Estimand meaning in one sentence
An estimand is “a precise description of the treatment effect reflecting the clinical question posed by the trial objective.” That is the definition in the ICH E9(R1) addendum, which FDA finalized as guidance for industry in May 2021 and which took effect in the EU on July 30, 2020.
Put simply, the estimand states exactly what number the trial is trying to estimate, before anyone decides how to analyze the data. FDA describes the goal as helping everyone involved agree on the treatment effect of interest, because without that clarity “the reported treatment effect will be misunderstood.”
Why the estimand framework exists
Trials rarely run exactly as planned. Patients stop the drug, take rescue medication, switch therapy or die. Each of these events changes what an outcome measurement means. Before E9(R1), many trials handled them implicitly through missing-data methods, so two teams could analyze the same trial and report two different “treatment effects” without saying they were answering different questions.
The addendum calls these intercurrent events: events after treatment starts that affect either the interpretation or the existence of measurements. Its examples include discontinuing assigned treatment, using additional or alternative therapy, and death. It also clarifies that study withdrawal and administrative censoring are not intercurrent events by themselves; they are missing-data problems.
The attributes of an estimand
The addendum builds an estimand from:
| Attribute | Question it answers | Example |
|---|---|---|
| Treatment | What is compared with what? | Drug X 10 mg vs placebo, both on background therapy |
| Population | In whom? | Adults with type 2 diabetes inadequately controlled on metformin |
| Variable (endpoint) | Measured how, and when? | Change in HbA1c from baseline to week 26 |
| Intercurrent events | How are events such as rescue medication handled? | Treatment policy for discontinuation; hypothetical for rescue therapy |
| Population-level summary | Summarized as what? | Difference in means between arms |
The addendum lists treatment, population, variable and population-level summary as attributes and handles intercurrent events through strategies that shape those definitions. Many teams and training materials present the intercurrent-event handling as a fifth attribute, which is why “five attributes” is a common shorthand.
The five strategies for intercurrent events
| Strategy | What it means | Typical use |
|---|---|---|
| Treatment policy | The event is part of the treatment regimen; use the observed outcome whether or not the event occurred | Effect of being prescribed the drug in practice |
| Hypothetical | Estimate the outcome in a defined scenario where the event would not occur | Effect if rescue medication were not available |
| Composite variable | Build the event into the outcome, for example counting death or discontinuation as failure | Responder analyses, mortality-sensitive endpoints |
| While on treatment | Use only responses observed before the event | Symptom relief while taking the drug |
| Principal stratum | Target the subgroup in which the event would or would not occur | Effect in patients who would tolerate the drug |
A single estimand often combines strategies, one per type of intercurrent event.
Efficacy estimand vs treatment policy estimand
Sponsors often report two estimands side by side. Novo Nordisk’s REDEFINE 1 announcement (December 20, 2024) reported 22.7% weight loss for CagriSema under the trial product estimand, “if all people adhered to treatment”, and 20.4% under the treatment policy estimand, “regardless of treatment adherence.” Same trial, same data, two different questions. Terms such as “efficacy estimand”, “trial product estimand” and “treatment regimen estimand” are sponsor labels, not ICH terms, so always check which strategy sits behind them.
Estimand vs estimate vs estimator vs endpoint
- Estimand: the target, defined by the clinical question.
- Estimator: the analysis method used to compute an estimate of the estimand from trial data.
- Estimate: the number the estimator produces.
- Endpoint (variable): one attribute of the estimand, not the estimand itself.
The addendum also defines sensitivity analysis as analyses that test whether the main estimator’s conclusions hold when its assumptions are violated, and supplementary analyses as additional analyses that generally carry less weight.
Where estimands must appear
E9(R1) says the protocol should explicitly define the primary estimand, and the protocol and analysis plan should pre-specify the main estimator and a suitable sensitivity analysis. Secondary estimands likely to support regulatory decisions need the same treatment.
Protocol templates are catching up. The ICH Assembly adopted ICH M11, the Clinical electronic Structured Harmonised Protocol (CeSHarP), at Step 4 in November 2025. In the EMA-published M11 template (effective in the EU from June 11, 2026), Section 3 is titled “Trial Objectives and Associated Estimands” and asks for an estimand table covering population, treatment, endpoint, population-level summary and intercurrent-event strategies for each primary and secondary objective.
Common weaknesses
- Estimand stated in the protocol but not carried into the statistical analysis plan.
- Intercurrent events listed without a strategy for each one.
- Missing-data methods chosen first, with the estimand reverse-engineered to fit.
- Sponsor labels (“efficacy estimand”) used without defining the strategy.
- Sensitivity analyses that do not target the same estimand as the primary analysis.
- Clinical study report text that describes a different estimand from the one planned.
Checklist: estimand review
- Each primary and key secondary objective has a matching estimand
- Treatment, population, variable and population-level summary are explicit
- Every anticipated intercurrent event has a named strategy and rationale
- Main estimator is aligned with the estimand, not just with the endpoint
- Sensitivity analyses target the same estimand under different assumptions
- Protocol, SAP and CSR use identical estimand wording
- Labels such as “efficacy estimand” are mapped to ICH strategies
How AskDona helps
Estimands drift. The protocol says one thing, the SAP another, and the CSR a third, especially after amendments. AskDona reads the protocol, amendments, SAP and CSR side by side, extracts each estimand’s attributes and intercurrent-event strategies, and flags where documents disagree, with citations to the page and paragraph. Run a sample estimand consistency check across protocol and SAP on a current study. AskDona shows the evidence; your biostatistics and regulatory teams decide.
Editor’s note: Definitions are quoted from the EMA-hosted ICH E9(R1) text and FDA’s guidance page. The REDEFINE 1 figures come from the sponsor’s press release, not a peer-reviewed paper, and are used only to illustrate how two estimands can differ. Several academic reviews of estimand uptake could not be opened during research and are not cited.
FAQ
What is an estimand in simple terms? It is a precise statement of the treatment effect a trial aims to estimate: for whom, compared with what, measured how, and how events such as stopping treatment are handled.
What is the estimand framework? The structured approach in ICH E9(R1) for aligning trial objectives, design, analysis and interpretation around a clearly defined treatment effect.
What is a treatment policy estimand? One that uses observed outcomes regardless of intercurrent events such as treatment discontinuation, reflecting the effect of the treatment policy as assigned.
What is the difference between an estimand and an estimate? The estimand is the target; the estimate is the number produced by the estimator from the data.
Is an endpoint the same as an estimand? No. The endpoint, or variable, is one attribute of the estimand.
Does FDA require estimands? FDA adopted E9(R1) as final guidance in May 2021. Guidance is not binding, but reviewers expect estimands to be defined.
References
All pages opened and checked by October 19, 2026.
| Source | Type | Used for |
|---|---|---|
| ICH E9(R1) addendum (EMA-hosted) | Regulatory guideline | Definitions, attributes, strategies, documentation, EU dates |
| FDA E9(R1) guidance page | Regulator | US status and purpose |
| Novo Nordisk REDEFINE 1 release (via BioSpace) | Sponsor press release | Two-estimand example |
| RQA: ICH Assembly, Nov 2025 | Professional association news | ICH M11 Step 4 |
| ICH M11 CeSHarP template (EMA) | Regulatory template | Section 3 estimand structure, EU effective date |