Statistical methods

All main analyses will be performed on the intention to treat population and followed according to randomisation. The primary analysis and selected secondary efficacy analyses will also be performed on the ‘per protocol’ population. For comparison between the two randomised groups Fisher’s non-parametric permutation test will be used for continuous variables, Mantel-Haenszel Chi² test for ordered categorical variables and Fisher’s exact test for dichotomous variables and Chi² test for non-ordered categorical variables. For dichotomous outcome variables two-sided 95% confidence intervals (CIs) for the difference in proportions between groups will be calculated as well as risk ratios (RRs) with 95% CI. For continuous outcomes two-sided 95% CI for the difference in means between groups will be calculated based on Fisher’s non-parametric permutation test. The main analysis will be the above unadjusted analyses.

We will perform our primary analysis with the composite primary outcome. If group imbalances in important baseline characteristics are found, complementary analyses adjusted for these variables will be performed with analyses of covariance for continuous variables and with generalised linear models with binomial distribution and log link function in order to calculate adjusted RRs with 95% CI for dichotomous variables. The distribution of the variables will be given as mean, standard deviation (SD), median quartiles 1 and 3 (Q1, Q3) for continuous variables and as numbers and percentages for categorical variables. Imputations for missing data will be performed, when applicable. Missing data will be imputed using fully conditional multiple imputation in the main analysis. A sensitivity analysis will be performed with the full analysis dataset without imputation. All tests will be two-sided and conducted at the 5% significance level.