Predictive value of serum high‐mobility group box 1 levels for checkpoint inhibitor pneumonitis.

Background and Objective: Checkpoint inhibitor pneumonitis (CIP), caused by the anti‐programmed cell death‐1 (PD‐1)/programmed cell death ligand‐1 (PD‐L1) antibody, can be a fatal adverse event in cancer patients. However, no predictive biomarkers for CIP have been identified. Because high‐mobility group box 1 (HMGB1) can aggravate lung injury and potentially increase the immune response, it was investigated as a predictive blood marker. Methods: Blood samples, prospectively stored before anti‐PD‐1/PD‐L1 monotherapy between December 2015 and October 2020, were obtained at two university hospitals from 87 and 43 non‐small cell lung cancer (NSCLC) patients (discovery and validation cohorts, respectively). We retrospectively evaluated the association of serum HMGB1 levels with the incidence of CIP developed within 3 months of initiating anti‐PD‐1/PD‐L1 therapy. Results: CIP was observed in 9 (10.3%) and 6 (14.0%) patients in the discovery and validation cohorts, respectively. In each cohort, serum HMGB1 levels were significantly and reproducibly higher in patients with CIP. In the discovery cohort, an HMGB1 cut‐off level of 11.24 ng/ml was identified by receiver operating characteristic analysis. CIP incidence in the HMGB1high subgroup was significantly higher than that in the HMGB1low subgroup in the discovery (41.2% vs. 2.9%) and validation cohorts (36.4% vs. 6.3%). In an exploratory pooled analysis, three patients died of grade 5 CIP; a 19.29 ng/ml HMGB1 cut‐off level detected grade 5 CIP with 100% sensitivity and 96.85% specificity. Conclusion: Our results suggest that HMGB1 may be a potential blood marker to predict the development and severity of CIP in NSCLC patients. CIP incidence in the patients with high levels of serum HMGB1 (≥ 11.24 ng/ml) was significantly and reproducibly higher than in those with its low levels (< 11.24 ng/ml) in patients with NSCLC. Our results suggest that HMGB1 may be a potential blood marker for the prediction of CIP. [ABSTRACT FROM AUTHOR]