利用次級資料庫評估藥物罕見副作用風險之研究-fluoroquinolones之不良反應為例

Using secondary database to evaluate rare drug risk–Adverse reaction of fluoroquinolones

周煦文(Hsu-Wen Chou)

指導教授 ： 賴美淑

國立臺灣大學/公共衛生學院/流行病學與預防醫學研究所/博士(2014年)

https://doi.org/10.6342/NTU.2014.11259

摘要

研究背景與目的：
為偵測藥物罕見副作用風險，隨機臨床對照試驗有其限制，因此如何適切的應用次級資料庫及多組傾向性分數，同時控制為數眾多之干擾因子，且考量多組治療間之比較，實為重要的議題。研究文獻雖業已指出fluoroquinolone類藥物具引起嚴重甚至致命的血糖異常及心律不整的可能性，然而過去研究多受限於研究樣本數不足，結果未有一致性，亦無針對fluoroquinolone類藥物間不同藥品進行相對風險比較的研究。
因此，本研究以台灣的全民健康保險大型資料庫，利用多組傾向性分數分析fluoroquinolone與嚴重心室心律不整及血糖不良反應之風險。

研究材料與方法：
本研究資料來源為全民健康保險資料庫及死亡登記檔的世代研究，選取研究期間內於門診接受口服研究抗生素的新使用者做為研究對象。以接受macrolides治療之病患做為血糖不良反應之參考對照組；接受amoxicillin/clavulanic acid治療之病患做為心室心律不整及心血管死亡反應之參考對照組；另選取第二代cephalosporins做為負控制組。
利用多組傾向性分數，以多變量邏輯式迴歸模型進行校正，而得各組抗生素對研究事件之粗勝算比和校正勝算比(aOR)及其 95% 信賴區間(confidence interval, CI)。再根據基本特性與重要共病進行分層分析，並以不同追蹤區間進行敏感度分析，分別以7天及14天之不同追蹤區間進行額外分析。

研究結果：
在心室心律不整研究中，以多組傾向性分數校正後之結果顯示，與amoxicillin-clavulanate類藥物相較，moxifloxacin及azithromycin顯著增加嚴重心室心律不整(aORs (95% CI), 分別為2.62 (1.78-3.86)；2.88(2.05-4.04))及心血管死亡之風險(2.23 (1.60-3.11)；1.98(1.43-2.75))，clarithromycin及ciprofloxacin組則並無觀察到有顯著風險增加。
同類藥物間之相互比較分析顯示，azithromycin對嚴重心室心律不整及心血管死亡之風險皆顯著高於clarithromycin，而moxifloxacin及levofloxacin亦顯著高於ciprofloxacin。在敏感性分析部分，分析結果顯示不論心室心律不整或心血管死亡，azithromycin、moxifloxacin及levofloxacin皆為於服藥後7天內風險最高，風險皆隨追蹤時間增加而遞減。
於樣本數影響力之模擬校正分析，研究結果顯示，在事件數較多時，分析結果較為穩健，故多組傾向性分數校正、pairwise配對及減少樣本數之抽樣校正分析結果和誤差皆相近。然而於發生率小於萬分之一之罕見事件時， pairwise配對及抽樣校正分析會損失大量對照參考組之事件數，導致統計檢定力下降，降低研究結果之精確(precision)程度，亦增加對風險估計之標準誤。
在嚴重血糖不良反應研究中，以多組傾向性分數校正後之hyperglycemia adjusted ORs (95% CI)分別為moxifloxacin 2.48 (1.50-4.12)、levofloxacin 1.75(1.12-2.73)、ciprofloxacin 1.87 (1.20-2.93)；而hypoglycemia adjusted ORs (95% CI)分別為moxifloxacin 2.13 (1.44-3.14)、levofloxacin1.79 (1.33-2.42)、ciprofloxacin 1.45 (1.07-2.00)，cephalosporins組則並無觀察到有顯著風險增加。Fluoroquinolones類藥物間的相互比較分析中，和ciprofloxacin相較，moxifloxacin顯著增加hypoglycemia之風險。
在同時合併使用胰島素或sulfonylurea之病患中，moxifloxacin皆顯著增加hypoglycemia之風險，而對接受glinides, metformin, 或thiazolidinediones治療之病患，所有fluoroquinolones皆顯著增加hyperglycemia的風險。在病患合併有慢性腎臟病，和同時接受sulfonylurea治療病患，levofloxacin皆顯著增加hypoglycemia的風險。在敏感性分析部分，結果顯示dysglycemia的效應並未隨觀察時間改變而有所不同

結論：
本研究結果顯示，適當的使用多組傾向性分數，可在同時考慮多組治療選擇下，犧牲最少統計檢定力，同時校正所有潛在的干擾因素，得到較精確之罕見事件風險估計。且於稀少事件數之下，依重要危險因子執行分層分析，以提供臨床治療選擇之重要依據。
經由多組傾向性分數校正分析，和amoxicillin-clavulanate相較，moxifloxacin及azithromycin顯著增加心室心律不整及心血管死亡風險。雖fluoroquinolones具群組效應，但在不同種類之fluoroquinolones存在程度上之差異，尤以moxifloxacin具最高的風險。
在糖尿病患中，和macrolides相比，fluoroquinolones顯著增加高血糖及低血糖之風險，且在不同種類之fluoroquinolones存在程度上之差異。在fluoroquinolones中，moxifloxacin有最高的風險造成低血糖，其次為levofloxacin及ciprofloxacin。合併有慢性腎臟病，以及正在接受insulin或sulfonylurea治療之糖尿病患，更為容易有血糖代謝不良反應。

