Abstract Rationale Checkpoint-inhibitor pneumonitis (CIP) remains a leading cause of ICU utilization, and outcomes are poorest in the steroid-refractory subset. Diagnosis is often delayed because clinical and radiographic features overlap with infection and radiation injury, leading to unnecessary antibiotics, treatment interruptions, and late escalation of immunosuppression. Circulating extracellular vesicle (EV) microRNAs (miRNAs) are stable, biologically plausible readouts of lung immune-injury and are amenable to standardized pre-analytic and analytic pipelines under MISEV2023. Emerging CIP cohorts already show multi-miRNA panels with high diagnostic discrimination (AUC ≥0.84-0.90) and independent prognostic value, especially when combined with simple hematologic indices. A rigorous synthesis focused on EV-miRNA signatures for grade ≥2 CIP and steroid-refractoriness can therefore define the current performance envelope and highlight standardization gaps. Methods Following PRISMA-2020, we searched MEDLINE, Embase, Web of Science, Scopus, and Cochrane (Jan 1, 2020-Oct 31, 2025) for studies evaluating circulating/EV-miRNAs in patients on immune checkpoint inhibitors with CIP endpoints; related EV-miRNA pneumonitis cohorts were included to inform biological plausibility. Two reviewers screened/extracted; diagnostic studies were appraised with QUADAS-2 principles and EV reporting against MISEV2023. Primary outcomes were diagnostic performance (AUC) for grade ≥2 CIP and association with severity or steroid-refractoriness; secondary outcomes included differentiation from infectious/radiation pneumonitis and survival. Results Eighteen studies met inclusion: one CIP-focused discovery/validation cohort; one RT-ICI pneumonitis cohort; one ICI-toxicity EV-miRNA study (non-lung); and fifteen guideline/mechanistic/adjacent EV-miRNA papers. In the largest CIP-focused study (JITC 2025), a three-miRNA signature—EV miR-193a-5p plus serum miR-193a-5p and miR-378a-3p—distinguished CIP from non-CIP with AUC 0.870 (training) and 0.837 (validation); adding lymphocyte count improved AUCs to 0.900 and 0.932, and the three-miRNA score predicted overall survival (HR 2.827; p = 0.040). In RT-ICI pneumonitis, an exosomal miR-148b/301a/423 cluster predicted symptomatic pneumonitis and worse survival, implicating PTPN14-YAP signaling. For radiation pneumonitis, plasma exosomal miR-200b-5p integrated with dosimetry achieved AUC 0.84 for symptomatic RP, supporting lung-injury EV-miRNA plausibility. Heterogeneity in EV isolation/normalization and CIP adjudication precluded meta-analysis. Conclusion EV-miRNA signatures—anchored by a validated three-miRNA panel—demonstrate clinically meaningful discrimination for grade ≥2 CIP, differentiate infection/radiation pneumonitis, and provide prognostic separation, particularly when paired with simple clinical indices such as lymphocyte count. Given persistent heterogeneity in EV workflows and CIP adjudication, the next step is harmonized, multicenter validation using MISEV-aligned preanalytics, prespecified thresholds, and blinded comparators, with enrichment for steroid-refractory phenotypes and reporting of decision-curve and time-to-treatment metrics. If replicated, these panels could enable acceleration in correct diagnosis, curbs unnecessary antibiotics, and guides earlier immunomodulator escalation in high-risk patients. This abstract is funded by: none
Basta et al. (Fri,) studied this question.