隨著癌症的議題越來越被重視，探討癌症風險的世代研究也逐漸地增加，因此對於進行流行病學研究的效率也應該要被關注。至今還未曾有一個同時討論一種疾病與多種癌症關聯性的分析系統。本論文旨在建立一個自動化分析系統，以台灣人口為基底來分析疾病及各癌症之間的關聯性。我們建立一個網頁來引導研究員循序漸進的填寫欲分析的資料，完成後系統會檢查其完整性並將查詢資料傳送到系統核心。接著將查詢條件轉換成相對應的SAS程式碼，藉由SAS統計軟體進行資料分析。分析結束後系統會將輸出的結果轉換成常見的統計圖表並呈現於網頁上。建立完分析系統後，我們選了七篇刊登在不同期刊上的論文來驗證我們的分析結果，我們主要分成三個部分來比較結果，其中包含病人的基本資料表、疾病及所有癌症的關聯性和疾病及特定癌症部位的關聯性。雖然我們的分析結果跟參考文獻上的結果有些差異，但相似的部分還是占大多數的。本論文完成一個能自動化分析疾病及癌症關聯性的系統，且經過驗證後其結果顯示我們的系統是可信的，也能產生出和期刊上癌症研究的論文相似的結果。;Introduction: With the issue of cancer is being taken seriously, more and more cohort studies are conducted to investigate the risk of developing cancer. Therefore, the efficiency of research on the epidemiologic studies ought to be concerned. Currently, none of analysis systems have considered the correlations between a disease and all cancer types simultaneously. This study aims at implementing an automatic analysis system to analyze correlations between a disease and some of site-specific cancers in a large Taiwanese cohort. Methods: We build a website to instruct researchers to make a query step by step. After each step is performed, the query is check whether it’s complete by the system and sent to the system core to be transformed into corresponding SAS codes. Then, our system keeps executing the SAS file by the statistical software until the end of analysis. Finally, the system core converts the SAS output files into result tables and displays them on the website. Results: In order to verify our results by journals, we choose seven studies from different journals based on NHIRD or LHID. After that, we divided the results into three partitions to compare with them, including baseline of characteristics among patients with main disease, correlation between main disease and all cancers and correlation between main disease and specific cancer types. Although during verifying, there are some difference between the analytical results and the previous works, the similarities between them are in majority. Conclusions: In this thesis, an automated analysis system is implemented to analyze correlation between a disease and some of site-specific cancers. The results verified by the journals illustrate that our system is reliable and can provide similar results in comparison with journals in cancer-related cohort study.
