3Rs-Centre Utrecht Life Sciences

Faculteit Dierengeneeskunde

 
March 2018
 
An appeal to improve the quality of animal experiments

How does better design and statistical analysis of experiments using laboratory animals contribute to the 3Rs? What problems result from the use of inbred strains? Dr. Hein van Lith is associate professor at the Utrecht department of Animals in Science and Society (Faculty of Veterinary Medicine, Utrecht University). His expertise lies in the quality of animal experiments, i.e. the 3Qs: Quality of animal experimentation; Quality of laboratory animals; Quality of statistical design and analyses of animal experiments.He explains why a good experimental design is so essential, and what points of attention regarding study quality are most important in the field of reproducible preclinical biomedical research. He explains how his work contributes to the 3Rs.

Dr. Van Lith has been part of the department almost from the start (35 years ago). “When I started as a PhD candidate at this department (formerly known as department of Laboratory Animal Science) in 1987, the quality of animal studies was already highly valuated,” says Van Lith. “That means that not only the experiment itself should be standardized, i.e. controlling exogenous factors, but also the animals used in the study, i.e. controlling endogenous factors.” For an overview of the factors that can influence animal research see Claassen (1994) and Baker & Lipman (2015). “So I studied the standardization of the genetic background, feeding regimes and later in my research trajectory I also contributed to the microbiological standardization of laboratory animals.” Within a standardized experiment, the results comprise less variation, which is the argument for standardization in the first place. Further, such animal experiments meet a basic rule of experimental design: all variables should be controlled except that due to the treatment. Less variation means fewer animals are needed to obtain reliable results (Festing, 2018). In addition, precisely defining laboratory animals and their environment supports reproducibility. Furthermore, when researchers decide on a relevant animal model for their study, genetics, microbiology and nutrition can contribute to the reduction of laboratory animals as well. An alternative view is that standardization can identify idiosyncratic effects and hence causes poor reproducibility (Karp, 2018). The problem with standardization, however, is the fact that the results of standardized experiments are difficult to generalize and to extrapolate. “There is always a trade-off between standardization and variation,” says Van Lith. “If you standardize every factor in your study, this will be at the expense of the ability to generalize and to extrapolate. Therefore, you might have to offer variation to increase the translational value of the animal model, but the offering of variation should happen in a standardized manner.” Van der Staay et al. (2010) elaborate on this topic.

In silico and databases

Van Lith stresses the importance of databases in biomedical research. For his research Van Lith very often consults the databases of the Mouse Genome Informatics (MGI) consortium (Eppig, 2017), which provides the most comprehensive information about the characteristics of mouse inbred strains. There is even a database of mouse databases: the Mouse Resource Browser (Zouberakis et al., 2010). “There is so much to find on the internet, like data obtained in previous preclinical studies that have not yet been fully utilized.” One of his PhD candidates uses data from earlier studies to synthesize new evidence regarding themes such as the biological value of individual behavioral variation within inbred strains of mouse models. Although new theories often need verification in an animal model, much of the preliminary work can already be achieved by investigating existing data. Van Lith: ”If I would do a PhD in this field of research again, I would choose a PhD-study fully in silico. Although synthesis of evidence methods like systematic reviews are upcoming (Ritskes-Hoitinga & Wever, 2018), I think this type of research is highly underestimated, as well as the publication of negative results, for that matter.” According to Van Lith, there is much to be gained from negative results and unpublished data. “If you cannot publish your results in an article, at least publish them in a database. That way, the data can still be very valuable to other studies.” There are also journals that aim specifically for papers with negative results, like the Journal of Negative Results in BioMedicine (started in 2002, but unfortunately ceased last year). Such attempts reduce the positive bias in the scientific literature. These steps towards open science and transparency also help to prevent the unnecessary repetition of experiments that we have seen before.

Reproducibility crisis

On the latter topic, Van Lith mentions a few other points of attention. First of all, he stresses that – of course – the findings of experimental science using laboratory animals should be reproducible (independent from time, location and experimenter), but that such experiments should only be reproduced when something new is added. For example, when a new version of drug 2.0 is tested, the 1.0 version of the drug should be included in the study. That way, the study will help to check the reproducibility of the results from testing drug 1.0, while the therapeutic (and side-)effects of drug 2.0 (compared to drug 1.0 and the placebo) can also be evaluated. The new addition also adds value to the study, which will increase the possibility of getting the data published. A substantial area of concern in the context of reproducibility, however, is the fact that there should be more uniformity of animal experimentation between laboratories (harmonization). For example, there is a variation in the design of the apparatuses of anxiety tests (e.g. elevated plus maze, light–dark box, open field; Hogg, 1996; Kulesskaya & Voikar, 2014) which may influence the behavioral results of the study, and thus also makes it more difficult to compare the outcomes of multiple studies. Van Lith encourages efforts to reach an international consensus on these types of design aspects, since it affects the reproducibility.

