The review defines corticosteroid-refractory checkpoint inhibitor pneumonitis (CIP) as persistent or worsening clinical or respiratory manifestations despite approximately 48–72 hours of guideline-concordant systemic glucocorticoid therapy, after appropriate evaluation for infection, tumor progression, and other causes. Investigators noted that ASCO, ESMO, and NCCN use similar response-based approaches but apply slightly different reassessment windows—ASCO and NCCN advising reassessment at ~48 hours and ESMO at ~72 hours. The authors emphasize that refractory disease is distinct from corticosteroid-dependent, recurrent, or chronic CIP, each carrying different therapeutic implications and timelines.
Reported CIP incidence varies by regimen and population; trial-based estimates in NSCLC are lower than some real-world series. The review reports that approximately 10% or more of patients with CIP respond poorly to glucocorticoids and may progress to refractory disease. Mortality in refractory cohorts is consistently higher than in steroid-responsive cases, though absolute rates are heterogeneous across single-center retrospective studies. A multicenter retrospective analysis described an acute, subacute, and chronic temporal framework for CIP, with an apparent shift from inflammatory lesions toward organizing or fibrotic patterns over weeks to months; however, that cohort was not restricted to refractory disease, and the natural history of steroid-refractory CIP specifically remains incompletely defined.
Clinically, patients with corticosteroid-refractory CIP commonly present with dyspnea, cough, fever, and hypoxemia; many develop persistent hypoxemia or acute respiratory failure and more frequently require intensive care and mechanical ventilation than steroid-sensitive patients. High-resolution CT imaging is central to diagnosis and follow-up. Patterns are heterogeneous: organizing pneumonia (OP) is often the most frequent pattern, while diffuse alveolar damage (DAD) is associated with extensive bilateral ground-glass change, consolidation, and worse outcomes. No laboratory marker reliably predicts steroid refractoriness. Some patients show elevated CRP, lactate dehydrogenase, neutrophilia, lymphopenia, or lung injury markers such as KL-6 and surfactant protein D, but their predictive value for glucocorticoid response remains unproven. The review states that tocilizumab has been used in a small number of cases where IL-6 elevation was suspected to play a role, yet response prediction is uncertain.
Bronchoalveolar lavage (BAL) is useful primarily to exclude infection and obtain lower respiratory specimens, particularly in grade ≥2 disease when imaging is atypical or infection cannot be ruled out. BALF cellular patterns are heterogeneous: several studies reported lymphocytosis, but proportions varied and cohort sizes were small. T-cell subsets showed inconsistent skewing—some series reported CD8 predominance, others reported different CD4/CD8 ratios linked to relapse. BALF cytokine profiling yielded discordant results; IL-6, IL-17A, and CXCL10/IP-10 were variably reported as elevated across studies. Single-cell sequencing of BALF in CIP has demonstrated T-cell enrichment and activated cytotoxic phenotypes along with pro-inflammatory monocyte/macrophage expansion. The authors caution that BALF data are best viewed as exploratory mechanistic tools rather than validated diagnostic classifiers for corticosteroid-refractory CIP, given retrospective designs, small samples, variable timing relative to steroid exposure, and assay heterogeneity.
The review frames corticosteroid-refractory CIP as biologically heterogeneous. Candidate mechanisms include sustained activation of tissue-resident T cells (notably CD8+ populations), inflammatory T-cell–myeloid interactions, cytokine network redundancy that can bypass single-cytokine blockade, and innate immune dysregulation. In a subset of patients, failure of epithelial repair and transition toward maladaptive organizing or fibrotic remodeling may perpetuate respiratory dysfunction despite immunosuppression. The authors are careful to state that most mechanistic data compare CIP versus no CIP rather than steroid-sensitive versus steroid-refractory disease, so these pathways should be treated as plausible contributors rather than proven determinants of glucocorticoid resistance.
Available second-line and rescue options include mycophenolate mofetil, infliximab, intravenous immunoglobulin (IVIG), tocilizumab, cyclophosphamide, and calcineurin inhibitors. Evidence for these agents derives predominantly from case reports, series, and retrospective cohorts; randomized controlled trial data are absent. The review highlights that there is no established optimal treatment sequence and that decisions have largely been empirical, guided by local practice and individual patient characteristics. Investigators cited small series reporting benefit with various agents in select patients, but they warned that heterogeneous case selection, concomitant therapies, and lack of comparative data preclude definitive recommendations.
To move beyond empirical escalation, the authors propose a hypothesis-generating framework that maps candidate biological phenotypes to mechanism-informed therapies. The framework distinguishes at least five overlapping phenotypes: lymphocyte-dominant, cytokine-dominant, fulminant acute lung injury/DAD, innate/myeloid-dominant, and organizing/fibrotic-remodeling. Each phenotype suggests different potential therapeutic priorities—for example, lymphocyte-dominant disease might prioritize targeted lymphocyte-directed agents, whereas cytokine-dominant presentations could prompt consideration of cytokine blockade. The review emphasizes that phenotypes may overlap and evolve over time, and that accurate phenotype assignment requires longitudinal, multimodal assessment including imaging, BALF profiling, and systemic biomarkers.
Clinical practice currently relies on guideline-based early steroid use for CIP and timely reassessment within the 48–72 hour window used by ASCO, ESMO, and NCCN. For patients who fail to improve, the review recommends targeted evaluation for infection and other causes before escalation. The authors identify key research priorities: harmonized, prospectively applied definitions of corticosteroid-refractory CIP; longitudinal multimodal phenotyping to capture temporal evolution; development of treatment-predictive biomarkers; and prospective comparative studies to test mechanism-informed interventions. They note that translating the phenotype-to-therapy concept into practice will require these foundational steps.
Limitations and unanswered questions
The review acknowledges substantial evidence gaps: most clinical data are retrospective and single-center, mechanistic studies often lack direct comparison between steroid-sensitive and steroid-refractory populations, BALF and cytokine studies are heterogeneous and small, and no randomized trials exist to define optimal second-line therapy. The authors call for prospective multicenter validation of proposed risk factors and phenotype definitions, standardized timing and methodology for BALF and cytokine assays, and randomized or comparative effectiveness studies to evaluate mechanism-matched treatment strategies. Management, they argue, should move toward identifying the dominant biological process at a given disease stage and matching treatment intensity and mechanism accordingly, but considerable translational and clinical research is required before this approach can be widely implemented.