應用資料探勘技術與統計方法於併用化療藥物與標靶藥物之藥物交互作用不良反應預測

Application of Data Mining and Statistical Approaches to Adverse Drug-Drug Interaction Prediction in Chemo Drug and Target Drug

林永欣(Yung-Hsin Lin)

指導教授 ： 謝銘鈞

國立臺灣大學/工學院/醫學工程學研究所/碩士(2012年)

https://doi.org/10.6342/NTU.2012.10566

摘要

惡性腫瘤，又名癌症，根據世界衛生組織的報導[1]，已經成為全球人口的主要死因，在2008年間因癌症死亡的人數高達760萬人次，占全球死亡人口總數的13%[2]。過去在治療癌症的選擇上主要包括外科手術、化學治療及放射線治療，而標靶藥物的出現已成為許多癌症治療的另一個選擇。標靶藥物(Target Drug)，具有專一性，只作用在腫瘤生長相關的「標靶接受體(Ligand)」以抑制腫瘤。由於癌症細胞才會大量表現這些專一性標靶接受體，因此相較於傳統化療藥物(Chemo Drug)，使用標靶藥物進行治療被認為對正常細胞的副作用較小。然而根據2011年的文獻[3]指出，發生致命不良反應(Fatal Adverse Reaction) 於使用標靶藥物-癌思停(Avastin)的病患，機率為2.9％；相較於單獨只使用傳統化療藥物的患者，接受化療藥物合併使用癌思停的患者，發生致命性不良反應的相對風險卻增加1.33倍。只是閱讀大量文獻回顧的研究方法曠日廢時，因此近年來的資料探勘[4–6]技術提供了一個快速找尋藥物不良反應的方法。但基於單一藥物不良反應幾乎已受到各期臨床研究所發掘，因此藥物交互作用不良反應將是本論文嘗試要研究的重點。我們使用卡方測試法(Chi-square Test)與先演演算法(Apriori Algorithm)嘗試尋找標靶藥物與化療藥物間可能引發的藥物交互作用不良反應(Drug-Drug Interaction)。由於標靶藥物通常用於化療病患，因此我們從Chemocare.com網站[7]中挑出現有的化療藥物對應美國藥物不良反應資料庫[8]後，篩選出使用化療藥物的病例，並進一步挑選使用標靶藥物-癌思停(Avastin)的不良反應記錄，嘗試在這群紀錄中找出標靶藥物-癌思停與其他併用的化療藥物下，可能產生的藥物交互作用不良反應。然而因標靶藥物使用在癌症病患非常常見，並且有部分通報資料可能來自於癌症第四期(Phase IV)的臨床研究(這些病人的癌細胞已經轉移至全身，存活率極低)，死亡(Death)是這些癌症化療病患極易產生的不良反應，因此我們藉由過去探討致命性不良反應的4篇文獻[3], [9–11]中整理48種(對應回美國藥物不良反應資料庫中癌症病例，實際存在的致命性不良反應為23種) 致命性不良反應[12]，試圖找出因併用標靶藥物與化療藥物後所可能產生的致命性藥物交互作用不良反應，為醫學研究人員在研究相關議題上，提供一個快速且有效的篩選方法。

關鍵字

藥物不良反應 ； 資料探勘 ； 化療 ； 標靶藥物 ； 癌思停 ； 致命性不良反應

並列摘要

Malignant tumors, also known as cancer, according to the World Health Organization[1] announced that cancer is a worldwide major cause of death, accounting for 7.6 million deaths in 2008, around 13% of all deaths[2]. The choice of the treatment of cancer including surgery, chemotherapy and radiation therapy, the emergence of drug targets, has become a cancer treatment is another option. Target drug is a specific chemical role in tumor growth and “target receptor (ligand)” to suppress tumors. The “target receptor”, including tumor growth related receptor (ligand), genes, or messaging pathway, and tumor angiogenesis factor as well. Due to target drug action on special location compared to traditional chemical drugs on normal cells. Target drug is believed to improve chemotherapy efficacy while reducing side effects.
However, according to the Journal of the American Medical Association (JAMA) in 2011 article[3] found that the use of Bevacizumab (Avastin) patients get fatal adverse reactions (FAEs) probability is 2.9%. Compared to the patients using of chemo drugs alone, the cancer patients receiving chemo drug and target drug Avastin, the relative risk of getting fatal adverse reactions is increased by 1.33 times.
In addition, the research methodology of reviewing past journal articles is time-costing. In recent years, data mining techniques to provide a quick look for adverse drug reactions [10–12]. But almost of the single adverse drug reactions had been discovered by clinical researchers. Consequently, our ambition of this study is to find the suspicious adverse drug-drug interactions under the usage of target drug and chemo drug.
We use the Chi-square test and Apriori Algorithm to find the suspicious adverse drug-drug interactions under the usage of target drug and chemo drug. Because target drugs are usually used on patients receiving chemotherapy, we single out the existing chemotherapy drugs (including target drugs and chemo drugs) from the Chemocare.com website[7] corresponds to the Adverse Events Reporting System (AERS) database[8], screen out the cancer patients receiving chemotherapy drugs. Furthermore, we select the usage record of target drug - Avastin from cancer patients as our mother group, try to identify the suspicious adverse drug-drug interactions under the usage of Avastin and other chemo drugs.
However, target drugs are usually used in cancer patients, and there are some experimental data may come from the Phase IV of clinical research (the patient's cancer cells had metastasized throughout the body, the survival rate is very low), death is the adverse reaction prone to be produce in cancer patients with cancer chemotherapy.
Therefore, we look up four of previous studies [3], [9–11] about fatal adverse events, totally screen out 48 kinds of fatal adverse reactions[12] (mapping to the cancer patients in AERS database, the actual existence of a fatal adverse reactions are 23), try to find out the suspicious fatal adverse drug-drug interactions in the cancer patients receiving target drug and chemo drug, to provide a efficiency way and effective screening method to filter out the suspicious fatal adverse drug-drug interactions for researchers study the related issues.

並列關鍵字

Adverse drug reaction ； chemotherapy ； chemo drug ； target drug ； data mining ； Bevacizumab (Avastin) ； fatal adverse events ； drug-drug interaction

參考文獻

[3] V. Ranpura, S. Hapani, and S. Wu, “Treatment-related mortality with bevacizumab in cancer patients: a meta-analysis,” JAMA, vol. 305, no. 5, pp. 487–494, Feb. 2011.

[4] M. Hauben, D. Madigan, C. M. Gerrits, L. Walsh, and E. P. Van Puijenbroek, “The role of data mining in pharmacovigilance,” Expert Opin Drug Saf, vol. 4, no. 5, pp. 929–948, Sep. 2005.

[5] A. M. Wilson, L. Thabane, and A. Holbrook, “Application of data mining techniques in pharmacovigilance,” Br J Clin Pharmacol, vol. 57, no. 2, pp. 127–134, Feb. 2004.

[6] J. Almenoff, J. M. Tonning, A. L. Gould, A. Szarfman, M. Hauben, R. Ouellet-Hellstrom, R. Ball, K. Hornbuckle, L. Walsh, C. Yee, S. T. Sacks, N. Yuen, V. Patadia, M. Blum, M. Johnston, C. Gerrits, H. Seifert, and K. Lacroix, “Perspectives on the use of data mining in pharmaco-vigilance,” Drug Saf, vol. 28, no. 11, pp. 981–1007, 2005.

[7] C. H. Bedell, “A changing paradigm for cancer treatment: the advent of new oral chemotherapy agents,” Clin J Oncol Nurs, vol. 7, no. 6 Suppl, pp. 5–9, Dec. 2003.