Design Review Hearing
Questions That Must Be Answered Before Enrollment
A clinical trial does not begin when the first subject signs consent. It begins when the organization declares that its design can withstand scrutiny.
Before enrollment, the following questions must be answered — not casually, not implicitly, but in a form that could be defended before regulators, investors, and patients.
I. Definition of the Claim
What precisely is the treatment effect this trial intends to estimate?
In which population does that effect apply? Relative to what comparator? Over what time horizon? Under what handling of discontinuation, switching, rescue therapy, or death?
If this quantity cannot be stated without referencing the statistical model, it has not yet been defined.
If different members of the design team describe the objective differently, the design is not yet stable.
II. Ownership of Assumptions
What effect size is being assumed — and who takes responsibility for that assumption?
What event rate or variance assumption anchors the sample size? What happens if those assumptions are materially wrong?
Is dropout being treated as a logistical inconvenience to inflate for, or as a change in the scientific question?
In a non-inferiority trial, who determined the margin? On what historical and clinical basis? Is assay sensitivity defensible under current practice? Does the constancy assumption plausibly hold?
If these answers depend on optimism rather than justification, the risk has not been acknowledged.
III. Allocation of Error and Early Stopping
Is an interim analysis planned? For what purpose?
How is the probability of false positive redistributed across analyses? What are the operating characteristics under plausible true scenarios?
When someone states that “the stopping rate is too high,” which probability are they referring to — stopping under null, stopping under benefit, or stopping under harm?
Who will see interim data? Who has authority to recommend stopping? Is that authority documented?
If governance is unclear, the design is unstable.
IV. Protection Against Bias
How does the randomization structure protect against selection bias?
Is allocation concealment operationally secured?
Does the design rely on balance in ways that cannot be guaranteed?
Has the randomization plan been independently validated to a level commensurate with its risk?
If the trial is open-label, what bias pathways have been anticipated, and how are they mitigated?
If these protections are assumed rather than demonstrated, interpretation will later fracture.
V. Discipline of Claims
What constitutes trial success?
Is success defined by a single estimand, or multiple?
If multiple objectives exist, what logical rule governs success?
How is multiplicity controlled across endpoints, timepoints, and subgroups?
Which findings will support labeling claims? Which are exploratory?
If a subgroup contradicts the overall result, what principle governs interpretation?
If these boundaries are not defined in advance, interpretation will drift.
VI. Adaptation and Structural Change
Will any aspect of the design adapt after initiation?
Could adaptation alter the estimand?
How is type I error preserved under that adaptation?
Does enrichment redefine the population midstream?
If the answer to these questions is deferred until after interim data are seen, adaptation becomes revision rather than design.
VII. Pre-Specification and Accountability
Which elements of the design are fully locked prior to enrollment?
Which may adapt under rule-based procedures?
Is there a documented record of design decisions and their justifications?
Could an external reviewer reconstruct the logic of the design without oral explanation?
If not, the design exists in conversation, not in structure.
Final Condition
A trial design is ready for enrollment only when:
The treatment effect is precisely defined.
The assumptions are owned.
The risks are acknowledged.
The governance is explicit.
The claims are bounded.
Enrollment converts assumptions into exposure. After that point, ambiguity is no longer theoretical.
If these questions cannot be answered clearly and consistently, the trial is not yet ready to begin.