Clinical Stage Normalization

1. Product rule

Model what is measured. Annotate what is described. PhaseFolio preserves exact sponsor or source labels, but the rNPV engine does not treat every label as a new probability transition. A row called Phase 2a maps to the canonical Phase II PoS gate. The phrase Phase 2a does not add an uplift, haircut, or second survival gate by itself.

The stage-normalization model separates three records: work packages for cost and timing, PoS gates for survival math, and mapping decisions explaining how an exact label became a valuation input. This strengthens the assumption, evidence, decision, and export records without adding unsupported clinical precision.

2. Why labels are not ordinary stages

Large clinical-success datasets are phase-level, not stable subphase-level benchmarks. BIO, QLS Advisors, and Informa Pharma Intelligence report transition probabilities across standard development phases; Wong, Siah, and Lo similarly estimate phase-level clinical-trial success parameters. Those sources support Phase I, Phase II, Phase III, and regulatory-review gates. They do not support a universal Phase 2a versus Phase 2b PoS split.

Therefore, Phase 2a and Phase 2b can be modeled as separate work packages when their costs or durations differ, but they map to one Phase II PoS gate by default. If a user explicitly creates an additional custom gate or a human-confirmed override, the override is visible, rationale-backed, and versioned.

3. What can change rNPV

A label alone does not change PoS. However, a normalized stage plan can still change rNPV through explicit assumptions: cost, duration, launch timing, post-approval obligations, regional bridging cost, or an analyst-confirmed route such as accelerated approval. Splitting one 24-month Phase II cost into two 12-month work packages changes cost timing because each package is discounted at its own midpoint. The disclosure is therefore: PoS unchanged; discounted cost and timing may change rNPV.

4. Default mappings

5. Combined and adaptive designs

NCI definitions of Phase 1/2 and Phase 2/3 trials support preserving combined-phase labels: these designs intentionally answer adjacent questions in a single protocol. They do not justify a proprietary blended PoS. FDA adaptive-design and master-protocol guidances support treating adaptive, basket, umbrella, and platform labels as evidence-generation structure. They can affect trial operations and uncertainty; they do not create a universal approval-probability modifier.

6. Human override boundary

AI import, deck extraction, or ClinicalTrials.gov review may propose exact labels and mappings. Durable scenario math changes only when the user accepts visible assumptions: a cost, duration, launch timing, custom gate, route selection, or override rationale. PhaseFolio agents may propose; the human-confirmed scenario remains the source of truth.

7. Engine and export behavior

Engine 2.8.0 adds an optional development_stage_plan. If absent, legacy stages[] scenarios run the untouched legacy branch. If present, the engine computes from work packages and canonical PoS gates, and emits a pos_gate_breakdown so UI, exports, and agents can see which gates were actually used. Signed exports should disclose the exact label, canonical gate, quantitative treatment, rationale, methodology version, and human overrides.

8. Limitations

• Phase-level success benchmarks still assume independence between canonical gates. Combined protocols may violate that independence, so the methodology discloses the assumption rather than pretending a more granular label solves it.
• Phase 2a/2b conventions vary by sponsor, modality, and indication. Source context and reviewer judgment matter.
• Accelerated approval and pivotal Phase 2 routes need explicit scenario structure. A label alone cannot remove Phase III from the model.
• Regional approvals and bridging studies are not clinical survival gates; their economics belong in geography-specific revenue and launch assumptions.