Table 2.

The percentage of simulations in which an intervention effect was observed for neither outcome, one outcome or both outcomes when analysing two outcomes, using a variety of methods to control the FWER

Method	Pairwise correlation between outcomes	Number of outcomes an intervention effect was observed on
		0	1	2
Bonferroni	0	16.1	48.4	35.5
	0.2	18.6	43.2	38.2
	0.4	20.6	37.7	41.7
	0.6	23.4	32.7	43.9
	0.8	26.3	26.3	47.5
Holm	0	16.1	35.6	48.3
	0.2	18.6	31.0	50.4
	0.4	20.6	26.4	53.0
	0.6	23.4	22.0	54.6
	0.8	26.3	16.0	57.7
Hochberg	0	15.1	35.6	49.4
	0.2	17.6	31.0	51.5
	0.4	19.3	26.4	54.3
	0.6	22.0	22.0	56.0
	0.8	24.8	16.1	59.1
Hommel	0	15.1	35.6	49.4
	0.2	17.6	31.0	51.5
	0.4	19.3	26.4	54.3
	0.6	22.0	22.0	56.0
	0.8	24.8	16.1	59.1
Stepdown minP	0.0	23.7	37.5	38.8
	0.2	25.6	33.6	40.8
	0.4	29.6	27.1	43.4
	0.6	32.2	20.2	47.6
	0.8	33.8	13.8	52.4

In these simulations there was missing data in the outcomes (15% in one outcome and 25% in the other outcome). The tables display various degrees of correlation between the outcomes, ranging from no correlation (ρ = 0.0) to strong correlation (ρ = 0.8)