Early-phase development combines limited information with consequential decisions. Small cohorts, evolving dose levels, and short follow-up make precision especially important: a few participants can change a summary substantially. The appropriate response is careful description and explicit uncertainty, rather than presenting exploratory signals as established effects.
ICH E8(R1) is a general study-design reference. The working approach below focuses on how biometrics teams can support early decisions while preserving the limits of the available evidence.
1. Describe the decision being supported
Dose escalation, expansion, feasibility, and preliminary activity are different objectives. Identify which decision a review package supports and who is authorized to make it. A descriptive efficacy plot should not quietly become a formal success criterion if the study was not designed for that purpose.
For a hypothetical dose cohort, report the participants contributing to each summary, their follow-up, and the data cutoff. A denominator should not include newly enrolled participants who have not yet had an opportunity for the relevant assessment unless that is the specified rule.
2. Keep participant context close
In small cohorts, individual histories can be highly informative. Review dose exposure, interruptions, relevant concomitant treatment, and assessment timing alongside aggregate summaries. A patient profile may help explain an outlying value, but it does not establish causality by itself.
The FDA’s premarketing safety guidance is relevant background. Our recommendation is to use a compact review packet with both cohort-level summaries and recoverable participant-level evidence, rather than relying on a single dashboard score.
3. Resist false precision
Report estimates with an appropriate account of uncertainty and avoid strong comparative language when the design cannot support it. A large percentage difference based on very few participants can be unstable. Exploratory subgroup patterns deserve particular caution.
For rare-disease programs, the FDA’s rare-disease development guidance is useful additional reading. Small populations require deliberate evidence planning; their scarcity does not make weak endpoint definitions or undocumented assumptions acceptable.
4. Build foundations that can travel
Keep data definitions, derivation decisions, coding versions, and output specifications organized from the beginning. The program may later need integrated analyses or a larger confirmatory study. Early consistency can reduce reconstruction without forcing a small study into an unnecessarily elaborate operating model.
Establish a short release checklist: correct cutoff, known pending data, reviewed participant cases, approved output status, and unresolved questions. Distinguish rapidly refreshed exploratory material from controlled deliverables.
A strong early-phase workflow supports judgment under uncertainty. It makes the available evidence easier to understand, identifies what remains unknown, and preserves the reasoning behind a decision so the next phase can build on it responsibly.
