Animal Experiments And Animal ‘Attrition’
The use of animals in biomedical experimentation is something that many animal advocates find both outrageous and seemingly intractable. Animals are used by researchers in their attempts to study a range of medical conditions, from the serious (cancers, Parkinson’s, and other diseases) to the superfluous (such as impotence). Throughout all of this, animal advocates question the acceptability and usefulness of these experiments on both ethical and predictive grounds. Meanwhile, scientists who want to demonstrate ethical practices claim to abide by the “3Rs” — “replacement, reduction, and refinement.” Some research institutions spend billions of dollars every year to develop new drugs, but the “failure to translate laboratory findings into clinical applications” leads to many questions. Much of this criticism revolves around “selection of animal models, internal validity, statistical power, reporting, and publication bias.”
One essential element of assessing any preclinical study is to get have accurate numbers. If done properly, researchers note, the reporting of animal numbers should give a full account of all animals lost during the experiment. This “attrition” number is important not only for ethical reasons but also because it can have an impact on the statistical validity of the study and may “be indicative of side effects or toxicity of new treatments. Unreported loss of these animals, therefore, is a potentially harmful form of selection bias.” In this meta-study, researchers demonstrated “the consequences and prevalence of attrition in preclinical research” by simulating data and comparing results “with and without animal loss” to look at two kinds of attrition: random loss and biased removal. Then they applied this method to a sample of studies relating to stroke and cancer to understand the frequency of animal attrition reporting and determine if attrition influenced the study results.
Not surprisingly, in their simulated datasets the researchers found that randomly excluding animals from the sample decreases the sample size and statistical power of the study. But they also showed that when animals are excluded in a targeted fashion, it “can have extreme consequences with regard to false positives and skewed interpretation.” In one such simulation, the removal of outliers of the sample size increased the false positive rate from 5% to a startling 46%. When they applied their meta-analysis to a set of cancer and stroke studies, they found that “47%, or 25/53 of experiments with attrition reported 25% animal loss or more.” This was equal to or worse than what the researchers considered the “worst case scenario.” They note that these rates could lead to effect sizes “inflated by 25% to 175% amongst experiments with statistically significant results.”
What do these results mean for animal experimentation? The researchers are careful to qualify that “attrition of animals is often unforeseen and does not reflect willful bias.” They also note that there are simple steps that the scientific community can use to try to ameliorate the situation. For animal advocates, however, the study provides yet more evidence that the use of animal experiments is (and will continue to be) plagued by procedural and statistical problems, which is to say nothing of the ethical implications. This study can be used by animal advocates to demonstrate some of the statistical flaws in particular animal experiments, as well as animal experimentation more broadly.