The antibiotic-immunotherapy interaction in NSCLC: a systematic review and meta-analysis of 54,250 patients across three outcomes

Background: Approximately 20-28% of NSCLC patients receiving ICIs also receive systemic antibiotics. Whether this reflects a true microbiome-mediated pharmacological interaction or confounding by indication remains unresolved. Methods: We searched PubMed, Scopus, and CENTRAL (2013-2025) for NSCLC studies reporting OS, PFS, or ORR by antibiotic exposure. Random-effects meta-analysis was performed, with mostly non-pre-specified subgroup analyses by study design, ICI class, antibiotic timing, and line-of-therapy, plus leave-one-out, exposure-window, and adjustment-status sensitivity analyses and meta-regression of heterogeneity sources. Risk of bias was assessed with ROBINS-I and ROB 2, and certainty of evidence was rated with GRADE. Results: Forty-one studies including 54,250 patients were analyzed, 27.8% of whom were antibiotic-exposed. Antibiotic exposure was associated with worse OS (HR 1.47, 95% CI 1.30-1.66; p < 0.00001; I2 = 79%) and PFS (HR 1.32, 95% CI 1.18-1.47; p < 0.00001; I2 = 65%), but not ORR (OR 0.95, 95% CI 0.67-1.35; p = 0.77; I2 = 88%). RCT post-hoc estimates were non-significant (OS HR 1.20, p = 0.07; PFS HR 1.09, p = 0.41; I2 = 0%), whereas observational estimates were larger and driven by less-adjusted cohorts. Subgroup, sensitivity, and meta-regression analyses indicated that heterogeneity was explained mainly by study design and line of therapy. In chemo-free ICI analyses, only mixed ICI regimens remained significant for OS and PFS. GRADE certainty was very low for observational outcomes and low for RCT-based estimates. Conclusions: The strongest available evidence does not support a clear antibiotic-related reduction in ICI efficacy in NSCLC. The observational signal is more compatible with confounding and exposure heterogeneity than with a uniform pharmacological interaction, although a context-specific biological effect cannot be excluded.