Statistical Considerations When Planning Phase I Clinical Trials in Oncology - From the Safety Perspective — PMDA Early Consideration Summary

In Plain Language

The document addresses how to assess the safety of dose-escalation plans for Phase I oncology trials, which often study patients because of concerns about strong toxicity. It asks sponsors to explain how the trial will move between doses, decide which dose can be selected, and stop if no dose is appropriate. It also highlights simulations—computer-based evaluations of how a proposed design might behave under different assumed dose-toxicity patterns—as a way to examine safety before a trial begins. The simulations should include realistic situations, including situations where every planned dose is too toxic to select. The goal is to support patient protection and a clear review of the proposed plan, not to prescribe one dose-escalation design for every trial.

What This Guidance Covers

The document focuses on statistical considerations for the safety evaluation of dose-escalation designs in Phase I clinical trials of oncology drugs. In this setting, the purpose of a Phase I trial often includes evaluating tolerability, safety and pharmacokinetics in patients with cancer, as well as investigating and determining dosage and administration. This differs from the general pattern described in the document for drug development outside oncology, where Phase I trials are generally conducted in healthy volunteers and dosage and administration are explored in later Phase II trials with patients.

The discussion is particularly relevant when a clinical trial plan is subject to a 30-day-CTN review. The document explains that when an investigational drug is a new active ingredient, a new route of administration, or a new combination drug, and it is the first time the drug has been administered to Japanese participants, the clinical trial plan is subject to that review. It describes the review as a process intended to prevent healthcare and hygiene hazards before a trial starts. The source refers to a Clinical Trial Notification (CTN) and calls this process the “30-day-CTN Review.”

The document is about evaluating operating characteristics from a safety perspective; it does not set out to compare the general strengths and weaknesses of dose-escalation designs or establish which design should be chosen. Its points are intended for dose-finding in Phase I trials involving Japanese participants. The document cautions that these points may not necessarily apply when planning a Phase I trial in Japan based on dose-finding results obtained overseas. It also notes that the approaches reflect tolerability-evaluation methods at the time of publication and may change in light of new findings and other information.

Why It Matters

In traditional oncology Phase I trials, the maximum tolerated dose (MTD) is understood as the highest dose within an acceptable range of toxicity. The traditional approach described in the document assumes that efficacy and toxicity increase with dose, and that maximum efficacy is expected at the MTD. The MTD is estimated while tolerability is assessed, and the recommended dose for later trials is then determined based on that estimate.

The document notes that a variety of dose-escalation designs and MTD-estimation methods are now used. Their operating characteristics—the way a design behaves across different possible dose-toxicity situations—therefore need to be considered in the context of the proposed trial. A design may select doses, expose participants at particular dose levels, or stop the trial differently depending on the assumed toxicity pattern. PMDA’s review considers whether the plan is taking appropriate measures to protect participants, not just whether a design produces an MTD estimate.

The document also responds to a practical review issue: the information needed to explain why a dose-escalation design is appropriate may not be sufficiently clear in a clinical trial plan. The document builds on the published checklist for inquiries on an initial clinical trial notification by describing statistical factors and considerations that can affect a design’s operating characteristics. It aims to reduce the need for inquiries and responses by clarifying the information and explanations to provide.

Key Recommendations

1. Describe the dose-escalation design in enough detail to understand its decisions. Explain design-specific settings, such as prior distributions and parameters, as well as the number of dose levels and cohort size. State the rules for escalation, staying at a dose, de-escalation and eliminating a dose. The explanation should include the target dose-limiting toxicity (DLT) rate, the boundaries for escalation and de-escalation, and the number of participants with DLT experience relevant to decisions. Tables or flowcharts may be useful.

2. Explain the rules for stopping tolerability evaluation and selecting the MTD. State rules such as the maximum number of participants overall and at each dose, and the cutoff probability used to control overdose. Clearly identify any specific doses with stopping rules that differ from those for other doses. Explain the MTD definition and the minimum number of cases needed to select it, including whether isotonic regression is performed. The document gives examples of MTD definitions, without limiting the definition to these examples: the highest dose not determined to require de-escalation; the highest dose with an estimated DLT rate below the target DLT rate; or the dose with an estimated DLT rate closest to the target DLT rate. If a dose determined to require de-escalation—especially the starting dose—could nevertheless be selected as the MTD, explain why that selection is appropriate from a safety perspective.

