Helicobacter pylori, the stomach bacterium infamous for causing gastric cancer, has long been suspected of also fueling tumors in the colon. A new narrative review published in Frontiers in Cellular and Infection Microbiology pulls together the clinical, microbial, experimental and treatment evidence on that question, and its verdict is deliberately measured: many observational studies do link the infection to colorectal neoplasia, but a human causal relationship — and any preventive benefit of eradication — remains unestablished. The review, by Jian Xu of the People’s Hospital of Danyang, Affiliated Danyang Hospital of Nantong University, imposes a four-question framework on a sprawling literature: which infection phenotype was actually measured, whether the exposure preceded the colorectal outcome, whether a proposed intermediate was experimentally perturbed, and whether a defined intervention changed the outcome. Applying those questions exposes a striking gap between the consistency of associations and the fragility of causal claims.

The observational signal is real and repeatedly reproduced, though with high heterogeneity. A meta-analysis of 55 studies pooled an odds ratio of 1.47 for colorectal adenoma (95 percent confidence interval 1.36 to 1.59), 1.77 for advanced adenoma (1.56 to 2.00), and 1.59 for colorectal cancer (1.39 to 1.82), with heterogeneity statistics of 88, 65 and 89 percent respectively. Longitudinal data add a temporal dimension but diverge widely: a Taiwanese claims-based cohort reported an adjusted hazard ratio of 1.87 for colorectal cancer, while a US Veterans cohort found a much smaller hazard ratio of 1.18 among people who underwent infection testing. Neither estimate can be read as a universal effect, because populations, exposure records, baseline lesions and surveillance practices differ. A retrospective cohort also associated infection with metachronous advanced neoplasia over a median 41 months of follow-up, with an adjusted hazard ratio of 1.74. The review stresses that pooling does not remove the exposure misclassification, selection processes or residual confounding baked into the original studies, and that overlapping primary studies prevent successive syntheses from counting as independent replications.

One of the review’s sharpest points concerns what ‘H. pylori positivity’ even means. Serology detects antibodies that can persist long after bacterial clearance, whereas urea breath and stool antigen tests more closely track active infection, and histology depends on tissue sampling. A single measurement cannot establish infection duration, timing relative to a lesion, or sustained eradication. Protein-specific serology adds resolution but yields inconsistent results: overall seropositivity was not associated with colorectal cancer in the Southern Community Cohort Study or in a large prospective elderly population, even though selected antigen responses — including antibodies against the vacuolating cytotoxin A — showed associations in some cohorts, with European fully adjusted estimates spanning the null. Antigen responses mark exposure and immunity, not the amount of bacterial product reaching the colon. Direct detection of bacterial material in colorectal tissue is equally unsettled: some histochemical and immunohistochemical studies report enrichment in adenoma tissue, while a polymerase chain reaction study found neither H. pylori nor Streptococcus gallolyticus markers in any of 210 colorectal samples. Local detection alone cannot show whether bacteria initiate a lesion or merely adhere to an altered niche after it develops.

Gastric physiology offers an alternative indirect route. In a cross-sectional study of 6,018 health-check participants, infection accompanied by atrophic gastritis was associated with adenoma and advanced adenoma, while atrophy alone was not significant after adjustment. Korean and Japanese endoscopy studies reported similar links between infection-related atrophy and advanced neoplasia or polyps. But concurrent measurements of atrophy and colorectal lesions cannot establish causal mediation, and prospective counterpoints exist: the German ESTHER cohort found no association of infection or atrophy with incident colonic cancer over roughly ten years. The endocrine story is similarly incomplete. Hypergastrinemia has been associated with distal adenomas, yet prospective analyses within adenoma chemoprevention trials found no increased recurrent-adenoma risk with higher gastrin. Prediagnostic metabolomics provides the strongest temporal human evidence: in one study, infection-associated metabolites statistically mediated part of the colorectal cancer association in males, though the confidence intervals were wide, and the accompanying genetic analysis concerned methionine sulfone rather than infection itself — leaving the origin and causal role of the metabolite unresolved.

The gut microbiome supplies the most experimentally persuasive — but still limited — mechanistic link. Stool profiling shows that infected individuals differ in bacterial and fungal composition, yet such cross-sectional profiles establish neither directionality nor mediation, since cancer itself can reshape the microbiota. The decisive perturbation comes from mice: in animals carrying a mutation in the adenomatous polyposis coli (Apc) gene, the tumor-promoting effect of H. pylori infection was attenuated under germ-free conditions, and stool-transfer experiments supported a contribution of infection-associated microbiota. Some immune changes, however — increased T-cell infiltration and reduced regulatory T-cell abundance — persisted without microbes, indicating that not all effects were microbiota-dependent. Profiling in infected mice also identified temperate-phage expansion before tumor onset, though predicted bacterial hosts and inferred gene functions fall short of proof that a phage drives tumor promotion. Crucially, the review notes that these animal findings do not identify a necessary organism, do not demonstrate the same mediation in humans, and do not reproduce decades-long human infection with its changing diet, treatment histories and microbial communities.

