Strategy for intention to treat analysis in randomised trials with missing outcome data.