Table 2.

Cox univariate regression comparing the random forest clustering (RFC) thresholds for the oestrogen receptor (ER) and the progesterone receptor (PR) with both the manual analysis and the 10% cutoff thresholds of the quantitative data

	Manual analysis threshold			10% automated threshold			RFC threshold
	HR	95% CI	P value	HR	95% CI	P value	HR	95% CI	P value
Cohort I (n = 179)
ER
OS	0.259	0.127 to 0.530	0.001	0.154	0.067 to 0.355	0.002	0.235	0.116 to 0.473	0.001
PR
OS	0.307	0.158 to 0.597	0.001	0.361	0.183 to 0.712	0.01	0.386	0.197 to 0.756	0.02
Cohort II (n = 564)
ER
OS	0.62	0.465 to 0.826	0.006	0.68	0.509 to 0.907	0.009	0.631	0.480 to 0.830	0.001
BCSS	0.677	0.494 to 0.928	0.006	0.664	0.492 to 0.898	0.008	0.592	0.445 to 0.787	< 0.001
RFS	0.714	0.540 to 0.943	0.009	0.797	0.603 to 1.054	0.11	0.701	0.538 to 0.914	0.009
PR
OS	0.647	0.483 to 0.867	0.004	0.719	0.538 to 0.961	0.026	0.705	0.524 to 0.947	0.020
BCSS	0.602	0.445 to 0.814	0.001	0.7	0.518 to 0.947	0.020	0.673	0.495 to 0.916	0.012
RFS	0.73	0.549 to 0.972	0.03	0.797	0.603 to 1.053	0.790	0.791	0.593 to 1.054	0.109

HR, hazard ratio; CI, confidence interval; OS, overall survival; BCSS, breast cancer-specific survival; RFS, recurrence-free survival.