A payload- and target-informed toxicity atlas of antibody–drug conjugates: a disproportionality analysis of the FDA Adverse Event Reporting System

Background Antibody–drug conjugate (ADC) toxicity reflects interactions among target biology, payload chemistry, linker properties and individual drug characteristics. Whether agents sharing a target or payload also share a toxicity profile is unclear. We aimed to construct a real-world toxicity atlas of ADC-associated adverse-event reporting spanning drug, payload and target levels, and to test whether ADCs sharing a target antigen or payload platform remain distinguishable in their reporting profiles. Methods We conducted a disproportionality analysis of individual case safety reports in the publicly available US FDA Adverse Event Reporting System (FAERS), 2012Q4–2026Q1. Primary-suspect reports were identified using a curated ADC dictionary and classified by drug, payload class and target antigen. Medical Dictionary for Regulatory Activities Preferred Terms were grouped into nine clinical toxicity clusters, while contextual and non-toxicity terms were excluded. Reporting odds ratios (RORs) with 95% confidence intervals (CIs) were calculated against all remaining reports in the database; a signal required at least three exposed event reports and a lower 95% confidence bound above 1. Prespecified pairwise ADC comparisons, serious-only and report-handling sensitivity analyses, Benjamini–Hochberg false-discovery-rate correction, quarterly reporting patterns and outcome-reporting proportions were also evaluated. No case-by-case review or individual causality assessment was performed. Results Among 15 screened ADCs, 13 met the prespecified threshold of 100 primary-suspect reports, comprising 62,079 reports across 11 targets and six payload classes. Of 113 evaluable drug–cluster combinations, 71 met the signal criterion. Belantamab mafodotin showed the strongest ocular signal (ROR 36.77, 95% CI 34.01–39.76), trastuzumab deruxtecan showed prominent interstitial lung disease/pneumonitis enrichment (22.61, 21.46–23.82) and gemtuzumab ozogamicin showed marked neutropenia/hematologic enrichment (21.78, 19.78–23.99). Enfortumab vedotin showed neuropathy (9.79, 9.14–10.48), skin toxicity (7.47, 7.08–7.89) and hyperglycemia (4.74, 4.28–5.24) signals. Payload- and target-level analyses identified convergent platform patterns, whereas direct comparisons demonstrated heterogeneity among ADCs sharing targets or payloads. All 71 signals persisted after false-discovery-rate correction and 68 persisted in serious-only analyses. Conclusions ADC adverse-event reporting is structured by overlapping drug-, payload- and target-associated patterns rather than uniform class effects. This toxicity atlas supports hypothesis generation and clinically focused signal prioritization, while emphasizing that FAERS disproportionality cannot establish incidence, causality or comparative safety.