Every two weeks, a critical review of the most significant colorectal cancer research indexed in PubMed, with key findings, methodological appraisal, and clinical implications.
1. Clinicogenomic map of evolution and organ-specific metastatic tropism in advanced colorectal cancer (MSK cohort, Cancer Cell)
Paolo Manca (Memorial Sloan Kettering Cancer Center, New York; Fondazione IRCCS Istituto Nazionale dei Tumori, Milan), Henry Walch, Jierui Xu, Andrea Cercek, Karuna Ganesh, Julio Garcia-Aguilar, Nikolaus Schultz, et al.; senior authors Walid K. Chatila and Rona Yaeger (Memorial Sloan Kettering Cancer Center)
Key points
- Integrated clinical and genomic analysis of 7,238 consecutively sequenced colorectal cancer samples from 6,789 patients (5,174 primary tumors and 2,063 metastases), with an average sequencing depth of 622x.
- The cohort comprised 5,946 microsatellite-stable (MSS), 775 MSI-high and 68 ultramutated (pathogenic POLE/POLD1) cases; 87.6% of primary-tumor oncogenic mutations were shared with the matched metastasis across 139 primary-metastasis pairs.
- Evolutionary features such as clonality and oncogenic mutant allelic imbalance amplified the effect of recurrent driver alterations.
- In MSS disease, metastasis followed a sequence: median time to liver metastasis was 1.2 months versus 23.2 months for brain metastasis; lung, bone and brain metastases tended to co-occur, and median overall survival after a brain or adrenal metastasis was under one year.
- RAS pathway activation increased the risk of spread to every metastatic site examined, whereas WNT pathway activation was broadly protective, including lower rates of peritoneal spread.
Key limitations
- Single-institution, tertiary-referral cohort of patients who underwent clinical sequencing, which may over-represent advanced or complex disease.
- Observational associations between genomic events and metastatic sites do not establish causation, and treatment exposures differed between subgroups.
- The organ-specific risk model has not yet been tested prospectively for its ability to change surveillance or treatment decisions.
Future directions
- Prospective validation of genomics-informed, organ-specific metastasis risk prediction at diagnosis.
- Trials of site-tailored surveillance (for example, brain or lung imaging) in genomically defined high-risk subgroups.
- Functional studies of how RAS activation and WNT pathway status shape organotropism.
Metastatic colorectal cancer remains a leading cause of cancer death, with five-year survival for metastatic disease below one in five patients, and colorectal cancer is now the leading cause of cancer death among people younger than 50. Although the liver is the dominant site of spread, a substantial minority of patients develop metastases to the lung, peritoneum, bone, brain or adrenal glands, and these sites carry very different prognoses and management implications. Clinicians currently have few tools to predict, at diagnosis, where an individual tumor is likely to spread, and surveillance protocols are largely uniform regardless of tumor biology.
Colorectal cancer typically requires three to six driver alterations to become invasive, and large sequencing efforts have catalogued its recurrent mutations. What has been less clear is how the timing and clonal architecture of those alterations, and features such as allelic imbalance at oncogenic loci, modify their clinical effect, and whether specific genomic events predispose to particular metastatic sites. Prior studies linking genotype to organotropism were generally limited in size, focused on one or two sites, or lacked detailed clinical annotation of metastatic timing, leaving the field without a comprehensive map relating tumor evolution to the pattern and sequence of spread.
Manca, Yaeger, Chatila and colleagues at Memorial Sloan Kettering Cancer Center assembled clinical and genomic data from 7,238 consecutively sequenced samples from 6,789 patients, including 5,174 primary tumors and 2,063 metastases, sequenced to an average depth of 622x. The cohort included 5,946 MSS, 775 MSI-high and 68 ultramutated tumors with pathogenic POLE or POLD1 alterations. The investigators analyzed clonality and mutant allelic imbalance, examined 139 matched primary-metastasis pairs, and modeled the timing and co-occurrence of metastasis across seven organ sites. Findings were cross-checked against external resources including TCGA, GENIE BPC, GDSC and Caris Life Sciences data.
