Deviation from intention-to-treat analysis alters treatment effect

A recent article in the BMJ by Abraha et al, using large amounts of data from RCTs, has shown that deviation from intention-to-treat analysis (ITT) can lead to apparently larger treatment effects. When compared to standard intention-to-treat analysis there was a significant difference in the apparent treatment effect. When a deviation from ITT was compared to no ITT, there was no significant difference.