Analysis of statistical problems in clinical research articles of oncology / 肿瘤研究与临床

ABSTRACT

Objective To investigate the application of statistical methods in clinical research articles of oncology in China. Methods The core journals in Chinese oncology journals were retrieved in "Wanfang Data", and the application of statistical methods in the clinical research articles published in the first issue of 2017 were analyzed,including the selection of statistical indicators,the selection of statistical methods and the interpretation of statistical results. Results A total of 176 clinical research articles from 24 journals were taken into the study.In the selection of statistical indicators, 12.6 %(16 articles) used the description method of normal distribution data for the skewness distribution measurement data; 37.9 % (47 articles) described the count data with samples<20 as a percentage; 14.5 % (9 articles) replaced the constituent percentages with composition ratios. In the selection of statistical methods, 10.3 % (12 articles) did not perform analysis of variance when comparing among multiple groups, and the t test was used to compare each other directly. 15.1 % (8 articles) disregarded the preconditions of the use of the χ 2test. In the interpretation of statistical results, 18.4 % (27 articles) confused the statistical significance with biological significance. Conclusions The statistical problems are widespread in Chinese oncology clinical research articles,involving the entire process from study design to interpretation of results. Scientific research management departments, clinical researchers, and journal editorial board need to work together to improve the scientificity of clinical scientific research results of oncology in China.