關鍵字

多組傾向性分數校正分析 ； 藥物罕見副作用風險評估 ； 氟化奎諾酮抗生素 ； 血糖不良反應 ； 心室心律不整 ； 藥物流行病學 ； 世代研究 ； 健保資料庫

並列摘要

Background and objective:
Randomized control trial was limited to detect rare adverse drug effect. Therefore, how to appropriate using secondary database and multiple propensity score is an important issue. Several studies have indicated that a relationship exists between the use of fluoroquinolones and severe dysglycemia and arrhythmia. However, epidemiologic data regarding dysglycemia and arrhythmia risk with different fluoroquinolones were inconsistent. Therefore, to investigate the risk of severe ventricular arrhythmia and dysglycemia resulting from fluoroquinolones treatments, we performed a population-based study, using National Health Insurance database and multiple propensity score method.

Materials and Methods:
A population-based inception cohort study was conducted. Data was obtained from National Health Insurance claims data and the National Death Registry. We selected patients who had been prescribed amoxicillin/clavulanate as the reference group for severe ventricular arrhythmia risk assessment, and patients who had been prescribed macrolides as the reference group for severe dysglycemia risk assessment. Additionally, we selected patients who had been prescribed second generation cephalosporins as negative control group.
We estimated multiple propensity scores for each of the studied drugs using multinomial logistic regression by considering covariates including comorbidities, health resource utilization, concomitant medication, and indication for antibiotics.
The estimated multiple propensity scores were included in the multivariable logistic regression to estimate the crude and adjusted odds ratio (aOR) as well as the 95% confidence interval (CI) for the association between antibiotics and the risk of ventricular arrhythmia or dysglycemia. Stratified analysis was performed to evaluate potential effect modification. Furthermore, we performed sensitivity analysis for follow-up periods of 14 or 30 days to evaluate whether the effects were substantial.

Results:
The use of moxifloxacin and azithromycin showed a significant increase in the risk of both ventricular arrhythmia and cardiovascular death compared with amoxicillin-clavulanate treatments. The adjusted ORs (95% CI) for ventricular arrhythmia were 2.62 (1.78-3.86) for moxifloxacin and 2.88(2.05-4.04) for azithromycin. For cardiovascular death, the adjusted ORs (95% CI) for moxifloxacin and azithromycin was 2.23 (1.60-3.11), and 1.98(1.43-2.75) respectively. Neither clarithromycin nor ciprofloxacin were noted significant association with ventricular arrhythmia and cardiovascular death.
In sample size simulation analysis, the results were more robust with sufficient number of events. However, pairwise match and sampling adjust analysis were lose enormous events of reference group, which can lead lower statistical power, less precision estimate.
The adjusted odds ratios of levofloxacin, ciprofloxacin, and moxifloxacin compared with macrolides were 1.75 (1.12-2.73), 1.87 (1.20-2.93), 2.48 (1.50-4.12) for hyperglycemia and 1.79 (1.33-2.42), 1.46 (1.07-2.00), 2.13 (1.44-3.14) for hypoglycemia, respectively. No significant association was noted between cephalosporins and dysglycemia. Patients taking moxifloxacin faced a significantly higher risk of hypoglycemia than those receiving ciprofloxacin.
When the results were stratified according to various anti-diabetic agents, the use of moxifloxacin increased the risk of hypoglycemia significantly among patients receiving concomitant insulin or sulfonylurea treatments. All types of fluoroquinolones significantly increased risk of a hyperglycemia event among patients treated with diet therapy and those receiving glinides, metformin, or thiazolidinediones. For levofloxacin, increased risk of hypoglycemia was evident among patients suffering from chronic kidney disease and those receiving concomitant sulfonylurea treatment. We obtained similar results in the sensitivity analysis using follow up periods of 7 and 14 days.

Conclusion:
Our study indicated that using multiple propensity score can adjust potential confounding factors and to analyze multiple treatment choice simultaneously with rare events.
Moxifloxacin and azithromycin were significantly increase risk of severe ventricular arrhythmia and cardiovascular death. Diabetics using oral fluoroquinolones faced greater risk of severe dysglycemia. The risk of severe ventricular arrhythmia, cardiovascular death and hypoglycemia varied according to the type of fluoroquinolone administered, and was most commonly associated with moxifloxacin.

並列關鍵字

multiple propensity score adjust ； rare adverse drug reaction risk assessment ； fluoroquinolones ； dysglycemia ； ventricular arrhythmia ； pharmacoepidemiology ； cohort study ； National Health Insurance Research Database

參考文獻

54. Lin CC, Lai MS, Syu CY, Chang SC, Tseng FY. Accuracy of diabetes diagnosis in health insurance claims data in Taiwan. J Formos Med Assoc 2005;104(3):157-63.

1. WHO. Pharmacovigilance: Ensuring the Safe Use of Medicines 2004.

2. Freemantle N, Marston L, Walters K, Wood J, Reynolds MR, Petersen I. Making inferences on treatment effects from real world data: propensity scores, confounding by indication, and other perils for the unwary in observational research. BMJ 2013;347:f6409.

3. Cepeda MS, Boston R, Farrar JT, Strom BL. Comparison of logistic regression versus propensity score when the number of events is low and there are multiple confounders. Am J Epidemiol 2003;158(3):280-7.

4. Council for International Organizations of Medical S. Guidelines for preparing core clinical-safety information on drugs : report of CIOMS Working Group III. Geneva, Switzerland: CIOMS, 1995.