Diagnostic value of bronchoalveolar lavage fluid cytokine and lymphocyte profiles for lower respiratory tract infections in patients with lymphoma

Objective Distinguishing lower respiratory tract infections (LRTIs) from noninfectious pulmonary infiltrates (NPIs) in patients with lymphoma is challenging. We assessed the diagnostic value of bronchoalveolar lavage fluid (BALF) cytokine and lymphocyte profiles (CPs and LPs). Methods This retrospective study included 147 participants: 55 lymphoma patients with LRTIs, 52 with NPIs, and 40 non-lymphoma patients with LRTIs. Fourteen BALF cytokines and ten BALF and peripheral blood lymphocyte measures were assessed by multiparameter flow cytometry. LASSO-selected logistic regression models were developed using five-fold cross-validation. BALF models incorporated CPs and LPs, whereas Base models used routine inflammatory markers, with or without sputum production. Models were compared using DeLong tests, NRI, IDI, and 1,000-resample bootstrap optimism correction. Results LASSO retained BALF IL-8 and CD3 + CD4 + for BALF models and CRP and SAA for Base models. Test-set AUCs were 0.845 and 0.853 for BALF-A and BALF-B, vs. 0.973 and 0.969 for Base-A and Base-B, respectively. Base models significantly outperformed corresponding BALF models. Optimism-corrected AUCs were 0.872 for both BALF models and 0.980 and 0.977 for Base-A and Base-B. Adding sputum production improved BALF-model reclassification without significantly increasing AUC and provided no benefit to Base models. Eight immune parameters differed by lymphoma status among patients with LRTIs, with additional infection-type-specific patterns. Conclusions BALF IL-8/CD4-based models distinguished LRTIs from NPIs in patients with lymphoma, although CRP/SAA-based models performed better. BALF immune profiling may complement conventional biomarkers and reveal immune signatures associated with lymphoma status and infection type.