blog

This is a call for researchers working on pain behaviours in rodents to publish an independent re- analysis of the blinded data in Morland et al. (2015) and to update their article following an unblinded analysis.

One of the best safeguards against experimenter bias muddying results is to make sure the researcher doesn’t know which individuals belong to what group. Although this is a standard component of clinical trial design, blinding is much rarer in preclinical animal studies: a report by the NC3Rs in 2010 found that 86% of the 271 articles involving animal research analysed did not report any attempt to blind the experiment. Blinding should be just as important in preclinical research as in clinical because it helps protect ethical as well as scientific integrity; more unbiased reporting may also help reduce the animals used in research as results are likely to be more valid.

Researchers using animals also have an extra responsibility for minimizing unnecessary duplication of animal experiments by sharing all of their results and data. This usually means making the full dataset available, including group information, so others can access all information needed for reuse or reanalysis. However, full data release may occasionally undermine attempts to reanalyse work in a blinded fashion.

Andrew Rice (Imperial College London) and colleagues have published an article that reconciles blinding with data sharing, so others can attempt to replicate the findings. Their video files of locomotor behaviours in rats following bladder inflammation are directly accessible from the article, but group information is not; the only way to differentiate each animal is by their ID. As explained in the article, the ‘unblinding’ lists, which contain the group codes, are only available upon request from F1000Research. As researchers in this field likely have prior expectations as to how rats typically behave following inflammation, splitting up the data in this way makes it easier for them to remain unbiased, whilst still ensuring that the data is accessible.

Call for participants

We want to showcase an example of an article that reports a blind re-analysis that is subsequently updated after unblinding; this would allow others to compare the pre- and post-unblinding analyses, and prove that the initial analysis was indeed conducted blind. This is the idea:

1. Participating researchers download the blinded video data files to further explore them using behavioural analytical software.
2. Once complete, the authors submit a paper composed of: Introduction, Methods, and Data.
3. Upon submission F1000Research will provide the unblinding lists, so that the authors can conduct the final statistical analyses to assess differences between groups. In the meantime the initial submission will be published and reviewers invited.
4. Once the unblinded analysis is complete the authors submit their Results and Discussion, which will be published along with the previous sections in a new version of the article. Reviewers will be re-invited.

Although we encourage researchers to participate in this reproducibility initiative, every request for the unblinding lists will be honoured, no matter the reason or intended venue for publication.

If you are interested, please email us at research@f100.com or share this post in case any of your friends and colleagues might be.