Allocation concealment and blinding are characteristics that prevent bias in randomised controlled trials and experimental studies. However, these concepts are often confused. Using a randomised controlled trial as an example, the statistician Philip Sedgwick explains the differences between allocation concealment and blinding, and why these characteristics are important:

Researchers investigated whether a nutritious meal and food packages was more effective at encouraging completion of tuberculosis treatment compared to nutritional advice alone. Participants were 270 adults with untreated and newly diagnosed tuberculosis recruited from primary care clinics in Dili, Timor-Leste. Participants were randomised to treatment or control group using a computer-generated randomisation sequence. Participants were allocated to either group with the randomisation sequence concealled in sequentially numbered, opaque sealed envelopes. An outcome assessor blinded to participants’ allocated group assessed the primary outcome.

In a clinical trial, random allocation is needed so that participants in treatment and control groups are similar at baseline. This allows us to conclude whether results observed are due to differences in treatment, rather than differences in baseline characteristics between groups. Consequently, random allocation of participants to treatment or control groups is needed to infer causal effects of the treatment on outcomes.

Allocation concealment means participants and investigators (eg. investigators who provide the treatment, assess the outcome, or analyse the data) are not aware of the allocation sequence before random allocation occurs. That is, before randomisation, neither participants nor investigators know which group a participant will be allocated to. Blinding means (some) participants and investigators are not aware of which group a participant is allocated to after random allocation occurs.

Allocation concealment is needed to prevent the random allocation sequence from being undermined during recruitment and treatment allocation. If participants and investigators know the allocation sequence before random allocation occurs, investigators may select (unconsciously or otherwise) which participants are recruited, or the order in which participants are recruited. For example, investigators might think some participants would not accept or would be unsuitable for the next allocated group in the sequence. Without allocation concealment, participants might also “self-select” or choose to particpate or not participate based on whether they favour or dislike the next allocated group in the sequence. Staying faithful to the randomised sequence can be difficult. Indeed, the physician Kenneth Schulz writes about the lengths some investigators go to in order to break the allocation concealment.

Allocation concealment prevents selection bias, or a systematic difference between participants who are recruited for a trial and those who are not. This results in a sample that does not represent the population. Allocation concealment also prevents allocation bias, that is, a systematic difference between participants in how they are allocated to treatment or control groups.

Allocation concealment is always possible in clinical trials and is needed to achieve blinding, but blinding is not always possible. In the example above, allocation concealment was achieved by concealing the allocation sequence in opaque sealed envelopes. However, participants and investigators providing the treatment (nutritious meal or nutritional advice) could not be blind to allocated group. This means it was still possible for participants to demonstrate a response bias without blinding. However, the outcome was assessed by an investigator blind to group, which would help minimise assessor bias.

Summary

Allocation concealment and blinding prevent different types of biases in randomised controlled trials and other studies. Allocation concealment is always possible in clinical trials but blinding is not always possible. Readers need to be cognizant of whether and how these biases were prevented in order to judge the credibility of research findings.

References

Schulz KF and Grimes DA (2002) Allocation concealment in randomised trials: defending against deciphering. Lancet 359:614-618.

Sedgwick P (2013) Allocation concealment vs blinding in randomised controlled trials BMJ 347:f5518.