Epidemiological Study On Risk Factors And Risk Assessment Model Of Breast Cancer In Chinese Women

BackgroundBreast cancer is one of the most common malignancies in women, accounting for a quarter of new cases and 15% of deaths caused by cancer in female. Being the leading cause of cancer death, breast cancer threatens the life and health of female and cause enormous impact on the economy, society, families and psychologic health of women. In 2012,1,677,000 women were newly diagnosed with breast cancer, representing 25.2% of all newly diagnosed cancers according to Cancer Statistics,2014. Among these new cases, more than 18,7000 were from China and this number increased at twice the speed of gloabal growth rate. The incidence of breast cancer in China is lower than global rate, but recently it has been rising sharply. According to the latest China Cancer Registration Report, breast cancer incidence rate among Chinese women were 42.55/100,000 in 2012. Although this was a low incidence rate compared with the whole world, its growth rate has been twice the velocity of global since 1990s, especially in urban areas. Currently, breast cancer is the most frequent cancer in Chinese women and the sixth leading cause of death from cancer in Chinese women.It has been proved to be an effective way to carry out disease prevention by exploring risk factors. As to breast cancer, screening and monitoring of high-risk population with risk assessment tools based on established risk factors has been indicated as an effective way. With numbers of studies on the risk factors of breast cancer have been conducted, knowledge of the risk factors of breast cancer goes depth and physiological basic study, clinical features and treatment of breast cancer have made great progress. However, breast cancer associated risk factors vary with races and regions, that is the reason why results from foreign studies can’t accurately predict the risk level of breast cancer in Chinese women. And most studies which evaluate the risk factors of breast cancer carried in China are based on small samples in every single region, so these results can’t be applied to predict incidence risk of breast cancer in our country. Therefore, it’s very necessary to conduct studies about risk factors of breast cancer among Chinese women in larger samples and wider areas. Amounts of studies have shown that obesity is a systemic inflammatory disease. Not only as the storage room, fat tissues also bear endocrine function. And in the studies concerning relationship between obesity and malignancies, adipokines are considered to be the molecular connections. Adiponectin (ADPN), an important adipokine, is believed to be a key factor of carcinogenesis by differentiation induction of fat cells and anti-angiogenesis effects. Subsequently many investigators conduct many case-control studies to assess the connections between adiponection and risk of breast cancer, but discordance exists among these studies.The prevention pattern based on the evaluation of risk of breast cancer had decreased the incidence of breast cancer in western countries. This reminds us that the pattern of analysing and assessing risk factors based on the study of them, then screening the population at high risk and giving them effective preventive intervention, followed by continuous monitoring is an effective and reasonable way to prevent breast cancer. Series of foreign studies have been carried out to explore risk assessment tools of breast cancer, and several risk prediction models have been established. However, few domestic studies have been done by now and risk assessment tools were deficient.It is necessary to conduct large-scale population based studies to gather more data into building, testing and appling breast cancer risk models which suit the incidence characteristic among women in our country. Identifying accurate risk factors of breast cancer, building risk assessment models and giving specific preventive intervention according to the assessment will significantly affect the prevention work of breast cancer.This study aims to explore the environmental risk factors which can influence the incidence of breast cancer, so that we can present preventive measures and suggestions, and we can estalish an assessment model to evaluate the risk level of breast cancer in Chinese women, eventually provide evidence to establish strategies to prevent and control breast cancer.Objective:1. Investigate environmental factors in association with breast carcinogenesis in China, and analyse the interaction effects between these factors, by