建構ShinyNMA網絡統合分析及視覺化互動雲端工具

黃韋銘；劉人瑋

網絡統合分析 ； R ； ShinyNMA ； Network Meta-analysis ； R ； ShinyNMA

台灣實證醫學學會會刊

14卷2期（2023 / 12 / 01）

64 - 76

繁體中文;英文

背景：網絡統合分析（network meta-analysis, NMA）是一種讓研究者可以評估或合併直接與間接證據的統計分析方法，這項分析方法可以針對多重介入進行分析，因此相較於傳統配對統合分析方法，可以提供更多的資訊。目前，許多機構，如考科藍組織（Cochrane Collaboration）已經將NMA用為整合證據的重要方法之一，採用這種分析方法的相關研究也迅速增加。然而，進行NMA需要編寫程式語言的基礎知識，對於沒有相關經驗的使用者，並不容易執行。目的：建置一個雲端運算之網路統合分析工具，讓使用者上傳數據，不需安裝統計軟體也不需編寫程式語言，即可取得網路統合分析結果。方法：作者開發的線上互動網站是以R（版本4.1.0）及其套件shiny、metaphor（版本2.4-0）、netmeta（版本1.3-0）、BUGSnet（版本1.0-4）製作。為了使每個使用者能夠盡可能得知並盡可能推廣此網站，作者上傳ShinyNMA至shinyapps.io，無需任何網站開發語言例如HTML，使用者也無需安裝任何軟體就能使用操作。可用性（usability）以系統可用性量表（System Usability Scale, SUS）測試。結果：共發放174份網路問卷，共回收34份（回收率23.13%），經過計算後平均為73.3分（標準差16.9分），相當於B級，整體滿意程度為60.1%。次組分析發現，有沒有網絡統合分析經驗與SUS分數有顯著相關，有分析經驗使用者的SUS平均分數顯著高於無相關經驗使用者。結論與應用：本研究以R結合shiny等套件，開發以免編寫程式語言的網絡統合分析雲端工具ShinyNMA，在可用性評估中屬於使用者體驗良好，有網絡統合分析經驗，使用ShinyNMA的體驗會優於無相關經驗使用者。

Background: Network meta-analysis (NMA) is a statistical method that enables researchers to evaluate or combine direct and indirect evidence. It allows for the analysis of multiple interventions, thus offering more information compared to traditional pairwise meta-analysis. Today, many institutions, such as the Cochrane Collaboration, have adopted NMA as one of their critical evidence-synthesis methods, with the number of related studies rapidly increasing. However, conducting an NMA requires foundational knowledge in programming languages, making it challenging for inexperienced users. Objective: To build a cloud-based network meta-analysis tool that enables users to upload data and obtain results of NMA without installing statistical software or writing programming code. Methods: The authors developed an interactive online website using R (version 4.1.0) and its packages shiny, metaphor (version 2.4-0), netmeta (version 1.3-0), and BUGSnet (version 1.0-4). To ensure maximum outreach and usage, the authors uploaded ShinyNMA to shinyapps. io, requiring no web development language such as HTML, or software installation by users. Usability was tested using the System Usability Scale (SUS). Results: A total of 174 online questionnaires were distributed, of which 34 were returned (return rate 23.13%). The mean SUS score was 73.3 (SD of 16.9), equivalent to a B grade, and the overall satisfaction level was 60.1%. Subgroup analysis found a significant correlation between the experience in network meta-analysis and SUS scores. Users with prior experience had a significantly higher average SUS score than those without. Conclusion and Applications: This study utilized R and the shiny package to develop ShinyNMA, a programming-free cloud tool for network meta-analysis. The usability assessment classified it as having a good user experience. Users with prior experience in network meta-analysis were found to have a better experience using ShinyNMA than those without.