How flexible can a clinical trial become without compromising the reliability of its results? The ability to evaluate multiple treatments under a single protocol, introduce new treatment arms or discontinue existing ones offers significant opportunities for medicines development. However, it also raises methodological questions that become particularly important when trial results are intended to support a marketing authorisation application.
These issues are addressed in the European Medicines Agency’s (EMA) Reflection paper on methodological aspects in confirmatory platform trials, published on 19 October 2026 and open for public consultation until 31 January 2027.
Developed by the Methodology Working Party and adopted for consultation by the Committee for Medicinal Products for Human Use (CHMP), the document examines the conditions under which platform trials can generate sufficiently robust evidence for regulatory decision-making.
Particular attention is given to randomisation, control groups, blinding, multiplicity and the planning of statistical analyses. The reflection paper remains a draft and is intended to complement, rather than replace, existing regulatory guidance.
These issues are addressed in the European Medicines Agency’s (EMA) Reflection paper on methodological aspects in confirmatory platform trials, published on 19 October 2026 and open for public consultation until 31 January 2027.
Developed by the Methodology Working Party and adopted for consultation by the Committee for Medicinal Products for Human Use (CHMP), the document examines the conditions under which platform trials can generate sufficiently robust evidence for regulatory decision-making.
Particular attention is given to randomisation, control groups, blinding, multiplicity and the planning of statistical analyses. The reflection paper remains a draft and is intended to complement, rather than replace, existing regulatory guidance.
What are platform trials?
In a conventional randomised clinical trial, an investigational treatment is generally compared with a control treatment according to a design established before the trial begins.
Platform trials, by contrast, allow several treatments to be evaluated within a common trial structure. They may include multiple experimental arms, a shared control group and, in some cases, the addition or discontinuation of treatment arms while the trial is ongoing.
Their main potential advantage is the ability to use a common infrastructure to address several clinical questions, rather than organising entirely separate trials for each treatment.
The EMA adopts a sufficiently broad definition to include multi-arm trials in which all treatments start and finish at the same time. Adaptive modifications during the course of the trial are therefore not a prerequisite for a study to qualify as a platform trial.
A distinction must also be made between platform trials and adaptive trials more generally. An adaptive trial design allows prospectively planned modifications based on accumulating trial data, according to predefined rules. Platform trials may incorporate adaptive features, but the two concepts are not interchangeable.
Similarly, a master protocol provides a common framework for coordinating multiple investigational activities. However, this is a broader concept and does not necessarily imply a platform trial.
These distinctions become particularly relevant in the confirmatory setting. When trial results are intended to provide pivotal evidence for regulatory assessment, organisational flexibility must be supported by a statistical design capable of preserving the validity of the conclusions.
Planning Must Come Before Results
One of the central messages of the EMA reflection paper concerns the need to define trial objectives and analysis strategies prospectively.
In platform trials, the presence of multiple treatments and the possibility of modifying the trial structure increase the risk that certain decisions may be influenced by emerging results.
The EMA therefore emphasises the importance of specifying primary objectives, estimands – the treatment effects the trial is intended to estimate – statistical hypotheses and methods for controlling the type I error rate before the relevant data become available.
Where possible, the conditions for adding or discontinuing treatment arms, the data to be used for individual comparisons and the timing of interim and final analyses should also be defined in advance.
The Agency acknowledges that not every aspect can be fully specified before a platform intended to evolve over time is launched. Nevertheless, the fundamental rules should be established prospectively, and subsequent modifications must be appropriately justified.
Insufficiently detailed planning may undermine the confirmatory status of a trial, potentially limiting the role of its findings to exploratory evidence.
Randomisation and the Control of Bias
Randomisation remains essential to achieving reliable comparisons between treatments. In platform trials, however, its implementation can become more complex.
The introduction or discontinuation of treatment arms may affect the probabilities of participants being allocated to different groups. Changes over time in patient characteristics and available treatment options may also influence the comparability of results.
The draft reflection paper expresses a preference for maintaining constant allocation ratios for the comparisons of interest, even when the overall number of treatment arms changes.
In this context, the EMA does not support the use of response-adaptive randomisation, whereby allocation probabilities are modified in response to outcomes observed during the trial.
The concern is methodological: changes in allocation ratios can complicate statistical analyses and make it more difficult to distinguish treatment effects from changes occurring during the trial.
Re-randomising participants who have already entered the trial is also generally discouraged in confirmatory settings, as it may affect the target population and introduce additional statistical dependencies.