The future of animal studies

Although dr. Van Lith uses – besides the in silico methods – for his research laboratory animals, he also hopes that we will come to a future in which animal studies are no longer needed. As far as animal studies are still an irreplaceable part of doing research, however, it is of outmost importance that these studies are of high quality in all possible manners. “We owe that to the animals, who would otherwise have been sacrificed for nothing”. Besides, Van Lith points out that most of his expertise in quality of animal experiments is also relevant in other disciplines, like in vitro methods that have the potential to replace animal studies. In vitro experiments have to meet the same quality requirements as animal experiments (Festing, 2001). For example, the cells you use must be of good genetic quality (i.e. no misidentified and/or contaminated cell lines; Uchio-Yamada et al., 2017), or else it would jeopardize the extrapolation and reproducibility of the study. Of course in silico research should also be of high quality. Again the 3Qs: quality of i) in vivo animal studies, ii) in vitro studies and iii) in silico studies.

 
References and further reading
  • Baker, D.G., Lipman, N.S. (2015). Chapter 33 – Factors that can influence animal research. In: Laboratory Animal Medicine, Third edition (Fox, J.G., Anderson, L.C., Otto, G., Pritchett-corning, K.R., Whary, M.T., eds.) pp. 1441-1495, Academic Press (an imprint of Elsevier Inc.), Cambridge, USA, http://dx.doi.org/10.1016/B978-0-12-409527-4.00033-X. 
  • Claassen, V. (1994). Neglected factors in pharmacology and neuroscience research – Biopharmaceutics, Animal characteristics, Maintenance, Testing conditions. Techniques in the Behavioral and Neural Sciences 12 (Huston, J.P., series ed.), pp. 1- 486, Elsevier Science B.V., Amsterdam. 
  • Eppig, J.T. (2017). Mouse Genome Informatics (MGI) Resource: Genetic, Genomic, and Biological Knowledgebase for the Laboratory Mouse. ILAR Journal 58:17–41, http://dx.doi.org/1093/ilar/ilx013. 
  • Festing, M.F.W. (2001). Guidelines for the design and statistical analysis of experiments in papers submitted to ATLA. Alternatives to Laboratory Animals - ATLA 29:427-446. 
  • Hogg, S. (1996). A review of the validity and variability of the elevated plus-maze as an animal model for anxiety. Pharmacology Biochemistry and Behavior 54: 21-30.
  • Karp, N.A. (2018). Reproducible preclinical research – Is embracing variability the answer? PLOS Biology 16(3):e2005413, https://doi.org/10.1371/journal.pbio.2005413. 
  • Kulesskaya, N., Voikar, V. (2018). Assessment of mouse anxiety-like behavior in the light–dark box and open-field arena: Role of equipment and procedure. Physiology & Behavior 133: 30–38, http://dx.doi.org/10.1016/j.physbeh.2014.05.006. 
  • Löscher, W., Ferland, R.J., Ferraro, T.N. (2017). The relevance of inter- and intrastrain differences in mice and rats and their implications for models of seizures and epilepsy. Epilepsy & behavior 73: 214-235, http://dx.doi.org/10.1016/j.yebeh.2017.05.040. 
  • Ritskes-Hoitinga, M., Wever, K. (2018) Improving the conduct, reporting, and appraisal of animal research - All stakeholders must act decisively to fix endemic problems. British Medical Journal 2018;360:j4935, http://dx.doi.org/1136/bmj.j4935. 
  • Uchio-Yamada, K., Kasai, F., Ozawa, M., Kohard, A. (2017). Incorrect strain information for mouse cell lines: sequential influence of misidentification on sublines. In Vitro Cellular & Developmental Biology – Animal 53:225-230, http://dx.doi.org/1007/s11626-016-0104-3. 
  • Van der Staay, F.J., Arndt, S.S., Nordquist, R.E. (2010). The standardization-generalization dilemma: a way out. Genes, Brain and Behavior 9: 849-855, http://dx.doi.org/1111/j.1601-183X.2010.00628.x. 
  • Zouberakis, M., Chandras, C., Swertz, M., Smedley, D., Gruenberger, M., Bard, J., Schughart, K., Rosenthal, N., Hancock, J.M., Schofield, P.N., Kollias, G., Aidinis, V. (2010). Mouse Resource Browser – a database of mouse databases. Database – The journal of Biological Databases and Curation 2010, Article ID: baq010, doi:10.1093/database/baq010.