Observer variation and relationship of computed tomography to severity of beryllium disease

Although high resolution computed tomography (HRCT) is commonly used to assess interstitial lung disease (ILD), relatively little is known about interrater reliability and construct validity of HRCT-reported nodules, ground-glass opacity, or other typical findings. We determined the interobserver and intraobserver variability of HRCT findings and correlated HRCT abnormalities with physiologic measures in 57 patients with chronic beryllium disease (CBD). Reliability of HRCT scan measurements were assessed using weighted kappa (K(W)) and intraclass correlation coefficients. We correlated HRCT with spirometry, body plethysmographic lung volumes, diffusing capacity for carbon monoxide (DL(CO)), maximal exercise testing with measurement of arterial blood gases, and bronchoalveolar lavage (BAL). Interobserver agreement for three of the HRCT abnormalities found in CBD was moderate: the K(W) for nodules, septal lines, and ground-glass attenuation were 0.53, 0.44, and 0.53, respectively. Agreement was poor for bronchial wall thickening (K(W) = 0.15). HRCT scores correlated significantly with DL(CO), gas exchange at rest and at maximal exercise, and lung volume. This study demonstrates that HRCT has good interrater reliability and correlates with indices of the severity of granulomatous lung diseases such as CBD.