In this article, the authors analysed potential sources of bias in 149 randomly selected biomedical research papers published between 2015 and 2017. The study is a follow-up of a previous analysis of 441 biomedical articles by John Ioannidis and colleagues for the period between 2000 and 2014. In both studies, similar reproducibility and transparency indicators were evaluated.
In comparison to the first evaluation, the authors found that more studies provide information on funding (69%) and conflicts of interest (61%). Substantial progress has also been made in data sharing. In contrast to papers published between 2000 and 2014, where raw data were not directly available, one out of five papers published in 2015-2017 include a data sharing statement. However, no improvement was found regarding the availability of full protocols. Minor progress was identified regarding the number of replication studies (i.e. 5% compared to 10 % of articles containing replication studies). In addition, the authors pointed out that there is little, if any, information on conflicts of interest or data sharing on PubMed.
In summary, although the analysis revealed some improvement regarding raw data sharing as well as funding and conflicts of interest disclosure, there is still a need for improving reproducibility and transparency practices in the biomedical literature.