- 學位論文
男性飲酒與多階段大腸腫瘤性狀探討：統計流行病學導向機器學習分析
The Effect of Male Alcohol Drinking on the Multi-state Colorectal Neoplasia: Statistic and Epidemiological Oriented Machine Learning Analysis
若您是本文的作者，可授權文章由華藝線上圖書館中協助推廣。
摘要
背景：飲用酒精後，身體當中產生的乙醛、氧化壓力、DNA甲基化的影響等與致癌機轉相關，故飲酒行為與大腸直腸瘜肉及大腸直腸癌可能存在著因果關係。然而，飲酒對於大腸直腸瘜肉及大腸直腸癌風險的影響程度仍未有定論，此外過往台灣未有充分相關文獻資料。本研究主要乃針對台灣男性飲酒行為作用與糞便潛血陽性、大腸直腸瘜肉、大腸直腸癌之風險進行研究與評估。 材料與方法：本研究以前瞻性世代設計納入台灣三個地區(基隆、彰化、台南)共計80339位40歲以上男性資料，包含19102(23.8%)名飲酒者，5377(6.7%)名戒酒者，與55355(68.9%)名未曾飲酒者，收案及追蹤的時間為2001-2009。本研究運用機器學習隨機森林(Random Forest)及貝氏網絡(Bayesian Network)等方式預測糞便潛血陽性、大腸直腸瘜肉、大腸直腸癌之風險並作模式評估，亦以Random Forest來評估影響糞便潛血陽性、大腸直腸瘜肉、大腸直腸癌、及大腸直腸瘜肉對於大腸直腸癌之變數重要性，並以貝氏深度學習進行因果關係評估。本研究亦運用單變項與多變項羅吉斯迴歸，以勝算比評估糞便潛血陽性、大腸直腸瘜肉、大腸直腸癌之風險。運用邏吉斯迴歸分析方法對飲酒者及戒酒者相較於未曾飲酒者之糞便潛血陽性勝算比、大腸直腸瘜肉勝算比、大腸直腸癌勝算比進行評估，並針對年齡、BMI 30以上與否、大腸直腸癌家族史、運動習慣、飲食習慣(平時食用蔬果數量、食用肉品數量)、抽菸、代謝症候群等干擾因子進行校正。本研究亦運用寇斯迴歸分析評估不同飲酒行為間大腸直腸瘜肉、大腸直腸癌、大腸直腸瘜肉加總大腸直腸癌之存活分析的死亡風險對比值。 結果：針對飲酒在不同階段大腸直腸癌癌化過程之作用，經機器學習統計分析發現飲酒非作用於腺癌，非直接與腺癌相關，但飲酒可能作用於腺腫，直接影響腺腫發生或經由高糞便潛血再影響腺腫，其次透過腺腫至腺癌，而飲酒也作用於高糞便潛血，與高糞便潛血發生有相關。相對於未曾飲酒者，較多的飲酒量、較長的飲酒時間、較短的戒酒時間會有較高的糞便潛血陽性風險(Trend test: P < 0.001)與大腸直腸瘜肉風險(Trend test: P < 0.001)，而大腸直腸癌之風險微幅上升但未達統計顯著。此外，相對於未曾飲酒者，較多的飲酒量、較長的飲酒時間，皆會有較高的大腸直腸瘜肉存活分析之死亡風險(Trend test: P < 0.001)，而大腸直腸癌之死亡風險僅微幅上升，除了每週飲酒8杯以上之死亡風險明顯上升(aHR: 1.51, 95%CI: 1.02-2.22)，其餘未達統計顯著。在機器學習方面，透過Random Forest得以檢視影響糞便潛血、大腸直腸瘜肉及大腸直腸癌相關因子之重要性排序，結果發現飲酒為僅次年齡的重要影響因子。 結論：本研究運用機器學習與迴歸方法釐清飲酒在不同大腸直腸癌癌化階段的可能作用，確立其與高糞便潛血值、腺腫發生有關，而非與腺癌直接相關。本研究亦證實了台灣男性飲酒行為與糞便潛血陽性及大腸直腸瘜肉存在著劑量反應效應，較大的飲酒量或較長的飲酒時間將產生較高的風險，並且大量飲酒也會造成大腸直腸癌之死亡風險上升。而飲酒乃是影響糞便潛血、大腸直腸瘜肉及大腸直腸癌之重要相關因子。
並列摘要
Background: According to current evidence, there may be a causal link between alcohol consumption and the development of colorectal neoplasia. Acetaldehyde, alcohol-induced oxidative stress and modulated DNA methylation are the possible mechanisms that contribute to how alcohol consumption leads to cancer. However, the impact of alcohol consumption on the risk of colorectal neoplasia remains elusive. Besides, few Taiwanese studies have investigated the influence of alcohol drinking on the multi-state colorectal neoplasia. This study aimed to examine the association between alcohol intake and fecal immunochemical test, colorectal neoplasia, as well as colorectal cancer in the male population of Taiwan. Materials and Methods: Total of 80339 male participants aged 40 or older from three regions of Taiwan (Keelung, Changhua, and Tainan) were enrolled in this prospective cohort study, including 19102(23.8%) current drinkers, 5377(6.7%) former drinkers, and 55355(68.9%) non-drinkers. The follow-up period was nine years (during 2001 to 2009). Machine learning was applied while variable importance checked using a Random Forest model for the studies of three phenotypes including fecal immunochemical test, colorectal neoplasia, and colorectal cancer. Random Forest and Bayesian Network were applied for the risks of three phenotypes. Causal influence for three phenotypes was assessed by Bayesian Network Deeping Learning. The risks of colorectal neoplasia and colorectal cancer were estimated by adjusted odds ratios (aORs) using an univariable and multivariable logistic regression model making allowance for a constellation of confounding factors including age, BMI, family history, tobacco consumption, physical activity, consumption of vegetables and meats, metabolic syndrome, and fecal immunochemical test. Cox