利用健康資料庫探索疾病的發展歷程：以大腸直腸癌全人口病例對照研究為例
No Thumbnail Available
Date
2019
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
研究目的：使用健康資料庫作為研究工具已成為一種趨勢，且具有很大的潛力。近年來，大腸直腸癌一直是主要的健康問題，其疾病的發展歷程尚未清楚。因此，本文利用健康資料庫探索與大腸直腸癌相關的過往疾病以及發展歷程。
研究方法：本研究使用台灣全民健康保險研究資料庫以及重大傷病證明明細檔的資料庫進行病例對照研究。研究納入2010年1月1日至2013年12月31日期間31,380位年齡在50歲或以上患有大腸直腸癌的患者，再根據性別，年齡，居住地區和保險費進行配對，納入31,380位非癌症病患為對照組。研究設定病患的第一次大腸直腸癌的診斷日期為研究基準日，篩選研究基準日前的過去疾病。除了分析全部疾病外，另進行病例組人數大於對照組人數的分析。病例組與對照組的人口統計學特徵使用卡方檢定或t檢定進行分析。相關疾病首先使用單變數條件邏吉斯回歸模型進行分析，並採用偽發現率(False discovery rate, FDR)的方法調整p值以避免誤判。再使用逐步篩選法進行分析，找到具有顯著相關的疾病。最後，對研究基準日4年以前的相關疾病進行路徑分析，以瞭解研究基準日4年前的疾病與大腸直腸癌之間的關係網絡。
研究結果：研究基準日4年前有8種疾病與大腸直腸癌有顯著相關，包括糖尿病、高血壓、痔瘡、腹部和骨盆症狀、創傷、腸道疾病、胃腸道出血和慢性肝病。根據路徑分析，糖尿病是大腸直腸癌疾病發生歷程的源頭節點。
結論：本研究使用健康資料庫並提出一個新的程序進行分析，所得的結果與前人研究相符。此外，我們提供了一個更全面的結果來進行解釋大腸直腸癌的過往疾病。 以大腸直腸癌為例，本研究證實了所使用的分析程序可展示大腸直腸癌的可能疾病病因，這一分析程序可能也用於其他癌症的研究，特別是發病機制未明確的疾病。
Objective: The use of health databases as medical research tool has become a trend, which has great potential for clinical research. Colorectal cancer has become a major health issue concern, and its development pathway is not fully understood. We provided a new study procedure to explore the prior diseases and developmental pathway of colorectal cancer by using health big data. Methods: Data were retrieved from Taiwan’s National Health Insurance Research Database (NHIRD) and Registry for Catastrophic Illness Patients Database (RCIPD). We conducted a case‒control study by using a population‒based medical database. A total of 31,380 patients with colorectal cancer, diagnosed from January 1, 2010, to December 31, 2013, aged 50 years or older, were included and matched with 31,380 controls using four variables: sex, age, residence, and insurance premium. Index date was set as the date of first colorectal cancer diagnosis. We explored prior diseases 1, 2, 3, ..., 9 year before the index date. Chi‒square and t test were used to examine the demographic characteristics difference between colorectal cancer patients and controls. Conditional logistic regression model was used to screen prior diseases through a univariate analysis, and the false discovery rate method was employed to avoid false positive. Then, prior diseases associated with colorectal cancer were further analyzed by a multivariate conditional logistic regression model with stepwise selection. Finally, path analysis was conducted to reveal developmental pathway of colorectal cancer. Results: We found 8 prior diseases or diagnoses were associated with colorectal cancer 4 years before the index date, namely diabetes mellitus, hypertension, hemorrhoids, symptoms involving the abdomen and pelvis, open wounds, disorders of the intestine, gastrointestinal hemorrhage, and chronic liver disease. According to the path analysis, diabetes served as an original node on the developmental pathway for colorectal cancer. Conclusions: Most of the associations between prior diseases and colorectal cancer identified in this study, which are confirmed previous studies’ findings. Besides, we provided more comprehensive results to explain the prior diseases of colorectal cancer. This new analysis procedure explains the possible disease’ etiology, which has confirmed by an example of colorectal cancer, and may be used in the other cancer, especially ones whose etiologies are incompletely understood.
Objective: The use of health databases as medical research tool has become a trend, which has great potential for clinical research. Colorectal cancer has become a major health issue concern, and its development pathway is not fully understood. We provided a new study procedure to explore the prior diseases and developmental pathway of colorectal cancer by using health big data. Methods: Data were retrieved from Taiwan’s National Health Insurance Research Database (NHIRD) and Registry for Catastrophic Illness Patients Database (RCIPD). We conducted a case‒control study by using a population‒based medical database. A total of 31,380 patients with colorectal cancer, diagnosed from January 1, 2010, to December 31, 2013, aged 50 years or older, were included and matched with 31,380 controls using four variables: sex, age, residence, and insurance premium. Index date was set as the date of first colorectal cancer diagnosis. We explored prior diseases 1, 2, 3, ..., 9 year before the index date. Chi‒square and t test were used to examine the demographic characteristics difference between colorectal cancer patients and controls. Conditional logistic regression model was used to screen prior diseases through a univariate analysis, and the false discovery rate method was employed to avoid false positive. Then, prior diseases associated with colorectal cancer were further analyzed by a multivariate conditional logistic regression model with stepwise selection. Finally, path analysis was conducted to reveal developmental pathway of colorectal cancer. Results: We found 8 prior diseases or diagnoses were associated with colorectal cancer 4 years before the index date, namely diabetes mellitus, hypertension, hemorrhoids, symptoms involving the abdomen and pelvis, open wounds, disorders of the intestine, gastrointestinal hemorrhage, and chronic liver disease. According to the path analysis, diabetes served as an original node on the developmental pathway for colorectal cancer. Conclusions: Most of the associations between prior diseases and colorectal cancer identified in this study, which are confirmed previous studies’ findings. Besides, we provided more comprehensive results to explain the prior diseases of colorectal cancer. This new analysis procedure explains the possible disease’ etiology, which has confirmed by an example of colorectal cancer, and may be used in the other cancer, especially ones whose etiologies are incompletely understood.
Description
Keywords
大腸直腸癌, 病例對照研究, 全民健康保險資料庫, 路徑分析, 共病症, 邏吉斯迴歸, colorectal cancer, case‒control study, National Health Insurance Database, path analysis, comorbidities, logistic regression