Points to Consider for Small Clinical Trials — PMDA Early Consideration Summary

In Plain Language

A small clinical trial, as defined in this document, is intended to provide primary evidence about a drug’s efficacy and safety, but has fewer participants than would be needed for adequate statistical power in the primary analysis. The document addresses circumstances where recruiting enough people for a conventional randomized controlled trial (RCT) is difficult, and where a smaller RCT or, in some situations, a single-arm trial may be considered. It asks developers to investigate recruitment realistically, choose a design and methods that fit the disease, drug and available evidence, and set success criteria in advance. The design should still support scientifically valid and interpretable conclusions despite the limited sample size. Sponsors should also be prepared to explain why the design was selected, what it can establish, and what its limitations are, including in advance discussions with PMDA.

What This Guidance Covers

The document concerns clinical trials intended to constitute the primary evidence of a drug’s efficacy and safety and expected to support a marketing authorization application, when the sample size is smaller than that required for adequate statistical power for hypothesis testing in the primary analysis. It focuses on situations where an RCT would ordinarily be considered desirable, but statistical power for a between-group comparison is insufficient or extremely limited.

It describes two broad situations. In one, an RCT is still expected to be feasible, but the number of participants that can be enrolled is likely to leave the between-group comparison underpowered. In the other, even after considering possible design and analysis choices, the power remains so limited that an RCT may be difficult to conduct and a single-arm trial may be considered. The document discusses potential ways to plan such trials and the information sponsors should explain to PMDA.

This is not a blanket endorsement of small trials or of any one design. Whether a small trial can serve as major evidence in a marketing application depends on the disease and drug characteristics and on information from the overall development program, including the small trial. The document strongly recommends advance consultation through PMDA clinical trial consultation meetings about an individual trial and the clinical data package intended to demonstrate sufficient evidence of efficacy and safety.

Why It Matters

The document starts from the general approach of obtaining evidence of efficacy and safety through an RCT, in which the investigational treatment is compared with a control within the same study. Planning such a trial usually involves setting a participant number sufficient to give adequate statistical power for the primary analysis. But a trial may not be feasible to initiate and complete within a reasonable timeframe. The document gives orphan drug development, where the target patient population is limited, and other barriers to enrolling participants as examples.

A small sample makes it more difficult to draw clear conclusions, particularly from a comparison between treatment groups. A single-arm trial presents a different interpretive challenge: without a randomized control, it is in principle open-label and can be affected by bias. The document therefore emphasizes that feasibility constraints do not remove the need for a scientifically valid design, predefined success criteria aligned with the study objectives, and results that can be interpreted appropriately.

Key Recommendations

1. Establish recruitment feasibility before settling on a design. Assess the planned study regions, the number of people with the target disease there, the availability of medical institutions able to participate, and how many patients are likely to meet the inclusion and exclusion criteria. Use this information to estimate the number of participants who can actually be enrolled and assess feasibility. Disease registries may be useful sources for this assessment.

2. Consider an RCT when it is feasible, even if it may be underpowered. If an RCT is expected to be feasible but a sufficient number of participants is difficult to recruit, consider the design and analysis elements described in the document. These may help increase statistical power and make results interpretable. The plan should explain the conclusions the trial can support, including the power and precision of its results.

3. Consider a single-arm trial only in the relevant circumstances. If the power for an RCT comparison remains extremely insufficient after considering design and analysis options, a single-arm trial may be needed. The document says such a trial may be possible where the disease course can be predicted sufficiently accurately and adequate information is available on responses to placebo or standard of care. A comparison with an external control or a clinically meaningful threshold may be appropriate in practice. Any threshold should be based on sources such as prior clinical trials, natural history studies or disease registries, and defined so the drug’s efficacy can be appropriately demonstrated against it. Where a placebo response is expected, for example, the threshold should be set to demonstrate efficacy exceeding that level.

4. Protect the information needed for evaluation. Because small trials have few participants, pay particular attention to minimizing, while maintaining scientific validity, situations where necessary evaluation data cannot be obtained. The document gives intercurrent events and failure to collect data corresponding to the estimand as examples.

5. Choose and prespecify the primary endpoint carefully. It should provide the most clinically relevant and convincing evidence directly related to the primary objective. If the achievable sample size cannot provide adequate power for the originally intended endpoint, a surrogate endpoint, a closely related endpoint or a biomarker evaluation may be considered when efficacy can reasonably be evaluated with it and its surrogacy can be clinically justified. Consider the risk that it may not predict the clinical outcome of interest or be directly weighed against safety, making benefit–risk assessment difficult. Other endpoints may be included as secondary endpoints to support data collection and interpretation. The primary endpoint should be prespecified. When an RCT is difficult and a single-arm trial is conducted, the primary endpoint should be objective, given the potential for bias in the results of subjective endpoints due to expectations regarding the drug under development.

6. Use suitable design and analysis methods with their limitations in view. The document discusses longitudinal data, analyses accounting for factors, significance levels for hypothesis testing, N-of-1 trials, adaptive designs and Bayesian approaches. The considerations for each are described below.

7. Explain the selected plan and alternatives. Present the anticipated enrollment, the feasibility of an RCT, the design and elements considered, and the conclusions the selected trial can support. If an RCT is judged difficult, explain the alternative design considered, why the RCT is not feasible, the alternative’s characteristics and limitations, and the conclusion it can yield. Justify an external control or threshold if one is used. Comparing the advantages and limitations of multiple possible designs may also be useful.

