This is a ready-to-use estimand specification for the protocol and statistical analysis plan of a clinical trial, built on the ICH E9(R1) framework. It forces the treatment effect to be defined precisely and in advance, with a chosen strategy for every anticipated intercurrent event, and it makes the data-collection consequences explicit so the estimand is not quietly broken during conduct. Replace every <<FILL: ...>> placeholder. A filled specimen follows. This is educational reference content, not regulatory advice.
Why this template earns its place
Without an estimand, the same data can yield different defensible answers depending on how intercurrent events are handled. E9(R1) removes that ambiguity by requiring the trial to define, before unblinding, the exact treatment effect being estimated. The specification also tells the data-management team which data to chase and the monitoring team which completeness to watch, which is how a statistical definition becomes a data-integrity requirement. See ICH E8(R1) and E9 trial design and statistics.
Header
| Field | Entry |
|---|---|
| Document number | <<FILL: DOC-ID>> |
| Protocol / SAP reference | <<FILL>> |
| Estimand role | Primary / Key secondary (<<FILL>>) |
| Version / date | <<FILL>> |
1. Objective (plain clinical language)
<<FILL: e.g. Does adding drug X to standard of care improve glycemic control in this population?>>
2. The five attributes
| # | Attribute | Specification |
|---|---|---|
| 1 | Treatment | <<FILL: intervention and comparator, dose, background therapy, duration>> |
| 2 | Population | <<FILL: the population by key eligibility criteria>> |
| 3 | Variable (endpoint) | <<FILL: the per-patient measurement, e.g. change from baseline in HbA1c at week 24>> |
| 4 | Intercurrent event handling | See the table in section 3 (one strategy per anticipated event) |
| 5 | Population-level summary | <<FILL: e.g. difference in means, odds ratio, hazard ratio>> |
3. Intercurrent events and chosen strategy
List every anticipated intercurrent event (ICE) and the single strategy chosen for it, with a clinical justification. Strategies: treatment policy, hypothetical, composite, while on treatment, principal stratum.
| Anticipated ICE | Strategy | Clinical justification | Data-collection consequence |
|---|---|---|---|
<<FILL: e.g. discontinuation for adverse event>> | <<FILL>> | <<FILL>> | <<FILL: e.g. collect endpoint at week 24 regardless>> |
<<FILL: e.g. use of rescue medication>> | <<FILL>> | <<FILL>> | <<FILL>> |
<<FILL: e.g. death>> | <<FILL>> | <<FILL>> | <<FILL>> |
4. Estimator and analysis
| Item | Specification |
|---|---|
| Primary estimator (method targeting the estimand) | <<FILL: e.g. MMRM / multiple imputation with stated assumption>> |
| Missing-data assumption, aligned to ICE strategy | <<FILL: e.g. missing at random, consistent with treatment policy>> |
| Analysis population | <<FILL: e.g. Full Analysis Set>> |
| Multiplicity handling | <<FILL: hierarchy / gatekeeping / alpha allocation>> |
5. Sensitivity and supplementary analyses
| Type | Analysis | What it tests |
|---|---|---|
| Sensitivity (same estimand, relaxed assumption) | <<FILL: e.g. missing-not-at-random tipping-point>> | Robustness of the primary estimate |
| Supplementary (additional question) | <<FILL>> | <<FILL>> |
6. Data-integrity consequences (hand-off to operations)
- Data that are critical to this estimand:
<<FILL>>(these become High CtQ data, see the CtQ factor register). - The EDC must not close the casebook on discontinuation where post-ICE data are required.
- Monitoring watches completeness of the estimand-critical data against a numeric tolerance.
Sign-off
| Role | Name | Signature | Date |
|---|---|---|---|
| Clinical lead | <<FILL>> | ||
| Biostatistician | <<FILL>> | ||
| Clinical QA | <<FILL>> |
Filled specimen
Example primary estimand for a phase 3 glycemic-control trial.
Objective: Estimate the effect of drug X 50 mg once daily versus placebo, each added to standard of care, on glycemic control.
Five attributes:
| # | Attribute | Specification |
|---|---|---|
| 1 | Treatment | Drug X 50 mg once daily plus standard of care vs placebo plus standard of care, over 24 weeks |
| 2 | Population | Adults with type 2 diabetes meeting eligibility, HbA1c 7.5 to 11.0 percent at screening |
| 3 | Variable | Change from baseline in HbA1c (percent) at week 24 |
| 4 | ICE handling | See table below |
| 5 | Summary | Difference between arms in mean change from baseline at week 24 |
Intercurrent events:
| ICE | Strategy | Justification | Data-collection consequence |
|---|---|---|---|
| Discontinuation of randomized treatment | Treatment policy | Effect of the regimen as used in practice | Collect week-24 HbA1c regardless of discontinuation |
| Initiation of rescue antihyperglycemic therapy | Treatment policy | Rescue is part of the real-world regimen | Collect week-24 HbA1c regardless of rescue |
| Death | Composite | A meaningful bad outcome; counts as non-response | Record; classified as non-responder |
Estimator: MMRM on change from baseline, FAS, missing at random assumption consistent with treatment policy. Multiplicity: fixed hierarchy, primary then key secondary. Sensitivity: missing-not-at-random tipping-point. Data-integrity consequence: week-24 HbA1c after discontinuation is High CtQ data; the EDC keeps the casebook open past discontinuation; monitoring escalates if the missing rate exceeds 8 percent.
That single specification tells data management to chase week-24 HbA1c even after discontinuation, tells monitoring that week-24 completeness is critical, and tells the statistician exactly what to estimate. It is the line from statistics to data integrity made concrete.
Common inspection findings this template prevents
- An objective stated (“demonstrate superiority”) with no estimand, leaving discontinuation, rescue, and death unhandled.
- Intercurrent events treated as missing data by default, breaking a treatment-policy estimand.
- A missing-data method that does not match the chosen ICE strategy, so the analysis answers a different question.
- The EDC closing the casebook on discontinuation when the estimand needs post-discontinuation data.
How to adapt this template
- Write the objective in plain clinical language first, then derive the five attributes.
- List every anticipated ICE and pick one strategy per event, justified clinically.
- Choose an estimator that targets the estimand and a missing-data assumption that matches the ICE strategy.
- Hand the data-collection consequences to data management and monitoring, and link the CtQ register.