Essential 10

4. Randomisation State whether randomisation was used to allocate experimental units to control and treatment groups. If done, provide the method used to generate the randomisation sequence. explanation

4a State whether randomisation was used to allocate experimental units to control and treatment groups. If done, provide the method used to generate the randomisation sequence.
4b Describe the strategy used to minimise potential confounders such as the order of treatments and measurements, or animal/cage location. If confounders were not controlled, state this explicitly.
Explanation
Examples

Using appropriate randomisation methods during the allocation to groups ensures that each experimental unit has an equal probability of receiving a particular treatment and provides balanced numbers in each treatment group. Selecting an animal ‘at random’ (i.e. haphazardly or arbitrarily) from a cage is not statistically random as the process involves human judgement. It can introduce bias that influences the results, as a researcher may (consciously or subconsciously) make judgements in allocating an animal to a particular group, or because of unknown and uncontrolled differences in the experimental conditions or animals in different groups. Using a validated method of randomisation helps minimise selection bias and reduce systematic differences in the characteristics of animals allocated to different groups [1-3]. Inferential statistics based on non-randomised group allocation are not valid [4,5]. Thus, the use of randomisation is a prerequisite for any experiment designed to test a hypothesis. Examples of appropriate randomisation methods include online random number generators (e.g. https://www.graphpad.com/quickcalcs/randomize1/), or a function like Rand() in spreadsheet software such as Excel, Google Sheets, or LibreOffice. The EDA has a dedicated feature for randomisation and allocation concealment [6].

Systematic reviews have shown that animal experiments that do not report randomisation or other bias-reducing measures such as blinding, are more likely to report exaggerated effects that meet conventional measures of statistical significance [7-9]. It is especially important to use randomisation in situations where it is not possible to blind all or parts of the experiment but even with randomisation, researcher bias can pervert the allocation. This can be avoided by using allocation concealment (see item 5 – Blinding). In studies where sample sizes are small, simple randomisation may result in unbalanced groups; here randomisation strategies to balance groups such as randomising in matched pairs [10-12] and blocking are encouraged [13]. Reporting the precise method used to allocate animals or experimental units to groups enables readers to assess the reliability of the results and identify potential limitations.

Report the type of randomisation used (simple, stratified, randomised complete blocks, etc.; see “Considerations for the randomisation strategy” below), the method used to generate the randomisation sequence (e.g. computer-generated randomisation sequence, with details of the algorithm or programme used), and what was randomised (e.g. treatment to experimental unit, order of treatment for each animal). If this varies between experiments, report this information specifically for each experiment. If randomisation was not the method used to allocate experimental units to groups state this explicitly and explain how the groups being compared were formed. 

 

Considerations for the randomisation strategy

Simple randomisation 

All animals/samples are simultaneously randomised to the treatment groups without considering any other variable. This strategy is rarely appropriate as it cannot ensure that comparison groups are balanced for other variables that might influence the result of an experiment.

Randomisation within blocks  

Blocking is a method of controlling natural variation among experimental units. This splits up the experiment into smaller sub-experiments (blocks), and treatments are randomised to experimental units within each block [5,13,14]. This takes into account nuisance variables that could potentially bias the results (e.g. cage location, day or week of procedure).

Stratified randomisation uses the same principle as randomisation within blocks, only the strata tend to be traits of the animal that are likely to be associated with the response (e.g. weight class or tumour size class). This can lead to differences in the practical implementation of stratified randomisation as compared to block randomisation (e.g. there may not be equal numbers of experimental units in each weight class).

Other randomisation strategies 

Minimisation is an alternative strategy to allocate animals/samples to treatment group to balance variables that might influence the result of an experiment. With minimisation the treatment allocated to the next animal/sample depends on the characteristics of those animals/samples already assigned. The aim is that each allocation should minimise the imbalance across multiple factors [15]. This approach works well for a continuous nuisance variable such as body weight or starting tumour volume.