3. Report safety-relevant operating-characteristic metrics. The document highlights the proportion of trials selecting each dose as the MTD; the average number of participants at each dose; the average number of participants with DLT at each dose; and the proportion of trials that terminate without selecting any dose as the MTD. For designs that use a target interval of DLT rates, several doses may correspond to the interval. Even so, report the selection proportion for each dose according to the MTD definition, including the selection proportion for doses in the range of excessive toxicity, as well as the participant and DLT averages at each dose.

4. Simulate multiple dose-toxicity scenarios, including excessive toxicity. From a safety perspective, assess what happens when all doses should not be selected as the MTD. The document calls these “excessive-toxicity scenarios.” In such a situation, it is important that the trial can be appropriately terminated without selecting a dose. For example, if an MTD is to be selected only when its estimated DLT rate is 33% or less, a scenario in which the true DLT rate exceeds 33% at every dose is an excessive-toxicity scenario. The document says it is desirable to include a moderate dose-toxicity relationship in which the true DLT rate at the lowest dose is slightly higher than the excessive-toxicity threshold. If the excessive-toxicity scenarios are limited to extreme toxicity or a sharp rise in toxicity, the proportion of trials terminating without an MTD may be overestimated.

5. Assess the design against the document’s three safety-focused aims. The selection proportion for the dose that should be chosen as the MTD should be the highest; the number of participants receiving highly toxic doses should be kept low; and the trial should terminate appropriately when no dose should be selected as the MTD. The review evaluates safety comprehensively, considering the whole trial plan and drug-specific toxicological and clinical perspectives. The document therefore says it is difficult to set general numerical criteria for these aims. For the first and third aims, however, at least the proportion of appropriate decisions—dose selection or trial termination—in each scenario should be the highest.

6. Evaluate design options and changes, not only the basic design. Options that may be used during the trial should be specified in advance and included in the operating-characteristic evaluation. Even if an option was not prespecified, it is necessary to evaluate and consider the operating characteristics when it is added, including whether safety can be ensured. Comparing simulations with and without the option can help show its impact. The document identifies several situations requiring this attention:

   • Changing cohort size or the minimum number of participants evaluated: If accelerated escalation is used and either quantity may change, explain the effect based on simulations. Also clarify the cohort size when participants are added at the same dose. The document illustrates that adding two more participants after one participant has already been enrolled makes three at that dose in total, whereas adding a new cohort of three makes four in total.
   • Adding a dose: Explain the conditions for moving to a dose below the prespecified minimum or to an intermediate dose, and assess the impact through simulations. If the detailed plan is not yet decided, set tentative doses and evaluate both situations in which an added dose can be selected as the MTD and situations in which it should not be selected.
   • Changing a dosing interval or schedule: Explain the change and measures taken, and evaluate its impact on operating characteristics. Possible measures include ending dose escalation before the change and restarting from the starting dose, or suspending escalation before the change and treating the continued escalation as an additional dose. These are examples, not an exhaustive list. The same consideration applies when multiple intervals or schedules are planned and information from different regimens is combined to make escalation decisions.
   • Using a backfill cohort: Backfilling means adding participants at a dose after it has been confirmed as tolerable and escalation from that dose has been decided, to gather more information for determining or optimizing the recommended dose while escalation continues. If DLT information from those additional participants will inform escalation decisions, explain the impact of the settings using simulations. If a maximum number of additional participants is fixed but the actual number may vary, clarify the simulation settings.
   • Moving from monotherapy dose-finding to combination therapy: Evaluate the planned combination-therapy dose-escalation design by simulation when possible at the time the clinical trial notification is submitted, even if the detailed plan is not yet decided. If the final design differs from preliminary assumptions, further investigation of the impact on operating characteristics will be needed, with efforts to ensure safety. When the doses or schedules of both drugs are considered together, rather than fixing one drug’s dose while escalating the other, consider the necessary scenarios based on the order of the combinations and explain the simulated impact.

7. Explain the safety case in the context of the whole plan. The document says it is also useful to consider the pros and cons of the selected design compared with other dose-escalation designs, such as the 3+3 design, and to explain what measures can be taken to ensure safety if there are safety-related disadvantages. It does not establish a general preference for one design.