The analysis showed that genomic evolutionary events, such as clonal status and oncogenic mutant allelic imbalance, amplified the consequences of recurrent driver alterations. Metastases were genomically close to their primaries: 87.6% of oncogenic mutations in primary tumors were shared with the matched metastasis. In MSS disease, metastatic spread followed a recognizable sequence, with liver metastases appearing early (median 1.2 months) and brain metastases late (median 23.2 months); adrenal metastases were also late events. Lung, bone and brain metastases tended to occur together, suggesting a shared biological route or permissive phenotype.
Pathway-level analyses provided the most clinically suggestive results. RAS pathway activation, including KRAS mutation, was associated with increased risk of spread to all metastatic sites examined, whereas WNT pathway activation, including APC alteration, was overall protective and was linked to lower rates of peritoneal spread and better prognosis. Outcomes after late-site metastasis were poor: median overall survival after diagnosis of brain or adrenal metastasis was under one year. More than 90% of patients with metastatic MSS tumors received chemotherapy, and about half of those with metastatic MSI-high disease received immune checkpoint blockade, providing context for the treated population.
The work draws its strength from scale, consecutive sampling and detailed clinical annotation, but it remains an observational analysis from a single tertiary center, where referral patterns and the decision to sequence may enrich for advanced or unusual disease. Associations between genomic events and metastatic sites are correlative and may be influenced by treatment, survival time and imaging practices, since late sites can only be detected in patients who live long enough. The predictive approach has not yet been tested prospectively, and validation relied on external datasets with differing composition and annotation.
The study offers a large, clinically annotated framework linking colorectal cancer evolution to where and when metastases arise, and the authors propose that combining primary tumor location with genomic data could help estimate organ-specific metastatic risk at diagnosis. If validated, such models could support tailored surveillance, for example earlier brain or lung imaging in high-risk genotypes, and inform selection for more intensive therapy. An accompanying commentary in the same issue of Cancer Cell (Verhagen et al.) was titled "Tumor evolution shapes metastatic competence in colorectal cancer." Functional work on RAS- and WNT-driven organotropism is a logical next step.
Source: Manca et al., "Integrated clinicogenomic analysis reveals the evolution and metastatic tropisms of advanced colorectal cancer," Cancer Cell, published online September 1, 2026. Read the paper · PubMed
2. Computer-aided quality feedback to improve colonoscopists' adenoma detection rate: stepped-wedge cluster-randomised trial (NCT04862793, The Lancet Digital Health)
Kristoffer Mazanti Cold (Copenhagen Academy for Medical Education and Simulation, Rigshospitalet, and University of Copenhagen), A. Vamadevan, A. Heen, A.S. Vilamann, M. Rasmussen, M.B.S. Svendsen; senior author Lars Konge (Copenhagen Academy for Medical Education and Simulation, Rigshospitalet, and University of Copenhagen)
Key points
- Stepped-wedge, cluster-randomised trial at three Danish hospital sites in FIT-positive screening patients aged 50-74: 1,109 colonoscopies in intervention periods and 1,001 in control periods, performed by 17 endoscopists acting as their own controls.
- Adenoma detection rate rose from 43.4% to 48.6% with computer-aided quality (CAQ) feedback after each procedure (OR 1.24; 95% CI 1.02-1.52; P=.032).
- Polyp detection rate was 54.2% versus 50.7% (OR 1.22; 95% CI 1.00-1.49; P=.048), and adenocarcinoma detection rate 8.9% versus 6.5% (OR 1.51; 95% CI 1.06-2.15; P=.027).
- Withdrawal time lengthened from 915 to 960 seconds (IRR 1.11; 95% CI 1.06-1.17; P=.003).
- Adenomas per colonoscopy (1.00 vs 0.85; IRR 1.15; 95% CI 0.99-1.33; P=.069) and polyps per colonoscopy did not differ significantly.
Key limitations
- Only three sites and 17 endoscopists, with unequal contributions across sites and periods.
- Conducted in a FIT-positive population with high baseline detection; effects may differ in primary colonoscopy screening or surveillance.
- Detection metrics are surrogates; post-colonoscopy colorectal cancer was not assessed.
Future directions
- Studies linking CAQ feedback to post-colonoscopy colorectal cancer and interval cancer rates.
