跳到主要內容

臺灣博碩士論文加值系統

(3.231.230.177) 您好!臺灣時間:2021/07/27 10:58
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

我願授權國圖
: 
twitterline
研究生:簡欣怡
研究生(外文):Shin-Yi Chien
論文名稱:癌症風險評估的小樣本統計方法之實證研究:以臺灣輻射污染建物居民追蹤研究為例
論文名稱(外文):Cancer Risk Assessment with Sparse Data:Taiwan Radiation Contaminated Buildings Cohort Study
指導教授:林逸芬
指導教授(外文):I-Feng Lin
學位類別:碩士
校院名稱:國立陽明大學
系所名稱:公共衛生研究所
學門:醫藥衛生學門
學類:公共衛生學類
論文種類:學術論文
論文出版年:2009
畢業學年度:97
語文別:中文
論文頁數:74
中文關鍵詞:Firth’s penalized likelihood method貝氏分析最小潛伏期Cox Model
外文關鍵詞:Cox ModelFirth’s penalized likelihoodbasedBayesian AnalysisMinimum Latent Period
相關次數:
  • 被引用被引用:0
  • 點閱點閱:322
  • 評分評分:
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:0
背景與目的:
自1992年在台北市爆發國內第一起輻射鋼筋暴露污染事件起,追蹤調查發現台灣自1983年起已有超過200棟建築物受到鈷六十污染的鋼筋所建築成之建物(輻射屋)。至2005年底為止,臺灣輻射屋暴露族群已有178人發生癌症,其中有128人從暴露到發生癌症的時間超過國際幅射防護委員會(ICRP)第六十號出版物建議之最小潛伏期(血癌:2年,實質癌:10年)。台灣輻射污染建物研究團隊已發表的研究顯示,臺灣輻射累積暴露劑量(Taiwan Cumulative Dose;TCD)較高者罹患血癌及乳癌的風險也有較高的傾向。前項研究以大樣本的統計方法評估罹癌風險,然而該族群在個別癌症的罹癌人數很少(血癌:6人,乳癌:17人),低劑量暴露癌症的最小潛伏期亦尚未有定論之標準。本研究目的是利用較適合小樣本的統計方法及定義不同的最小潛伏期分析臺灣輻射屋暴露族群之乳癌、血癌、及所有癌症之風險,並與已發表的研究結果作比較。

方法:
以臺灣輻射污染建物居民追蹤研究為例,比較四種Cox models的估計方法: 包括Wald-based、Partial likelihood ratio-based、Firth’s penalized likelihood method、及貝氏分析(Bayesian)等分析結果是否一致。並比較在不同最小潛伏期(minimum latent period)的定義下,分析結果是否有差異。

結果與結論:
在血癌的部份,Firth方法所估計的風險比(HR per 100mSv)為1.19(90%信賴區間: (1.05, 1.30)),Bayesian方法依不同先驗分布(prior) 風險比約在1.13至2.03;若將TCD分組分析,TCD >50mSv這組罹癌風險為≦50mSv組的2.6倍(90%信賴區間: (0.62, 9.25))。在乳癌的部份,Firth方法所估計的風險比(HR per 100mSv)為1.11(90%信賴區間: (1.00, 1.19)),TCD >50mSv這組罹癌風險為≦50mSv組的3倍(90%信賴區間: (1.34, 6.59))。這些都與前人研究一致, Firth方法所估計的HR比以partial likelihood-based或Wald-based的估計結果較為顯著。

整體來說,無論是血癌或乳癌,使用較適合小樣本的統計方法Firth’s penalized likelihood method及Bayesian analysis with normal priors時,即使設定不同最小潛伏期,累積輻射暴露劑量越高,則罹患血癌及乳癌的風險也有越高的趨勢。Firth’s的估計結果與Bayesian analysis (Normal prior)的結果相似,且信賴區間會比大樣本方法的Wald-based CI窄。Partial likelihood ratio-based所估計信賴區間與Bayesian analysis (Uniform prior)的結果較為相似。
Background:
The recent studies from the cancer risk assessment of the Taiwan radio-contaminated buildings cohort have shown a trend of increased risks in breast cancer and in leukemia as the cumulative radiation exposure increased (Hwang et al. 2008). A total of 178 cancer cases were identified in the cohort between 1983 and 2005; however, only 128 of these cases, including 6 Leukemia cases and 17 breast cancer cases, occurred after the minimum latent periods recommended by the International Commission on Radiological Protection (ICRP), which were 2 years for leukemia and 10 years for solid cancers. Large sample methods were used to estimate the relative risk in the previous studies. The objectives of this study were to evaluate the effect of the radiation exposure on cancer risk using methods more appropriate for small sample and applying different latent periods.

Methods:
The cancer risks were estimated by hazard ratio (HR) using Cox models with Wald-based and partial likelihood ratio-based methods, by the model with Firth’s penalized likelihood-based estimator, and by Bayesian methods. The relative risks assuming different latent periods were also estimated.

