Full Analysis Set (FAS) in clinical trials refers to the analysis population that comes as close as possible to the intention-to-treat (ITT) principle. The aim is to include all randomised study participants who make at least one relevant contribution to the data basis (e.g., receive at least one dose or have at least one efficacy assessment) in the primary efficacy analysis. The FAS approach is intended to reduce bias that can arise when participants are excluded after randomisation.
Distinction: Full Analysis Set, ITT and Per Protocol
The ITT principle generally requires that all randomised participants are analysed according to their original treatment allocation, regardless of protocol deviations or treatment discontinuations. In practice, however, evaluable data are not always available for all individuals. The Full Analysis Set is therefore often a pragmatic implementation of ITT, in which pre-defined minimum criteria determine who is included in the analysis.
This should be distinguished from the per-protocol population, which typically includes only participants who have adhered sufficiently to the protocol. Per-protocol analyses are often used as sensitivity analyses, but they may be more susceptible to selection bias if the reasons for exclusion are related to treatment success.
Why is the Full Analysis Set important?
The definition of the analysis population directly influences the interpretation of efficacy and safety. If the FAS is defined too narrowly, unintended optimism may result because problematic courses (e.g., early discontinuations or missing assessments) are underrepresented. If it is defined too broadly, analyses may be diluted if key outcomes are missing or assignment to treatment arms cannot be meaningfully represented.
In regulatory approval and assessment processes, the traceability of the population definition is crucial. Reviewers expect the criteria to be set before database lock and documented in the statistical analysis plan. In addition, it should be clearly described how missing data are handled (e.g., using appropriate imputation methods or model-based approaches) and which sensitivity analyses assess the robustness of the conclusions.
Typical criteria for defining the FAS
Which minimum criteria apply depends on the indication, endpoint and study design. Participants are often included if they were randomised and received at least one dose of the study medication. In some studies, at least one post-baseline efficacy assessment is additionally required in order to be able to evaluate the primary endpoint at all. It is important that these rules are consistent and justified, and are not adjusted retrospectively to improve results.
Particular attention is required for cases such as incorrect randomisation, missing consent, double randomisation or fundamental violations of inclusion criteria. Such special cases should be clearly addressed in SOPs and in the analysis plan so that the sponsor, CRO, biostatistics and medical writing apply the same logic. In studies with complex endpoints (e.g., time-to-event), it must also be clarified what minimum information is required to correctly represent censoring and events.
FAS in conjunction with data quality and monitoring
In operational conduct, FAS and data quality are closely linked. If key time points are missed or data are collected inconsistently, the proportion of missing endpoints increases, and the FAS definition becomes more practically relevant. Robust data management, clear query processes and a consistent monitoring concept (e.g., risk-based monitoring) help ensure that the primary analysis is based on a stable data foundation.
Even with decentralised elements or remote visits, it must be ensured that outcome assessments are documented in a standardised and auditable manner. Otherwise, the distinction between “participants without an evaluable contribution” and “participants with missing data” can become blurred, which complicates interpretation and may lead to regulatory queries. In practice, early alignment between Clinical Operations, Data Management and Statistics has proven effective to ensure that critical data fields are prioritised.
Handling missing data and sensitivity analyses
Because the Full Analysis Set often reflects nearly all randomised participants, many studies face the question of how to methodologically handle missing efficacy data. Common reasons include study discontinuations, missed visits or incomplete assessments. The chosen method (e.g., multiple imputation, mixed models for repeated measures, or conservative assumptions) should fit the endpoint type and the missing-data mechanisms.
It is important that the primary analysis does not rely solely on implicit assumptions. Sponsors and biostatistics therefore usually pre-define sensitivity analyses to assess whether the result remains stable when alternative assumptions about missing data are applied. In regulatory reviews, such analyses are often a focus, particularly when the proportion of missing data is relevant or when discontinuations are unevenly distributed between treatment arms.
Regulatory classification and documentation requirements
In the EU, clinical trials must be planned, conducted and analysed in accordance with the principles of Good Clinical Practice. This includes defining analysis populations in advance, reporting them transparently in the Clinical Study Report, and discussing the impact of alternative populations. Under EU Regulation 536/2014, consistency between the protocol, statistical analysis plan, dataset definitions and reporting is particularly relevant, as deviations can impair the verifiability of the results.
For benefit–risk assessment, it is often helpful to present FAS and per-protocol results side by side and justify why the primary conclusion is robust. Particularly for non-inferiority or equivalence questions, sensitivity analyses often receive special attention in practice.
FAQ
Does the Full Analysis Set always have to include all randomised participants?
Ideally, the goal is to implement ITT as completely as possible. In practice, however, the FAS may include minimum criteria if a meaningful endpoint analysis is not possible without certain data. These criteria must be defined and justified in advance.
Which analysis is more “conservative”: FAS or per protocol?
That depends on the context. FAS is often considered less prone to bias because exclusions after randomisation are avoided. Per protocol can provide additional perspectives in certain designs, but it is more susceptible to selection effects.
Where is the FAS definition documented?
In the study protocol and, above all, in the statistical analysis plan. In the Clinical Study Report, the population should be clearly described and the participant flow presented transparently.
Regulatory References (Selection)
- EU Regulation (EU) No 536/2014 (Clinical Trials Regulation, CTR)
- ICH E9: Statistical Principles for Clinical Trials
- ICH E6(R3): Guideline for Good Clinical Practice