Purpose Accurate phenotypes of chronic lung allograft dysfunction (CLAD) are critical to study disease mechanisms and ensure homogenous populations for clinical trials, but the optimal classification method is unknown. A restrictive CLAD phenotype involves parenchymal disease similar to pulmonary fibrosis. The Center for Computer Vision and Imaging Biomarkers at UCLA has established CT based computer aided diagnosis (CAD) scores for pulmonary fibrosis. We applied CT CAD scores to CLAD and compared 2 previously described methods for CLAD classification. Methods The cohort includes 73 lung recipients; 44 no CLAD and 29 with a CT done within 90 days of CLAD onset. The FVC loss method defines restrictive-CLAD (RCLAD) as a ≥20% decline in FVC from baseline at CLAD onset. The spirometric change method uses relative change in FVC divided by relative change in FEV1 from baseline to CLAD onset, and a value >0.5 is classified as restrictive allograft syndrome (RAS). For both methods, when RCLAD or RAS is not established, the default is bronchiolitis obliterans syndrome (BOS). We compared the quantitative CT CAD scores for QLF (fibrosis), QGG (ground glass), and QILD (QLF + QGG), expressed as a percentage of total lung involvement at TLC, between CLAD phenotypes using Dunn's multiple comparisons tests. Results Of the 29 CLAD cases, 15 were classified as RCLAD and 19 as RAS. The classification methods agreed in 21 cases: 13 RCLAD/RAS and 8 BOS/BOS. The methods disagreed in 8 cases: 2 RCLAD/BOS and 6 BOS/RAS. RCLAD was not associated with differences in QLF, QGG, or QILD. However, RAS was associated with significantly increased QLF, QGG, and QILD scores as compared to BOS and to no CLAD (Figure). Conclusion The spirometric change method for CLAD phenotype classification is associated with radiographic parenchymal lung disease, suggesting it is a better method to distinguish CLAD phenotypes. Our study is limited by small sample size and we have not yet incorporated features of air trapping from RV series.
Read full abstract