Experimental complexity deepens when different models disagree. Infected Apc-mutant mice showed tumor promotion accompanied by goblet-cell loss, STAT3 activation and regulatory T-cell changes, with related findings in human biopsies that support plausibility but not temporal sequence. Yet in an azoxymethane/dextran sodium sulfate model of colitis-associated cancer, infection actually reduced tumor formation and tumor-associated macrophage infiltration. Genetic susceptibility and chemically induced epithelial injury therefore create contexts in which infection-associated immune changes have opposing effects — the studies are not interchangeable replications of one mechanism. At the cellular level, CagA-positive bacterial coculture and CagA overexpression enhanced proliferation, invasion and autophagy in colon cancer cell lines in association with reduced microRNA-125b-5p, and restoring the microRNA counteracted these effects in vitro. That supports a molecular mechanism within the tested system, but it does not establish that CagA reaches human colorectal cells at a biologically meaningful exposure, nor that such exposure initiates lesions. Computational shared-gene analyses prioritize inflammatory pathways but do not demonstrate an infection-to-tumor sequence within individuals.

Eradication evidence illustrates how many different questions a single treatment can answer. Comparing successful clearance with persistent infection does not preserve randomization, because clearance depends on adherence, antibiotic resistance and follow-up testing. A US Veterans cohort associated untreated versus treated infection with higher hazards of incident and fatal colorectal cancer, but treatment was not randomized and differences in healthcare use and surveillance confound the estimates. Population-standardized comparisons give mixed pictures: a Hong Kong study found a standardized incidence ratio of 1.03 relative to the general population after eradication, while a Korean cohort of 931,585 treated individuals reported lower incidence and mortality than the general population over a mean 12.4 years, with a standardized incidence ratio of 0.66 and a mortality ratio of 0.51 — figures that must be interpreted alongside healthcare use and the exclusion of first-year cancers. A 2025 meta-analysis reported a treatment odds ratio of 0.44 from three retrospective studies, but with a 93 percent heterogeneity statistic and a wide interval. Most tellingly, two randomized trial cohorts diverged: the Shandong Intervention Trial found assigned treatment associated with lower colorectal cancer risk over 29.4 years (hazard ratio 0.47), while the Mass Intervention Trial in Linqu found no overall reduction after 13.8 years (hazard ratio 1.17). Both are post hoc cancer analyses, and assignment estimates the effect of offering treatment, not of bacterial clearance itself.

Genetic and confounding analyses further temper the causal case. Mendelian randomization studies found no clear association of genetically predicted seropositivity, CagA antibodies or VacA antibodies with colorectal cancer, though antibody instruments imperfectly represent active infection — so null genetic estimates do not exclude every infection effect, just as a metabolite instrument cannot validate an entire upstream pathway. Residual confounding remains a serious competitor: socioeconomic circumstances and early-life living conditions shape infection acquisition, while diet, smoking, adiposity, medication use and screening access influence colorectal risk and detection, and adjustment for age and sex alone cannot account for these differences. Reported effect modification — such as a Korean finding that the association with advanced neoplasia concentrated among younger smokers — requires formal interaction testing, since significance in one stratum and not another does not establish a biological subgroup effect. The review concludes that association and experimental plausibility are supported, but the magnitude, pathway specificity and reversibility of a human causal effect remain insufficiently established for broad colorectal recommendations.

On clinical translation, the review is explicit: infection status alone, on this evidence, does not justify earlier colonoscopy, shortened surveillance intervals, or eradication prescribed specifically to prevent colorectal cancer — a limitation that does not argue against treatment for established H. pylori-related indications such as peptic ulcer disease. Prediction models that incorporate serology show internal-validation promise but need external validation and demonstrations of clinical utility, and combined screening approaches, such as measuring H. pylori stool antigen from fecal immunochemical test tubes, are service-delivery opportunities distinct from etiological proof. Even tumor-targeting applications, like fluorescent Helicobacter outer membrane protein Q labeling colorectal tumors in mouse models, do not imply that infection caused the tumors. The research priorities the review proposes are concrete: repeated active-infection, antibody, gastric pathology and microbial or metabolomic sampling within the same individuals before adjudicated lesions arise; serial mediation analyses accounting for socioeconomic and lifestyle determinants; perturbation and rescue experiments of candidate microbial functions at biologically relevant exposures; and long-term treatment analyses that preserve randomized assignment while reporting clearance-based contrasts separately. Until those tests are done, the H. pylori–colon cancer hypothesis remains a compelling, well-mapped candidate pathway — not a settled cause.

Subject of Research: The association and potential causal mechanisms linking Helicobacter pylori infection to colorectal neoplasia

Article Title: Helicobacter pylori and colorectal neoplasia with emphasis on microbial mechanisms and causal evidence

Article References: Xu, J. (2026). Helicobacter pylori and colorectal neoplasia with emphasis on microbial mechanisms and causal evidence. Frontiers in Cellular and Infection Microbiology, 16, Article 1996461. https://doi.org/10.3389/fcimb.2026.1996461

Image Credits: AI Generated

DOI: 10.3389/fcimb.2026.1996461

Keywords: Helicobacter pylori, colorectal cancer, colorectal adenoma, gut microbiota, gastric atrophy, eradication therapy, CagA, Mendelian randomization, residual confounding, mucosal immunity, metabolomics, causal inference