- Evaluation in centres and programmes that do not use FIT-based screening.
- Head-to-head or combined testing with real-time computer-aided detection (CADe) systems.
The protective value of colonoscopy depends heavily on the skill and diligence of the endoscopist. The adenoma detection rate (ADR) is the main quality indicator for colonoscopy, and higher ADR is associated with lower risk of post-colonoscopy colorectal cancer, yet ADR varies widely between endoscopists. Most artificial intelligence work in colonoscopy has focused on real-time computer-aided detection (CADe), which flags lesions during the procedure. A different approach is to give endoscopists objective, procedure-level feedback on how well they examined the colon, aiming to change behaviour rather than to point out individual polyps.
Quality improvement programmes have traditionally relied on periodic audit and report cards, often delivered months after procedures and based on manually recorded metrics. Automated systems that assess aspects of the examination, such as mucosal inspection and withdrawal behaviour, could deliver immediate and consistent feedback. However, it has been uncertain whether such feedback produces measurable gains in detection in routine practice, and randomised evidence for computer-aided quality (CAQ) feedback, as opposed to CADe, has been limited.
Cold, Konge and colleagues at the Copenhagen Academy for Medical Education and Simulation, Rigshospitalet, conducted a multicentre, stepped-wedge, cluster-randomised trial (NCT04862793) in which three hospital sites served as clusters and crossed from control to intervention at randomised times. Patients aged 50-74 years referred after a positive faecal immunochemical test (FIT) were enrolled between February 23, 2022, and April 19, 2024. During control periods, colonoscopy was performed as standard care; during intervention periods, endoscopists received CAQ feedback after each colonoscopy. The primary outcome was ADR; secondary outcomes included polyp detection rate, adenomas and polyps per colonoscopy, adenocarcinoma detection rate and withdrawal time.
In total, 1,109 patients were examined in intervention periods and 1,001 in control periods; 44.3% were female. Seventeen endoscopists participated and served as their own controls. ADR was 48.6% (539 of 1,109) with CAQ feedback versus 43.4% (434 of 1,001) under standard care, corresponding to an odds ratio of 1.24 (95% CI 1.02-1.52; P=.032). The absolute gain of about five percentage points was achieved in a FIT-positive population that already had a high baseline ADR.
Secondary outcomes moved in the same direction. The polyp detection rate was 54.2% versus 50.7% (OR 1.22; 95% CI 1.00-1.49; P=.048), and the adenocarcinoma detection rate was 8.9% versus 6.5% (OR 1.51; 95% CI 1.06-2.15; P=.027). Withdrawal time increased from 915 to 960 seconds (IRR 1.11; 95% CI 1.06-1.17; P=.003), consistent with more careful inspection. Per-procedure counts did not differ significantly: adenomas per colonoscopy were 1.00 versus 0.85 (IRR 1.15; 95% CI 0.99-1.33; P=.069) and polyps per colonoscopy 1.36 versus 1.27 (IRR 1.10; 95% CI 0.95-1.27; P=.22), with no clear differences for tubulovillous or sessile serrated adenomas.
The stepped-wedge design allows every site to contribute both control and intervention data, but with only three clusters it is vulnerable to secular trends and to imbalance in case mix; site contributions were uneven, with one site providing most intervention-period procedures. The small number of endoscopists limits generalisability across operators with different baseline skill, and a Hawthorne effect from being observed cannot be excluded. Detection rates are surrogate outcomes, and the higher adenocarcinoma detection rate in a FIT-positive cohort may partly reflect case mix rather than earlier detection.
The trial provides randomised evidence that automated, post-procedure quality feedback, distinct from real-time lesion detection, can raise ADR and other detection metrics in routine screening colonoscopy, offering a scalable tool for quality assurance in organised screening programmes. The authors call for studies on whether these gains translate into fewer post-colonoscopy colorectal cancers and for evaluation in centres that do not use FIT-based screening. Combining CAQ feedback with CADe, and assessing durability after feedback stops, are natural next questions.
