Skip to main content

Table 3 ROC Curve Analysis for Model Predictions and total B-line numbers in UIP and NSIP

From: Lung ultrasound for assessing disease progression in UIP and NSIP: a comparative study with HRCT and PFT/DLCO

State Variable

Input Variable

AUC

CI (95%)

p-value

Optimal Cut-Off Point

Sensitivity

Specificity

Involvement of 50% of lungs in HRCT

predicted probabilities of model 1

0.991

0.807, 1.000

< 0.001

0.39

84.2%

84.5%

Severe impairment in DLCO value

predicted probabilities of model 2

0.865

0.736, 0.994

< 0.001

0.50

94.7%

84.5%

Crude number of B-line

0.822

0.687, 0.958

0.001

18

22

23

94.7%

84.2%

73.7%

37.5%

62.5%

68.7%

HRCT involvement > 50%

Crude number of B-line

UIP

0.786

0.633,

0.940

< 0.001

19 (18.5)

21

24(23.5)

94.7%

89.5%

73.7%

37.5%

50%

68.7%

NSIP

0.634

0.436,

0.832

0.185

-

-

-

Total number of regions with pleural fragmentation > 2

UIP

0.919

0.824,

1.000

< 0.001

2 (1.5)

68.4%

87.5%

NSIP

0.526

0.314,

0.739

0.808

-

-

-

Severe impairment in DLCO value

Crude number of B-line

UIP

0.822

0.687, 0.958

< 0.001

19(18.5)

21

23(22.5)

24(23.5)

94.7%

89.5%

84.2%

73.7%

37.5%

50%

62.5%

68.7%

NSIP

0.715

0.523,

0.907

0.028

20(19.5)

75%

72.2%

Total number of regions with pleural fragmentation > 2

UIP

0.821

0.671,

0.971

< 0.001

2(1.5)

82.4%

84.6%

NSIP

0.713

0.520,

0.906

0.031

2(1.5)

3(2.5)

58.3%

25%

72.2%

94.4%

  1. AUC: Area Under the Curve, CI: Confidence Interval, DLCO: Diffusing Capacity of the Lung for Carbon Monoxide, HRCT: High-Resolution Computed Tomography, UIP: Usual Interstitial Pneumonia