臺灣慢性骨髓性白血病病人使用酪氨酸激酶抑制劑與血管栓塞事件之相關性

Abstract

研究背景: 近期研究顯示使用新一代tyrosine kinase inhibitors (TKIs)會增加慢性骨髓性白血病(chronic myeloid leukemia, CML)病人發生血管栓塞副作用(vascular adverse events, VAEs)的風險，但目前多數觀察性研究因為人數較少、沒有排除曾有血管栓塞病史之病人等限制，無法進行穩固的統計推論。部分研究間之結果不一致，且缺少亞洲族群之相關研究。 研究目的: 本研究目的為分析臺灣CML病人，探討其血管栓塞事件發生率與一般病人族群之差異，並以incident new user design探討使用不同種類TKIs是否會影響血管栓塞副作用的發生率，再進一步使用羅吉斯迴歸分析可能存在之危險因子。另外，本研究亦使用prevalent new user design探討曾經使用過一線imatinib而後改用新一代TKI之CML病人，其血管栓塞副作用發生率之差異。 研究方法: 本研究為一回溯性世代研究，使用衛生福利部資料科學中心臺灣健保資料庫以及癌症登記檔做為資料蒐集來源。本研究納入癌症登記檔中，於2008年至2016年間新診斷為CML，且新使用TKI藥物之病人作為研究族群。本研究將分三部分進行：第一部分中，每位CML病人將依照出生年份、性別及處方年份，與五位一般病人族群做配對。接著，本研究將針對兩組間之血管栓塞事件發生率進行統計分析。本研究之第二部分及第三部分，將著重於探討使用不同種類的TKI對於日後血管栓塞副作用發生率之影響，並分別使用incident new user design及prevalent new user design之研究設計進行配對及統計分析。於incident new user design中，CML病人將依照一線所使用的TKI種類分為三組(imatinib、nilotinib或dasatinib)，接著每位nilotinib使用者及dasatinib使用者再分別以傾向分數與n位imatinib使用者做配對(n ≤ 2)。於prevalent new user design當中，所有曾經使用過nilotinib或dasatinib者，包含一線或二線使用者，會分別以先前imatinib暴露狀態與傾向分數，與imatinib使用者做配對。 於前三部分配對成功之病人則進入研究結果之分析。本研究將使用multivariable logistic regression，探討CML病人與一般病人族群對於血管栓塞事件之影響，並使用Cox proportional hazard ratio model以及Kaplan-Meier存活曲線，分析不同種類之TKI對於血管栓塞副作用發生率之影響。最後，再使用univariable及multivariable logistic regression，分析可能存在之危險因子。 研究結果: 本研究共收納1,111位CML病人，追蹤時間中位數約為2.2年。在本研究的第一部分，共有5,555位一般病人族群成功與1,111位CML病人進行配對，但在multivariable logistic regression當中，CML病人族群並不會增加血管栓塞事件的風險，反而有降低的趨勢，與一般病人族群相比，風險比為0.63 (95% CI 0.47-0.86, p < 0.01)。 本研究incident new user design中，經過傾向分數配對後，imatinib與dasatinib之發生率分別為7.4/千人年及15.3/千人年，dasatinib之風險比為1.71 (95% CI: 0.71-4.26, p = 0.23)，顯示dasatinib可能增加發生血管栓塞副作用之風險，但未達統計上的顯著差異。imatinib與nilotinib之發生率則分別為5.8/千人年與19.2/千人年，nilotinib之風險比為3.13(95% CI: 1.30 – 7.51, p = 0.01)，顯示nilotinib確實會顯著增加CML病人發生血管栓塞副作用之風險。 本研究prevalent new user design中，經過兩階段配對後，imatinib與nilotinib之粗發生率則分別為10.2/千人年及21.0/千人年，經校正後之nilotinib發生率比為0.79 (95% CI: 0.15-4.25, p = 0.78)，nilotinib prevalent new user似乎有較低的血管栓塞副作用發生風險，但並未達統計上的顯著差異。imatinib與dasatinib之粗發生率分別為8.0/千人年及15.2/千人年，經校正後之dasatinib發生率比為0.04 (95% CI: 0.00-∞, p = 0.98)，但因人數及血管栓塞事件數過少，導致信賴區間過寬。 在multivariable logistic regression當中，除了nilotinib之外，僅年齡及腦血管疾病相關病史會顯著增加發生血管栓塞副作用之風險，年齡每增加一歲之風險比為1.04 (95% CI 1.01-1.06, p<0.01)，而具有腦血管病史者之風險比為3.49 (95% CI 1.19-10.30, p=0.02)。 研究結論: CML病人族群之血管栓塞副作用發生率較一般病人族群顯著較低。但在incident new user design當中，使用新一代TKI，尤其是nilotinib，確實會顯著增加發生血管栓塞副作用的風險，dasatinib也會增加血管栓塞副作用發生的風險，但並未達顯著差異。在prevalent new user design中，nilotinib與dasatinib使用者的血管栓塞副作用粗發生率，皆較imatinib使用者高，但經校正後的發生率比及風險比則因為血管栓塞事件數過少，無法進一步進行穩固的統計推論。危險因子的部分，本研究發現除了nilotinib之外，年齡及腦血管疾病病史會顯著增加血管栓塞副作用的發生風險。

