Journal of Cancer and Tumor International

Microbial communities associated with malignant tissue and with the gastrointestinal tract have been proposed as determinants of response to immune checkpoint inhibitors, and the proposition has moved rapidly from murine models into randomised clinical testing. The speed of that translation has outpaced agreement on what the underlying evidence actually supports. This critical narrative review examines the relationship between microbiota and checkpoint-inhibitor outcomes by keeping two ecologically and methodologically distinct compartments separate: the intestinal microbiome, which acts at a distance through systemic immune and metabolic conditioning, and the intratumoural microbiome, which occupies the tumour microenvironment itself at very low biomass. Evidence published between 2015 and August 2026 was assembled from biomedical literature databases and open scholarly indexes, appraised qualitatively, and synthesised around mechanisms, reproducibility, interventional design and methodological risk rather than by individual study. Three conclusions emerge with differing confidence. Causal contribution of intestinal commensals to checkpoint-inhibitor efficacy is well supported by germ-free and gnotobiotic experiments, by interventional faecal microbiota transplantation trials and, most recently, by randomised placebo-controlled testing, although effect sizes remain modest and endpoints inconsistently met. Predictive taxonomic signatures, by contrast, have repeatedly failed to replicate across geographically distinct cohorts, and the field has accordingly shifted towards ecological, functional and metabolite-based scoring. Claims for a pan-cancer intratumoural microbiome have been substantially revised: the most influential pan-cancer analysis has been retracted, and independent decontaminated reanalyses of large whole-genome datasets indicate that robust tumour-resident communities are concentrated in orodigestive sites rather than distributed across all cancers. Unresolved problems include the absence of validated causal biomarkers, uncertain donor selection logic for microbial therapeutics, poorly characterised interactions with antibiotics and diet, and marked geographical narrowness of the evidence. Progress now depends less on additional descriptive profiling than on contamination-controlled study design, mechanistically anchored endpoints and adequately powered randomised trials.