Allocation Concealment

Keeping the next treatment assignment unknown to recruiters and clinicians until the moment a patient is irreversibly entered. It prevents anyone from steering sicker or more favourable patients into a preferred arm. Achieved by central/remote randomisation, sequentially-numbered opaque sealed envelopes, or pharmacy-controlled packs. It happens at allocation — distinct from blinding, which happens after.

A barrier between the randomisation list and the person enrolling patients. The allocation is only released once a patient is irreversibly entered — so selection cannot be gamed at the point of entry.

Trace one patient: the recruiter decides to enrol and commits before the arm is disclosed. With central randomisation or a sealed opaque pack there is no way to peek ahead. Poor methods — an open list, or a translucent envelope held to the light — let the recruiter see what’s next and decline or delay an “inconvenient” patient.

Broken concealment lets baseline prognosis differ between arms before treatment even starts — that is selection bias, and it inflates apparent effects. Trials with inadequate or unclear concealment report exaggerated treatment effects on average, so it is a core risk-of-bias domain.

• Concealment = before/at allocation → prevents selection bias
• Blinding = after allocation → prevents performance/detection bias
• Gold standard: central/remote randomisation or sequentially-numbered opaque sealed envelopes

CRASH-2 (Lancet 2010) — 20,211 trauma patients, TXA vs placebo. Allocation was concealed: treatment packs were identical apart from a number, with central telephone randomisation (Oxford CTSU) where practicable and a sequentially-numbered local pack system otherwise, so the recruiter never knew the next assignment. Concealment is not blinding: here the identical packs delivered both — recruiters could not subvert entry, and staff could not tell TXA from placebo afterwards.

• Equating allocation concealment with blinding — concealment is before/at allocation, blinding is after.
• Accepting “sealed envelopes” without checking they were opaque, sequentially numbered and tamper-evident.
• Missing that broken concealment causes selection bias and inflates the effect estimate.