Practical Implications

Interpretation: For a sponsor, these considerations make it useful to treat the simulation plan as part of the explanation of the trial design, rather than as a separate technical appendix that cannot be connected to actual trial decisions. The results are most informative when readers can see which assumptions produced them, what the design would do under each scenario, and how the outcomes relate to participant exposure and stopping decisions. The document does not prescribe a universal set of numerical pass/fail thresholds.

Interpretation: Statisticians and trial designers may need to map each design rule to its simulation counterpart. For example, cohort-size changes, additional doses, schedule changes and backfill rules should not be treated as operational details outside the design if they can influence escalation, MTD selection or stopping. The document’s emphasis is on evaluating the impact of these options, including by comparing scenarios with and without an option where useful.

Interpretation: Clinical operations and investigators can contribute by identifying which adaptations might realistically arise during conduct, how they would be implemented, and what participant information would inform a decision. Regulatory teams can use the document’s requested explanations as a way to check whether the plan clearly describes its escalation and stopping rules and whether the simulation scenarios represent the planned trial. These are practical readings of the document’s points, not additional PMDA requirements.

The document also says that where a specific dose-escalation design needs discussion, a pre-Phase I consultation meeting at PMDA may be used. It does not assume that design selection should be based solely on MTD-selection performance or safety assurance; it notes that selection for an individual trial is considered in light of development status.

Technical Details

The document’s simulation logic can be understood as a sequence of linked questions. This is an explanatory outline of the considerations it describes, not a separate PMDA algorithm:

The MTD definition is a consequential design setting because different definitions can select different doses from the same estimated toxicity pattern. Isotonic regression, if used, is a method for estimating toxicity rates across ordered dose levels while preserving their non-decreasing order. The document specifically asks sponsors to explain whether it is used; it does not state that every design must use it.

When the design is based on a target interval for the DLT rate, a group of doses may fall within that interval. The document nevertheless calls for dose-by-dose reporting: a summary that combines all doses in the interval may not show how often a dose in a range of excessive toxicity is selected under the specified MTD definition. Likewise, average enrollment and average DLT counts at each dose help describe participant exposure, alongside the probability of selecting or stopping.

Who Should Read It

This document is most relevant to sponsors planning oncology Phase I trials involving Japanese participants, particularly when the plan is subject to a 30-day-CTN review. It will also be useful to biostatisticians and clinical pharmacologists designing dose-escalation simulations; regulatory affairs staff preparing or reviewing a clinical trial notification; and clinical operations teams and investigators who need to understand how escalation, stopping and possible adaptations are planned. Researchers considering monotherapy-to-combination dose-finding may also find its discussion of preliminary assumptions and later design changes relevant.

Official PMDA Source

• Official title: Statistical Considerations When Planning Phase I Clinical Trials in Oncology - From the Safety Perspective （Early Consideration）
• Issuing body: Pharmaceuticals and Medical Devices Agency; Center for Product Evaluation
• Date on cover: December 4, 2024
• PMDA PDF: Official PMDA document (PDF)
• PMDA list page: PMDA Early Consideration documents

Source: Pharmaceuticals and Medical Devices Agency (PMDA) website, PMDA Early Consideration documents. This page was created by editing and summarising the PMDA document.

Please consult the official document; this summary does not replace it. PMDA provides the English version for reference purposes, and the Japanese original prevails if the two differ.

Related Clinical Biostats Resources

• Time-to-Event BOIN (TITE-BOIN) Design Calculator: a calculator resource relevant to dose-escalation design methods.
• Modified Toxicity Probability Interval (mTPI) Design Calculator: a tool related to interval-based dose-escalation designs.
• Continual Reassessment Method (CRM) Calculator: a calculator resource for another dose-escalation approach.
• 3+3 Design Simulator: a simulation resource for a design named in the PMDA document as a possible comparison.
• 3+3 Design Explained: an introduction to the 3+3 design, which the document mentions as an example for comparison.
• Maximum Tolerated Dose (MTD) Determination: a resource related to the document’s discussion of MTD definitions and selection.
• 3+3 Design Probabilities Table: a table resource for examining dose-escalation probabilities in the 3+3 design.