Warfarin併用一級交互作用藥品與消化道出血之風險分析

Abstract

研究背景 消化道出血是warfarin最常見的嚴重副作用，交互作用亦提高出血的風險。目前國內並沒有完整針對warfarin併用交互作用藥品與消化道出血間關聯的研究，因此希望透過本研究提供醫療人員藥品流行病學資料，以保障用藥安全。 研究目的 利用具有全國代表性之健保資料庫，分析warfarin使用者的消化道出血發生率、併用一級交互作用藥品是否增加消化道出血風險，以及併用交互作用藥品之出血病人群基本特性。 研究方法 以1997年至2007年臺灣全民健保資料庫100萬人抽樣歸人檔為材料，進行回溯性世代研究；研究族群為1998年1月1日至2007年12月31日新使用warfarin病人，從初次warfarin開方日追蹤至研究終點（第一次消化道出血急診或住院事件）或2007年12月31日前最後一筆申報資料，收集warfarin、56種一級交互作用藥品、潛在增加消化道出血用藥紀錄（包含天數及劑量）及共病症。 計算完全未使用一級交互作用藥品（warfarin alone）及併用個別一級交互作用藥品病人群之消化道出血粗發生率，並分析併用交互作用藥品病人中出血族群特性。 利用time dependent Cox proportional hazard model進行研究終點事件分析，並納入年齡、性別、消化道出血相關用藥及共病症等變項校正。 研究結果 10年研究期間共有5251位warfarin新使用者，平均年齡為63.3歲，男性佔多數（53.8%），warfarin使用期間或處方結束日之14天內第一次消化道出血住院事件共有189人，粗發生率為每100人年2.9人年，出血部位以消化道潰瘍相關出血佔大多數共60人（31.7%），warfarin alone病人群出血發生率每100人年4.6人。在本研究定義之56種一級交互作用藥品中，共有43種（76.8%）一級交互作用藥品於研究期間與warfarin併用，其中19種在併用期間發生消化道出血事件，出血個案數大於3人之個別藥品出血發生率分為ketoprofen每100人年51.5人、naproxen 23.6.人、fenofibrate 12.6人、cilostazol 5.5人、aspirin 4.7人、celecoxib 4.1人、ginkgo biloba 3.6人、amiodarone 2.1人及clopidogrel 2.4人，而併用交互作用藥品病人群中，出血族群最常見的特性為年齡大於70歲，warfarin劑量並未依年齡調降，消化道出血相關病史較多，同時使用增加消化道出血風險的藥品。 Cox model分析結果顯示， warfarin使用者在出血前14天，若併用heparin，每增加一個DDD (defined daily dose）的heparin (10000U)，消化道出血風險（hazard ratio, HR）增為2.614倍（1.779-3.842）；併用levofloxacin，每增加一個DDD (500 mg)的levofloxacin，HR=1.232 (1.017-1.494)；併用aspirin每增加一個DDD (100 mg)的aspirin，HR=1.027 (1.010-1.044)；併用carboplatin，每增加一毫克的carboplatin，HR=1.025 (1.016-1.033)；併用ketoprofen，每增加一天併用ketoprofen，HR=1.187 (1.035-1.362)。此外，病人本身具有共病症（如鬱血性心衰竭、凝血病變、消化道潰瘍或出血、消化道發炎、食道靜脈曲張、馬魏氏症候群及洗腎）、使用非類固醇抗發炎藥、皮質固醇及利尿劑（spironolactone）皆會增加出血風險。 結論 Warfarin使用者併用43種一級交互作用藥品中，只有其中5種（11.6%）與消化道出血住院事件有顯著相關性。部分風險藥品或共病症造成消化道出血風險反而比warfarin及一級交互作用藥品高，但本研究受限於健保資料庫，無法將病人服藥順從性、warfarin與食物、中草藥之交互作用一併分析。雖然研究結果顯示併用一級交互作用藥品與消化道出血風險不如預期的高，但如果病人年紀大於70歲，使用多種增加消化道出血風險之藥品及有消化道潰瘍出血病史時，必須更積極監測及調整warfarin使用劑量，保障用藥安全。

