============================================================================== STATCLINIC STATISTICAL ANALYSIS PLAN (SAP) TEMPLATE A fill-in-the-blank SAP template for medical research Free tools for every analysis: https://statclinic.net ============================================================================== HOW TO USE THIS TEMPLATE -------------------------- Fill in every [bracketed] placeholder with your own study's details. Finalize and date-lock this document BEFORE data collection begins (or, at minimum, before database lock and unblinding). Any change made after finalization must be documented as a dated, justified amendment -- see Section 12. Full guide with a complete worked example at: https://statclinic.net/blog/statistical-analysis-plan-sap-guide-medical-research STATISTICAL ANALYSIS PLAN -- VERSION [1.0] Study Title: [Full study title] Protocol Number: [If applicable] SAP Finalized Date: [Date -- must be before database lock/unblinding] Prepared By: [Name(s), role(s)] 1. STUDY DESIGN ------------------ [Study type: e.g., two-arm parallel-group randomized controlled trial / prospective cohort / case-control / cross-sectional] [Randomization method and stratification factors, if applicable] [Blinding: none / single / double] 2. PRIMARY OUTCOME --------------------- Variable: [Exact variable name] Definition: [Exact measurement method and unit] Time point: [Exact time of assessment] Derivation: [Formula if calculated, e.g., change score = follow-up - baseline] 3. SECONDARY OUTCOMES ------------------------- List each with the same precision as the primary outcome: (1) [Variable, definition, unit, time point] (2) [Variable, definition, unit, time point] (3) [Variable, definition, unit, time point] Note: Secondary outcomes are exploratory/hypothesis-generating unless the study was specifically powered for them. State this explicitly. 4. SAMPLE SIZE / POWER CALCULATION -------------------------------------- Target sample size: [N per group / total N] Assumptions: effect size = [value], alpha = [value], power = [value] Method/formula or software used: [state] Attrition adjustment: [percentage and resulting inflated N] 5. VARIABLE DEFINITIONS (CODEBOOK) -------------------------------------- For every variable used anywhere in the analysis: Variable | Definition | Coding/Unit | Type (Nominal/Ordinal/Continuous) [repeat row for every variable: outcomes, predictors, covariates, subgroup variables] 6. ANALYSIS POPULATION -------------------------- Primary population: [Intention-to-treat / Per-protocol / As-treated -- state which is primary] Definitions: [Exact criteria for inclusion in each population] 7. STATISTICAL METHODS -------------------------- Primary analysis: Test/model: [exact test or model name] Covariates (if adjusted): [list] Significance threshold: [alpha, one- or two-tailed] Secondary outcome analyses (list each): Outcome 1: [test/model] Outcome 2: [test/model] Outcome 3: [test/model] Decision rule for parametric vs non-parametric choice (if not yet known): [e.g., "t-test if Shapiro-Wilk p > 0.05, otherwise Mann-Whitney U"] 8. MISSING DATA STRATEGY ---------------------------- Expected missingness: [percentage, based on pilot data or prior studies] Assumed mechanism: [MCAR / MAR / MNAR -- state justification] Primary handling method: [e.g., multiple imputation, m = 20, chained equations] Sensitivity check: [e.g., complete-case analysis for comparison] 9. OUTLIER HANDLING STRATEGY --------------------------------- Detection rule: [quantitative definition, e.g., beyond 1.5x IQR from nearest quartile] Verification process: [how flagged values are checked against source data] Action: [correct confirmed errors; retain genuine extreme values] Sensitivity check: [e.g., Winsorization at 1st/99th percentile] 10. MULTIPLICITY / MULTIPLE COMPARISONS -------------------------------------------- Primary outcome: [no correction needed if single pre-specified outcome] Secondary/subgroup outcomes: [correction method if any, e.g., Bonferroni, Benjamini-Hochberg FDR -- or explicit statement that none is applied and results are exploratory] 11. SUBGROUP ANALYSES -------------------------- Pre-specified subgroups (list each, 3-5 recommended, each clinically justified): (1) [Subgroup variable and categories] (2) [Subgroup variable and categories] (3) [Subgroup variable and categories] Method: [e.g., treatment-by-subgroup interaction term in primary model] Interpretation: [state explicitly as hypothesis-generating] 12. SENSITIVITY ANALYSES ----------------------------- List each pre-specified sensitivity analysis and its purpose: (1) [e.g., per-protocol population instead of intention-to-treat] (2) [e.g., complete-case analysis instead of imputed data] (3) [e.g., unadjusted model instead of adjusted model] (4) [e.g., outliers excluded/Winsorized instead of retained] 13. SOFTWARE ---------------- Primary software: [Name, version, manufacturer] Additional software (if any): [Name, version, manufacturer, and which analysis it performs] 14. SAP AMENDMENT LOG -------------------------- Any change made after this SAP's finalization date must be logged here: Date | Section Changed | Description | Justification | Made Before/After Unblinding [repeat row for each amendment] ============================================================================== PRE-FINALIZATION CHECKLIST ============================================================================== [ ] Primary outcome precisely defined (variable, unit, time point, formula) [ ] Secondary outcomes listed with the same precision, marked exploratory [ ] Every variable used anywhere has a codebook entry [ ] Statistical method specified for every outcome, including decision rules where the test cannot yet be named [ ] Missing data mechanism assumed and handling method pre-specified [ ] Outlier detection rule is quantitative, not subjective [ ] Multiple comparisons approach stated explicitly [ ] Subgroup analyses limited to a small, clinically justified list [ ] Sensitivity analyses pre-specified, not added after seeing results [ ] Software, version, and manufacturer named [ ] This document is dated and finalized BEFORE database lock/unblinding ============================================================================== Full guide with a complete worked example, common mistakes, and FAQ: https://statclinic.net/blog/statistical-analysis-plan-sap-guide-medical-research Run your analysis free at: https://statclinic.net ==============================================================================