which etiological studies and establishment of risk assessment tools could be carried out.2. Explore the relationship between adiponection and breast cancer risk.3. Establish a premilinary model for evaluation of risk of breast cancer, and provide examples for community prevention of breast cancer.Methods1. A hospital-based case-control study was conducted in 22 three class hospitals across 11 provinces and cities to explore the risk factors in relation to incidence of breast cancer in female, and analyse the distribution rule of these risk factors in order to provide information, prevention measures for breast cancer etiological studies.2. Perform a meta-analysis to comprehensive search for articles related to relationship between circulation adiponection and breast cancer, Fixed or random effect pooled measure was selected on the basis of the results of homogeneity test.12 was used to evaluate the heterogeneity among studies. Meta regression and subgroup analysis were used to explore potential sources of between-study heterogeneity. An analysis of influence was carried out, which describes how robust the pooled estimator is to removal of individual studies. Publication bias was estimated using funnel plot and Egger’s test.3. Exam the total adiponectin and high-molecular-weight (HMW) adiponectin with the enzyme-linked immunosorbent assay (ELISA), analyse the connections between plasma total adiponectin, HMW adiponectin and risk of breast cancer in Chinese women. Explore if adiponectin can become a molecular marker included in the risk model of breast cancer.4. On the basis of screening out main risk factors for breast cancer, build a model with the help of Waikato Environment for Knowledge Analysis (WEKA), which integrated many machine learning methods.80 methods from 7 types of classifier (Bayesian classifier、tree、rule、function、lazy classifier、meta algorithms、Miscellaneous Classifiers) were built.First, using several evaluation metrics (accuracy rate, true positive rate, false-positive rate, precision rate, recall rate, F-measure, ROC area and the composite score) to select candidate statistical pattern recognition models.Results1. Baseline Characteristics:1613 pairs of 1:1 matched cases and controls were recruited.1489 pairs were eligible to enrollment and 124 pairs were excluded after logical check, in which 16 pairs with of benign diseases in case group,46 pairs with malignant diseases in control group,10 pairs with inclusion of non-Han ethnic,7 pairs with no-matched age,13 pairs with repeatedly enrollment,22 pairs with relapse diseases,18 pairs with incomplete information.(1) Pathological characteristics of cases:Among 1489 breast cancer patients, there are 1128 cases with invasive ductal carcinoma which account for 75.8% of whole population,127 cases with intraductal carcinoma which account for 8.5%,24 cases with invasive lobular carcinoma which accounts for 1.6%, and 194 cases with other type of cancer (including mucinous breast carcinoma, neuroendocrine carcinoma, comedocarcinoma and medullary carcinoma) which account for 10.7%. Luminal A type, luminal B type, HER-2 type and triple negative type account for 10.7% (n=159), 49.9% (n=743),8% (n=119) and 8.5% (n=126) of all.322 cases were estrogen receptor negative diseases which account for 21.63%,1018 estrogen positive cases which account for 68.37%,149 cases lost estrogen receptor status which account for 10.01%.417 cases were progesterone receptor negative which account for 28.01%, 951 cases were progesterone receptor positive which account for 63.09% and 121 cases lost progesterone receptor status which account for 8.13%.(2) Demographic characteristic:Among 1489 cases,92 cases (6.2%) aged between 25 to 34 years,451 cases (30.3%) aged between 35-44 years,588 cases (39.5%) aged between 45-54 years,315 cases (21.2%) aged between 55-64 years and 43 cases (2.9%) aged over 65 years. In our study patients aged over 45 years account for 63.5% of all cases, and no significant difference between two groups were observed (x2= 5.172, P= 0.222). The proportions of primary school and lower, junior high school, senior high school, college and postgraduate degree were 19.4%,33.8%,30.9%, 15.3% and 0.6%. As to education status, significant differences between two groups were obsereved (%2= 65.333, P<0.001). Occupation distribution of the cases was farmer (68.3%), worker (15.4%), teacher (1.7%), cicil servant (2.4%), businessman (2.4%), company employee (2.4%), housewife (5.7%), health-care workers (1.7%).