IPF patients have a median survival of 2-5 years. Diagnosis is based on clinical, histologic, and radiologic findings. Honeycombing is a key predictor of usual interstitial pneumonia on biopsy. We compared the HRCT diagnoses of IPF and of honeycombing by medical center radiologists (MCRs) at 58 sites and those by 3 subspecialized thoracic radiologists (STRs).
In a large trial of IFN-gamma 1b, CT images were evaluated by MCRs for either definite (honeycombing and other criteria) or probable (no honeycombing but other criteria present) IPF for study entry. After study completion, 2 STRs independently evaluated the CT images for consistency with an imaging diagnosis of IPF (in case of a disagreement, the third STR independently read the CT to develop consensus of 2 STRs). The STRs also scored extent of honeycombing. The relationship between baseline CT findings, histology, and clinical findings will be presented.
Of 330 patients, 315 had complete data. All patients had definite or probable IPF according to MCRs. The STRs agreed with the diagnosis of IPF in 283 (90%) patients, assessed the CT as inconsistent with IPF in 30 (10%), and had no consensus in 2 (<1%). Agreement on the presence or absence of honeycombing (n=314) occurred in 266 (85%) patients (Table 1Table 1
HRCT Diagnosis by Absence/Presence of HoneycombingHoneycombingSTRsSeen*Not seen*MCRsSeen251 (79.9%)12 (3.8%)Not seen36 (11.5%)15 (4.8%)*
By both STRs
Simple kappa coefficient testing for agreement between MCRs and STRs = 0.31).
While agreement between MCRs and STRs for honeycombing was moderate, 90% of cases diagnosed as IPF by the MCRs were assessed as consistent with IPF by the STRs.
In most cases MCRs and STRs agree on the imaging diagnosis of IPF. Agreement on the presence or absence of honeycombing was lower, suggesting that assessment of the presence of honeycombing is an area for reader improvement.
J.D. Godwin, Funding provided by InterMune, Inc.