Background Gastrointestinal (GI) bleeding was the most frequent complication of warfarin. Drug-drug interactions may increase this risk. In Taiwan, there is no comprehensive study to investigate the risk of GI bleeding associated with potentially significant warfarin drug interactions. Objective Using National Health Insurance Database of Taiwan to evaluate the incidence rate of GI bleeding in warfarin users, risk of GI bleeding associated with potentially significant warfarin drug interactions, and characteristics of warfarin users who were hospitalized due to bleeding. Methods This retrospective cohort study analyzed National Health Insurance Database with one million randomly selected individuals from January, 1997 until December, 2007. The study population (warfarin new users) started warfarin after January, 1998. They were followed from the first day of warfarin prescribing until the firtst GI bleeding complication requiring hospitalization or the last claim data befor December 2007. We collected all claim data of warfarin, 56 potentially significant interacting drugs (PIDs), risk medication and comorbidity. We calculated the crude incidence of GI bleeding events during warfarin exposure alone period (without PIDs) and periods of concomitant exposure (with PIDs), and also analyzed characteristics of patients with bleeding events. The time dependent Cox proportional hazard model was performed to determine the hazard ratio (HR) of risk factors associated with GI bleeding, adjusted with age, sex, risk medication and comorbidity. Results There were 5251 warfarin new users in the 10 year study period. Their mean age was 63.3 years old. More than half (53.8%) were male. A total of 189 patients were hospitalized for GI bleeding. The major etiology, peptic ulcer bleeding, was found in 60 patients (31.7%). The crude incidence rate of total warfarin users was 2.9 per 100 person-year (per 100 PY). During warfarin exposure alone, the incidence rate was 4.6 per 100 PY. Forty-three (76.8%) out of 56 PIDs were taken concomitantly by warfarin users. Among 43 PIDs, GI bleeding events occurred during combinationof 19 PIDs and warfarin. The incidence rate of the PIDs with more than 3 bleeding events were ketoprofen (51.5 per 100 PY), naproxen (23.6 per 100 PY), fenofibrate (12.6 per 100 PY) , cilostazol (5.5 per 100 PY), aspirin (4.7 per 100 PY) , celecoxib (4.1 per 100 PY), ginkgo biloba (3.6 per 100 PY), clopidogrel (2.4 per 100 PY) and amiodarone (2.1 per 100 PY). Patients who suffered from bleeding were usually older than 70 years old without dosing adjustment (reduction), more comorbidities prone to GI bleeding and taking risk medication. The Cox model analysis revealed that adjusted HRs of warfarin-PID for GI bleeding were 2.614 (1.779-3.842), 1.232(1.017-1.494), 1.027(1.010-1.044), 1.025 (1.016-1.033), 1.187 (1.035-1.362), for increasing 1 defined daily dose (DDD, 10000U) of heparin, increasing one DDD (500 mg) of levofloxacin, increasing one DDD (100 mg) of aspirin, increasing 1 mg of carboplatin, increasing one day of ketoprofen, respectively. In addition, patients with comorbidity (congestive heart failure, coagulopathy, peptic ulcer or bleeding, esophagitis, gastritis, duodenitis, esophageal varices, Mallory-weiss tear, renal dialysis), or using risk medication (NSAIDs, corticosteroid, spironolactone) had a higher risk of bleeding. Conclusions Warfarin users in this study used 43 PIDs and only 5 PIDs (11.6%) significantly increased the risk of GI bleeding. On the other hand, some risk medication or comorbidities caused higher risk than warfarin and PIDs. This study, however, was limited by not able to access data regarding patient adherence, drug-food or herb interactions. Although the risk of PIDs was not as high as we expected, physicians or pharmacists should pay attention to elderly (age>70 years old), patients taking risk medications or with history of peptic ulcer or bleeding. It may ensure medication safety in these patient populations by close monitoring and adjusting warfarin dosing.