应用分类树模型研究乳腺癌危险因素

目的应用分类树模型筛选乳腺癌的危险因素，并预测其发病风险，为乳腺癌的干预提供科学依据。方法用问卷调查及实验室检测等方式采集2010年7月至2012年6月就诊于黑龙江省牡丹江市第一、二人民医院及牡丹江医学院附属红旗医院的1023名女性的相关数据，利用分类树模型分析乳腺癌的影响因素，采取ROC曲线对模型进行评价。结果乳腺癌的危险因素为乳腺癌家族史、长期精神压抑、流产次数（≥3次）、初潮年龄（≤12岁）、平均行经时间（〉7d）、足月妊娠、乳腺良性疾病史、腰臀比（30．8）、职业和体质指数（325），食用豆类食品（1周不少于2次）、体育锻炼和哺乳情况（母乳喂养）为乳腺癌发生的保护因素（P〈0．050）。结论利用分类回归树模型可以快速、有效的从众多数据中挖掘出影响乳腺癌发病的主要因素并预测人群乳腺癌的发病风险，在流行病学研究中具有较高的应用价值。

Taxonomic tree model to study risk factors of breast cancer

Objective To analyse the risk factors of breast cancer using taxonomic tree model and predict the risk of the patients, and provide references for the intervention of the breast cancer. Methods The relevant data of 1023 women admitted in the First and Second People＇ s Hospital of Mudanjiang City and the Affiliated Hongqi Hospital of Mudanjiang Medical College from July 2010 to June 2012 were collected using the questionnaire survey, laboratory test, etc. The risk factors of breast cancer were screened using the taxonomic tree model, and ROC curve was used to evaluate this model. Results Family history of breast cancer, long time depression, times of abortion （ ≥3 times）, menarche age（ ≤ 12 years）, mean time of menstruation（ 〉7 d）, term pregnancy, case history of breast benign diseases, waist-to-hip ratio（≥0. 8）and body mass index （ ≥25 ） were risk factors for breast cancer; diet of legume food （ ≥2 times per week）, breast-feeding and physical exercise were protective factors（ P〈0. 050）. Conclusion The taxonomic tree model can screen out fast and effectively the major risk factors of breast cancer and predict the risk of breast cancer in population, indicating good application value in epidemiological studies.