Identification and validation of a threshold for early posttransplant bronchoalveolar fluid hyaluronan that distinguishes lung recipients at risk for chronic lung allograft dysfunction

BACKGROUND: Few tools exist for the early identification of patients at risk for chronic lung allograft dysfunction (CLAD). We previously showed hyaluronan (HA), a matrix molecule that regulates lung inflammation and fibrosis, accumulates in bronchoalveolar lavage fluid (BALF) and blood in CLAD. We aimed to determine if early posttransplant HA elevations inform CLAD risk. METHODS: HA was quantified in 3,080 BALF and 1,323 blood samples collected over the first posttransplant year in 743 adult lung recipients at 5 centers. The relationship between BALF or blood HA and CLAD was assessed using Cox models with a time-dependent binary covariate for "elevated" HA. Potential thresholds for elevated HA were examined using a grid search between the 50th and 85th percentile. The optimal threshold was identified using fit statistics, and the association between the selected threshold and CLAD was internally validated through iterative resampling. A multivariable Cox model using the selected threshold was performed to evaluate the association of elevated HA with CLAD, considering other factors that may influence CLAD risk. RESULTS: BALF HA levels > 19.1 ng/ml (65th percentile) had the largest hazard ratio (HR) for CLAD (HR 1.70, 95% confidence interval [CI] 1.25-1.31; p < 0.001), optimized fit statistics, and demonstrated robust reproducibility. In a multivariable model, the occurrence of BALF HA > 19.1 ng/ml in the first posttransplant year conferred a 66% increase in the hazards for CLAD (adjusted HR 1.66, 95% CI 1.19-2.32; p = 0.003). Blood HA was not significantly associated with CLAD. CONCLUSIONS: We identified and validated a precise threshold for BALF HA in the first post- transplant year that distinguishes patients at increased CLAD risk. J Heart Lung Transplant 2025;44:320-328 (c) 2024 International Society for Heart and Lung Transplantation. All rights are reserved, including those for text and data mining, AI training, and similar technologies.