請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/659
標題: | 評估大腸鏡轉介對於大腸直腸癌效益之統計方法 Statistical methods to evaluate the efficacy of referral colonoscopy on colorectal cancer |
作者: | Kuan-Chih Chen 陳冠至 |
指導教授: | 陳秀熙(Hsiu-Hsi Chen) |
關鍵字: | 大腸直腸癌,大腸鏡,傾向分數分析,閥值卜瓦松模型,閥值寇斯多相模型,不治療意向分析,遵從性調整, Colorectal cancer,Colonoscopy,Propensity score analysis,Hurdle Poisson model,Hurdle Coxian phase-type model,Intention-to-untreat estimate,Compliance adjustment, |
出版年 : | 2019 |
學位: | 碩士 |
摘要: | 背景
雖然許多研究已經證實大腸鏡檢查能有效降低大腸直腸癌的死亡及晚期癌的發生,但此效益僅能及於糞便潛血檢查陽性個案中後續接受大腸鏡確診者。先前的研究已顯示糞便潛血濃度(fecal hemoglobin concentration)為大腸直腸癌之替代指標,因此陽性個案若未接受轉介將會造成大腸直腸癌死亡或進展為晚期癌之風險,而這些危害,若個案能接受大腸鏡轉介卻可達到有效預防之效益。雖然不轉介之個案對於大腸鏡效益具有重要的影響,但此一情境在倫理考量以及實際執行上並不適合運用傳統的隨機分派試驗進行評估。因此必須透過一系列的包含傾向分數方法(propensity score-based approach)及閥值(hurdle-based approach)模型之統計方法來評估未接受轉介的行為對於大腸直腸癌死亡及晚期癌發生的影響。此外,不治療意向分析(intention-to-untreat)亦需要用來調整自我選擇偏差並量化未轉介之有害的影響,而遵從性調整(compliance adjustment)則用來評估當所有陽性個案皆接受大腸鏡確診時,其所增加的效益。 目的 本研究欲發展一系列的統計方法,來評估在糞便潛血檢查篩檢計畫下,轉介所帶來的影響。 材料與方法 資料來源 基隆社區整合式篩檢計畫 基隆社區整合式篩檢資料從1999年至2010年,共計有2591為陽性個案,並追蹤至2016年底。本研究利用傾向分數調整分析及傾向分數配對分析探討接受大腸鏡轉介與否其後續死於大腸癌或發生晚期癌之風險。 台灣全國組織性大腸癌篩檢計畫 台灣於2004年起開始提供兩年一次糞便潛血檢查之全國組織性大腸癌篩檢,目標族群為50歲至69歲之民眾,並在2013年上修至74歲。篩檢結果為陽性者將轉介並接受以大腸鏡為最主要之診斷工具加以確診。在2004年至2009年,計有49830位陽性個案,本研究追蹤此族群至2014年底,收集研究個案之大腸癌的發生及死亡。 統計方法 本研究運用傾向分數調整分析及傾向分數配對分析以評估大腸鏡確診對於後續大腸直腸癌死亡或晚期癌發生的影響。本研究並運用閥值卜瓦松模型(Hurdle Poisson model)及閥值寇斯多相模型(Hurdle Coxian phase-type model)以釐清未轉介之傾向及等待大腸鏡確診時間之過程的對於大腸鏡效益之影響。前述運用於評估轉介對於大腸鏡影響之效益統計方法並與不治療意向分析之結果進行比較。利用遵從性調整分法可評估當轉介率為100%的情境時,大腸鏡確診的效益為何。 結果 利用傾向分數配對分析於基隆資料之結果顯示,未接受轉介對於大腸直腸癌死亡的風險增加108%,而晚期癌則為61%,此結果相較傳統的寇斯回歸模型來得高。 利用閥值寇斯多相模型並結合傾向分數配對分析於台灣全國資料之結果顯示,未接受轉介對於大腸直腸癌死亡的風險增加52%,而晚期癌則為28%。而等待時間較長的人,其後續發生晚期癌的風險達統計顯著,但對於大腸直腸癌死亡並無顯著差異。與傳統的寇斯回歸模型相比,將等待時間納入模型中考量,此變項可多解釋7%的死亡風險。 不治療意向分析結果顯示,未接受大腸鏡轉介對於大腸直腸癌死亡的風險增加66%。若所有陽性個案皆接受大腸鏡確診,大腸鏡檢查將可降低48%的死亡風險。 結論 本研究發展了一系列的統計方法來評估未接受確診對於後續大腸直腸癌死亡及晚期癌發生的影響,更評估當轉介率為100%的情形下,大腸鏡確診在台灣全國大腸癌篩檢之效益。 Background Although there is a body of the evidence supporting the efficacy of colonoscopy in reducing advanced-stage colorectal cancer (CRC) and its death, the benefit of FIT-based screening programme followed by colonoscopic confirmation for FIT-positive subject can only be attained for those who comply with referral. As the Faecal haemoglobin concentration (FHbC) is the surrogate for colorectal neoplasm, the non-referral behavior among FIT-positive subjects thus results in the harmful effect of CRC death and the progression of colorectal neoplasm to advanced stage that can be averted through the referral on colonoscopy. Assessing the impact of referral on colonoscopy based on the conventional randomized controlled design is obviously infeasible and unethical. A series of statistical methods are required to evaluate the impact of non-referral on the advanced-stage CRC and its death, including propensity score-based approach and the Hurdle-based approach. The intention-to-untreat method is also required to quantify the detrimental effect of non-referral on colonoscopy making allowance for self-selection referral. The compliance adjustment is also required to assess the additional benefit when noncompliance is improved with full compliance. Aims We aimed to develop a series of statistic methods to quantify the impact of referral on colonoscopy under the context of FIT-based service screening programme. Material and methods Data Sources Data on Keelung Community-based Integrated Screening Programme Data on Keelung Community-based integrated Screening programme (KCIS) spanned from 1999 to 2010 was used to explore the impact of referral on colonoscopy using CRC mortality and the risk of advanced CRC among the FIT-positive subjects (n=2,591) by using the propensity score adjustment and propensity score matching analysis. This