運用藥物效能的統合分析（meta-analysis）針對同一藥物研究問題來綜述各個研究療效結果所進行的定量綜合統計分析方法，而廣義線性模型（generalized linear model, 簡稱GLM）的建立是將更多的統合分析結果資料再利用，客觀去推論、了解反應變數（outcome） 與控制變數（control variable）之間相關性。以糖化血色素改變量為反應變數(outcome)，選取藥物因子、病程因子、體重因子為控制變數（control variable），反應變數母體選擇伽瑪(gamma)分配，連結線性涵數(identity function)，並以p<0.05為具統計顯著意義。本研究屬次級資料分析，由加拿大一篇以長效胰島素類似物為主題隨機控制臨床實驗的整體回顧與統合分析之文獻資料，摘錄其中11篇第2型糖尿病使用長效胰島素類似物研究報告之數據，以電腦統計軟體R程式 3.03 for Windows進行數據分析，以描述性分析呈現各變項分佈情形，納入研究對象為第2型糖尿病患，其總人數共4824人。本研究結果顯示糖尿病人的身體質量指數是影響最明顯的控制變數；初期的肥胖的第2型糖尿病病人開使用藥延長治療時間對糖化血色素值有明顯影響；提早使用合併用藥比單獨用藥對初期肥胖第2型糖尿病患者的糖化血色素是有明顯影響。 Drug effectiveness of meta-analysis for the same drug problems to reviews conducted by various research results of quantitative statistical analysis methods. The aim of this meta-analysis and generalized linear model was to evaluate the clinical efficacy of long-acting insulin analogues compared with intermediate-and long-acting unmodified human insulin and oral antidiabets drug for the treatment of type 2 diabetes mellitus (DM). Comparative studies of the long acting analogues with these approaches are required to determine which is most effective. This study is an analysis of secondary data. This data is based on a comprehensive health technology assessment available from CADTH’s. A total of 11 randomized controlled trials on type 2 DM were eligible for review . Randomized controlled trials (RCTs) comparing long-acting insulin analogues with conventional human insulin were identified through electronic databases, web sites, and industry. We will introduce generalized linear models to provide drug effectiveness more convenient statistic tools in rating plan. Outcomes of interest included glycemic control (glycosylated hemoglobine, A1C) or hypoglycemia. Control variables of included body weight, quality of life, patient satisfaction, health care utilization, and long-term cmplications of diabetes.The research report data to computer r 3.03 for Windows statistical software for data analysis, descriptive analysis of variable distribution, study into type 2 diabetes patients, the total number of 4824 people. The results of this study showed diabetes patients ' body mass index is the most obvious impact of control variables initially obese of type 2 diabetes patients. There were statistically significant with respect to HbA1C levels using the drug combination and extend the treatment time in early obesity of type 2 diabetic patients.
