Allocation Concealment

Allocation concealment is the safeguard that prevents anyone recruiting a participant from knowing, or guessing, which arm that participant will be assigned to. It protects the moment of enrolment, which is when selection can still corrupt a randomised trial. Standard mechanisms are central randomisation by telephone or web service, pharmacy-controlled assignment, and sequentially numbered opaque sealed envelopes prepared by someone with no clinical role.

It is routinely confused with blinding, and the two operate at different times. Concealment applies before assignment and is always possible, even where masking the treatment afterwards is not, as in surgery versus physiotherapy. Cochrane risk-of-bias tools assess them as separate domains for that reason, and meta-epidemiological work going back to the 1995 analysis by Schulz and colleagues found that trials with inadequate or unclear concealment produced systematically larger effect estimates.

The consequence is that a paper describing itself as randomised has told you almost nothing until it says how the sequence was concealed. An investigator who can see that the next envelope is active treatment, and who believes the drug works, need not act dishonestly to bias the trial; a slight tendency to enrol the sicker patient when the next slot is active is enough to move the result.

The failure modes are mundane rather than exotic. Translucent envelopes read against a light, envelopes opened out of order, and alternation or assignment by date of birth described as randomisation have all been documented. When a trial reports only that participants were randomised, with no account of who generated the sequence and who held it, concealment is unknown and the effect estimate should be read as an upper bound.