Clinical characteristics, outcomes, and predictive modeling of patients diagnosed with immune checkpoint inhibitor therapy-related pneumonitis.

Purpose: The aim of this study is to better characterize the clinical characteristics and outcomes of patients diagnosed with Immune checkpoint Inhibitor (ICI) pneumonitis and propose predictive models.

Patients and methods: Patients diagnosed with ICI pneumonitis at Mayo Clinic from 2014 to 2022 were studied. All cases were independently reviewed by our pulmonology specialist (A.E.) to confirm the appropriate diagnosis. The grading of pneumonitis was defined in accordance with ASCO guidelines (Schneider et al. in J Clin Oncol 39(36):4073-4126, 2021. https://doi.org/10.1200/JCO.21.01440 ). Predictive modeling was performed using gradient boosting machine learning technology, XGBoost (Chen in 1(4):1, 2015), to conduct binary classification and model reverse engineering using Shapley statistics (Lundberg and Lee in Adv Neural Inf Process Syst 30, 2017).

Results: One hundred and seventy patients with ICI pneumonitis were included (median age 67; IQR 59, 75). Median overall survival was 2.3 years (95% CI: 1.8, NR). A higher grade of ICI pneumonitis was associated with inferior survival (HR 5.85, 95% CI: 2.27, 15.09; p < 0.001). Patients who were rechallenged with immunotherapy had significantly improved hazard of survival compared to patients not rechallenged (HR 0.37, 95% CI: 0.21, 0.68; p = 0.001). Risk of death from ICI pneumonitis prior to starting immunotherapy was modeled with an area under the curve of the receiver operator characteristic (AUC-ROC) of 0.79 with the most contributory features including peripheral blood lymphocyte count, oxygen dependence, pulmonary function testing, and PD-L1 expression.

Conclusion: The presentation of ICI pneumonitis is highly variable, and outcomes are dependent on severity, but favor grade 2 disease when patients are rechallenged with immunotherapy. However, using commonly available clinical data, we can accurately identify patients at high risk of death from ICI pneumonitis. Further effort is needed to produce clinical models able to provide clinician decision support when evaluating patients with ICI toxicities and considering ICI rechallenge.