Backgrounds: Tyrosine kinase inhibitors (TKIs) have shown long-term survival benefits in chronic myeloid leukemia (CML) patients. Nevertheless, significant concern has been raised regarding the long-term TKIs associated vascular adverse events (VAEs). Objectives: The objective of this retrospective cohort study is to investigate the incidence of VAEs in Taiwanese CML patients treated with different TKIs (imatinib, nilotinib, and dasatinib). Two different study designs, the incident new user design and the prevalent new user design, were adopted. The incidence of VAEs in CML patients compared to general patient population were also analyzed. The potential risk factors were identified by conducting multivariable logistic regression. We also discussed about the incidence of VAEs in prevalent new users. Methods: We conducted a retrospective cohort study using Taiwan Cancer Registry (TCR) and National Health Insurance Research Database (NHIRD). Adult patients diagnosed with CML during 2008 to 2016 were identified from TCR. In our first analysis, each CML patients was matched to 5 general patients by birth year, sex, and prescription year. We compared incidence of VAEs between CML patients and general patients. In the second and third analysis, incidence of VAEs between different TKI users were assessed using two study design: incident new user design and prevalent new user design. In the incident new user design, CML patients were categorized into three groups according to their first-line TKI treatment (imatinib, nilotinib, and dasatinib). Nilotinib users and dasatinib users were separately matched to imatinib users by 1:n (n≤2) propensity score matching. In the prevalent new user design, patients initiated nilotinib/dasatinib, including the incident new users and the prevalent new users, were separately matched to imatinib users according to previous imatinib exposure condition and propensity score. Successfully matched patients in every parts of our study were qualified for outcome analyses. Outcome were assessed by using Cox proportional hazard model and Kaplan-Meier survival curves. Logistic regressions were used to investigate potential risk factors. Results: A total of 1,111 CML patients treated with first-line imatinib, nilotinib, and dasatinib, were identified (mean age: 48.9 years; female: 41.7%) and followed for a median of 2.2 years. In the first part of our study, the incidence of VAEs in general patient population is significantly lower than that of CML patients, with hazard ratio of 0.63 (95% CI 0.47-0.86, p < 0.01). In the incident new user design, after propensity score matching, the incidence of VAEs for dasatinib (15.3 per 1000 person-years vs. imatinib 7.4 per 1000 person-years) and nilotinib (19.2 per 1000 person-years vs. imatinib 5.8 per 1000 person-years) were significantly higher than imatinib. The hazard ratio of dasatinib and nilotinib were 1.71 (95% CI: 0.71-4.26, p = 0.23) and 3.13 (95% CI: 1.30-7.51, p = 0.01) as compared to imatinib. In multivariable logistic regression, older age (IRR: 1.04 for each year increase in age, p<0.01) and history of cerebrovascular diseases (IRR: 3.49, p=0.02) were associated with increased risk of VAEs. In the prevalent new user design, after two-stage matching, the crude incidence of VAEs for nilotinib (21.0 per 1000 person-years vs. imatinib 10.2 per 1000 person-years) were non-significantly higher than imatinib, with adjusted incidence rate ratio of 0.79 (95% CI 0.15-4.25, p = 0.78). The crude incidence of VAEs in dasatinib (15.2 per 1000 person-years vs. imatinib 8.0 per 1000 person-years) is non-significantly higher than imatinib, with adjusted incidence rate ratio of 0.04 (95% CI 0.00-∞, p = 0.78). Conclusions: The risk of VAEs in CML patients treated with TKI is significantly lower than that of general patient population. Second generation TKIs, particularly nilotinib, were associated with higher incidence rate of VAEs as compared with imatinib in incident new user design. In prevalent new user design, nilotinib and dasatinib were associated with lower incidence rate of VAEs, although the number of VAEs were so low that a robust analysis was not available. Potential risk factors of TKIs-associated VAEs include older age and history of cerebrovascular diseases.