Association of baseline lower respiratory tract microbiota composition and bacterial load with the risk of PD-1/PD-L1 inhibitor-associated immune-related pneumonitis: A single-center prospective cohort study.

Immune checkpoint inhibitors improve outcomes in solid tumors, but immune-related pneumonitis (irP) can lead to treatment interruption and even death; current stratification mainly relies on medical history and imaging, lacking reproducible organ-specific biomarkers. This prospective cohort study evaluated the association of baseline lower respiratory tract microbiota features and bacterial load with the risk of PD-1/PD-L1-related irP. Bronchoalveolar lavage was performed before immunotherapy; 16S sequencing derived the Pathogen-Enrichment Score (PES), and quantitative PCR quantified total bacterial load. The primary endpoint was first occurrence of grade ≥2 irP within 6 months. Firth-corrected multivariable logistic regression adjusted for prior thoracic radiotherapy, baseline interstitial lung disease, and recent antibiotic use. Of 294 evaluable cases, 34 developed grades ≥2 irP. Baseline total bacterial load was significantly higher in patients who later developed irP, showing a monotonic dose-response relationship with risk (nonlinearity P = 0.183); adjusted odds ratio = 1.58 per 1 log10 copies/mL (P = 0.024). PES was also elevated (adjusted odds ratio = 1.43 per 1 SD, P = 0.033). The dysbiotic phenotype was characterized by reduced diversity and enrichment of Streptococcus, Veillonella, and Haemophilus, genera commonly associated with oral microbiota and aspiration. Adding PES and bacterial load to a clinical model improved discrimination (AUC 0.81 vs 0.73, P = 0.019) and reclassification (categorical NRI = 0.24, P = 0.004). Event rates across risk strata were 2.24, 13.46, and 32.65%. Baseline lower respiratory tract dysbiosis and elevated bacterial load indicate higher irP risk; adding these microbiologic measures to a clinical model improves prediction and risk stratification.