Table 1.

Likelihood χ² ratios by test and cohort.

	ER+/HER2− (N = 3284)
	Trinary classification			Binary classification
	df	LRχ2	p-value	df	LRχ2	p-value
Univariate models
Oncotype	2	118.0	<0.0001	1	109.87	<0.0001
Prosigna	2	146.9	<0.0001	1	127.31	<0.0001
Mammaprint	1	119.5	<0.0001	1	119.45	<0.0001
Bivariate models
Oncotype + Prosigna	4	177.9	<0.0001	2	164.47	<0.0001
Oncotype + Mammaprint	3	145.7	<0.0001	2	143.34	<0.0001
Prosigna + Mammaprint	3	168.8	<0.0001	2	155.11	<0.0001
Bivariate vs. univariate
Oncotype + Prosigna vs. Oncotype	2	59.97	<0.0001	1	54.60	<0.0001
Oncotype + Mammaprint vs. Oncotype	1	27.78	<0.0001	1	33.48	<0.0001
Prosigna + Oncotype vs. Prosigna	2	31.02	<0.0001	1	37.16	<0.0001
Prosigna + Mammaprint vs. Prosigna	1	21.89	<0.0001	1	27.80	<0.0001
MammaPrint + Oncotype vs. Mammaprint	2	26.28	<0.0001	1	23.89	<0.0001
Mammaprint + Prosigna vs. Mammaprint	2	49.34	<0.0001	1	35.65	<0.0001

LRχ² = likelihood ratio chi-squared value, all models run exiting at 10 years. Likelihood χ² ratios(LRχ²) for univariate(single test) or bivariate(two tests in sequence) derived using 10-year distant metastasis free survival as end point, ER+/HER2+ve cases = all ER+ve/HER2−ve cases (irrespective of nodal status and chemotherapy), ΔLRχ² = change in Likelihood χ² ratio when two tests are used sequentially. Trinary classification: results using results from Oncotype-Dx trained and Prosigna-trained tests categorized as low, intermediate, and high risk, binary classification: results using dichotomous results for all tests, see text for cut-points, ΔLRχ² = change in LRχ² for comparison of 2 tests versus a single test.