A power calculation16 indicated a target of 53 participants
in each group to detect a standardized effect size of 0.25
(medium) with 0.95% power at a 5% significance level. We
aimed to recruit a sample size of 126 participants after
allowing for 20% loss to follow-up