(3) Analysis of Influencing Factors:In the univariate analysis,15 variates with significant differences were identified (p<0.05), including area (rural or urban), highest education background, family income, economic status, social status, body mass index, waist-to hip ratio(WHR), family history of breast cancer, history of hypertension, dairy products, fruits and vegetables, sleep, life satisfaction at present, awareness of relevant knowledge and behavior to prevent score. After multivariate conditional logistic regression analysis, including area (OR= 1.269,95% CI= 0.984 1.638), and economic status (OR=1.237,95% CI=1.019-1.501), WHR (OR=1.329, 95% CI=0.983-1.797), menopause (OR=1982,95% CI=1.360-2.888), family history of breast cancer (OR=2.418,95% CI= 1.361-4.294), the current life satisfaction (OR=1.852,95% CI=1.436-2.390), satisfaction of sleep (OR-1.412, 95% CI=1.410-1.749), dairy products (OR=0.813,95% CI=0.716-0.923), relevant knowledge score(OR=0.685,95% CI=0.517 0.907), behavior prevention score (OR = 0.675,95% CI=0.520 0.876).(4) Adiponectin level in plasma:In case group level of circulation total adiponectin was 6.353±3.551μg/ml and HMW adiponectin level was 2.518±1.885.In control group, total adiponectin was 6.563±3.721μg/ml and HMW adiponectin was 2.583±1.877μg/ml.No statistical significant difference of total adiponectin or HMW adiponectin was found between case group and control group (p=0.164,0.401).2. Meta-analysis:According to the standard of screening and exclusion,17 articles referring to relations between circulation adiponectin level and breast cancer were included in our meta-analysis.(1)For the pooled OR analysis, a total of 3,578 cases and 4,363 controls were included for the meta-analysis performed on OR. No relation was found between high level of adiponectin and incidence of breast cancer (OR=0.902,95% CI=0.773-1.053). After exclusion of studies which had great influence on heterogeneity, we found that high level of adiponectin can decrease risk of breast cancer, the pooled OR was 0.838 (0.744-0.943), the same result was found in postmenopausal breast cancer women (OR=0.752,95%CI=0.604-0.936) while no relation was found in premenopausal breast cancer women (OR=0.895,95%CI=0.638-1.256). No significant influence and publication bias were observed before and after sensitivity analysis, Egger’s test did not found significant publication bias.(2) For pooled SMD meta-analysis,1407 cases of breast cancer and 1428 controls were included. After exclusion of studies which had great influence on heterogeneity, results showed that high level of adiponectin can decrease risk of breast cancer (SMD=-0.348,95% CI=-0.533- -0.614). No significant influence and publication bias were observed before and after sensitivity analysis, Egger’s test did not found significant publication bias.3. Evaluation model for risk of breast cancer and its assessment:We used the models built with 5 kind of statistic pattern recognition which include logistic regression, SVM (support vector machine), Random Forest, Bayesian network and PCA-LDA. For the model built by SVM, discriminant accuracy within the group was 90.2%, for 1000 pairs of training sample set, it was 89.8%. For 489 pairs of samples, Extrapolation forecast accuracy was 91.7%. concordance correlotlon coefficient according to testing sample set was 0.82,5 fold cross-validation accuracy was 72.3%. After cross validation and prediction of the result, models built by SVM was superior to the others, and its Discriminant accuracy within the group was stable. There was little difference of discriminant effect among the other four models.Conclusions1. Prevention comes first, and prevention based on adjustable risk factors is an effective way. According to our study, family history of breast cancer, waist-to-hip ratio, sleep satisfaction, life satisfaction and dairy production are associated environmental factors for incidence of breast cancer. It is necessary to carry out prospective studies as to those adjustable fators such as psychological intervention, sleep regulation, health guidance and exercise.2. According to meta-analysis, high level of total adiponectin is related to lowere breast cancer risk and always a protective factor of breast cancer among postmenopausal women. While according to case-control study, high level of HMW adiponectin is a protective factor of breast cancer among Chinese women.3. Risk evaluation models of breast cancer built in our study can be used in the screening of population at high risk of breast cancer. However, as data source was limited, our models need more verification and improvement before they can be popularized and applied.