Source: Cold et al., The Lancet Digital Health, published online September 2, 2026 (NCT04862793). PubMed
3. Diagnostic yield of germline multigene panel testing in unselected colorectal cancer: systematic review and meta-analysis (The Lancet Gastroenterology & Hepatology)
Alessandro Mannucci (Gastroenterology and Gastrointestinal Endoscopy Unit, IRCCS San Raffaele Hospital and Vita-Salute San Raffaele University, Milan), M. Puzzono, G. Marzocca, A. Goel, S. Ogino, S. Syngal, M.B. Yurgelun, S. Gupta; senior author Giulia Martina Cavestro (IRCCS San Raffaele Hospital and Vita-Salute San Raffaele University, Milan)
Key points
- 21 studies included (PROSPERO CRD42025649177); among 6,925 unselected patients with colorectal cancer, 15.2% carried at least one pathogenic or likely pathogenic germline variant (95% CI 12.8-18.1; I²=84.1%).
- High-penetrance variants were found in 7.9% (95% CI 5.9-10.5) and high-penetrance colorectal cancer predisposition variants in 5.4% (95% CI 3.5-8.6).
- In early-onset disease (diagnosed before 50), yields were 17.7% for any variant (95% CI 15.0-20.7), 14.2% high-penetrance and 12.2% high-penetrance CRC predisposition.
- In late-onset disease (50 or older), yields were 13.0% for any variant (95% CI 10.8-15.7), 6.5% high-penetrance and 3.8% high-penetrance CRC predisposition.
- Yield fell with age but the high-penetrance yield remained above 5% up to age 67 (95% CI 59-75); panel size and publication year did not affect yield, while a first-degree family history increased it.
Key limitations
- Substantial heterogeneity across studies (I² 73% to 93%), with differing panels, populations and variant classification.
- Eleven of 21 studies had moderate or high risk of bias, and only English-language studies were included.
- Diagnostic yield does not by itself show that testing changes management or outcomes.
Future directions
- Prospective evaluation of universal germline testing for all patients with colorectal cancer.
- Cost-effectiveness and implementation studies, including genetic counselling capacity.
- Better representation of diverse ancestries and harmonised variant interpretation.
Identifying inherited cancer predisposition in patients with colorectal cancer affects their own treatment and surveillance and enables cascade testing of relatives. Traditional strategies select patients for germline testing based on age at diagnosis, family history or tumor mismatch repair status, but these criteria are known to miss a proportion of carriers. With next-generation sequencing panels now inexpensive and widely available, a recurring question is whether every patient with colorectal cancer should be offered multigene germline testing, as is increasingly done for some other cancers.
Individual studies of unselected colorectal cancer cohorts have reported widely varying rates of pathogenic germline variants, reflecting differences in panel size, population, age distribution and variant interpretation. Without a pooled estimate, it has been difficult to judge whether the yield of universal testing clears the thresholds commonly used to justify genetic testing, particularly in patients diagnosed after age 50, who are often excluded by age-based criteria. Quantifying yield by age group and by penetrance category is therefore central to the debate about universal testing.
Mannucci, Cavestro and colleagues at IRCCS San Raffaele Hospital and Vita-Salute San Raffaele University in Milan conducted a systematic review and meta-analysis registered with PROSPERO (CRD42025649177). They searched PubMed, Embase, Scopus and CENTRAL to August 2, 2025, for English-language studies applying next-generation sequencing panels of at least five genes to unselected patients with colorectal cancer. The primary outcome was person-level prevalence of at least one pathogenic or likely pathogenic (P/LP) variant. Prevalence was pooled with random-effects models, heterogeneity was quantified with I², and meta-regression examined age, panel size, family history and publication year.
Of 2,707 records, 21 studies were included. Across 6,925 patients with colorectal cancer, the pooled prevalence of any P/LP germline variant was 15.2% (95% CI 12.8-18.1; 1,061 of 6,925; I²=84.1%). High-penetrance variants were present in 7.9% (95% CI 5.9-10.5), and high-penetrance variants in established colorectal cancer predisposition genes in 5.4% (95% CI 3.5-8.6). These figures indicate that roughly one in seven unselected patients carries a germline finding and about one in twenty carries a high-penetrance colorectal cancer predisposition variant.
