Open-Label Study

An open-label study is one in which nobody is masked: the participant, the treating clinician and usually the outcome assessor all know which arm the participant is in. The label describes blinding and nothing else. An open-label trial can still be randomised, controlled, prospectively registered and analysed by intention to treat, while a non-randomised study can be blinded, so the two axes have to be read separately rather than collapsed into good and bad.

Open design is routine wherever masking is impractical or unsafe. First-in-human dose escalation runs open-label so a safety monitor can react to the dose actually given. Extension phases that follow a blinded trial go open-label once every participant is moved onto active drug. Comparing a daily injection against a weekly one forces a choice between an open design and a double-dummy scheme in which each participant takes both schedules, one of them dummy.

How much this costs depends on the endpoint. Knowledge of assignment moves appetite ratings, pain scores, global impression scales and anything a clinician grades by eye. It moves a centrally analysed glycated haemoglobin, a DXA scan or an adjudicated death very little. The standard compromise is the PROBE structure, open treatment with blinded endpoint assessment, which protects the measurement in trials where the treatment itself cannot be hidden.

The misuse is quoting an open-label result as though it settled efficacy. A single-arm open-label series in which participants improved from baseline has measured the drug plus expectation plus regression to the mean plus everything else that changed over the period, and no feature of the design separates them. Grey-market copy that calls such a series clinical trial evidence is literally accurate and close to evidentially empty.