The evidentiary standard for a study in a small population depends on its clinical context, according to an overview of regulatory guidelines. It groups the available approaches into three broad tasks: maximize the feasible sample size, adapt the trial design to the number of participants available, and include external information, while also examining modeling.
This was a guidance review, not a trial that measured patient outcomes. The authors considered guidance from the ICH, EMA and FDA, searched the selected sources exhaustively, and added a keyword search of the FDA database using stated inclusion and exclusion criteria.
Making a small study count
One route is to make the pool larger. The review points to broader eligibility, a longer period for enrolling participants or collecting follow-up, and wider collaboration as ways to increase the sample a study can realistically assemble. For time-to-event outcomes, observing more events can also improve statistical power.
Another route is to get more information from each participant. The review discusses choosing informative endpoints, measuring baseline characteristics that can be used for adjustment, taking measurements more often, following outcomes over time, and considering alternative trial designs.
Some designs can change elements of a study after interim data are available. But the review notes that the EMA's definition of an adaptive design requires full control of type-I error, the statistical guard against false-positive findings.
The limits of outside information
The review treats external information as a possible part of the evidence package. It also warns that external controls lack randomization and blinding, so the design alone cannot establish that bias is absent.
Bayesian borrowing, a statistical approach that brings outside information into an analysis, is also examined. The review says it generally cannot strictly control type-I error.
Models and combined evidence
Model-informed evidence comes with its own safeguards. The framework described in the review starts with a defined question and context of use, followed by a risk and impact assessment and evaluation through verification, validation and an assessment of whether the model applies to its intended use.
Evidence synthesis offers another route. Meta-analysis can combine clinical studies, produce more precise treatment-effect estimates, assess how much results vary across studies, and help detect adverse events.
But the review finds it unclear how to combine different kinds of evidence quantitatively, including randomized trials and electronic health records. That gap is central to its conclusion that approaches for synthesizing evidence across modalities remain insufficiently structured.
The standard is not one-size-fits-all
The review also puts the evidentiary bar in context. Multiple controlled trials have historically been the norm, but one pivotal trial has sufficed in some settings. It therefore does not present one design as a universal answer; the acceptable standard depends on the clinical context.
The review's unresolved issues are specific: official guidance for Bayesian methods remains limited, and there is no sufficiently structured way to combine evidence from different modalities. Its modeling framework also makes clear that models need to be tested for verification, validation and applicability before their use can be judged.
Because this is an overview of methods and regulatory guidance, it does not supply treatment-effect or risk estimates. It also does not establish that any specific design, model or external-data strategy is superior, or that one approach is acceptable in every setting.
Funding and disclosures
INVENTS received funding under grant agreement 101,136,365 from Horizon Europe, SERI and UKRI Innovate UK under the Horizon Europe Guarantee scheme. The authors declared no competing interests for this work.
Paper data and sources
Original title: Overview of Regulatory Guidelines Related to Methods and Models for Small Populations
Authors: Sluijterman L, Li L, Karlsson K et al.
Journal/Repository: Clinical pharmacology and therapeutics
Status: Peer-reviewed
First online: 2026-10-13
DOI: Not available
Original paper · Full text