Piwi-interacting RNAs as novel prognostic markers in clear cell renal cell carcinomas

Table 2 Receiver-characteristic curve analyses of piRNAs to discriminate between non-malignant vs. malignant tissue and non-metastatic vs. metastatic primary tumor tissue

piRNA	AUC (95 % CI)	P value different to AUC = 0.5	Differentiating ability at the Youden index^a		Overall correct classification (%)
			Sensitivity (95 % CI)	Specificity (95 % CI)
Non-malignant/malignant tissue
piR-30924	0.61 (0.53–0.70)	0.008	75 (65–83)	47 (35–59)	57.3
piR-57125	0.87 (0.81–0.92)	<0.0001	90 (82–95)	79 (69–88)	83.6
piR-38756	0.71 (0.64–0.78)	<0.0001	43 (33–52)	95 (87–99)	58.5
All piRNAs combined^b	0.91 (0.86–0.95)	<0.0001	91 (83–95)	86 (76–93)	85.8
Non-metastatic/metastatic
piR-30924	0.63 (0.54–0.73)	0.028	30 (15–49)	93 (85–98)	71.7
piR-57125	0.68 (0.58–0.77)	0.004	40 (23–59)	92 (84–97)	71.7
piR-38756	0.64 (0.54–0.73)	0.013	77 (58–90)	61 (49–72)	71.7
All piRNAs combined^b	0.76 (0.67–0.84)	<0.0001	73 (54–88)	74 (62–83)	75.5
piR-30924 + piR-57125^c	0.76 (0.67–0.84)	<0.0001	73 (54–88)	74 (62–83)	75.5

Abbreviations: AUC area under the receiver-operating curve, CI confidence interval
^aThe Youden index as a measure of overall diagnostic effectiveness is calculated by (sensitivity + specificity) - 1. When equal weight is given to sensitivity and specificity of a test, the cutoff at the maximum value of this index, which graphically corresponds to the maximum vertical distance between the ROC curve and the diagonal line, is referred to as optimal criterion
^bCalculated by full binary logistic regression
^cCalculated by stepwise binary logistic regression with all three miRNAs (backward likelihood elimination; entry p = 0.05, removal p = 0.10). In case of the differentiation between non-malignant and malignant tissue, the backward elimination did not result in a reduced model

ISSN: 1756-9966