Allocation bias

Systematic difference in how participants are assigned to treatment and comparison groups in a clinical trial. Index of biases | Editing advice | 

Background
Allocation bias may result if investigators know or predict which intervention the next eligible participant is supposed to receive. This knowledge may influence the way investigators approach potentially eligible participants and how they are assigned to the different groups, thereby selecting subjects with good prognoses (i.e. anticipated good outcomes and treatment responses) into one group more than another.

In a trial of different blood pressure medications the use of sealed envelopes to conceal group allocation resulted in imbalances in baseline blood pressure between the treatment and control groups, This introduced bias, since those in the control group already had lower blood pressures at the outset. The observed imbalance could have arisen if the investigator opened the envelopes before allocation to groups.

Example


In the nine trials using good allocation concealment, there was no effect of hypothermia treatment on death, whereas in the trials that did not conceal allocation there was a 38% reduction in the risk of death. Hypothermia for traumatic head injury. Cochrane Database of Systematic Reviews 2009.

Impact
There is evidence that over 80% of trials may have unclear allocation concealment. Trials in which allocation was inadequately concealed reported estimates that were between 7% and 40% larger than effects in trials in which allocation was adequately concealed, although the size and direction of the effect were not predictable. A simulation of trials using realistic conditions for allocation concealment showed that up to about 20% of true null hypotheses could be rejected because of false positive effects.

Preventive steps
Some standard methods of ensuring allocation concealment prior to randomization include sequentially numbered, opaque, sealed envelopes (SNOSE); sequentially numbered containers; pharmacy controlled randomization; and central randomization.

When triallists do more to show that the allocation was concealed, then this gives more trust that allocation bias has not affected the study’s results. For example, an audit trail can be recorded for sequentially numbered envelopes; sequentially numbered containers should look identical to each other from the outside; study staff should have correct training for maintaining allocation concealment; the block sizes for randomization should not be known; and all these features of the trial should be reported in the trial publications.

Where cluster randomisation or minimisation are used in the allocation to treatment of control groups, it may not be possible to prevent future allocations from being predictable and thus concealment might not be adequate. In such cases, pre-defined statistical adjustments can be considered and again should be reported in trial publications.

Researchers should try to understand the reasons for allocation concealment and the problems that can arise if proper trial protocol including allocation concealment is not followed. Allocation concealment procedures should also be specified in a trial’s protocol, and reported in detail in any publication of the trial’s results.

Cite as
Catalogue of Bias Collaboration. Spencer EA, Heneghan C, Nunan D. Allocation bias. In: Catalogue of Bias 2017. https://catalogofbias.org/biases/allocation-bias/

PubMed feed
Retrieve dynamically from PubMed