A multi-database pharmacovigilance study reveals distinctive immunosuppressive and opportunistic infection disproportionality signals with bevacizumab and temozolomide combination therapy in glioblastoma

Background Bevacizumab (BEV) plus temozolomide (TMZ) is increasingly used for glioblastoma, yet the immunosuppression-related adverse event spectrum and safety signals of this combination—including potential interaction effects and time-to-onset patterns—have not been comprehensively characterized in real-world pharmacovigilance data. Methods A retrospective analysis was conducted using FAERS and CVARD; a multi-algorithm framework (Omega shrinkage model, PRR, and ROR) was applied for signal detection, and interaction analyses were conducted to explore potential drug–drug interaction signals (more-than-additive reporting disproportionality patterns) in spontaneous reporting data; the Weibull model was used to evaluate time-dependent onset-hazard patterns based on TTO, and multivariate logistic regression was performed to identify factors associated with IRAE reporting. Results A total of 1,076 reports in the BEV + TMZ combination therapy group, 2,633 reports in the BEV monotherapy group, and 3,156 reports in the TMZ monotherapy group were included. The combination regimen exhibited a unique IRAEs profile, showing strong reporting disproportionality signals for rare but serious immunosuppressive and opportunistic infection events, including hemophagocytic lymphohistiocytosis (HLH; Ω = 0.648), strongyloidiasis ( Ω = 0.744), Epstein–Barr virus infection ( Ω = 0.580), and cytomegalovirus pneumonitis ( Ω = 0.943). Drug interaction analysis suggested potential interaction signals between BEV and TMZ in spontaneous reporting data, with markedly elevated PRR/interaction ORs for hematological (e.g., pancytopenia) and non-hematological toxicities (e.g., enteritis), exceeding the additive effect expected from the two monotherapies in spontaneous reporting data. Temporal dynamic analysis indicated that combination therapy altered the time-to-onset pattern for some events; for instance, the median time to onset for lymphocytopenia shifted from 34 days (BEV monotherapy) to 23 days, and the onset-hazard pattern shifted from an increasing type to a decreasing type. Multivariable logistic regression adjusting for sex, age group, body-weight group, and reporter type showed higher odds of IRAE reporting for BEV monotherapy versus TMZ monotherapy (OR = 3.6037, 95% CI: 1.0378–12.652, p < 0.05), while the overall IRAE reporting odds for BEV + TMZ were not significantly different from TMZ (OR = 2.9677, 95% CI: 0.5018–15.2056, p = 0.1988) or BEV (OR = 0.8235, 95% CI: 0.1454–4.6642, p = 0.8263). Conclusion BEV + TMZ combination therapy shows disproportionality signals for severe immunosuppression and opportunistic infections. These signal patterns, including time-to-onset distributions, may help prioritize clinical vigilance and hypothesis generation, but do not estimate incidence or establish causality and require prospective confirmation.