What is a statistical analysis plan?
A statistical analysis plan (SAP) is the document that spells out, in technical detail, how a clinical trial’s data will be analyzed. The protocol states the objectives, design and principal analyses; the SAP expands them into exact definitions, models, handling rules and table shells so the analysis can be reproduced and is not shaped by the results.
ICH E9 sets the basic rules. The SAP “may be written as a separate document” from the protocol, it “should be finalised before breaking the blind,” and “formal records should be kept of when the statistical analysis plan was finalised as well as when the blind was subsequently broken” (ICH E9, Section 5.1). If a blind review suggests changes to principal features in the protocol, those go in a protocol amendment.
Why the SAP matters more in 2026
- Estimands. The ICH E9(R1) addendum expects the protocol and analysis plan to pre-specify the main estimator and a suitable sensitivity analysis for each primary estimand. A modern SAP is organized around estimands, not just endpoints. See our companion article on estimands.
- Multiplicity. FDA’s final Multiple Endpoints in Clinical Trials guidance (October 2022) warns that as endpoints multiply, so does “the likelihood of making false conclusions,” and describes grouping and ordering strategies to control it. The SAP is where that strategy is fixed.
- Transparency. For applicable clinical trials with a primary completion date on or after January 18, 2017, the protocol and SAP must be submitted with results to ClinicalTrials.gov as PDF/A files and are made public. Anyone can compare what you planned with what you reported.
- Structured protocols. ICH M11 (the CeSHarP template) puts estimands in Section 3 and statistical considerations in Section 10 of the protocol, which raises the bar for keeping the SAP consistent with a more structured protocol.
What a statistical analysis plan should contain
The most widely used reference is Gamble et al., “Guidelines for the content of statistical analysis plans in clinical trials” (JAMA, 2017), which reached consensus on a minimum of 55 items after a Delphi survey, an expert meeting, review by 51 statisticians and piloting in five trials. The table groups the typical content.
| Section | What to include |
|---|---|
| Administrative | Title, trial registration number, SAP version and date, protocol version it is based on, roles and signatures, revision history |
| Introduction | Background, objectives, and the estimand for each primary and key secondary objective |
| Trial methods | Design, randomization, blinding, sample size, interim analyses and stopping rules, timing of final analysis |
| Populations | Analysis sets (for example full analysis set, per-protocol, safety) with exact definitions |
| Outcomes | Definitions, derivation rules, visit windows, timing |
| Statistical methods | Models, covariates, handling of intercurrent events, missing data approach, sensitivity and supplementary analyses, subgroup analyses, multiplicity control |
| Safety | Adverse event coding, summaries, laboratory analyses |
| Outputs | Table, listing and figure shells; software and versions |
| References | Data management plan, trial master file references, standard operating procedures |
Oncology-specific points
Oncology SAPs typically need extra detail on time-to-event endpoints (censoring rules for progression-free and overall survival), tumor response assessment criteria, treatment switching and crossover, and interim efficacy analyses. These are exactly the places where intercurrent events, such as starting a new anticancer therapy, change the estimand, so state the strategy explicitly.
How to write a statistical analysis plan: a practical sequence
- Start from the final protocol and list every objective, endpoint and estimand.
- Draft analysis-set definitions and outcome derivations with the data management team, using the actual CRF and data standards.
- Specify the main estimator, sensitivity and supplementary analyses for each estimand.
- Fix the multiplicity strategy and interim analysis rules.
- Build table and figure shells and check they cover every pre-specified analysis.
- Review with clinical, regulatory and data management, then approve and version-control.
- Finalize before database lock and unblinding, and keep the dated record ICH E9 asks for.
- Document any post-finalization change, the reason and whether it was made with knowledge of the data.
Common weaknesses
- SAP based on an outdated protocol version after amendments.
- Estimands in the protocol but only endpoints in the SAP.
- Vague missing-data rules (“appropriate methods will be used”).
- Sensitivity analyses that test a different question from the primary analysis.
- Multiplicity strategy covering primary but not key secondary endpoints that will appear in labeling claims.
- No dated finalization record, making it hard to show the plan preceded unblinding.
- Clinical study report that reports analyses not in the SAP without labeling them post hoc.
Checklist: SAP review before finalization
- SAP references the current protocol version and all amendments
- Every objective maps to an estimand, endpoint, analysis set and estimator
- Intercurrent events and missing data have explicit rules
- Sensitivity analyses target the same estimand
- Multiplicity control covers all claims-relevant endpoints
- Interim analyses and stopping rules match the protocol and DMC charter
- Table shells cover every pre-specified analysis
- Version history, approvals and finalization date are recorded before unblinding
- Document is ready for PDF/A posting with appropriate redactions
How AskDona helps
A protocol, its amendments, the SAP and the CSR can run to hundreds of pages, and inconsistencies between them are what reviewers and journal editors notice first. AskDona reads these documents together, maps each objective to its estimand, endpoint and planned analysis, and flags mismatches with citations to the exact page. Run a sample protocol–SAP consistency review on a study approaching database lock. AskDona shows the evidence; your biostatistics team decides.
Editor’s note: ICH E9 wording is quoted from the EMA-hosted step 5 text; its glossary and Section 7 were not fully readable in the version accessed. The Gamble guideline details come from the published abstract in the University of Liverpool repository; the full 55-item checklist should be consulted directly. Oncology points reflect common practice rather than a single cited guidance.
FAQ
What is a statistical analysis plan in clinical trials? A document that specifies in technical detail how trial data will be analyzed, expanding on the protocol so analyses are pre-specified and reproducible.
When should the SAP be finalized? ICH E9 says before breaking the blind, with formal records of when it was finalized and when the blind was broken.
Is the SAP part of the protocol? It can be, but ICH E9 allows it to be a separate document. Many sponsors keep a short statistical section in the protocol and a detailed separate SAP.
Do SAPs have to be made public? For applicable clinical trials with a primary completion date on or after January 18, 2017, the protocol and SAP are submitted with results to ClinicalTrials.gov and made public.
Is there a standard SAP template? There is no single regulatory template. The Gamble et al. 2017 guideline sets 55 minimum items, and many institutions publish templates based on it.
References
All pages opened and checked by October 19, 2026.
| Source | Type | Used for |
|---|---|---|
| ICH E9 (EMA step 5) | Regulatory guideline | SAP timing and records |
| FDA Multiple Endpoints guidance | Regulator | Multiplicity |
| NLM PRS Document Upload Introduction | Government (NIH/NLM) | 42 CFR 11.48 posting rules |
| Gamble et al., JAMA 2017 (Liverpool repository) | Peer-reviewed guideline | 55 minimum SAP items |
| ICH E9(R1) addendum (EMA) | Regulatory guideline | Estimator and sensitivity pre-specification |
| ICH M11 CeSHarP template (EMA) | Regulatory template | Protocol Sections 3 and 10 |