From inflammatory phenotypes to clinical decisions: a stratified diagnostic and therapeutic framework for CIP

Checkpoint inhibitor pneumonitis (CIP) is an uncommon but potentially life-threatening immune-related adverse event that may interrupt effective anticancer therapy and require prolonged immunosuppression. Its clinical presentation, radiographic appearance, inflammatory biology, and response to corticosteroids vary substantially, while infection, radiation pneumonitis, tumor progression, pulmonary embolism, and pre-existing interstitial lung disease may produce overlapping findings. Conventional severity grading remains important for communication and triage but does not fully capture this heterogeneity.

This narrative review synthesizes clinical, radiographic, immunological, and treatment-response evidence to develop a framework for phenotype-guided CIP management. We conceptualize CIP as a dynamic pulmonary immune syndrome and organize its assessment across four dimensions: clinical severity, radiographic phenotype, inflammatory and immune phenotype, and corticosteroid responsiveness. Emerging evidence involving CD4+ and CD8+ memory and tissue-resident T cells, Th17.1 and Tfh-like programs, pro-inflammatory monocytes and macrophages, cytokine networks, and humoral responses is linked to clinically relevant questions while remaining hypothesis-generating. Three patterns are emphasized: low-inflammatory-burden, steroid-responsive disease; high-risk progressive disease; and persistent immune-activation or steroid-refractory disease. The framework connects these patterns with diagnostic reassessment, multidisciplinary evaluation, corticosteroid treatment, recognition of treatment failure, immunomodulatory escalation, chronic pulmonary follow-up, and individualized immune checkpoint inhibitor rechallenge.

CIP should not be managed as a single toxicity defined by grade alone. Joint interpretation of clinical trajectory, imaging, inflammatory signals, and corticosteroid response may improve risk stratification and support more proportionate diagnostic and therapeutic decisions. This framework is a clinically grounded heuristic rather than a validated algorithm and requires prospective multicenter evaluation.