Identification of Prognostic Clinical Features in Grade 4 Immune-Related Adverse Events: A Triangulation Study

Area of Science:

• Oncology
• Immunology
• Critical Care Medicine

Background:

• Grade 4 immune-related adverse events (irAEs) are severe complications of immune checkpoint inhibitor therapy.
• Limited data exists on factors predicting poor prognosis in these rare, life-threatening events.
• Systematic research is needed to identify key mortality predictors.

Purpose of the Study:

• To identify clinical features robustly associated with mortality in patients experiencing grade 4 irAEs.
• To address the scarcity of data and rarity of these events.
• To leverage a triangulation approach for reliable feature selection.

Main Methods:

• Employed a
• triangulation
• framework integrating univariate statistics and machine learning.
• Analyzed data from 26 cancer patients admitted to the ICU for grade 4 irAEs.
• Validated features consistently identified by both statistical and machine learning approaches.

Main Results:

• Univariate analysis identified 21 mortality-associated features; machine learning refined this to 11.
• Eight features were consistently validated: inverse association with mortality (body mass index, VEGF-inhibitors) and positive correlation (vasopressor therapy, oxygen therapy, lactate levels, pneumonia, neutrophil percentage).
• Identified key roles of shock, respiratory failure, and inflammation in mortality.

Conclusions:

• Identified 8 routinely available early ICU clinical features associated with mortality in grade 4 irAEs patients.
• Findings highlight the importance of shock, respiratory failure, and inflammation.
• These features may aid early risk stratification and guide future research validation.