Skip to main content

Table 3 Receiver operating characteristic (ROC) curves results obtained for selected biomarkers and logistic regression models

From: CD44, TGM2 and EpCAM as novel plasma markers in endometrial cancer diagnosis

Model

Sensitivity (%)

Specificity (%)

ROC AUC

95% CI

PPV

NPV

Difference in AUC

p-value

EC vs. non-EC

 8-marker

93

83

0.945

0.86–0.99

 5-marker

84

83

0.895

0.79–0.96

0.05

0.11

EC vs. control*

 CD44/EpCAM/TGM2

84

100

0.937

0.84–0.96

100

99.9

 CD44/EpCAM

82

75

0.842

0.73–0.92

0.85

99.9

0.09

0.067

 CD44/TGM2

81

100

0.911

0.8–0.97

100

99.9

0.03

0.209

 EpCAM/TGM2

84

94

0.909

0.79–0.97

3.52

99.9

0.02

0.212

 CD44

49

100

0.834

0.71–0.92

100

99.8

0.11

0.027

 EpCAM

42

95

0.667

0.54–0.78

2.14

99.8

0.3

< 0.001

 TGM2

78

100

0.901

0.79–0.97

100

99.9

0.04

0.206

FIGO1A vs. control

 CD44/TGM2

69

94

0.847

0.68–0.95

2.9

99.9

FIGO1A vs. FIGO1B-4

 Mesothelin/G1

95

78

0.911

0.79–0.98

Endometriosis vs. control

 CD44/TGM2

91

100

0.98

0.85–1.00

100

99.1

EC vs. endometriosis

 ALDH1A1/Midkine

100

73

0.897

0.76–0.97

  1. *prevalence of EC (0.26%) used for calculation of PPV and NPV was based on data from https://gis.cdc.gov/Cancer/USCS/DataViz.html
  2. 8-marker model: ALDH1A1, CA9, CD44, Hepsin, Kallikrein 6, L1CAM, Midkine, TGM2
  3. 5-marker model: ALDH1A1, CD44, EpCAM, Midkine, TGM2