Statistical Considerations When Planning Phase I Oncology Trials in Japan (Safety Perspective)

PMDA’s Early Consideration document (dated December 4, 2024) focuses on how sponsors should justify the dose-escalation design in Phase I oncology trials from a patient-safety perspective, especially in the context of Japan’s 30-day Clinical Trial Notification (CTN) review for trials involving first administration to Japanese participants (e.g., new active ingredient/route/combination).

Reference (PDF): Statistical Considerations When Planning Phase I Clinical Trials in Oncology – From the Safety Perspective (Early Consideration).

---

1. Why PMDA focuses on operating characteristics

The document explains that, during the 30-day CTN review, PMDA may need to evaluate whether the planned dose-escalation approach appropriately prevents health hazards, and that this can require reviewing the operating characteristics of the escalation design (often via simulation).

A core message is that sponsors should be able to explain safety assurance even when some trial specifics are not fully decided at initiation, and should consider how any built-in flexibility (adaptive measures) affects safety.

---

2. Minimum statistical description PMDA expects for dose-escalation designs

When describing operating characteristics, the paper asks sponsors to clearly specify (at minimum):

2.1 Escalation rules (including de-escalation and elimination)

Sponsors should explain the criteria for escalation/stay/de-escalation/elimination, including the target DLT rate, boundaries used for escalation vs de-escalation, and DLT-count thresholds. Tables or flowcharts may be helpful.

2.2 Stopping rules for tolerability evaluation

Sponsors should specify rules such as maximum overall sample size, maximum per-dose sample size, and any probability cutoffs used to control overdosing; dose-specific exceptions must be stated explicitly.

2.3 How MTD is defined and selected

The paper expects a clear statement of how the MTD is selected (including minimum data requirements and whether isotonic regression is applied), and notes that MTD definitions can differ (e.g., highest dose not requiring de-escalation; highest dose with estimated DLT below target; dose with estimated DLT closest to target). If there is a possibility that a dose that would normally trigger de-escalation (especially the starting dose) could still be selected as MTD, the sponsor should justify that choice from a safety perspective.

---

3. Safety-focused operating-characteristic metrics to report

From a safety standpoint, the paper highlights reporting these metrics by dose level:

• Proportion of MTD selection (how often each dose is selected as MTD)
• Average number of participants treated at each dose
• Average number of participants experiencing DLT at each dose
• Proportion of trial termination where no dose is selected as MTD

For designs using a target-interval approach, the paper cautions that reporting only an interval may hide important safety behavior; sponsors should still present dose-level selection and dose-level exposure and DLT counts (including within excessively toxic ranges).

---

4. Scenario planning: include excessive-toxicity scenarios

The document asks sponsors to simulate multiple plausible dose–toxicity relationships and explicitly include scenarios in which no dose should be selected as MTD (excessive-toxicity scenarios), because safe termination behavior is critical.
It also notes a practical pitfall: if the assumed excessive-toxicity scenario is unrealistically extreme (e.g., toxicity rises too sharply), termination may look artificially high; a more moderate relationship (e.g., lowest dose slightly above the excessive-toxicity threshold) is desirable.

---

5. Three safety questions PMDA wants your simulations to answer

The paper frames safety evaluation around three key points:

1. The design most often selects the dose that should be selected as MTD.
2. The number of participants exposed to highly toxic doses is kept low.
3. The trial appropriately terminates when all doses are too toxic to select as MTD.

It also emphasizes that PMDA evaluates safety comprehensively (design + toxicology + clinical context), so it is difficult to set one universal numeric threshold; however, sponsors should show that appropriate decisions (dose selection or termination) dominate within each simulated scenario.

---

6. If your protocol allows design changes, simulate those options too

A major practical section explains that if certain operational options may be used, sponsors should pre-specify them and evaluate operating characteristics including those options; even if not pre-specified, adding them later requires considering whether safety can still be ensured.

Examples explicitly called out:

6.1 Changing cohort size / minimum evaluable participants (e.g., accelerated escalation)

If cohort size or minimum evaluable participants may change for accelerated escalation, the impact must be explained via simulation; even details like how many additional participants are added after a first-participant DLT at the starting dose should be clarified.

6.2 Adding doses (lower-than-planned minimum dose or intermediate doses)

If additional doses may be introduced, sponsors should explain transition conditions and simulate the impact. If details are not fixed, set tentative doses and simulate both cases where the added dose could be selected as MTD and where it should not be selected.

6.3 Changing dosing interval or schedule

If dosing interval/schedule may change, sponsors should describe the change, mitigation measures, and simulate the impact; suggested approaches include terminating escalation and restarting from the starting dose, or suspending escalation and treating the new regimen as an additional dose.

6.4 Backfill cohorts

If backfilling is planned and DLT information from backfill participants feeds escalation decisions, the settings and their impact on operating characteristics should be evaluated by simulation; variability in actual backfill sample size should be reflected in simulation assumptions.

6.5 Dose exploration in combination therapy

Even if combination dose-finding details are not fully decided at trial start, the paper recommends evaluating operating characteristics by simulation to the extent possible in advance, and revisiting simulations if the final combination design differs from assumptions. It also highlights the need to consider multiple scenarios when both drugs’ dose/schedule may vary rather than holding one fixed.

---

7. Practical submission template (what to include in your CTN package)

To align with the paper’s expectations, many sponsors will benefit from including:

• A concise design specification: dose levels, cohort size rules, escalation/de-escalation/elimination rules, stopping rules, MTD definition and selection procedure.
• A simulation plan: scenario set (including excessive-toxicity scenarios), number of simulations, decision rules implemented, and any adaptive options simulated (backfill, added doses, schedule change).
• A standardized results table by scenario showing: MTD selection proportions by dose, average treated by dose, average DLT by dose, and termination probability.
• A short safety narrative interpreting results through the three key safety questions and explaining any trade-offs versus simpler designs (e.g., 3+3) and risk-mitigation actions.

---

References (PDF)

• Statistical Considerations When Planning Phase I Clinical Trials in Oncology – From the Safety Perspective (Early Consideration).