Regression was applied for the survival of colorectal polyp and colorectal cancer. Results: It has been found that alcohol does not have a direct effect on colorectal cancer. However, alcohol may have an effect on adenomas, directly influencing their occurrence or through high fecal occult blood leading to the development of adenomas. Furthermore, alcohol also affects high fecal occult blood, which is associated with its occurrence. Compared with non-drinkers, risks of positive fecal immunochemical test and colorectal neoplasia increased (Trend test: P < 0.001) as to larger amount or longer periods of drinking as well as shorter periods after quitting alcohol. Risk of colorectal cancer was not significant in such manner. Mortality risk of colorectal neoplasia increased as to larger amount or longer periods of drinking (Trend test: P < 0.001), which was not noticed regarding to the mortality risk of colorectal cancer except heavy drinking (For >= 8 drinks / week, aHR: 1.51, 95%CI: 1.02-2.22). Machine learning was applied using a Random Forest model for variable importance for the studies of fecal immunochemical test, colorectal neoplasia, and colorectal cancer. Alcohol drinking was found to be the second most important factor while the most important one was age. Conclusions: The study has revealed the possible mechanism of the occurrences of high fecal hemoglobin and adenoma induced by alcohol. We also found the dose-response effect of alcohol drinking on the risk of positive FIT and colorectal neoplasia in Taiwanese male population as the effects size increased with large quantity or longer durations of alcohol consumption. Besides, alcohol consumption was found to be one of the most important factors for positive FIT, colorectal neoplasia, and colorectal cancer.
參考文獻
Kyujin Lee, Yong Hwan Kim Colorectal Polyp Prevalence According to Alcohol Consumption, Smoking and Obesity Int. J. Environ. Res. Public Health 2020, 17, 2387
Song-Yi Park, Lynne R. Wilkens, Veronica Wendy Setiawan, Kristine R. Monroe, Christopher A. Haiman, and Loïc Le Marchand Alcohol Intake and Colorectal Cancer Risk in the Multiethnic Cohort Study Am J Epidemiol. 2019;188(1):67–76
I-Chen Wu, Chien-Hung Lee, Chao-Hung Kuo, Fu-Chen Kuo, Deng-Chyang Wu, Ying-Chin Ko, Fang-Jung Yu Consumption of Cigarettes but not Betel Quid or Alcohol Increases Colorectal Cancer Risk J Formos Med Assoc 2009 Vol 108 No 2
Sarah McNabb, Tabitha A. Harrison, Demetrius Albanes, Sonja I. Berndt, Hermann Brenner, Bette J. Caan, Peter T. Campbell, Yin Cao, Jenny Chang-Claude, Andrew Chan, Zhengyi Chen, Dallas R. English, Graham G. Giles, Edward L. Giovannucci, Phyllis J. Goodman, Richard B. Hayes, Michael Hoffmeister, Eric J. Jacobs, Amit D. Joshi, Susanna C. Larsson, Loïc Le Marchand, Li Li, Yi Lin, Satu Männistö, Roger L. Milne, Hongmei Nan, Christina C. Newton, Shuji Ogino, Patrick S. Parfrey, Paneen S. Petersen, John D. Potter, Robert E. Schoen, Martha L. Slattery, Yu-Ru Su, Catherine M. Tangen, Thomas C. Tucker, Stephanie J. Weinstein, Emily White, Alicja Wolk, Michael O. Woods, Amanda I. Phipps, and Ulrike Peters Meta-analysis of 16 studies of the association of alcohol with colorectal cancer Int. J. Cancer: 146, 861–873 (2020)
A Pedersen, C Johansen, M Grønbæk Relations between amount and type of alcohol and colon and rectal cancer in a Danish population based cohort study Gut 2003;52:861–867