Disproportionality signals, phenotypic patterns, and time-to-onset of immune checkpoint inhibitor-associated autoimmune encephalitis: a cross-database analysis of FAERS and JADER

Background Immune checkpoint inhibitor (ICI)-associated autoimmune encephalitis (irAIE) is a rare but clinically significant neurological immune-related adverse event. Cross-database phenotypic and temporal patterns remain incompletely characterized. We evaluated disproportionality, phenotypic clusters, time-to-onset (TTO), and severe-outcome reporting features using FAERS and JADER. Methods Public FAERS and JADER reports from 2004 Q1 through 2025 Q4 were analyzed using a prespecified primary AIE definition. Sensitivity analyses examined broader definitions, exclusion of infectious CNS events, alternative comparators, and positive and negative controls. Disproportionality was assessed using ROR, PRR, IC, and EBGM/EB05. Phenotypic clusters, TTO distributions, and Weibull reporting-time shapes were also evaluated. Results FAERS contained 195,793 ICI-related reports with 586 irAIE reports; JADER contained 68,835 with 395 irAIE reports. In FAERS, any ICI exposure showed concordant disproportionality across metrics versus all other reports (ROR 47.16, 95% CI 42.75–52.02) and using a fixed non-ICI cancer comparator (ROR 23.25, 95% CI 20.28–26.66). JADER showed the same direction (ROR 37.19, 95% CI 30.31–45.63; fixed non-ICI cancer-comparator ROR 54.34, 95% CI 34.67–85.17). Internal ICI comparisons did not support stable class-level differences. The apparent limbic/typical predominance disappeared after exclusion of AIE case-defining index terms. TTO was calculable for 239 FAERS and 153 JADER irAIE reports, with median TTOs of 43.0 and 41.0 days. Overall Weibull β estimates were below 1 in both databases; in analyses restricted to reports from 2021 onward, this pattern persisted in FAERS but was attenuated in JADER, where the 95% CI included 1. Exploratory severe-outcome models were inconsistent across databases; the FAERS association between TTO ≤90 days and death or life-threatening outcome was not reproduced in JADER. Conclusion irAIE reports showed directionally reproducible disproportionality across FAERS and JADER, including under conservative comparator analyses. Apparent limbic/typical predominance depended on case-defining terminology, while median reported TTO was similar across databases among date-complete reports. Weibull estimates were reporting-period sensitive, and date missingness limited temporal interpretation. Because individual reports could not be clinically adjudicated, these findings represent pharmacovigilance reporting associations rather than confirmed diagnoses, incidence estimates, or causal effects.