Cross-database thromboembolic reporting signals for ovarian cancer therapies: a focus on bevacizumab

Thromboembolic events (TEs) can disrupt ovarian cancer treatment and adversely affect outcomes. Post-marketing reporting patterns across systemic therapies and regulatory databases remain poorly characterized. We analyzed the FDA Adverse Event Reporting System (FAERS; 2004Q1-2025Q1), Canada Vigilance (2002–2025), and Japanese Adverse Drug Event Report database (JADER; 2004–2025). Reports involved ovarian, fallopian tube, or primary peritoneal cancer and prespecified systemic therapies. MedDRA preferred term-validated definitions identified venous thromboembolism (VTE), arterial thromboembolism (ATE), and their union (any_TE). We assessed ROR, PRR, chi-square, information component, and IC025 within each database; cross-database replication was the primary criterion. Pooled and covariate-adjusted analyses were supportive. Bevacizumab subgroup, co-medication, time-to-onset (TTO), fatal-outcome, network, and drug-target Mendelian randomization (MR) analyses provided descriptive or exploratory context. After excluding male-coded records, 55,271 reports met inclusion, including 1,296 TE cases. Seven of 39 drug-event combinations replicated in at least two databases. Bevacizumab alone replicated for any_TE, VTE, and ATE in all three systems. In FAERS, RORs were 2.28 (95% CI 1.96–2.65; IC025 0.81) for any_TE and 2.38 (1.94–2.93; IC025 0.81) for VTE. The ATE ROR was 2.34 (1.91–2.87; IC025 0.79). Paclitaxel replicated for any_TE and ATE, whereas doxorubicin replicated for any_TE and VTE. Rucaparib appeared in one database and was not assessable for replication. Only 1,757/45,025 (3.9%) FAERS reports had usable dates. Median recorded TTO was 90 days for bevacizumab and 13 days for carboplatin and paclitaxel; these selected intervals were interpreted descriptively. Genetically proxied VEGFA inhibition was associated with stroke (OR 1.17, 95% CI 1.09–1.26; P = 1.33e-05), but not with VTE or pulmonary embolism. Bevacizumab showed the most reproducible reporting signals across all three databases for composite, venous, and arterial thromboembolism. Together with greater MR concordance for arterial outcomes, this phenotype-resolved pattern motivates separate biological hypotheses for VTE and ATE. Clinically, it reinforces vigilance for both phenotypes during bevacizumab-containing ovarian cancer treatment and supports phenotype-specific prospective validation.