Yields were higher in early-onset disease: 17.7% for any P/LP variant (95% CI 15.0-20.7), 14.2% for high-penetrance variants (95% CI 11.7-17.3) and 12.2% for high-penetrance colorectal predisposition variants (95% CI 9.5-15.5). In late-onset disease, corresponding yields were 13.0% (95% CI 10.8-15.7), 6.5% (95% CI 4.4-9.5) and 3.8% (95% CI 2.0-7.5). Yield declined with age at diagnosis, but the high-penetrance yield stayed above 5% up to age 67 years (95% CI 59-75). Panel size and publication year did not affect yield, while a first-degree family history of colorectal cancer was associated with higher overall and colorectal-specific yields, but not with non-colorectal yields.
Heterogeneity was high across nearly all estimates (I² 73.3% to 93.4%), reflecting differences in panels, populations and variant classification over time. Risk of bias was low in 10 studies but moderate in 9 and high in 2, and restriction to English-language publications may have omitted relevant cohorts. Funding came from Italian and European public sources and the US National Cancer Institute. Importantly, the analysis measures diagnostic yield rather than the downstream clinical impact of testing on patients and relatives.
The meta-analysis shows that germline testing yields substantial numbers of actionable findings in unselected patients with colorectal cancer, including well beyond the early-onset group, and that high-penetrance yield remains above commonly cited testing thresholds into the late 60s. The authors conclude that the data support further evaluation of universal germline testing for colorectal cancer. Prospective implementation studies, cost-effectiveness analyses and attention to genetic counselling capacity will determine whether universal testing becomes standard practice.
Source: Mannucci et al., systematic review and meta-analysis of germline multigene panel testing in colorectal cancer, The Lancet Gastroenterology & Hepatology, published online September 3, 2026. PubMed
4. Heterozygous germline MSH3 and MLH3 mutations as colorectal cancer predisposition alleles (Gut)
Ignacio Soriano (Department of Oncology, University of Oxford), K. Sherwood, J.C. Ward, J. Fernandez-Tajes, J.F. Wilson, B. Glimelius, T. Sjöblom, C. Palles, et al.; senior author Ian Tomlinson (Department of Oncology, University of Oxford)
Key points
- Genetic study of about 12,000 colorectal cancer and multiple-polyp cases and 460,000 controls, with cancer genome sequencing in 2,023 patients.
- One case carried biallelic MSH3 mutations and another biallelic MLH3 mutations, resembling some patients with constitutional mismatch repair deficiency.
- Heterozygous carriers had increased colorectal cancer risk: about 2.2-fold for MSH3 (P=6.6×10^-5) and 1.6-fold for MLH3 (P=0.028), mediated by somatic second hits on the wild-type allele.
- All MSH3- or MLH3-deficient cancers were microsatellite-stable but hypermutant, with roughly 12-fold more deletions of 2 bp or longer and usually mutational signature ID4 (P<0.0001).
- Cancers in heterozygotes without a second hit were not hypermutant; single second hits sometimes inactivated both MSH3 and the nearby APC gene.
Key limitations
- Risk estimates are modest and penetrance is incomplete, so absolute risk for individual carriers remains uncertain.
- Confidence intervals were not reported in the abstract, and biallelic findings rest on single cases.
- Ancestry composition of the case and control resources may limit generalisability to other populations.
Future directions
- Deciding whether MSH3 and MLH3 belong on clinical colorectal cancer germline panels and how carriers should be managed.
- Using signature ID4 and long-deletion hypermutation as tumor-based screens for possible germline carriers.
- Studying immunotherapy responsiveness in MSS hypermutant tumors driven by MSH3 or MLH3 loss.
Lynch syndrome, caused by germline defects in the canonical mismatch repair genes MLH1, MSH2, MSH6 and PMS2, is the most common hereditary colorectal cancer syndrome, and its tumors are typically microsatellite-unstable. MSH3 and MLH3 are non-canonical mismatch repair genes whose products help repair insertion-deletion errors. Their role in inherited colorectal cancer risk has been less clear, and they are not routinely interpreted as cancer predisposition genes in clinical practice.
Biallelic MSH3 mutations had previously been linked to colorectal cancer and adenomas, and biallelic MLH3 mutations to colorectal cancer in very few carriers, suggesting recessive inheritance. