Before data exist
The question, stated precisely enough to be answerable. The design, chosen because it can answer that question. The primary outcome, singular. The analysis plan. The sample size, with the assumptions it rests on stated.
Registration where the design allows it, which puts these on record and is what makes later departures visible. This is the point of maximum leverage and minimum cost.
During conduct
Ethical approval before the first participant. Consistent measurement. Data management that will still make sense in two years. A record of protocol departures as they happen, since reconstructing them later is unreliable.
Recruitment is where timelines fail; feasibility work exists to find that out before the definitive study rather than during it.
Analysis
The pre-specified analysis first, reported whatever it shows. Exploratory work labelled as exploratory. Missing data handled by a stated, justified method rather than by deletion without comment.
Writing and submission
The reporting checklist for the design, completed as the paper is written. A journal chosen for scope fit before anything else. The submission package assembled against the journal’s requirements — cover letter, title page, declarations, reporting checklist.
Review and after
A response letter that reproduces every comment and points to every change. After acceptance, the finding has to reach people who will act on it, which means different documents for different audiences and the same uncertainty preserved in each.
In short
What to take from this.
- Register before data collection — maximum leverage, minimum cost.
- Record protocol departures as they happen.
- Report the pre-specified analysis whatever it shows.
Sources
- EQUATOR Network reporting guidelines
- ICMJE Recommendations, on registration and reporting
Related