Network meta-analysis of 10 phase III trials and over 8,000 advanced gastric cancer patients ranks trastuzumab and tislelizumab highest for delaying quality-of-life deterioration

The study systematically searched MEDLINE, CENTRAL and Scopus from database inception to December 2025 and manually screened ASCO and ESMO conference abstracts from 2016 to 2025, ultimately including 10 phase III randomized controlled trials with over 8,000 patients and covering six health-related quality-of-life domains. Prior network meta-analyses of first-line advanced gastric cancer therapy focused mainly on survival benefit and did not systematically synthesize patient-reported quality-of-life data; this study used patient-reported outcomes as the primary assessment tool and brought quality of life into the first-line treatment comparison. Included studies were restricted to phase III randomized controlled trials, assessed with the Cochrane Risk of Bias Tool and reported as overall low risk of bias; screening and selection were performed independently by two reviewers with disagreements resolved by consensus.

SUCRA rankings showed trastuzumab ranked first for global health status/quality of life (89%), physical function (86%), appetite loss (83%) and pain (76%), with tislelizumab close behind for global health status, physical function and pain; pembrolizumab ranked highest for nausea/vomiting (90%) and nivolumab ranked highest for the EQ-5D utility index (95%). The study ranked treatments separately across six quality-of-life domains, revealing differential advantages on specific symptoms rather than only an overall survival ranking. Each domain was based on quantitative synthesis of four to eight trials; the Bayesian network meta-analysis used platinum-based chemotherapy as the common reference, and frequentist random-effects sensitivity analysis was consistent with the primary analysis.

In the exploratory composite analysis integrating overall survival and quality of life via the minimum distance criterion, trastuzumab showed the most favorable point-estimate performance across most quality-of-life domains, with MDC values ranging from 0.287 (global health status) to 0.362 (nausea/vomiting). The study attempted to build a composite evaluation framework incorporating both survival and patient-reported outcomes, responding to the ASCO value framework recommendation to integrate survival benefit and toxicity to quantify net health benefit. The authors explicitly state the composite metric is exploratory, that no established methods exist to generate uncertainty estimates, that it assumes independence and equal weighting of overall survival and quality of life, and that it has not been broadly validated in oncology.

In sensitivity analyses excluding four post-hoc exploratory quality-of-life datasets and restricting to trials applying a 10-point or greater deterioration threshold, treatment rankings were broadly consistent with the primary analysis, with only minor rank fluctuations in individual domains. In response to potential reporting bias from post-hoc analyses and inconsistent TTD definitions across trials, the study prespecified sensitivity analyses to test robustness of the conclusions. The study also notes that random-effects models can only handle statistical heterogeneity and cannot fully resolve measurement-driven methodological heterogeneity.

The results apply to first-line systemic treatment decisions in unresectable advanced gastric cancer and must be interpreted within each biomarker-eligible population: trastuzumab-related results derive solely from HER2-positive patients, zolbetuximab data are limited to CLDN18.2-positive tumors, and immune checkpoint inhibitor trials enrolled mixed or PD-L1-enriched populations. For clinicians, this provides a ranking reference that considers patient-reported symptom burden alongside survival; for trial designers, the integrated survival and quality-of-life framework can serve as a reference for outcome assessment in future gastric cancer clinical trials and real-world studies.

The composite metric assumes independence and equal weighting of overall survival and quality of life, which may not reflect real-world differences in patient priorities, and no established method exists to generate confidence intervals or uncertainty measures for its point estimates. SUCRA reflects the probability of being top-ranked and does not directly correspond to the magnitude of clinical difference. Trials used inconsistent operational definitions and thresholds for time to deterioration, and this endpoint is subject to censoring from disease progression and death, so quality-of-life score deterioration may stem from tumor progression itself rather than treatment toxicity. In addition, the available text is an incomplete version: Figures 1 to 3 and Supplementary Tables 1 to 11 are not included, so the effect sizes and confidence intervals for each domain comparison cannot be verified here.