The Application Of Correspondence Analysis And Trend Surface Analysis In Malignant Tumor Death Situation

Objective: Understand the death situation of malignant tumor in some county of Shandong Province, and discuss the relations between different areas and 10 malignant tumor mortality rate and discover the high incidence area of tumor and the low;study the geographical distribution of mainly malignant tumor mortality , accordingly discover the cause of cancer, provide a theoretical basis of prediction, prevention, treatment of malignant tumors ; research the correct application of correspondence analysis and trend surface analysis in the concrete practice, and provide reference in technology for the same research.Methods: The datas are from the third investigation of death carried out by the department of Clinical Oncology Institute and the county Health Bureau in Shandong Province during 2000 and 2002.Three methods involving factor analysis, correspondence analysis and the trend surface analysis are used to find the relation between different areas and different malignant tumors by correspondence analysis, as well as the geographical trend of malignant tumors by trend surface analysis. The whole process of statistic analysis were performed by SAS 8.2 and SPSS 13.0.Results: 1. Results of factor analysis: 10 kinds of cancer (factor) can be expressed by four comprehensive factors, these cumulative contribution rate is 74.54%. There are common etiological factors between different tumors.Thus, by factor analysis ,10 malignant tumors can be divided into two major groups: one is gastric cancer, esophageal cancer, colon cancer, leukemia, breast cancer ,nasopharyngeal cancer and anothor is lung, liver, bladder and cervical cancer. Through the correlation matrix, we can find there is a strong correlation between the tumors which belongs to the same group.so we can divide 10 tumors into 2 groups for correspondence analysis,so as to the full expression of initial data.2. Results of correspondence analysis: There is a strong aggregation between gastric cancer (X₁), esophageal (X₂) and colon cancer (X₅) thorugh the first correspondence analysis . All the township (except i)in this county are concerned with the three kinds of cancer, and the mortality of this three kinds of malignant tumors is high throughout the county .At the same time ,suggest that these three kinds of cancer may have certain common etiological factor.Leukemia (X₃) and breast cancer (X₄) have a greater load in the first factor, while nasopharyngeal carcinoma (X₆) has a larger negative load in the first factor,suggesting there may be a common etiological factor in leukemia and breast cancer which may be a protective factor in nasopharyngeal carcinoma.High incidence region of leukemia and breast cancer is located in the county's northern township (especialy f,g ) where nasopharyngeal carcinoma is a low. Nasopharyngeal Carcinoma has a close relationship with township i, suggesting it may be a high incidence area of nasopharyngeal carcinoma, but a low of leukemia, breast cancer and stomach cancer, esophageal cancer, colon cancer.Through the second correspondence analysis, lung cancer (X₇) and hepatocellular carcinoma (X₈) have a relation in the second factor, suggesting there maybe a common etiological factor between the two kinds of cancer, and the county's majority area are relatively close to the two kinds of cancer. At the same time lung cancer (X₇) and bladder cancer (X₉) have a strong correlation in the first common factors, suggesting the two kinds of cancer may also have a common pathogenic factors. The northwestern area of the county may be a high incidence region of lung cancer and bladder cancer.3. Results of trend surface analysis: gastric cancer and nasopharyngeal cancer mortality have obvious geographical distribution regularity. Among them, gastric cancer mortality in the county is gradually increased from west to east, while the NPC is gradually decreased from west to east. The goodness of fit of gastric cancer and nasopharyngeal cancer's trend equations are 39.5% and 37.7%, admitting the trend equation can indicate 39% and 37.7% actual variation, also suggesting that the two kinds of cancer's mortality has certain geographical distribution regularity and the geographic and environment factors may play an important role in malignant tumors incidence. In addition, in some areas was significantly higher residual value (or less), showing a particularity, which also shows a basical concordence in the results with correspondence analysis which can find the areas of high incidence and the lows.Conclusion: By analyzing the residents' cancer mortality factor information offerred by a county in Shandong Province we can see there is a common etiological factor among different cancer. 10 kinds of cancer can be divided into two major groups through factor analysis. Through correspondence analysis we gained the corresponding relationship between malignant tumors mortality and different regions, from which we found the high incidence area and low incidence area. Furtherly, we can conduct a etiology study in those area to find the protective factors. By trend surface analysis we can find gastric cancer and carcinoma have a geographical distribution regularity. In these areas there may be some kinds of protection or risk factors at the same time to find abnormal points (region), with a view to further understanding the occurrence of cancer, incidence and death situation and their geographical distribution of malignant tumors, in order to providing a scientific basis for the prediction, prevention, treatment.Wo can use sub- correspondence analysis to solve the question when the first 2 factors' cumulative percentage are lower for the full expression of initial data. we can retrieve the shortage of correspondence analysis and trend surface analysis by multielement stepwise regression which can provide reference for the same kind of research.