Protocol freeze
Version the instructions, variables, controls, stop conditions, analysis plan, and permitted deviations before sites receive results.
CROSS-SITE TRIAL GATE · P11
Many sites do not automatically create independence.
A trial becomes informative when instruments, sites, people, custody, analysis, and failure reporting are characterized well enough to distinguish replication from shared bias.
Instrument diversity
Describe how detector substrates, calibration, resolution, environment, and operator coupling differ.
Site independence
Record institutional, financial, personnel, data, and AI-model overlap that could create false independence.
Blind exchange
Separate protocol distribution, raw data custody, analysis, and synthesis wherever the research question permits.
Failure symmetry
Publish nulls, failed runs, exclusions, dropouts, and non-comparability with the same identifiers as positive outcomes.
Review boundary
Independent reviewers approve protocol adequacy and safety; they do not pre-approve a scientific conclusion.
ADVANCE HOOK
Open only after independent protocol review
- P10 safety, reversibility, control, and stop gates pass.
- Roles and conflicts are named before data collection.
- Each site can preserve its own evidence envelope.
- Comparison rules retain detector and frame differences.
- A public path exists for nulls, disputes, and withdrawal.