Scientific reputations have been destroyed by a failure to use blinding. Lack of blinding can cause serious bias. Very often a scientist wants a particular outcome from an experiment, and it is quite possible to choose the material or interpret the data in such a way as to achieve the desired result. It should always be made clear in any scientific publication whether or not blinding was used in assessing the results.
What is blinding
Need for blinding
Blinding in practice
Blinding in assessing histological slide
A C57BL/6 mouse heterozygous for the yellow allele (Ay)
There may be a subjective element in assessing the outcome of an experiment and/or the investigator may want a particular result. In such cases it is possible, unintentionally, to bias the results towards that favored outcome, leading to incorrect conclusions. This can be avoided if the investigators do not know to which treatment group an individual belongs until the end of the experiment. Human clinical trials are usually done “double blind”; neither the patient nor the doctor assessing outcome knows which treatment the patient has received.
In 1988 Nature published a paper by Jacques Benveniste and colleagues who reported that human basophil degranulation could be triggered by homeopathic dilutions of one part in 10120 of an anti-IgE antiserum. As it is unlikely that even a single molecule would be present in such a dilute solution they proposed that water could retain memory of molecules no longer present. However, when a team from Nature visited the laboratory and the experiment was repeated “blind” no such effect was observed. This episode effectively destroyed Benvenist’s career, and demonstrated the danger of making important claims based on un-blinded or otherwise badly designed experiments.
Animals or other experiment units and treatments should be coded in such a way that it is impossible to identify the treatment or treatment group until after the experiment has been decoded. For example, whoever makes up drugs for dosing the animals could colour code them (red, green, black), and cage labels could be similarly colour coded so that they receive the correct drug, but the staff administering them would not know which was the placebo etc. Colours should be mixed on each shelf. However, care must be taken to ensure that blinding does not lead to mistakes.
Where sham operated animals are being compared with non-sham operations, the animals should be numbered in non-consecutive runs so that any person assessing the results will not know the treatment group.
There are some situations where blinding is impossible. If two mouse strains are being compared, say for a behavioural c haracter, and one is black and the other brown, it is obviously impossible to do the experiment blind with respect to strain. In this case reliance needs to be placed in randomisation of the order in which the animals are tested.
Blinding in the assessment of histological slides is of particular importance because of the large subjective element in assessing the results. Some pathologists claim that it is impossible to read slides blind with respect to treatment group. In such cases they may be prepared to look at a small sample from each treatment group un-blinded to get their eye in, and then look at the rest blinded.