Cell cycle correlated genes dictate the prognostic power of breast cancer gene lists

BMC Medical Genomics

Table 1 Multivariable Cox proportional hazards analysis for each principal component variable¹.

Data set	Variable	Hazard Ratio	p-value	Hazard Ratio	p-value	Hazard Ratio	p-value
		All tumors		ER positive tumors		ER negative tumors
NKI2
	PC1	1.8	0.0001	2.8	< 0.0001	0.9	0.61
	PC2	0.6	< 0.0001	0.7	0.004	0.7	0.03
	PC3	1.2	0.01	2.1	0.0002	1.1	0.37
	Age	0.9	0.0009	0.9	0.01	0.9	0.02
	Tumor size²	1.8	0.01	2.2	0.009	2.4	0.04
	Grade 2³	2.0	0.13	1.1	0.87
	Grade 3³	2.2	0.09	1.3	0.55
	ER+	0.9	0.67
KJX64/KJ125
	PC1	1.5	0.004	2.2	< 0.0001	0.9	0.74
	PC2	0.6	0.002	0.6	0.002	0.5	0.12
	Grade 2⁴	2.7	0.002	3.0	0.005	1.5	0.54
	Tumor size²	3.2	0.002	4.4	0.002	1.8	0.42
Wang
	PC1	1.8	< 0.0001	1.8	< 0.0001	1.7	0.06
	PC2	1.6	< 0.0001	1.7	0.0003	1.6	0.02
	PC3	2.0	< 0.0001	2.2	< 0.0001	1.5	0.05
	ER+	0.8	0.26

1. Up to three principal component variables (PC1-PC3), each derived from the top ranked probes for a set of correlated genes predictive of metastatic recurrence latencies, were included in the models. Hazard ratios for these PCs represent the change in hazards per standard deviation increase in the variable. A hazard ratio greater than 0 indicates that increasing levels of genes positively correlated with the PC variable are associated with an increased risk of metastasis.
2. Hazard ratios for tumor size represent differences in tumor sizes greater than 2 cm versus tumors less than 2 cm.
3. Hazard ratios for grades are relative to grade 1 tumors. Grade variables were not included in the analysis of ER negative tumors due to the low numbers of tumors representing the various grades.
4. Hazard ratios are for grade 2 versus other grades.

ISSN: 1755-8794