John Ioannidis, MD, DSc, has published a perspective piece in today's Science Translational Medicine about the difficulties of extrapolating results from preclinical studies in animals to humans. Although most preliminary safety and efficacy studies of medications are conducted first in animals, there have been many examples of promising drug candidates that either don't work in humans, or even do harm. Ioannidis comments in the article (subscription required):
Potential explanations for the failure of animal models to capture treatment effects in humans can be placed into two categories: First, both the human and animal results are accurate, but human physiology and disease are not adequately captured by animal models. Second, the animal literature is susceptible to biases in the study design, to reporting biases that distort the published evidence, or both. Indeed, although the scientific literature related to human clinical trials suffers from biases, data from preclinical animal studies appear to be associated with even greater bias, for a variety of reasons discussed below.
Specific problems include, among others, a lack of randomization of treatment and of blinding the researchers to the treatment or intervention the animal has received. From the article:
Because of these caveats, it is nearly impossible to rely on most animal data to predict whether or not an intervention will have a favorable clinical benefit–risk ratio in human subjects.
Essentially, that's why we do in human trials. But unfortunately there are specific cases when it's nearly impossible to conduct the necessary studies to verify a successful intervention in animals will also work in humans. In 2002 the Food and Drug Administration created the Animal Rule, for use when it would be unethical or impossible to conduct the required human research - such as for very rare diseases or for exposure to chemical or biological weapons. In his perspective piece Ionannidis praises another article in the journal that analyzed 21 studies on the anthrax vaccine to meet the requirements of the Animal Rule and show that the vaccine is likely to be effective in humans.
Ioannidis summarized his thoughts for me this morning:
Animal research is extremely important, and in theory it can offer valuable preclinical insights. However, currently published preclinical evidence from animals seems to have very limited concordance with what we see in humans. Almost everything seems to work in animals, and then almost nothing works in humans. The credibility and utility of animal research may be strengthened, if we can bolster the way animal experiments are designed, conducted, and published in the literature.