外阴触血、外阴肿物、外阴疼痛、绝经与否、中性粒细胞绝对值（GRA）、中性粒细胞与淋巴细胞的比值（NLR）、血清糖类抗原125（CA125）指标进入回归方程。回归模型最大似然比（LR）检验具有差异（P <0.05）。混淆矩阵回代分析结果发现该模型总体预测正确率为78.44%（302/385）。结论 应用临床资料和流行病学资料构建外阴上皮内瘤变危险度评价模型，能有效提高外阴上皮内瘤变正确筛查率。

Abstract:

Objective To risk assessment model of vulvar intraepithelial neoplasia established by clinical and epidemiological data. Methods The clinical and epidemiological data of 101 cases with intraepithelial neoplasia, 134 cases with benign lesions, and 150 cases with completely normal vulva were collected. The clinical data were analyzed by single factor analysis, the cut-off values of influencing factors were obtained by ROC curve, the logistic regression model was constructed, and the accuracy of the model was verified by using confusion matrix back analysis. Results Logistic regression analysis: Age, smoking, human papillomavirus (HPV) infection, pruritus of vulva, blood contact of vulva, tumor of vulva, pain of vulva, menopause or not, absolute value of neutrophils (GRA), ratio of neutrophils to lymphocytes (NLR), serum carbohydrate antigen 125 (CA125) index were concluded in the regression equation. The maximum likelihood ratio (LR) test of regression model showed significant difference (P < 0.05). The results of obfuscation matrix analysis showed that the overall prediction accuracy of the model is 78.44% (302/385). Conclusions The risk assessment model of vulvar intraepithelial neoplasia was established by using epidemiological and clinical data, and can effectively improve the correct screening rate of vulvar intraepithelial neoplasia.