Outcome pre-specification requires sufficient detail to guard against outcome switching in clinical trials: a case study

Table 1 Results for different outcome definitions of further bleeding in TRIGGER

Time-point	Method of assessment	Definition	Liberal policy (n (%)) (n = 403)	Restrictive policy (n (%)) (n = 533)	Odds ratio^c(95% CI)	P value
In hospital^a	Clinical judgement or visual inspection	Recurrent and persistent bleeding	31 (5.8)	18 (4.5)	0.94 (0.37–2.40)	0.89
In hospital^a	Clinical judgement or visual inspection	Recurrent bleeding only	21 (4.0)	8 (2.0)	0.46 (0.22–0.98)	0.04
In hospital^a	Visual inspection only	Recurrent and persistent bleeding	24 (4.5)	9 (2.2)	0.54 (0.22–1.33)	0.18
In hospital^a	Visual inspection only	Recurrent bleeding only	14 (2.6)	3 (0.7)	0.23 (0.11–0.50)	< 0.001
Day 28^b	Clinical judgement or visual inspection	Recurrent and persistent bleeding	42 (8.2)	27 (6.9)	0.83 (0.50–1.37)	0.47
Day 28^b	Clinical judgement or visual inspection	Recurrent bleeding only	32 (6.3)	17 (4.3)	0.58 (0.39–0.86)	0.007
Day 28^b	Visual inspection only	Recurrent and persistent bleeding	31 (6.1)	13 (3.3)	0.50 (0.32–0.78)	0.002
Day 28^b	Visual inspection only	Recurrent bleeding only	21 (4.1)	7 (1.8)	0.39 (0.20–0.76)	0.006

^aIn hospital: 1 patient was excluded from the analysis because of missing data on further bleeding (this left 532 patients in the liberal policy group and 403 in the restrictive policy group)
^bDay 28: 31 patients were excluded from the analysis because of missing data on further bleeding (this left 512 patients in the liberal policy group and 393 in the restrictive policy group)
^cAnalysis was conducted using generalised estimating equations, with an exchangeable correlation structure within clusters and robust standard errors. The model adjusted for the following covariates: presence of shock; age; the number of co-morbidities; and the presence of coagulation

ISSN: 1745-6215