Biostatistical Analysis on Recent Issues : A Sampling study

Abstract

This paper aims at examining the statistical methodology used in the research articles published in the journals of high repute. Conducting the correct statistical analysis is of paramount importance in every research, and hence, statistical methodology has often been checked predominantly in journals. In this paper, the original research articles published in the top 10 leading medical journals published in India, and indexed by PubMed in 2003 and a decade later in 2013 have been reviewed. The present article examined—study design types, frequencies of statistical methods, errors/defects in study design and statistical analyses, and also the implementation of CONSORT checklist in RCT’s (randomized clinical trials). Chi-square and Fisher’s exact tests, (wherever appropriate) were used to compare any proportional differences in use of statistical methods and various errors/defects during the 10 years in the study period. All the tests used in this article were conducted at 5% level of significance and were 2-tailed. Chi-square tests were conducted at one degree of freedom. Using arbitrary p thresholds (like p<0.01) instead of reporting exact p-values, reporting p value without test statistics, insufficient (or inappropriate) description of methods and p values without confidence intervals were most common of the errors present. In conclusion, this study indicates that Indian medical research seems to have made some progress regarding study design defects, but there remains ample room for improvement regarding statistical analyses. Most published studies continue to use a retrospective clinical design, with randomized clinical trials being quite rare. Those RCTs that are published often have serious methodological problems, including absence of sample size estimation and power calculations, as well as failure in (or in the reporting of) randomization.