Investigation Of The Risk Factors And The Initial Exploration Of Predictive Models Of Breast Cancer In Zhengzhou

BackgroundBreast cancer,which usually occurs in breast glandular epithelial tissue, is one of the most common malignant among women around the world.It has serious impacts on women’s physical and mental health and is even life-threatening, with the incidence occult and the prognosis poor, its incidence is increasing and ages of onset are becoming younger. The incidence of breast cancer in China was lower than that of Europe and the United States, but in recent years, the incidence increased year by year, especially in some economically developed areas, the monitoring data shows that:the incidence has leapt to the first female malignancy in Beijing、Guangzhou and Shanghai. However, the cause of breast cancer has not been fully understood. Previous studies suggest that breast cancer is the result of both genetic and environmental factors. However, in recent years, with the progress of society, culture exchange and blending, factors related to female exposure also change. Therefore, risk factors of breast cancer need to be further analyzed and assessed, and prediction model of breast cancer risk factors is a simple and effective way which can be used to screen high-risk groups, it can not only predict the risk of incidence, but also be intervened as early as possible, in order to achieve the primary prevention of the disease and to reduce the burden of disease.ObjectiveTo investigate the relationships between the general condition, education levels, marital status, health status, personal medical history, lifestyles, dieting habits, psychological factors, occupational exposure history and other factors and breast cancer in target population, and to establish the model of breast cancer risk prediction, which can be used to seek breast cancer risk factors, screen high-risk groups, and to provide a scientific basis for the breast cancer prevention and the development of interventions in Zhengzhou.MethodsQuestionnaires were used in the study, hospitalized patients,who had lived in Zhengzhou for over10years, were selected from the center of general surgery treatment in the Third Affiliated Hospital of Zhengzhou University from December2010to June2012, information was collected after the informed consent of research subjects, and81diagnosed cases of breast cancer were as the case group and90cases of non-breast cancer patients to the same period in the hospital as the control group, eventually. Epidata3.0and the double-entry verification and logical consistency verification were used to establish questionnaire database, and then data was exported to convert to SPSS format.Firstly,to compare the differences of the various factors between the case group and the control group, the chi-square test was used to compare the differences of the rates, and rank sum test was used for the intensity of ordered categorical variable levels. Secondly,Multi-Factor Dimensionality Reduction2.0software was used to analyze the interaction of various factors for the significant variables in the univariate analysis, and the optimal model of different combinations of factors was screened out, and multivariate unconditional logistic regression analysis was utilized, and then the establishment of a logistic regression model to verify the results of multi- dimensionality reduction analysis, and to evaluate the logistic regression model. Except Multi-Factor Dimensionality Reduction2.0software, other analysis were operated on software SPSS12.0.Results16risk factors of statistical significance of breast cancer were screened by univariate analysis,which are as follows:body mass indexs (P=0.004), education levels (P=0.005), occupation (P <0.001), dietary habits (P<0.001).smoking (P=0.001), alcohol consumption (P=0.006), physical exercise (P=0.002), menarche age(P<0.001), menstrual cycle (P <0.001), ages for the first time live birth(P=0.039), times of live Birth(P=0.020), the number of abortions (P<0.001), lactation (P <0.001), estrogen replacement therapy (P<0.001), family history of breast cancer (P O.001), history of benign breast disease (P <0.001).The optimal model of MDR is a third-order interaction model, which contains age of menarche, estrogen replacement therapy, family history of breast cancer,and the model has the highest test set balance accuracy and cross-validation consistency,0.813,8/10,respectively,and the permutation test was statistically significant (P<0.001).By logistic regression analysis,a total of eight variables were incorporated into the model,which are BMI X1, eating habits X4, the time of live birth X11, history of abortion X12, the lactation X13, estrogen replacement therapy X14, family history of breast cancer X15, history of benign breast disease X16, the final prediction model of0.323X11(2)+4.002X12-2.813X13+2.438X14+3.950X15+2.683X16.Logistic regression models were evaluated by using the ROC curve, and the area under the curve is0.814,with a95%confidence interval of0.738-0.889.the sensitivity and specificity of the model in breast cancer screening were72.6%,90.2%, respectively, which approaches that of the ROC curve optimal(sensitivity and specificity were75.0%,92.0%,respectively.). ConclusionThe risk factors for breast cancer may include a high body mass index, high-fat eating habits, less of the number of live births, numbers of abortion, no breastfeeding, estrogen replacement therapy,a family history of breast cancer, history of benign breast disease.