Immune checkpoint inhibitors (ICIs) improve the outcomes across solid tumours, although they can cause severe immune-related adverse events (irAEs). Due to this possibility of side effects, practical and low-cost predictors of severe irAEs are needed to guide patient monitoring and care.

We conducted a single-centre retrospective cohort study involving 593 patients who were treated with anti-PD-1/PD-L1 monotherapy or anti-PD-1/PD-L1 plus anti-CTLA-4 combination therapy from June 2016 to November 2024. The primary endpoint was the time to the first severe irAE (grade ≥ 3). Peripheral blood biomarkers were evaluated at baseline and immediately before the cycle 3. The cumulative incidence was estimated, and the associations were quantified using the Fine–Gray subdistribution hazards ratio (sHR) model in a competing risks framework.

Overall, 11.6% of patients experienced a severe irAE, with the median time to the first event being 12 weeks and the most frequent severe irAE being colitis (n = 21; 3.5%). Combination therapy was associated with a higher risk when compared with monotherapy (sHR 3.71, 95% confidence interval [CI] 2.25–6.13). Baseline eosinophils > 250/µL were associated with an increased risk (sHR 2.22, 95% CI 1.35–3.65). A lower red cell distribution width (RDW) was likewise associated with the risk at two timepoints: baseline RDW ≤ 15.8% (sHR 2.60, 95% CI 1.35–5.01) and pre-cycle 3 RDW ≤ 14.3% (sHR 2.71, 95% CI 1.44–5.09). The effects were directionally consistent across subgroups, and no interactions were detected. The other blood biomarkers tested were not significant (all p > 0.05).

A high baseline eosinophil count and a lower RDW early on during therapy identify patients at increased risk of severe irAEs. These accessible measures could support personalised monitoring and biomarker-guided patient selection. However, external validation is needed to confirm the robustness and validate the thresholds identified in this study prior to clinical use.