Br J Clin Pharmacol. 2026 Sep 7. doi: 10.1002/bcp.70810. Online ahead of print.
ABSTRACT
AIMS: Conventional pharmacovigilance identifies drug-event reporting signals but not heterogeneity among anti-HER2 cardiac dysfunction reports. We benchmarked this association using disproportionality analysis and assessed whether report-level clustering added information beyond cardiovascular co-reporting.
METHODS: We analysed OpenVigil/FAERS data. A separate openFDA analysis estimated overall, individual-drug and drug-group reporting odds ratios (RORs) and proportional reporting ratios (PRRs). The clustering cohort comprised 589 unique reports, grouped by cardiovascular co-reporting and clustered using Gower distance-based partitioning around medoids. Severe outcome comprised death, hospitalization, life-threatening events or disability. Multivariable logistic regression and a nested likelihood-ratio test compared conventional and cluster-based models.
RESULTS: Echocardiography-related cardiac dysfunction was disproportionately reported with anti-HER2 therapies overall (ROR 26.64, 95% CI 25.61-27.71; PRR 25.95, 95% CI 24.97-26.96) and across evaluated drugs and groupings. Among 589 reports, 383 (65.0%) had dysfunction only and 206 (35.0%) had cardiovascular co-reporting; severe outcomes occurred in 17.5% and 41.7%, respectively (adjusted OR 3.24, 95% CI 2.14-4.95). Five clusters were retained. The LV dysfunction with cardiovascular co-reporting cluster had the highest severe outcome proportion (52.0%) and higher adjusted odds than the ADC-associated EF-decline cluster (adjusted OR 4.18, 95% CI 1.55-11.68). Adding the cluster term improved model fit (χ2 = 11.34; p = 0.023).
CONCLUSIONS: Conventional analyses confirmed a strong anti-HER2 reporting signal and an association between cardiovascular co-reporting and severe outcomes. The retained clusters added complementary report-level resolution without constituting a validated clinical classification or prediction tool.
PMID:42705876 | DOI:10.1002/bcp.70810