cohort was followed till the occurrence of CRC with the stage information collected and also the occurrence of CRC death till the end of 2016. Data on Taiwanese nationwide population-based colorectal cancer screening Programme The Taiwanese organized service CRC screening program was launched in 2004 by using FIT at biennial basis. The target population was those aged between 50 and 69 years and later extended to 74 years in 2013. Screening attendants with positive FIT were referred to receive confirmatory examination using colonooscopy as a major tool. A total of 49,830 subjects with positive FIT during the period of 2004-2009 were enrolled. This cohort was followed to identify the occurrence of CRC and the stage information and CRC death by comparing the national cancer registry and national death registry till the end of 2014. Statistical Methods The propensity score approach including the adjustment and matching was applied to quantify to impact of referral on colonoscopy in terms of CRC mortality and the risk of advanced CRC. To elucidate the mechanism of non-referral including the propensity of being non-complaint and the waiting process including the waiting time and the tendency of being long-waiting for colonoscopy, the Hurdle Poisson and Hurdle Coxian phase-type model, respectively were applied. The estimated results based a series of statistical methods were compared with that derived from the principle of intention-to-untreat analysis. By using the non-compliance adjustment approach, the efficacy of colonoscopy under the scenario of 100% compliance was the assessed. Results The detrimental effect of non-referral results in the increase in the risk of CRC mortality and advanced CRC by 108% (95% CI: 32-223%) and 61%(95 % CI: 8-144%), respectively, by using the propensity score matching approach based on the data on KCIS cohort. The magnitude regarding the impact of non-referral derived by using the conventional method was lower compared with that derived by using the propensity score matching approach. Regarding the impact of non-referral at the nationwide scale, those who were non-referral were at increased risk for CRC mortality by 61%. The impact of non-referral behavior results in the increase in the risk of CRC mortality and advanced CRC by 52% (95% CI: 16-96%) and 28% (95% CI: 9-49%) by using the Hurdle Coxian-phase type in conjunction with the propensity score matching analysis. While the tendency of being the type of long-waiting showed a significant impact on the risk of advanced CRC (aHR: 4.02, 95% CI: 1.71-9.46), the detrimental impact for CRC mortality was not statistically significant (aHR: 1.74, 95% CI: 0.24-12.65). Compared with conventional Cox regression model, the consideration of transition from low to long waiting time waiting explained the additional 7% of risk of CRC mortality. The intention-to-untreat estimate gave the results on the impact of non-referral for colonoscopy in the increase of CRC mortality risk by 66% (95% CI: 46-87%). Given the scenario that all FIT-positive subjects compliant with referral, the efficacy of colonoscopy will reduce the risk of CRC mortality by 48% (95% CI: 39-56%), indicating 13% extra benefit given full compliance. Conclusions We developed a series of statistical methods to assess the impact of non-referral on CRC mortality and advanced CRC and further evaluated the efficacy of colonoscopy with 100% compliance under the context of Taiwan nationwide colorectal cancer screening programme. |
URI: | http://tdr.lib.ntu.edu.tw/handle/123456789/659 |
DOI: | 10.6342/NTU201903592 |
全文授權: | 同意授權(全球公開) |
顯示於系所單位: | 流行病學與預防醫學研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-108-1.pdf | 6.41 MB | Adobe PDF | 檢視/開啟 |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。