Table 2. Performance metrics of the genomic and clinical predictors.

The corresponding ROC and calibration curves are represented graphically in Figure 2.

			T/FAC (n=91)	FACx6 (n=87)
Genomic predictor	ROC	AUC	0.711 (95%CI:0.570–0.852)	0.584 (95%CI:0.353–0.815)
		PPV	38% (95%CI:21–56)	9% (95%CI:1–29)
		NPV	88% (95%CI:77–95)	92% (95%CI:83–97)
		Sensitivity	63% (95%CI:38–84)	29% (95%CI:4–71)
		Specificity	72% (95%CI:60–82)	75% (95%CI:64–84)
Clinical predictor	Discrimination (ROC)	AUC	0.89 (95%CI:0.85–0.93)	0.82 (95%CI:0.75–0.89)
	Calibration	p-value	0.21	0.03
		E max	10.5%	15.5%
		E average	4.8%	9.2%

Abbreviations: FAC = 5-Fluorouracil, Doxorubicin and Cyclophosphamide; T/FAC = weekly Paclitaxel and 5-Fluorouracil, Doxorubicin and Cyclophosphamide; AUC = Area Under the ROC Curve, ROC = receiver operating characteristic, CI = confidence interval; E = difference in predicted probabilities and observed frequencies; E max = maximal error, E average= average error.