Statistical Analysis Plan (SAP): A randomized controlled trial of intensive blood pressure control on cardiovascular risk reduction in patients with atrial fibrillation: the CRAFT trial
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
The CRAFT trial is an international, multicenter, randomized controlled, open-label, parallel-group trial with blinded outcome assessment, designed to evaluate whether intensive blood pressure (BP) control can reduce cardiovascular (CV) risks in patients with atrial fibrillation (AF). A total of 1,675 participants with documented AF are randomized to either intensive BP control (target home systolic BP [SBP] <120 mmHg) or standard BP control (target home SBP <135 mmHg). The primary outcome is a hierarchical composite of time to CV death, number and time to first stroke, number and time to first myocardial infarction (MI), and number and time to first heart failure hospitalization (HFH), analyzed using the win-ratio approach. The win ratio and its 95% confidence interval (CI) will be calculated assuming an approximately normal distribution of the log win ratio, with the standard error estimated from the Finkelstein–Schoenfeld Z-statistic. Secondary outcomes include each component of the composite endpoint, all-cause mortality, renal outcomes, quality of life, major bleeding, and other CV events, analyzed using Cox proportional hazards, Poisson, and logistic regression models as appropriate. Sensitivity analyses will use matched win-ratio methods based on CHA[DS[-VASc score components. The planned sample size provides 80% power to detect a win ratio of 1.50 for intensive versus standard BP control after a mean of 3 years of follow-up. Findings from CRAFT will provide robust evidence on optimal BP targets for reducing cardiovascular risk in patients with AF.
Revison Summary
This version contains this updated information.
1) We modified the protocol date was corrected to reflect the finalized version.
2) The method for estimating the 95% confidence interval of the win ratio was revised. Instead of using a bootstrap-based approach, the confidence interval is now calculated under the assumption that the log win ratio follows an approximate normal distribution, with its standard error estimated from the Z-statistic of the Finkelstein-Schoenfeld test.
3) The illustrative figure for the win ratio was updated.
4) The definition and analytical approach for subgroup analyses were updated. A detailed definition of ASCVD was added, and the analysis method was revised to use the win ratio stratified for each subgroup instead of including subgroup–treatment interaction terms in the main model.
5) The approach to handling missing data was revised — the previous plan for multiple imputation and sensitivity analyses was removed, and it was specified that no imputation will be conducted.
6) The matched win ratio analysis was updated to specify that matching is performed using the individual components of the CHA[DS[-VASc score and implemented through a nearest neighbour matching method.
7) No statistical test will be conducted among SAE and AESI.
