Why a Protocol Written After the Search Is Already Compromised

A protocol is a commitment device. Once the search has been run and the results skimmed, every subsequent methodological choice is contaminated by knowledge of what those choices would produce.

In practice, the most common failure we see is not fabrication. It is the quiet drift of eligibility criteria after the team notices which studies would change the pooled estimate.

Registering a protocol does not guarantee quality, and it certainly does not guarantee publication. What it does is make deviations visible, which is the only condition under which peer review can function.