Examples of nuisance variables that can be accounted for in the randomisation strategy 

  • Time or day of the experiment  
  • Litter, cage or fish tank 
  • Investigator or surgeon – different level of experience in the people administering the treatments, performing the surgeries, or assessing the results may result in varying stress levels in the animals or duration of anaesthesia 
  • Equipment (e.g. PCR machine, spectrophotometer) – calibration may vary 
  • Measurement of a study parameter (e.g. initial tumour volume) 
  • Animal characteristics (e.g. sex, age bracket, weight bracket) 
  • Location – exposure to light, ventilation and disturbances may vary in cages located at different height or on different racks, which may affect important physiological processes

Implication for the analysis  

If blocking factors are used in the randomisation, they should also be included in the analysis. Nuisance variables increase variability in the sample, which reduces statistical power. Including a nuisance variable as a blocking factor in the analysis accounts for that variability and can increase the power, thus increasing the ability to detect a real effect with fewer experimental units. However, blocking uses up degrees of freedom and thus reduces the power if the nuisance variable does not have a substantial impact on variability. 

 

References

  1. Schulz KF, Chalmers I, Hayes RJ and Altman DG (1995). Empirical evidence of bias. Dimensions of methodological quality associated with estimates of treatment effects in controlled trials. Jama. doi: 10.1001/jama.1995.03520290060030
  2. Schulz KF and Grimes DA (2002). Allocation concealment in randomised trials: defending against deciphering. Lancet (London, England). doi: 10.1016/s0140-6736(02)07750-4
  3. Chalmers TC, Celano P, Sacks HS and Smith Jr H (1983). Bias in treatment assignment in controlled clinical trials. New England Journal of Medicine. doi: 10.1056/NEJM198312013092204
  4. Greenberg BG (1951). Why randomize? Biometrics. doi: 10.2307/3001653
  5. Altman DG and Bland JM (1999). Statistics notes. Treatment allocation in controlled trials: why randomise? BMJ. doi: 10.1136/bmj.318.7192.1209
  6. Percie du Sert N, Bamsey I, Bate ST, Berdoy M, Clark RA, Cuthill I, Fry D, Karp NA, Macleod M, Moon L, Stanford SC and Lings B (2017). The Experimental Design Assistant. PLoS Biol. doi: 10.1371/journal.pbio.2003779
  7. Hirst JA, Howick J, Aronson JK, Roberts N, Perera R, Koshiaris C and Heneghan C (2014). The need for randomization in animal trials: an overview of systematic reviews. PLoS ONE. doi: 10.1371/journal.pone.0098856
  8. Vesterinen HM, Sena ES, ffrench-Constant C, Williams A, Chandran S and Macleod MR (2010). Improving the translational hit of experimental treatments in multiple sclerosis. Multiple Sclerosis Journal. doi: 10.1177/1352458510379612
  9. Bebarta V, Luyten D and Heard K (2003). Emergency medicine animal research: does use of randomization and blinding affect the results? Acad Emerg Med. doi: 10.1111/j.1553-2712.2003.tb00056.x
  10. Taves DR (1974). Minimization: a new method of assigning patients to treatment and control groups. Clinical pharmacology and therapeutics. doi: 10.1002/cpt1974155443
  11. Saint-Mont U (2015). Randomization does not help much, comparability does. PLOS ONE. doi: 10.1371/journal.pone.0132102
  12. Laajala TD, Jumppanen M, Huhtaniemi R, Fey V, Kaur A, Knuuttila M, Aho E, Oksala R, Westermarck J, Mäkelä S, Poutanen M and Aittokallio T (2016). Optimized design and analysis of preclinical intervention studies in vivo. Scientific reports. doi: 10.1038/srep30723
  13. Bate ST and Clark RA (2014). The design and statistical analysis of animal experiments. Cambridge University Press. https://www.cambridge.org/core/books/design-and-statistical-analysis-of-animal-experiments/BDD758F3C49CF5BEB160A9C54ED48706
  14. Kang M, Ragan BG and Park JH (2008). Issues in outcomes research: an overview of randomization techniques for clinical trials. J Athl Train. doi: 10.4085/1062-6050-43.2.215
  15. Altman DG and Bland JM (2005). Treatment allocation by minimisation. BMJ. doi: 10.1136/bmj.330.7495.843