Interobserver variability in the determination of upper lobe-predominant emphysema

Background: Appropriateness for lung volume reduction surgery is often determined based on the results of high-resolution CT (HRCT) scanning of the chest. At many centers, radiologists and pulmonary physicians both review the images, but the agreement between readers from these specialties is not known. Methods: Two thoracic radiologists and three pulmonologists retrospectively reviewed the HRCT scans of 30 patients with emphysema involved in two clinical studies at our institution. Each reader assigned an emphysema severity score and assessed upper lobe predominance, using a methodology similar to that of the National Emphysema Treatment Trial. In addition, the percentage of emphysema at -910 Hounsfield units was objectively determined by density mask analysis. Results: For the emphysema severity scores, (Spearman) correlation between readers ranged from 0.59 (p = 0.0005) to 0.87 (p < 0.0001), with generally stronger correlations among readers from the same medical specialty. Emphysema severity scores were significantly correlated with prebronchodilator and postbronchodilator spirometry findings, as well as with density mask analysis. In the assessment of upper lobe predominance, K statistics for agreement ranged from 0.20 (p = 0.4) to 0.60 (p = 0.0008). Examining all possible radiologist-pulmonologist pairs, the two readers agreed in their assessments of emphysema distribution in 75% of the comparisons. Readers agreed on upper lobe-predominant disease in 9 of the 10 patients in which regional density mask analysis clearly showed upper lobe predominance. Conclusions: In a group of patients with varying emphysema severity, interobserver agreement in the determination of upper lobe-predominant disease was poor. Agreement between readers tended to be better in cases with clear upper lobe predominance as determined by densitometry.