The Challenge of Shared Control Groups
Using a common control group is one of the most attractive features of platform trials.
When several experimental treatments are compared against the same control group, it may be possible to reduce the total number of participants required compared with conducting independent trials with separate control groups, while maintaining adequate statistical power for individual treatment comparisons.
However, this approach introduces correlations between statistical comparisons, which must be considered when assessing the risk of erroneous conclusions.
A particularly sensitive issue concerns the use of non-concurrent controls: participants enrolled in the control group during periods when the experimental treatment under evaluation was not yet available for randomisation.
The EMA expresses a strong preference for concurrent controls, as comparisons between participants randomised during the same period, when allocation to the relevant treatment arms was possible, provide stronger support for causal interpretation.
The use of non-concurrent controls may introduce time-related bias, for example through changes in clinical practice, patient characteristics or trial operating conditions.
The document acknowledges that, in specific and adequately justified circumstances, such data may contribute to statistical analyses. However, their inclusion in the primary analysis of the primary endpoint is generally considered unacceptable, whereas they may be more appropriate for supplementary analyses or certain secondary endpoints.
When the control treatment needs to change
Another methodological challenge arises when an experimental treatment produces positive results and consideration is given to using it as the new control for treatment arms that remain active.
A decision that may appear reasonable from a clinical perspective can have significant implications for the statistical interpretation of trial results.
The EMA notes that meeting a primary endpoint does not automatically demonstrate a favourable benefit–risk balance or imply that the medicinal product has received marketing authorisation.
Changing the control treatment therefore requires particular caution. For treatment arms already underway, the draft generally discourages such modifications except in exceptional circumstances.
The situation may differ for new arms that have not yet started, where a documented change in the standard of care could justify adopting a different control from the one originally planned.
This distinction illustrates the importance of considering methodological decisions alongside developments in clinical knowledge and the regulatory environment.
Statistical Multiplicity and the Reliability of Conclusions
When a trial evaluates multiple treatments and tests several hypotheses, controlling statistical errors becomes more complex.
The EMA document pays particular attention to type I errors, which occur when a statistically significant treatment effect is declared despite no true effect being present.
In platform trials, multiple comparisons do not automatically require a single overall control of the family-wise error rate (FWER) across every hypothesis.
According to the draft, the need for joint error control depends on how the hypotheses are related and the regulatory decisions they are intended to support.
Where several comparisons contribute to the same regulatory question, the type I error rate must be considered across the relevant family of hypotheses. Where decisions are genuinely independent, separate approaches may be justified.
A shared control group introduces further considerations. Although it may reduce the probability of at least one false-positive conclusion compared with independent control groups, it can increase the probability of multiple false-positive conclusions occurring together.
The EMA therefore stresses the importance of evaluating the overall statistical strategy and considering additional confirmatory evidence where necessary.
Maintaining Trial Integrity Over Time
In platform trials, analyses of different treatments may take place at different points in time. Results from some arms may therefore become available while recruitment continues in others.
This makes the management of blinding and access to trial data particularly sensitive.
Knowledge of interim or final results from one arm could influence decisions concerning other treatments, potentially compromising trial integrity.
The draft EMA reflection paper highlights the value of independent data monitoring committees, clearly defined rules governing access to information and appropriate separation between those analysing results and those making operational decisions.
In some circumstances, separate analysis teams or coordinated timing of assessments across treatment arms may be necessary.
These requirements call for close collaboration between clinical, biostatistical, operational and regulatory functions.
Implications for medicines development
The EMA’s reflection paper confirms regulatory interest in trial designs capable of evaluating multiple treatments within a common framework. At the same time, it identifies the methodological conditions necessary for their results to support confirmatory regulatory assessments.
For sponsors, designing a platform trial means considering not only recruitment efficiency and operational organisation from the outset, but also the interpretability of the evidence that the trial will ultimately generate.
Defining objectives, managing randomisation, selecting appropriate control groups and planning statistical analyses therefore become central elements of the clinical development strategy.
The reflection paper is currently open for public consultation. Pharmaceutical companies, researchers and other stakeholders can submit comments until 31 January 2027.
The outcome of this consultation will contribute to shaping the European regulatory approach to confirmatory platform trials. For the pharmaceutical industry, the document already provides a useful basis for assessing the methodological choices needed to support reliable regulatory decisions.
Reference
Vuoi trovare più facilmente MakingPharmaIndustry su Google?
Aggiungici alle tue fonti preferite
