Watch on YouTube

7:15

Running the trial: fidelity, monitoring, and reporting

Some trials fail because the intervention was never delivered as designed. Tracking reach, dose, fidelity and acceptability lets a team tell a theory that failed from an implementation that did. While the trial runs, the team checks data quality without looking at outcomes by arm. That job belongs to an independent Data Safety Monitoring Board, which can stop a trial for benefit, harm or futility, as boards did when the male circumcision trials showed 51 to 60% reductions in HIV acquisition. Every look raises the chance of a false positive, so early stopping rules are strict. CONSORT then makes the report checkable.

Also covered: implementation fidelity · Data Safety Monitoring Board · O'Brien-Fleming boundaries · CONSORT