Table 4 SSP prediction performance validated against Prosigna classification in the full external clinical series OSLO2-EMIT0 and ABiM.

SSP model vs. Prosigna^a	Accuracy	Accuracy Null	Kappa	Class value	Sensitivity	Specificity	Positive predictive value	Negative predictive value
OSLO2-EMIT0 cohort^b
Subtype, n = 103	0.83 (0.74, 0.89)	0.52	0.73 (0.62, 0.84)	Luminal A	0.85	0.90	0.90	0.85
				Luminal B	0.71	0.93	0.80	0.90
				Basal	0.86	0.99	0.92	0.98
				Her2-enriched	1.00	0.93	0.50	1.00
ROR risk classification, n = 103	0.68 (0.58, 0.77)	0.44	0.50 (0.37, 0.63)	Low	0.74	0.86	0.65	0.90
				Intermediate	0.29	0.90	0.56	0.75
				High	0.91	0.74	0.73	0.91
Binary ROR risk categorization, n = 103	0.82 (0.73, 0.89)	0.56	0.64 (0.49, 0.78)	Low/Intermediate	0.74	0.91	0.91	0.73
				High	0.91	0.74	0.73	0.91
ABiM cohort^c
Subtype, n = 100	0.80 (0.71, 0.87)	0.38	0.72 (0.60, 0.83)	Luminal A	0.82	0.92	0.86	0.89
				Luminal B	0.74	0.86	0.74	0.86
				Basal	0.93	1.00	1.00	0.99
				Her2-enriched	0.79	0.93	0.65	0.96
ROR risk classification, n = 98	0.84 (0.75, 0.90)	0.60	0.70 (0.57, 0.83)	Low	0.75	0.95	0.79	0.94
				Intermediate	0.58	0.95	0.73	0.90
				High	0.95	0.79	0.88	0.91
Binary ROR risk categorization, n = 98	0.89 (0.81, 0.94)	0.60	0.76 (0.63, 0.89)	Low/Intermediate	0.79	0.95	0.91	0.88
				High	0.95	0.79	0.88	0.91

Figures in parentheses represent 95% confidence interval.
^aProsigna reference in the OSLO2-EMIT0 cohort is the clinical test performed on FFPE as prescribed whereas Prosigna reference in the ABiM cohort is Nanostring data (not clinical test) generated in Lund on FFPE material and sent to Nanostring for calculating the Prosigna Subtype and ROR readout.
^bPerformance evaluated in the OSLO2-EMIT0 cohort (all available cases).
^cPerformance evaluated in the ABiM cohort (all available cases).

Quick links

Search