2015-08-012024-05-14https://scholars.lib.ntu.edu.tw/handle/123456789/660836摘要：研究背景：病人在接受藥物治療後可能因為反應不佳或產生不良反應而停藥，而這些病人往往是高危險群。因此，如果想研究長期使用藥物的療效或風險，但在研究對象發生藥物非順從性 (non-adherence)當下停止追蹤，會導致非獨立設限 (informative censoring)。藥物流行病學研究如果不能適當的處理這個情況很可能會產生偏誤的結果。研究目的：本計劃將延續利之前的研究，利用糖尿病患全民健保資料與臨床資料，發展以貝氏模型協助個別病人醫療決策之方法，進一步考慮因藥物non-adherence所產生informative censoring問題，建立初診斷第二型糖尿病患，長期接受不同口服降血糖藥物 (metformin, sulfonylureas、glinides、alpha-glucosidase inhibitor、pioglitazone、DPP4-inhibitors) 發生嚴重心血管事件，包括心肌梗塞、缺血性腦中風、及因心衰竭住院之機率。研究方法：本研究為回溯性追蹤研究。將以2009/1/1-2010/12/31年齡20歲以上之新診斷第二型糖尿病病患為研究對象，包含的變項包括性別、年齡、罹患糖尿病時間、糖尿病併發症、共病、使用藥物、體重、血壓、及實驗室檢查結果。發展出可以同時處理藥物 non-adherence之貝氏統計模型，以進行adherence-adjusted analysis，考慮不同的先驗分佈與censoring機轉的假設進行敏感度分析，並進一步以2012/1/1-2013/12/31病患資料驗證預測模型之準確度。預期目標：發展出可處理因drug non-adherence造成之informative censoring的貝氏統計模型，建立個別第二型糖尿病病患選擇不同口服降血糖藥物心血管風險之評估方法<br> Abstract: Background: Patients who discontinue or switch to other medications shortly after drug initiation are mostly due to suboptimal response or adverse reaction, and they are also at higher risks of adverse outcomes. Therefore, studies that evaluate long-term effect associated with drug use but stop following these patients at the date of medication non-adherence will cause informative censoring. Pharmacoepidemiological researches that fail to handle this issue appropriately will lead to biased results.Study goal: A Bayesian analysis model that taking into consideration informative censoring due to drug non-adherence will be developed. Large healthcare databases will be analyzed to build a model to predict 1 year incidence of severe cardiovascular events, including acute myocardial infarction, ischemic stroke, and heart failure that leads to hospitalization, among individual type 2 diabetic patient receiving different oral anti-diabetic agents, including metformin, sulfonylureas, glinides, alpha-glucosidase inhibitor, pioglitazone, and DPP4-inhibitors.Methods: A retrospective cohort study will be conducted by using the National Health Insurance claims database and clinical datasets in several hospitals in Taiwan. Participants will be newly diagnosed type 2 diabetic patients, aged ≥ 20 years old, who initiated the above medications during 2009-2011. Potential explanatory variables will include age, sex, diabetic duration, diabetes-associated complications, comorbidities, concomitant medications, body weight, blood pressure, and laboratory results. The unobserved outcome of interest due to drug non-adherence will be treated as outcome missing at random. Sensitivity analysis that using several assumptions about prior probabilities and mechanisms of missingness will be conducted. A prediction model for individual patient's risk while receiving different oral agents will be developed. Data during 2012-2013 will be used for validating the accuracy of the proposed prediction model.Expected results: This study will develop a Bayesian analysis model that can be used to evaluate individual patient's outcome associated with long-term drug use.Keywords: Informative censoring, medication non-adherence, Bayesian methods, anti-diabetic therapy, large healthcare database