Abstract
The advent of genetically engineered animal models of disease has notably advanced translational research beyond traditional animal models of disease (eg, spontaneous mutants, xenograft models, surgical models, experimentally induced disease). Because of these advances, many more scientists (including nonbiologists such as chemists and engineers) are involved with in vivo translational research. In a 2007 letter to the editor, Valli et al noted that more than 60% of National Institutes of Health extramural funding supports research utilizing animals, of which more than 80% involve the use of mice. 13 Despite this situation, consistent and detailed methodology related to the in vivo portions of the animal studies is not often included in the manuscripts.
A recent publication by Kilkenny et al 8 looked into the reporting of methodology in studies incorporating translational research. Their scrutiny of the quality of experimental design, analysis, and reporting in biomedical research utilizing laboratory animals made it clear that animal data are frequently not reported or are incompletely reported in the published literature. In their review, which used a systematic search of Medline and EMBASE from January 1999 to March 2005, they identified 170,000 publications using rats, mice, and nonhuman primates. From these publications, 271 articles were chosen for evaluation based on multiple criteria, including provisions that they had been publicly funded (either in the United Kingdom or United states or both), that they did not perform ex vivo studies to gain their conclusions (eg, Langendorff assay), and, finally, that the reporting laboratory had no more than two articles on the same subject. These works were from journals with a range of impact factors, including Nature and Science. From them, 48 studies were randomly chosen for detailed evaluation of study design and quality of data analysis.
The most startling finding in their review was that only 59% of in vivo study publications included a clear hypothesis, detailed three animal characteristics (sex, strain, weight, or age), and specified the number of animals used. In 13% of studies, no animal numbers were identified. The sex of the animals was not reported in 26% of the studies, and only 43% reported age. In addition, the authors pointed out that few researchers reported randomization of animals into treatment groups and that in several cases, the statistical design and reporting of statistical findings were not described.
Variation in animal strains and its impact on the outcome of translational studies—whether traditional or in genetically engineered mice—have been well documented. 1,5,9,10 Attention has also recently been given to the effect of mouse age and sex on study outcomes, which also vary by strain. 4,6,7,10,14 Given that there is significant strain variation in background lesions and susceptibility to disease and cancer, it is imperative that authors report this information 2 and that journals require it. A clear methodology in translational studies must include—at a minimum—the number of animals per group; the strain, stock, or line; the age; and the sex. Genetically engineered mice and mutant mice must also be reported with the appropriate nomenclature—for example, the correct nomenclature for NOD-SCID gamma (NSG) mice is NOD.Cg-PrkdcscidI12rgtm1Wjl (see http://www.informatics.jax.org/)—and the number and method of backcrossings in the mice. 2,3,11,12 Accurate interpretations of study outcomes and the ability to compare results between different laboratories require this information. Without it, conclusions cannot be validated, and differences in conclusions cannot be reconciled.
Careful experimental design is essential to the quality of in vivo studies. The expense of nonhuman primate studies obliges researchers to optimize the quality of information derived from a single study. Sadly, however, rodent studies may not be designed and performed in a manner that maximizes the quantity, quality, and reliability of the generated data. Researchers have a scientific, as well as an ethical and moral, responsibility to design appropriate animal studies. Beyond careful consideration of the impact of the background strain and backcrossings 3 on the outcome and validity of the study data, translational studies must have well planned in vivo molecular and histological assessments with enough animals to generate statistical reliability. The findings by Kilkenny et al8 remind us that as authors, reviewers, and editors, we are obliged to ensure the accurate reporting of study design in translational research; reporting must at a minimum include the number of animals used per group and the sex, age, and strain, stock, or line of the animals (with correct genetic nomenclature). This information will be essential for accurate and appropriate interpretations of scientific findings as we accumulate more and more translational data.