Statistical reanalysis of vascular event outcomes in primary and secondary vascular prevention trials

Table 3 Rating of statistical tests

Outcome	Levels^†	Comparator datasets	P	Rating of tests relative to each other
Outcome	Levels^†	Comparator datasets	P	MWU	Adj. OLR	WR	OLR	BS	Adj. MLR	Adj. BLR	Adj. CPH	TT	CAT	CPH	MT	CSB	CSO	CSF
Stroke	3	56	< 0.0001	2	4	3	7	9	10	1	5	13	12	6	11	8	14	15
	4	23	< 0.0001	2	3	1	6	10	12	4	5	11	13	7	8	9	14	15
	5	16	0.0002	2	3	1	6	8	12	4	5	11	13	7	10	9	14	15
	8	12	0.0115	3	2	1	7	8	10	4	5	13	14	6	9	11	12	15
Stroke/TIA	4	35	< 0.0001	1	3	5	2	4	8	10	12	7	6	11	9	13	14	15
	5	17	< 0.0001	1	3	9	7	2	4	10	11	6	5	12	8	13	14	15
	6	13	< 0.0001	1	2	9	3	6	4	10	11	7	8	12	5	13	14	15
	9	12	< 0.0001	1	2	7	3	4	6	11	10	8	9	12	5	14	13	15
MI	3	47	0.010	1	6	12	5	2	7	8	11	4	3	10	14	13	9	15
Bleeding	3	32	< 0.0001	4	9	3	11	7	1	8	6	5	2	10	15	14	14	13
	4	26	0.035	3	7	5	8	2	1	10	9	6	4	11	13	14	12	15
	5	13	0.032	6	8	5	3	2	1	11	7	4	10	12	15	14	9	13
Vascular	3	47	< 0.0001	2	3	5	6	10	8	1	4	11	12	7	13	9	14	15

The numbers in bold represent the tests that are the most efficient and do not differ statistically from one another. The P-value is from the results of the Friedman ANOVA. The order of the rating of test is based on the mean rank calculated by the Duncan’s multiple range test; the most efficient test (i.e. the test with the smallest mean rank) is rated the best with a score of 1 and the least efficient with a score of 15
Abbreviations
Adj.: adjusted; BLR: binary logistic regression; BS: bootstrapping; CAT: Cochran-Armitage trend test; CPH: Cox proportional hazards; CSB: Chi-square binary event outcome; CSF: Chi-square binary fatal event outcome; CSO: Chi-square ordinal event outcome; MLR: multiple linear regression; MT: median test; MWU: Mann-Whitney U test; OLR: ordinal logistic regression; TT: t-test; Vascular: combination of stroke and MI; WR: win ratio test
^†Defined in Supplementary Table 4.

ISSN: 1471-2288