Studying Biostatistics Data in Original Articles Published by the Journal of Biomedical and Laboratory Sciences

生物醫學暨檢驗科學雜誌原著論文的生物統計學資料之研究

背景：生物醫學暨檢驗科學雜誌（生檢雜誌）是為台灣醫檢師發行的專屬期刊，至今已出版35卷，而且論文中經常出現生物統計學資料。那些生物統計學資料的品質適當性如何？從來沒有人做過研究。方法：我們詳細閱讀最近10年生檢雜誌刊登的141篇原著論文，探討這些論文是否錯用統計學方法，以及數據分析是否有缺失的情形，並以Cochran-Armitage檢定和Chi-square檢定統計我們的研究結果。結果：（1）141篇論文中有90篇（63.8 %）使用統計學方法；（2）使用的統計學方法共有29種，排名前5名的是：Student’s t-test（27篇論文使用）、the Chi-square test（20篇）、sensitivity（specificity、predict value等）（18篇）、analysis of variance（ANOVA）（14篇）和Pearson’s correlation（9篇）;（3）90篇論文中有18篇（20.0%）被發現錯用統計學方法: Student’s t-test（11篇）、ANOVA（4篇）、Pearson’s correlation（2篇）和the Chi-square test（1篇）;（4）除了錯用統計學方法外，數據分析有缺失的有20篇（22.2%）: Student’s t-test（11篇）、Chi-square test（3篇）、the Mann-Whitney U test（2篇）、trend analysis（2篇）paired t-test（1篇）和Pearson’s correlation（1篇）;（5）刊登於2014-2018年和刊登於2019-2023年的論文之間，錯用統計學方法和數據分析有缺失的情形，都不具統計學差異（p值分別為0.796和0.614）。結論：生檢雜誌的生物統計學資料其品質適當性參差不齊。因此，吾等建議，是不是該期刊編輯委員會可制訂有效的策略，來改善這情況。

Background: The Journal of Biomedical and Laboratory Sciences (JBLS), exclusively for medical technologists in Taiwan, has published 35 volumes focusing on biostatistics data. However, the quality appropriateness of the data remains unexplored. Methods: We reviewed 141 original articles published by the JBLS from 2014 to 2023, assessing misused biostatistics methods and analytical defects with the Cochran-Armitage test and the Chi-square test. Results: Biostatistics data were featured in 90 (63.8%) of 141 identified articles. Among the 90 articles, incorrect biostatics methods namely Student’s t-test (11 articles), analysis of variance (4 articles), Pearson’s correlation (2 articles), and the Chi-square test (1 article) were used in 18 (20.0%). Analytical defects were identified in 20 (22.2%) of the 90 articles, with defects in Student’s t-test (11 articles), the Chi-square test (3 articles), the Mann-Whitney U test (2 articles), trend analysis (2 articles), paired t-test (1 article) and Pearson’s correlation (1 article). For both the misused biostatistics methods and analytical defects studied, no differences were observed in articles published in 2014-2018 and 2019-2023 [p = 0.796 and 0.614, respectively]. Conclusion: The quality of Some biostatistics data in the JBLS is suboptimal are inappropriate. We suggest that the editorial board of the JBLS design a strategy to improve data quality appropriateness.