Practical Implications

Interpretation for sponsors and study teams: the document’s sequence makes recruitment evidence an input to design selection, rather than a final operational check. Teams can use the anticipated number of eligible participants, study regions, participating sites and enrollment period to assess whether the planned RCT is realistic. That estimate should then be considered alongside the number needed for adequate power and the likely meaning of the results if enrollment falls short.

Statisticians and clinical teams will need to connect the objective, endpoint and analysis to the evidence the trial is expected to provide. For example, choosing a surrogate endpoint is not simply a way to reduce the participant number: the document ties that choice to whether it can reasonably evaluate efficacy and whether its surrogacy can be clinically justified. Likewise, the document does not present raising the significance level as a general solution to small enrollment: where evidence on efficacy beyond the trial is limited and the overall efficacy assessment relies heavily on it, setting the level above 5% is, in principle, not recommended.

For regulatory and clinical operations teams, the document points to a clear planning task: assemble the rationale in an order that makes the path from feasibility assessment to design choice understandable. If a single-arm design is selected, the team should also make clear how its results will be interpreted and, where relevant, why the external control or threshold is appropriate. The document strongly recommends discussing the individual trial plan and clinical data package with PMDA in advance.

Technical Details

Primary endpoint and data use. When the primary endpoint is continuous, analyzing measured values directly rather than converting them into categories (such as “effective” or “ineffective”) may retain more information and enable evaluation with a smaller sample. The document cautions that this approach should be selected with careful consideration of what change or state in the endpoint is clinically meaningful. Repeated measurements from the same participant are longitudinal data; using them can incorporate more information into the evaluation. The data selection and analysis approach should be prespecified.

Factors and randomization. In an RCT, stratified randomization based on factors considered to influence efficacy, and analyses that account for such factors (including factors used for stratification), may increase power. With few participants, however, stratified randomization may not achieve its purpose sufficiently. Consider the factors’ influence on efficacy and the anticipated participant number when deciding whether to use it.

Significance level. The document describes the general practice for primary between-group hypothesis testing in an RCT as setting the significance level to 5% or less for a two-sided test, to control the probability of a Type I error at 5% or less. A higher level, such as 10%, may be desired in some small trials. However, when evidence on efficacy beyond the trial is limited and the overall efficacy assessment is expected to rely heavily on its results, setting the significance level above 5% is, in principle, not recommended.

N-of-1 trials. An N-of-1 trial is one in which a single patient receives an investigational treatment and placebo or another control in a random order, and the results are evaluated. Results from multiple patients studied in a similar way may also be integrated. The document notes limitations similar to those of crossover trials and says these trials can generally be used where symptoms are stable, treatment effects appear relatively quickly, and effects disappear promptly after treatment ends.

Adaptive designs. An adaptive design allows prospectively planned changes to one or more trial aspects based on interim analysis of accumulating participant data. It can address planning-stage uncertainty by incorporating information obtained during the trial and may increase power for a given expected sample size. For a small trial, which aspects may be modified should be planned in light of the specific efficacy information needed. The document says to refer to discussions on E20 in the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH).

Bayesian approach. This is a broad range of methods that combine a prior distribution, representing pre-existing information, with trial data to derive a posterior distribution of the treatment effect. In small trials, this approach may make it possible to incorporate information outside the planned trial explicitly. The document gives examples: use in adaptive designs, using existing control-group data to reduce the number of participants in a concurrent RCT control group, and quantitative evaluation to support interpretation of a single-arm trial. Use may be considered when the reasons are clear and the resulting conclusions are sufficiently robust. The document identifies considerations including the rationale, appropriateness of the prior distribution, suitability of success criteria based on posterior probabilities, evaluation of operating characteristics (including simulations), and robustness. It is important to discuss a planned Bayesian approach in advance through PMDA clinical trial consultation meetings.

Who Should Read It

This document is relevant to sponsors and clinical development teams planning a trial with limited recruitment prospects, particularly when the trial is intended to provide primary evidence of efficacy and safety. It will also be useful to statisticians developing trial designs and analyses; clinical operations teams estimating enrollment; investigators and medical writers explaining the endpoint and design rationale; and regulatory teams preparing for PMDA consultations or a marketing authorization application. Its considerations are especially pertinent when deciding between an underpowered RCT and a single-arm trial, or when proposing one of the statistical approaches discussed in the document.

Official PMDA Source

• Official title: Points to Consider for Small Clinical Trials (Early Consideration)
• Issuing body: Pharmaceuticals and Medical Devices Agency, Center for Product Evaluation
• Date on cover: March 19, 2026
• PMDA PDF: Points to Consider for Small Clinical Trials (Early Consideration) (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.

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

• Bayesian Assurance Calculator — a calculator resource relevant to planning and evaluating Bayesian trial designs.
• Bayesian Sample Size Using Historical Borrowing (Power Prior) — a resource on sample-size planning with historical information.
• Prior Distributions: Choosing and Justifying Priors — explains prior distributions, a consideration in the document’s discussion of Bayesian methods.
• Understanding ICH E9 Statistical Principles — provides background on statistical principles referenced by the document.