Based on a retrospective ILD registry cohort, we have evaluated routine biomarkers from PBL and BAL fluid for their association with a set of visually semi-quantified typical HRCT finding scores.
Patient data used for this analysis were retrieved from the ILD registry of Kepler University Hospital Linz, Austria. The registry as well as the present evaluation have been conducted in concordance with the Declaration of Helsinki and were approved and re-assessed on a yearly basis by the ethics committee of the Federal State of Upper Austria (Study number I-26-17). All patients enregistered were subsequently discussed by the monthly local ILD-board after they had undergone a standardized ILD evaluation program including assessment of patient history, physical examination, HRCT imaging, pulmonary function tests and laboratory analyses with standard autoimmune serologies.[19, 20] Patients, in whom ILD board discussion resulted in no diagnosis of an ILD were excluded from this study.
HRCT images were acquired according to protocols suggested by the relevant guidelines.[18, 19, 21] If clinically feasible, prone imaging was preferred in order to differ opacities in dependent lung areas from true interstitial lung abnormalities.
Blood samples were analyzed with a Sysmex® XN-3000 hematology analyzer (Sysmex Europe GmbH, Norderstedt, Germany) for blood cell counts and a Cobas® 8000 modular analyzer (Roche Diagnostics International AG, Rotkreuz, Switzerland) for C-reactive protein (CRP) and lactate dehydrogenase (LDH).
Bronchoalveolar lavage was performed according to the relevant guidelines,[23, 24] when clinically indicated by the treating physician or when requested by the ILD-board. A total of 100 mL of 0,9% saline was instilled and retrieved in aliquots of 20 mL via flexible bronchoscopy under sedoanalgesia. The BAL location was a segmental bronchus of either one of the upper lobes including the lingula or the middle lobe at the discretion of the conducting physician according to the location of most active or extensive disease in HRCT. BAL samples were prepared using 100 µL of BAL fluid on a Tharmac® Cellspin I cytocentrifuge (Tharmac GmbH, Wiesbaden, Germany) at 700 rounds per minute for 5 minutes and Wright Giemsa staining. Cell counting was performed manually under 400-fold magnification, cell fractions were given as % of the total cell count, excluding epithelial cells or erythrocytes.
To allow for statistical analyses of HRCT scans, we have previously devised a semi-quantitative scoring system based on four elementary lesion types: nodular pattern (NDL), reticular abnormalities (interlobular septal and intralobular interstitial thickening and honeycombing – RET), increased lung attenuation (consolidations (CON), ground glass opacities(GGO)) and reduced lung attenuation (emphysema - EMP) findings. Besides, extent of mosaic attenuation (mosaic perfusion, air-trapping - MOS) and traction bronchi(-ol)ectasis (TBR) were assessed. [21, 25, 26]. For quantification, both lungs were separated in an upper-, middle- and lower lung area, as defined by thirds of the largest cranio-caudal diameter in the sagittal reconstructions, leading to six distinct lung areas. For each quantified pattern (RET, TBR, EMP, CON, GGO, NDL, MOS) the individual extent was calculated as the sum of all involved defined lung areas (0–6). The described HRCT scoring process was accomplished during the respective ILD-board session by a specialized ILD radiologist in a non-blinded fashion.
To evaluate the associations between the mentioned inflammation biomarkers with the standardized imaging features, correlation coefficients were calculated for each HRCT finding score and each blood and BAL biomarker. Direction, strength, and significance of these correlations were depicted in color-coded tables for visual analysis. To test for clinical relevance of the associations between HRCT finding scores and the different PBL and BAL biomarkers, groups with no or minimal involvement (0–1), medium (2–5) and extensive involvement (5–6) were compared using the Kruskal-Wallis test. All statistical analyses were performed using R (R: A Language and Environment for Statistical Computing; Version 3.6.0; https://www.R-project.org). For all tests performed, a p-value < 0.05 was regarded statistically significant.