This systematic review and single-arm meta-analysis pooled safety data for bispecific antibodies (BsAbs) that concurrently target PD-1/CTLA-4 or PD-1/VEGF in patients with non-small cell lung cancer (NSCLC). The primary objective was to separately quantify and descriptively synthesize pooled safety profiles for each BsAb class and to summarize class-specific patterns of adverse events observed across published prospective clinical trials through January 2026.
The authors conducted a PRISMA-guided systematic search of prospective clinical trials available through January 2026. Seventeen studies meeting inclusion criteria were identified and included, representing a combined total of 1,802 patients. Single-arm proportion meta-analyses were performed separately for the PD-1/CTLA-4 and PD-1/VEGF BsAb classes using random-effects models to generate pooled incidence estimates for key safety endpoints.
No formal statistical test comparing the two BsAb classes was performed. Given notable clinical heterogeneity among cohorts — including variation in study phase, treatment line, molecular subtype, individual BsAb agents, and concomitant treatment regimens — pooled estimates were interpreted descriptively rather than as definitive comparative evidence.
Exploratory subgroup analyses were planned for treatment regimen, treatment line, and individual BsAb agent. Because relatively few cohorts were available within individual phase strata, study phase effects were evaluated descriptively rather than through formal stratified meta-analysis.
Seventeen prospective trials contributed to the pooled analyses. The combined sample across studies totaled 1,802 patients with NSCLC. The included cohorts varied in trial phase, prior treatment exposure, molecular background, and use of combination therapies such as chemotherapy, which contributed to clinical heterogeneity when interpreting pooled safety estimates.
In the pooled analysis of PD-1/CTLA-4 BsAb cohorts, the incidence of grade 3 or higher treatment-related adverse events (TRAEs) was 41.79%. The pooled incidence of immune-related adverse events (irAEs) in this class was 46.68%, while pooled rates of grade 3 or higher irAEs were reported as low. The pooled treatment discontinuation rate due to TRAEs was 10.26% for PD-1/CTLA-4 studies.
The authors noted variability in AE rates across individual PD-1/CTLA-4 agents, an observation derived from exploratory subgroup assessment rather than from formal between-agent comparisons.
For PD-1/VEGF BsAb cohorts, the pooled incidence of grade 3 or higher TRAEs was 43.51%. The pooled incidence of irAEs was lower than in the PD-1/CTLA-4 analysis, at 27.10%, with pooled rates of grade 3 or higher irAEs also reported as low. Treatment discontinuation due to TRAEs was pooled at 4.44% for PD-1/VEGF studies.
Exploratory observations indicated that higher AE rates in PD-1/VEGF datasets were seen predominantly in regimens that included chemotherapy, suggesting regimen context influenced observed toxicity frequencies.
Subgroup analyses were described as exploratory. Across PD-1/CTLA-4 studies, AE rates appeared to differ by individual agent. Across PD-1/VEGF studies, higher AE rates were more commonly associated with chemotherapy-containing regimens. The authors emphasize these subgroup signals should be interpreted cautiously because they may reflect uneven distributions of agents, patient populations, study phases, and treatment strategies rather than intrinsic class-level toxicity mechanisms.
Clinical heterogeneity was explicitly acknowledged as a major limitation to causal or comparative inference. The authors therefore framed pooled estimates as descriptive, hypothesis-generating findings rather than definitive comparisons between BsAb classes.
Reported pooled treatment discontinuation rates due to TRAEs differed between the two classes: 10.26% for PD-1/CTLA-4 and 4.44% for PD-1/VEGF. Although the pooled incidence of any irAE was higher in the PD-1/CTLA-4 analysis (46.68% vs 27.10%), pooled rates of grade 3 or higher irAEs were low for both classes, according to the authors’ synthesis.
These tolerability metrics were presented with the caveat that differences may be attributable to cohort-level variation rather than inherent drug-class properties.
The review provides a descriptive synthesis of available safety data for PD-1/CTLA-4 and PD-1/VEGF BsAbs in NSCLC, but it stops short of claiming direct comparability between classes. Key limitations include substantial heterogeneity across included cohorts in study phase, treatment line, molecular background, individual agents, and concomitant therapies. The limited number of cohorts in some strata precluded robust stratified meta-analyses.
Consequently, the authors recommend treating the pooled AE estimates as hypothesis-generating signals that require validation in more homogeneous prospective studies or in head-to-head trials. They also note exploratory subgroup observations that warrant further investigation but should not be taken as definitive evidence of class-specific toxicity patterns.
This systematic review and single-arm meta-analysis pooled safety data from 17 prospective trials (1,802 patients) for BsAbs targeting PD-1/CTLA-4 and PD-1/VEGF in NSCLC. Pooled rates of grade ≥3 TRAEs were similar between classes (approximately 42–44%), with higher pooled incidence of any irAE in PD-1/CTLA-4 cohorts compared with PD-1/VEGF cohorts. Treatment discontinuation due to TRAEs was higher in PD-1/CTLA-4 pooled data than in PD-1/VEGF pooled data. Because of marked heterogeneity across studies, these pooled estimates are descriptive and intended to generate hypotheses for future, more controlled comparisons.
Competing interests: the authors declared no competing interests. Ethics approval and consent to participate were not applicable, and consent for publication was not applicable.