Table 5.

 Definitions of “intention to treat” analysis used in protocols and publications

	Protocols (n=30)	Publications (n=33)
Group assignment for analysis
According to original randomisation:	5	20
Explicitly described	5	3
Assumed on basis of participant flow diagram	0	8
Assumed on basis of same participant numbers at baseline and in results	0	9
According to intervention actually received*	2	0
No description	23	13
Exclusion of participants from analysis
Exclude losses to follow-up only	4	13
Exclude data from participants owing to†:	22	18
Inadequate adherence to treatment	19	12
Other protocol violations	6	5
Adverse events or lack of efficacy	10	1
Other reasons	6	1
Unclear reasons for exclusion	4	2

*Regardless of original randomised group assignment.

†Some trials had multiple reasons for exclusion.