Experimental design/statistics

The greater the number of animals used in an experiment, the greater will be the overall costs, in terms of animal suffering. Thus, the number of animals should be the  minimum which is consistent with the aims of the experiment.

Well designed and correctly analysed experiments can lead to a reduction in animal use without reducing scientific output. Indeed, poorly designed experiments may give the wrong results leading to a waste of both animals and scientific resources. A well designed experiment should have the following characteristics:

Lack of bias: Where there are two or more treatment groups, these should be in identical environments and be similar in every way apart from the applied treatments. Bias can be avoided by random allocation of animals to the treatment groups, and by ensuring that housing and all subsequent treatments are done in a random order. Ideally, researchers should be blinded with respect to treatment groups until the final statistical analysis.

High power: Powerful experiments are ones which have the maximum chance of detecting a true treatment effect. Power is achieved by controlling and eliminating inter-subject variation, and by increasing sample sizes. Variation is controlled by choosing animals of similar genotypes (e.g. use of isogenic strains ) which, as far as possible, are of the same weight and age and have had a similar environment throughout their lives. Measurement error should be minimised by careful technique and good instrumentation. Variation due to circadian rhythms or fluctuations in the environment can often be reduced by appropriate experimental design, such as the use of randomised block or Latin square experiments.

Although power is increased by increasing sample size, an unnecessarily large experiment will be subject to diminishing returns and will waste animals and scientific resources. So sample size should be determined using a formal method such as power analysis or the resource equation method.

The power analysis method for comparing two groups, for example, requires a specification of:

• the type of statistical test to be used, such as a t-test or a chi-squared test to compare two proportions
• the significance level to be used (often a 5% level)
• the required statistical power (usually 80-90%), the sidedness of the test (a 2-sided test is usual)
• the effect size of biological interest (i.e. how much of a difference in means or proportions would be of biological or clinical importance to be able to detect)
• an estimate of the standard deviation (when comparing means) - this must come from a previous experiment

These factors are discussed in more detail by Dell et al. (2002), Festing et al. (2002) and more briefly by Festing and Altman (2002) (see References above). The StatPages.net website offers online calculations of sample size combining the above factors.

The resource equation method depends on the law of diminishing returns and is most appropriate for small non-routine and more complex animal experiments likely to be analysed using the analysis of variance. This suggests that the total number of subjects, N, (e.g. animals or cages of animals if these are what is being assigned to treatments) minus the number of treatment combinations, T, should be approximately E = 10-20. For example, an experiment to compare four treatments with six rats per treatment would have N = 24 minus T= 4, so E would be 20. This is within the acceptable range. However, there may be good reasons for going above this upper limit. But if E is 30 or 40, the experiment may be too large and may waste resources.

A wide range of applicability: It is often useful to find out whether similar results are obtained in males and females or with different strains or as a result of different diets or environments. Similarly, the response to a drug may depend on prior treatment, the effects of other drugs, or routes of administration. These effects can be studied efficiently using factorial experimental designs. A common misconception is that the use of both sexes in an experiment means that twice as many animals need to be used. In fact the experiment can usually be done using the same total numbers, half males and half females. Such a factorial experiment also indicates whether the two sexes will respond in the same way, which is not possible if the two sexes are used in different experiments.

Simplicity: Experiments should not be so complicated that mistakes are made in their execution, or the statistical analysis becomes unduly complicated. Small pilot studies should be used before starting a major experiment to ensure that the experiment is logistically efficient and to give some preliminary indication of likely results. All experiments should be pre-planned, and should not be changed while they are in progress.

Indicate their range of certainty: When finished each experiment should immediately be statistically analysed so that the results can be used in planning future experiments. An appropriate statistical analysis should show the range of uncertainty in the conclusions, usually indicated by significance levels and confidence intervals.