請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/77571
標題: | 單純病例研究方法應用在以疫情監測資料估計流感疫苗保護力 Case-Only Method for Estimating Influenza Vaccine Efficacy with Surveillance Data |
作者: | 張佩儒 Pei-Ju Chang |
指導教授: | 方啟泰 |
關鍵字: | 單純病例方法,勝算比,基因與環境獨立,隨機分派試驗,流感疫苗保護力,主效應,交互作用, case-only method,odds ratio,gene-environment independence,randomized clinical trial,influenza vaccine efficacy,main effect,interaction, |
出版年 : | 2018 |
學位: | 碩士 |
摘要: | 前言:不同病例間的比較在傳染病實驗室監測過程中自然而然的發生。在傳染病領域的研究中早在二十世紀末就被提出了case-case comparison的概念,常見的方法有screening methods以及Broome method,但是這兩種方法在估計疫苗保護力上有所局限。我們亟需一個通用的單純病例研究方法應用在以疫情監測資料估計疫苗保護力。
方法:我們用羅吉斯迴歸來呈現出單純病例研究中勝算比的闡釋,並且發現單純病例研究中勝算比為暴露組與對照組風險比值的比率。本篇論文希望把單純病例研究法應用在非隨機臨床試驗的資料上,在羅吉斯迴歸模型中設置抵銷項對暴露風險的比例進行偏移調整,並且模型中風險的估計允許加入重要干擾因子的交互作用項作為主效應的修飾,但此干擾因子必須與環境暴露獨立。在傳染病流行病學中,一般常見病例對照的研究方法,此方法耗費大量資源與時間,因此我們針對疫情監測資料提出一種單純病例研究法來評估流感疫苗效力。 實例:我們進行蒙地卡羅模擬 (Monte Carlo Simulation) 流感疫情監測資料以及臺灣感染克雷伯氏肺炎桿菌(Klebsiella pneumoniae, KP)之追蹤資料。本研究單純病例研究法在流感疫情監測模擬資料中所估計出來的疫苗保護力與全人口模擬資料的點估計值有幾乎相同的結果,並且能夠有效的調整干擾因子交互作用的修飾。另外,探討臺灣糖尿病與感染克雷伯氏肺炎桿菌(Klebsiella pneumoniae, KP)關係的模擬追蹤資料中,抵銷項根據干擾因子分層設置不同的環境暴露比例值,其估計結果與預設值近乎相同。 結論:本篇單純病例研究補足了先前case-case comparison方法的限制,適用於罕見疾病,且干擾因子必須與環境暴露獨立,例如流感疫苗保護力的監測資料,並且建立在環境暴露盛行率外部資料可取得的情況,如果滿足所要求的假設條件,單純病例研究法和完整的世代追蹤研究幾乎有相同的估計值,在這樣的資料設置下,單純病例研究法的準確度是很高且有效的。而且在面臨受試者收案困難及龐大的成本下,可以減少受試者的數量,並且減少大量的時間與金錢成本的支出。 Introduction: Case-case comparison studies of infectious diseases often naturally arise in laboratory-based surveillance setting. In the field of infectious diseases, the concept of case-case comparison was proposed as early as the end of the twentieth century. The common methods are screening methods and Broome method, but these two methods have limitations in estimating vaccine efficacy. We need a common and easy case-only study method to estimate vaccine efficacy with surveillance data. Method: We use a logistic regression model to show the interpretation of odds ratio in case-only study is the ratio of risk ratio. We hope to apply the case-only method to data from non-randomized clinical trials. In the logistic regression model, offsets are set to adjust the proportion of exposure risk, and the model allows to adjust the interaction between confounders and exposures. The interaction acts as a modification of the main effect, but the confounding factor must be independent of environmental exposure. In the epidemiology of infectious diseases, case-control methods are commonly used. This method consumes a lot of resources and time. To estimate influenza vaccine efficacy, experts are used to gather cases and controls to analyze vaccine efficacy. Therefore, we present a case-only method to estimate influenza vaccine efficacy with surveillance data. Simulation: We conducted Monte Carlo Simulation to generate influenza surveillance data and cohort data of Klebsiella pneumoniae (KP) infection in Taiwan. We fit the case-only logistic model considering three types of influenza separately, and we compare the results to those of total population. For each position, the estimates of VE comparing VE in population are all numerically close. Another example, simulated cohort data of the relationship between diabetes mellitus and Klebsiella pneumoniae (KP) infection in Taiwan, the offset was stratified according to the confounding factor to set different environmental exposure ratios, and their estimated results are nearly the same. Conclusion: The niche for the case-only method is cohort data with assumptions that are gene-environment independence and rare disease, e.g. influenza vaccine efficacy we presented here. For such settings the case-only method is appealing for its statistical efficiency and for minimizing the number of subjects from whom the requisite expensive costs are measured. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/77571 |
DOI: | 10.6342/NTU201802527 |
全文授權: | 未授權 |
顯示於系所單位: | 流行病學與預防醫學研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-106-2.pdf 目前未授權公開取用 | 988.04 kB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。