Checking randomization

It is often helpful to check that randomization appears to have been done appropriately. Where dates of randomization are available, this can be explored by looking at plots of cumulative accrual over time; one would expect numbers enrolled to each intervention to be similar and for enrolment curves to cross frequently. It can also be informative to look at the distribution of randomizations by day of the week. Here, provided that reasonable numbers of individuals have been randomized, one would expect to see roughly the same numbers randomized to each intervention on any given weekday, and that trials randomizing during normal clinic hours have few, if any, participants enrolled on unexpected days. It is also useful to check that the intervention groups are balanced for important baseline characteristics and within important participant subgroups, but bearing in mind that statistically significant imbalances can occur by chance.