Purpose: To retrospectively analyze computed tomographic (CT) findings of chronic idiopathic interstitial pneumonia (IIP) and to determine which findings are most helpful for distinguishing IIP from usual interstitial pneumonia (UIP) with univariate and multivariate analyses.
Materials and methods: Institutional review board approval and informed consent were not required for this retrospective review of patient records and images. Two observers working independently and without knowledge of the diagnosis evaluated the extent and distribution of various thin-section CT findings (ground-glass opacity, consolidation, reticulation, and honeycombing) in 92 patients (51 men, 41 women; mean age, 56 years; age range, 29-81 years) with a histologic diagnosis of UIP (n = 20), cellular nonspecific interstitial pneumonia (NSIP) (n = 16), fibrotic NSIP (n = 16), respiratory bronchiolitis-associated interstitial lung disease (RB-ILD) (n = 11), desquamative interstitial pneumonia (DIP) (n = 15), or lymphoid interstitial pneumonia (LIP) (n = 14). Observers used univariate and multivariate statistical analyses to compare their findings with the extent and distribution of UIP.
Results: Observers made the correct diagnosis in 145 (79%) of 184 readings. Multivariate logistic regression analysis showed that the independent findings that distinguished UIP from cellular NSIP were the extent of honeycombing and the most proximal bronchus with traction bronchiectasis (odds ratio, 5.16 and 0.37, respectively); the finding that distinguished UIP from fibrotic NSIP was the extent of honeycombing (odds ratio, 2.10). CT features that distinguished UIP from RB-ILD and DIP included extent of ground-glass opacity (odds ratio, 0.76), thickening of bronchovascular bundles (odds ratio, 1.58), the most proximal bronchus with traction bronchiectasis (odds ratio, 0.22), and the number of segments with traction bronchiectasis (odds ratio, 3.64).
Conclusion: UIP has a characteristic appearance that usually facilitates distinction from other types of chronic IIPs at thin-section CT. The most useful finding when differentiating UIP from NSIP was the extent of honeycombing.
(c) RSNA, 2006.