A statistical analysis plan, or SAP, is the document that fixes how a study’s data will be analysed before anyone looks at the results. It turns the statistical intentions in a protocol into a precise, reproducible recipe: the endpoints, the populations, the methods, and the exact tables and figures that will be produced. If a trial is the experiment, the SAP is the pre-registered promise about how you will judge it.
Why write a SAP at all
The value of a SAP is credibility. When the analysis is specified in advance, no one can accuse you of fishing for a favourable result after seeing the data. Regulators expect it, journals increasingly expect it, and a data monitoring committee cannot do its job without one. A good SAP also saves time and money: the programming, the mock tables, and the reporting are all mapped out before the database locks, so the final analysis is execution, not improvisation.
What a statistical analysis plan contains
- Objectives and endpoints: primary, secondary, and exploratory, each tied to a specific analysis.
- Analysis populations: intention-to-treat, per-protocol, safety, and how each is defined.
- Statistical methods: the model for each endpoint, covariates, and how assumptions are checked.
- Handling of missing data: the primary approach and the sensitivity analyses that test it.
- Multiplicity: how the study controls error when there is more than one test.
- Interim analyses: timing, stopping rules, and who sees the data.
- Mock table shells: the empty layout of every table, listing, and figure the analysis will produce.
The international reference for this structure is the ICH E9 guideline on statistical principles for clinical trials, which is why a well-built SAP is often described as being written “to ICH E9.”
When do you need one
Write the SAP after the protocol is settled and before the database is locked or unblinded. For a confirmatory trial it should be final before any comparative analysis is run. If your study has a data monitoring committee, the SAP (or a DMC-specific version of it) needs to exist before the first interim look. Leaving it until after you have seen the data defeats its entire purpose.
A short answer to the common questions
Is a SAP the same as the statistics section of a protocol? No. The protocol states the statistical intent at a high level; the SAP expands it into the full, executable detail, including the table shells.
Who writes it? A statistician, working with the clinical and regulatory team. It is signed off before unblinding.
If you are planning a trial and want the statistics fixed properly before you start, that is exactly the work we do. See our clinical trial services or tell us about your study.