Importance: The use of immune checkpoint inhibitors (ICIs) is increasing. Little is known about the frequency of late-onset immune-related adverse events (irAEs) and the patient-specific risk factors associated with their development.

Objectives: To assess the incidence of persistent or de novo late-onset irAEs requiring hospitalization and identify patient factors associated with risk of late-onset irAEs.

Design, Setting, and Participants: This retrospective observational cohort study conducted from January 2011 to October 2022 included patients who received ICIs and were hospitalized with irAEs at an academic medical center. Exclusion criteria included ICI therapy outside of the hospital system and no irAE diagnosis during admission. Data were analyzed from November 15, 2022, to January 8, 2025.

Exposure: Late-onset irAEs.

Main Outcomes and Measures: The main study outcomes were (1) incidence of irAE hospitalization at 0 to 6 months (early), more than 6 to 12 months (intermediate), and more than 12 months (late) after ICI initiation and (2) patient factors associated with risk of late-onset irAEs.

Results: Among the 795 patients hospitalized with irAEs, the median age was 67.3 years (IQR, 58.3-74.8 years); 476 (59.9%) were male. Most patients (n = 517 [65.0%]) received anti–programmed death ligand 1 (PD-L1) and anti–programmed cell death 1 monotherapy, with the most common indications being melanoma (n = 335 [42.1%]) and lung cancer (n = 167 [21.0%]). The median time from start of ICI therapy to hospital admission was 2.7 months (IQR, 1.2-6.1 months), with 14.7% of patients (n = 117 of 795) presenting 6 to 12 months after initial ICI exposure and 10.8% of patients (86 of 795) presenting more than 12 months after initial exposure. The irAEs most likely to present late included those involving the kidney (10 of 32 [31.3%]) and hematologic (5 of 23 [21.7%]) organ systems. In univariate analysis, ICI type was significantly associated with the timing of hospital admission for irAEs; of the 517 patients receiving anti–PD-L1–based therapy, 13.5% (n = 70) presented late compared with 5.4% (9 of 167) receiving dual therapy with anti–cytotoxic T-lymphocyte–associated protein 4 (P < .001). Patients receiving perioperative ICI therapy were significantly more likely to be admitted at the intermediate interval (16 of 68 [23.5%]) compared with those with metastatic disease (87 of 678 [12.8%]) (P = .03). Timing of irAE was also significantly associated with active ICI exposure; among the patients presenting late, 7.4% (48 of 651) had received ICI therapy within the last 60 days compared with 26.4% (38 of 144) who had not had recent ICI exposure (P < .001).

Conclusions and Relevance: The findings of this retrospective observational cohort study suggest that late irAEs are possible, with a subset of patients presenting years after the start of ICI therapy. Clinicians must remain vigilant for irAEs regardless of elapsed time from ICI therapy, especially as patients live longer and ICIs become more widely used. Future investigations are needed to better understand the risk factors for late-onset irAEs and the distinct immunologic pathways that underlie such events.