A comparison of visual and quantitative methods to identify interstitial lung abnormalities

BMC Pulmonary Medicine

Table 2 Comparison of methods for the detection of interstitial lung abnormalities in COPDGene and Framingham Heart Study (FHS) participants

COPDGene	HAAs < 10 %^a	HAAs ≥ 10 %^a	HAAs < 95^th percentile (HAA < 6.44 %)^a	HAAs ≥ 95 percentile (HAA ≥ 6.44 %)^a
No ILA^a	1909	21	1850	80
ILA	158	5	139	24
	Kappa 0.03, p = 0.03		Kappa 0.13, p < 0.0001
Framingham Heart Study	HAAs < 10 %	HAAs ≥ 10 %	HAAs < 95^th percentile	HAAs ≥ 95 percentile
No ILA^a	1987	306	2188	105
ILA	104	46	132	18
	Kappa 0.11, p = <0.0001		Kappa 0.08, p < 0.0001
COPDgene Cohort:
HAAs at 10 %	Sensitivity 3 %	Specificity 99 %	PPV 19 %	NPV 92 %
HAAs at 95^th%	Sensitivity 15 %	Specificity 96 %	PPV 23 %	NPV 93 %
Framingham Cohort:
HAAs at 10 %	Sensitivity 31 %	Specificity 87 %	PPV 13 %	NPV 95 %
HAAs at 95^th%	Sensitivity 12 %	Specificity 95 %	PPV 15 %	NPV 94 %

^aHAAs high attenuation areas (defined by the percentage of the lung occupied by high attenuation areas between −600 and −250 Hounsfield Units) [11]
^bNumber of subjects grouped as “no ILA” that are classified as indeterminate: COPDgene: HAA <10 % - n = 739, HAAs ≥ 10 % - n = 18, HAAs < 95 % – n = 698, HAAs ≥ 95 % – n = 59
^cNumber of subjects grouped as “no ILA” that are classified as indeterminate: Framingham: HAA <10 % - n = 805, HAAs ≥ 10 % - n = 184, HAAs < 95 % – n = 926, HAAs ≥ 95 % – n = 63

ISSN: 1471-2466