Results and Conclusions:
A significant radiation risk for leukemia were observed ( HR per 100mSv ) is 1.19(90%CI: (1.05, 1.30)) by Firth’s method. If the TCD were treated as categorical, the people who received TCD >50mSv have 2.6 times (90%CI: (0.62, 9.25)) risk as those received TCD≦50mSv。The hazard ratio per 100mSv for breast cancer was 1.11(90%CI: (1.00, 1.19)) by Firth’s method。If the TCD were treated as categorical, the people who received TCD >50mSv have 3 times (90%CI: (1.34, 6.59)) risk as those received TCD≦50mSv。 In general, by using Firth’s method or by Bayesian’s method with normal prior, the cancer risk were increased as TCD increased, even applying different minimum latent periods.
The results showed that the hazard ratios and their inferences by Wald-based Cox model and by Bayesian analysis with uniform prior were in general more conservative than those by partial likelihood ratio-based methods, by Firth’s method, and by Bayesian analysis with normal prior. When the latent periods were shorten, the Hazard Ratio slightly decreased. However, the results were not dramatically changed from the previous studies based on these small samples.
第一章 緒論 1
第一節 研究背景與動機 2
第二節 研究目的 3

第二章 材料與分析方法 5
第一節 研究族群與資料來源 6
第二節 統計方法與分析策略 7
一、分析變項 8
二、統計分析模式 11

第三章 分析結果 20
第一節 不同統計模式的相對癌症風險估計(Hazard ratios estimates using different models) 21
第二節 定義不同最小潛伏期的相對癌症風險估計(Hazard ratios estimates using different minimum latent periods) 26
第三節 敏感度分析之結果 29

第四章 討論與建議 33
第一節 不同潛伏期的影響 34
第二節 分析模式方法之差異 34

參考文獻 38
附表 44
附圖 55
1.HWANG, S.-L., et al., Cancer risks in a population with prolonged low dose-rate gamma-radiation exposure in radiocontaminated buildings,1983-2002. International Journal of Radiation Biology, 2006. 82(12): p. 849-858

2.Hwang, S.-L., et al., Estimates of Relative Risks for Cancers in a Population after Prolonged Low-Dose-Rate Radiation Exposure: A Follow-up Assessment from 1983 to 2005. Radiation Research, 2008. 170: p. 143-148

3.林書儀, 額外風險與相對風險模式之比較: 以台灣輻射污染建物居民追蹤研究為例 (未發表的碩士論文). 2008
4.Therneau, T.M. and P.M. Grambsch, Modeling Survival Data : Extending the Cox Model. 2001: Springer

5.Firth, D., Bias reduction of maximum likelihood estimates. Biomelrika, 1993. 80(1): p. 27-38

6.Heinze, G. and M. Schemper, A Solution to the Problem of Monotone Likelihood in Cox Regression. Biometrics, 2001. 57: p. 114-119

7.Heinze, G. and M. Schemper, A solution to the problem of separation in logistic regression. Statistics in Medicine, 2002. 21: p. 2409-2419

8.Heinze, G. and M. Ploner, SAS and SPLUS programs to perform Cox regression without convergence problems. Computer Methods and Programs in Biomedicine, 2002. 67: p. 217-223

9.Preliminary Capabilities for Bayesian Analysis in SAS/STAT Software

10.Du, P. and C. Gu, Penalized likelihood hazard estimation: Efficient approximation and Bayesian confidence intervals. Statistics & Probability Letters, 2006. 76: p. 244-254

11.Ji-Xian Wang, J.D.B., Jr., Ben-Xiao Li, Jing-Yuan Zhang, Joseph F. Fraumeni, Jr., Cancer Among Medical Diagnostic X-Ray Workers in China. Journal of the National Cancer Institute, 1988. 80: p. 344-350

12.James H. Yiin, M.K.S.-B., Sharon R. Silver, Robert D. Daniels, Gregory M. Kinnes, Dennis D. Zaebst, James R. Couch, Travis L. Kubale and Pi-Hsueh Chen, Risk of Lung Cancer and Leukemia from Exposure to Ionizing Radiation and Potential Confounders among Workers at the Portsmouth Naval Shipyard. Radiation Research, 2005. 163: p. 603-613

13.John D. Boice, J., Sarah S. Cohen, Michael T. Mumma, Elizabeth Dupree Ellis, Keith F. Eckerman, Richard W. Leggett, Bruce B. Boecker, A. Bertrand Brill and Brian E. Henderson, Mortality among Radiation Workers at Rocketdyne (Atomics International), 1948–1999. Radiation Research, 2006. 166: p. 98-115

14.E. Cardis, et al., The 15-Country Collaborative Study of Cancer Risk among Radiation Workers in the Nuclear Industry: Estimates of Radiation-Related Cancer Risks. Radiation Research, 2007. 167: p. 396-416
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top