Statistical Methods

Before analysis, data cleaning of outlying observations was undertaken, and a random 5% sample of source data was reentered and checked. Missing data were not imputed. Analysis was conducted as “intention to treat,” with the participant as the unit of measurement. Baseline characteristics for each of the groups are presented using frequencies and percentages (categorical variables) and means and standard deviations (normally distributed continuous variables) or median and interquartile range (nonnormally distributed variables). For this pilot trial, the primary analysis tested the feasibility of the statistical analysis for a definitive trial. Comparability of groups at baseline was assessed using clinically significant differences. Frequency and IR (with 95% confidence intervals) of device occlusion were used to summarize the impact of the intervention. Primary end points were compared between groups using log-rank test for equality of survivor function and clinically significant differences. Statistical tests were considered significant at the 95% level (P < .05, 2-tailed). Stata (version 15; StataCorp, College Station, Texas) was used for statistical analysis.