An endpoint in a clinical trial is a pre-defined measurement criterion used to assess efficacy, safety, or another study objective. Endpoints form the basis for statistical analysis and the regulatory assessment of a medicinal product or medical device. A precise, pre-specified endpoint definition is a prerequisite for the scientific integrity of a clinical trial and is carefully reviewed by authorities and ethics committees as part of study approval.
Primary and secondary endpoint
Every clinical trial has exactly one primary endpoint—the central efficacy measure on which sample size planning (power calculation) and the confirmatory statistical analysis are based. It determines whether a study answers its main research question. Secondary endpoints complement the primary endpoint and serve to characterise additional efficacy or safety aspects. Their analysis is exploratory in nature and requires adjustment for multiplicity if formal conclusions are to be drawn. Tertiary or exploratory endpoints are analysed without adjustment and are used to generate hypotheses for future studies.
Surrogate endpoints versus clinical endpoints
A clinical endpoint directly measures how a patient feels, functions, or survives—for example, overall survival (OS), symptom-free interval, or quality of life. A surrogate endpoint is a laboratory value or imaging parameter considered a proxy for clinical benefit—for example, blood pressure as a surrogate for cardiovascular events, or progression-free survival (PFS) as a surrogate for OS in oncology. Surrogate endpoints enable shorter studies and faster approvals, but carry the risk that the correlation with clinical benefit is weaker than assumed. The EMA accepts validated surrogate endpoints for accelerated approval pathways, but often requires post-marketing studies with clinical endpoints for confirmation.
Patient-reported outcomes as endpoints
Patient-reported outcomes (PROs) are endpoints reported directly by the patient without interpretation by a physician—for example, pain intensity, fatigue, or quality of life measured using validated instruments such as EQ-5D or PROMIS. PROs are becoming increasingly important in regulatory procedures and health technology assessment (HTA) because they reflect treatment benefit that is relevant from the patient perspective. The EMA has published specific guidelines on the development and validation of PRO instruments. Digital collection via ePRO systems (e.g., via app or tablet) now enables more frequent, less burdensome data collection directly from patients.
Selecting the right endpoint is one of the most consequential decisions in clinical development. An overly broad composite endpoint can mix clinically irrelevant events with true outcomes and bias the study result. Conversely, overly narrow or rare single endpoints require very large sample sizes and long study durations. Early alignment of the endpoint concept with authorities through Scientific Advice or End-of-Phase II meetings reduces the risk that the chosen endpoint will not be accepted in the marketing authorisation procedure.
Composite endpoints
Composite endpoints combine multiple clinical events into a single endpoint—for example, MACE (major adverse cardiovascular events): cardiovascular death, non-fatal myocardial infarction, and stroke. They increase statistical power for rare individual events and reduce the required sample size. At the same time, they impose methodological requirements: the components should be clinically comparable, biologically plausibly related, and similar in frequency and severity. Differences in event rates between components can complicate interpretation of the composite endpoint, as a treatment effect may be driven by only one component. Regulatory authorities therefore always require results to be presented by component.
Regulatory authorities place great emphasis on the pre-specification of all endpoints in the study protocol. The concept of preregistration in public databases (ClinicalTrials.gov, EU CTIS) ensures that primary and secondary endpoints are defined and publicly documented before the study starts. Retrospective endpoint changes without regulatory approval are considered a serious breach of scientific integrity. Under ICH E9(R1), the estimand framework introduced a binding methodological tool that further specifies endpoint definitions and systematically addresses ambiguities in the analysis of treatment discontinuations or concomitant therapies.
Frequently Asked Questions
Can the primary endpoint still be changed from a regulatory perspective after the study has started?
Changing the primary endpoint after study start is considered a substantial amendment and must be submitted to authorities and ethics committees for approval. It is scientifically sensitive because it can raise suspicion of endpoint switching—i.e., retrospectively adapting the research question to observed data. In a blinded setting (without knowledge of treatment assignment), such a change is more acceptable from a regulatory perspective than after unblinding.
What is an estimand under ICH E9(R1), and how is it related to the primary endpoint?
An estimand (per ICH E9(R1)) precisely describes what is to be estimated: the target population, the intervention, the variable (endpoint), the handling of intercurrent events (e.g., treatment discontinuation), and the summary measure. The endpoint is only one component of the estimand. Explicitly defining the estimand before study start is now a regulatory standard and prevents ambiguity in the analysis.
What is the exact methodological difference between an endpoint and a clinical outcome?
The terms are often used synonymously in clinical research, but they are not methodologically identical. An outcome refers to the result or event itself—for example, death, myocardial infarction, or pain reduction. An endpoint is the operational definition of that outcome in the study protocol, including the time point of measurement, the measurement instrument, and the statistical analysis rule. An outcome can therefore give rise to multiple endpoints, depending on how and when it is measured.