The Reproducibility Crisis in Science Continues

Many members of the scientific community go to great lengths to defend experimenting on animals. But are such studies defendable when most aren’t even reproducible?

Scientific research is only of value when the results can be reproduced time and time again. Only then can researchers have confidence that the data they have generated is reliable. But the use of animal models is anything but reliable.

The problem of reproducibility is no secret to NAVS. In previous editions of Science First, we have reported that the rate of reproducibility of preclinical animal-based research is as low as 11%. That means as many as 9 out of 10 experiments using animals are simply not reproducible. In addition to wasting the lives of countless animals, this non-replicable research is wasting money—upwards of $28 billion every year in the U.S. alone.

A recent article in The Scientist covered this important issue, noting that “many poorly designed animal studies still seem to be falling through the cracks, accruing the associated costs without providing the potential benefits.”

This conclusion is based partly upon the work of one UK-based group, the Collaborative Approach to Meta-Analysis and Review of Animal Data from Experimental Studies (CAMARADES) team, which has tried to better understand the reproducibility problem through extensive literature searches and analyses. Emily Sena, leader of the CAMARADES group, notes that when the animal methodology sections of scientific publications are examined, “[v]ery few studies took simple measures to reduce bias.”

This is important, because studies lacking bias-reducing measures such as blinding (in which certain information is withheld from researchers so as not to influence their interpretation of results) as well as randomization of test subjects, tend to report seemingly better outcomes and larger treatment effects. The failure to include these bias-reducing measures can mask the latent defects that underlie the published data and conclusions.

Such sloppy science is unacceptable and wasteful. As Thomas Hartung of the Johns Hopkins Bloomberg School of Public Health notes, “Research that is not quality-controlled is unethical.” And that is especially true when that research involves animals.

While the article goes on to explain some efforts being made to improve the scientific rigor of animal experiments, NAVS’ perspective is that this misses the larger problem. Issues with animal experiments extend far beyond issues of study design—there are flaws inherent with the use of animals, period.  

From NAVS’ perspective, we need to be asking why the research community continues to have the “luxury” of working with animal models at all. The reality remains that bias or not, animal models simply lack human relevance. If the scientific community really wants to improve human health and well-being through research, then more effort needs to be made not on making animal models slightly better, but on developing and using human-relevant non-animal models that have a higher probability of improving human health.


Source: Kwon, D. “Many preclinical studies carried our in vivo are poorly designed and generate irreproducible data, but efforts to address the problem are on the rise,” The Scientist, September 1, 2019.

This entry was posted in News and tagged on September 13, 2019.
Comments are closed.