Trials and Error

Only about 50% of animal studies report design methods such as randomization and blinding, according to preliminary research by Michael Festing, former laboratory animal scientist at the UK??s Medical Research Council.

The right animal model needs to be matched to the test design and purpose. For example, Slattery et al. showed that a rat model for depression involving reward processes may not be effective in determining pharmacological intervention.¹

But choices for the right model are limited, and certain choices (e.g., primates) tend to limit the number of animals that can be used. The study has to be big enough to detect a clinically important effect, but not so big it wastes resources.

Without perfect matching of cellular processes and mechanisms of action, there is always the chance that one drug is acting differently on two animal species.

Even basic processes assumed to be similar in mice and men have been found to operate by completely different mechanisms, as reported by Odom et al that liver transcription factors bind differently in mice and men.² Many side effects can appear in humans that were not picked up by the corresponding animal models, even if administered at a fraction of the dose used in the animal tests.

Perhaps the most egregious instance of this involved TGN1412, an anti-CD28 antibody, which was tested without negative effect in non-human primates but caused multiple organ failure in six human subjects. It remains unclear whether an unexpected difference in the drug??s mechanism of action or other factor caused the opposing results .

These elements are often rarely reported. Differences have been reported among mice of different sexes (see "Why Sex Matters" on p. 61). Age and weight variants in animals can confound physiological differences between species and distort test results.

Even test groups in randomized studies can have statistically significant differences in factors such as weight, and thus need re-randomization, according to information written by Michael Festing on his website (http://isogenic.info/).

Some test animals, such as mice, have never left their cages before a procedure is performed. Being handled by the lab worker can be traumatic. Certain environments have been shown to produce stress in animals and seeing other animals suffering can cause "sympathy pains."

Lee et al. found that even prenatal stress can hinder rat social behavior.³ Large populations of lab animals are also subject to disease outbreaks.

Genetically Altered Animals Schalkwyk et al. reported that transgenic mice maintain remnants of their original genome and may express phenotypes that could be misinterpreted as treatment results.⁴

Knockout or knockin mice may also only represent a portion of manifestations of a disease, that is, some hereditary and sporadic versions of disease have different mechanisms of action.

Researchers often don??t report negative results in animal tests, leaving clinical trial development oblivious to potential outcomes of using a new drug. Published positive results are five times more likely to lack reported randomization and blinding, according to Bebarta et al.⁵