Trial Registration and Prospective Protocol

Registration means depositing a trial in a public registry before the first participant is enrolled, recording the population, the arms, the sample size, and above all the primary and secondary outcomes and when they are measured. The point is that the analysis becomes a matter of public record before any data exist to shape it, and that the trial's existence is known whether or not it ever produces a publishable answer. A dated protocol and statistical analysis plan serve the same function in more detail.

Journal editors, through the ICMJE, made prospective registration a condition of publication from 2005. In the United States the FDA Amendments Act of 2007, tightened by a rule effective in 2017, requires registration on ClinicalTrials.gov and posting of summary results, generally within twelve months of completion. The COMPare project in 2016 checked published trials against their registered protocols and found reported outcomes often differed from prespecified ones.

A registry entry is a tool for the reader. It lets you compare the endpoint that was published with the endpoint that was planned, see when a change was made and whether it preceded unblinding, and confirm that the trial and its enrolment dates exist as described. Where the reported analysis silently differs from the registered one, the resulting p-values no longer mean what they claim.

Registration is nonetheless routinely quoted as a badge of quality. Entries are self-reported, unvetted for scientific merit, and creating one costs nothing, so a registration number establishes only that someone announced a trial. It is also not results reporting: a large fraction of completed studies never post results, and a record created after enrolment closed provides none of the protection prospective registration exists to give.