Background Immune checkpoint inhibitors (ICIs) can cause adrenal insufficiency (AI) as an uncommon but potentially life-threatening immune-related adverse event (irAE). Early symptoms are often vague and may overlap with cancer-related fatigue or treatment side effects, contributing to diagnostic delays and under recognition. Methods We conducted a systematic review and meta-ethnographic synthesis of 64 published case reports (2015–2024) describing ICI-associated AI. Data on demographics, tumor type, ICI regimen, presenting symptoms, diagnostic process, hormonal and imaging findings, treatment, and outcomes were extracted. Narrative data were synthesized to explore contextual barriers to timely diagnosis. Quality assessment was performed using a modified CARE checklist. Results Fatigue (67.2%), nausea (28.1%), appetite loss (25.0%), and hypotension (21.9%) were the most frequently reported symptoms. Notably, most patients (73.7%) presented with three or more concurrent symptoms, suggesting a pattern of symptom clustering that warrants clinical attention. Despite this, only 40.6% underwent dynamic endocrine testing, and pituitary imaging was frequently normal. Meta-ethnographic analysis revealed five recurrent themes underlying diagnostic delay: nonspecific symptomatology, low clinical suspicion, delayed hormonal testing, misleading imaging, and fragmented specialty care. Hospitalization occurred in 60.9% of cases, and glucocorticoid therapy was initiated in all. Recovery of adrenal function was rare (9.4%). Conclusion ICI-related adrenal insufficiency often presents with multiple nonspecific symptoms that are easily misattributed, delaying diagnosis. Clinicians should maintain a high index of suspicion, particularly when multiple symptoms co-occur in ICI-treated patients, and initiate prompt endocrine evaluation. Standardized diagnostic pathways and interdisciplinary communication are essential to improve recognition and outcomes of this serious irAE. Systematic review registration https://www.crd.york.ac.uk/PROSPERO/view/ , identifier CRD420251035850.
Ohta et al. (2025) studied this question.