Meta-analysis — a systematic and quantitative review of animal experiments to maximise the information derived
Meta-analysis provides a tool to statistically aggregate data from existing randomised controlled animal experiments. The results can then be summarised across a range of conditions and an increased pool of experimental data can be subjected to statistical analysis. New information
can be derived, but most frequently the results are a refinement of existing knowledge. By designing experiments and reporting protocols, so that they have the capability of being useful to meta-analyses, maximum benefit can be derived from individual randomised controlled experiments, which
may individually have little statistical power, and new avenues for productive research identified. The methodology for meta-analysis is derived from clinical trials in the medical sciences. Now that there is substantial output from animal science experiments, there is an opportunity to apply
the technique to these and reduce the need for further experimentation. This paper describes the contribution of meta-analysis to the reduction of animals in research and provides details on data collection, analysis, the models used, and on interpreting and reporting the results. Three applications
of meta-analysis to the field of animal science are also briefly described. First, the impact of undernutrition on the production and composition of milk from dairy cows confirmed existing knowledge about partitioning scarce nutrients to milk yield and live weight. Second, increased absorption
of cadmium — a widespread toxic element — from organic sources was detected in sheep, which was previously untested. Third, no significant relationships were found between common indicators of undernutrition and weight, and condition score in cattle suggesting that the common indicators
used are not suitable as evidence of long term undernutrition. This paper concludes that opportunities exist to increase the information gained from animal experiments by subjecting the results to meta-analysis, particularly if this can be anticipated in advance of study protocols being constructed.