Table 4 Diagnostic performance of each pancreas CT protocol in predicting margin-negative (R0) resection according to the history of neoadjuvant therapy

Reader 1

AUC

0.739 (0.581, 0.862)

0.842 (0.695, 0.936)

0.157

AUC

0.520 (0.415, 0.799)

0.580 (0.376, 0.766)

0.569

Sensitivity

87.9% (29/33)

84.8% (28/33)

0.999

Sensitivity

55.6% (10/18)

50.0% (9/18)

0.999

Specificity

44.4% (4/9)

77.8% (7/9)

0.250

Specificity

55.6% (5/9)

66.7% (6/9)

0.999

Accuracy

78.6% (33/42)

83.8% (35/42)

0.625

Accuracy

55.6% (15/27)

55.6% (15/27)

0.999

Reader 2

AUC

0.744 (0.586, 0.866)

0.865 (0.724, 0.951)

0.162

AUC

0.596 (0.391, 0.779)

0.512 (0.314, 0.708)

0.383

Sensitivity

93.9% (31/33)

93.9% (31/33)

0.999

Sensitivity

72.2% (13/18)

61.1% (11/18)

0.500

Specificity

33.3% (3/9)

55.6% (5/9)

0.500

Specificity

33.3% (3/9)

44.4% (4/9)

0.999

Accuracy

81.0% (34/42)

85.7% (36/42)

0.500

Accuracy

59.3% (16/27)

55.6% (15/27)

0.999

Reader 3

AUC

0.835 (0.688, 0.931)

0.832 (0.684, 0.929)

0.894

AUC

0.552 (0.350, 0.742)

0.512 (0.314, 0.708)

0.560

Sensitivity

84.8% (28/33)

81.8% (27/33)

0.999

Sensitivity

55.6% (10/18)

33.3% (6/18)

0.125

Specificity

55.6% (5/9)

77.8% (7/9)

0.500

Specificity

33.3% (3/9)

66.7% (6/9)

0.250

Accuracy

78.6% (33/42)

81.0% (34/42)

0.999

Accuracy

48.1% (13/27)

44.4% (12/27)

0.999

Pooled analysis

AUC

0.766 (0.682, 0.837)

0.827 (0.749, 0.888)

0.119

AUC

0.541 (0.427, 0.652)

0.533 (0.419, 0.645)

0.856

Sensitivity

88.9% (82.6%, 95.2%)

86.9% (80.1%, 93.6%)

0.625

Sensitivity

59.3% (45.7%, 72.8%)

48.2% (34.4%, 61.9%)

0.061

Specificity

44.4% (24.4%, 64.5%)

70.4% (52.0%, 88.8%)

0.016

Specificity

44.4% (24.4%, 64.5%)

59.3% (39.5%, 79.1%)

0.219

Accuracy

79.4% (72.2%, 86.5%)

83.3% (76.7%, 89.9%)

0.227

Accuracy

54.3% (43.2%, 65.4%)

51.9% (40.7%, 63.0%)

0.774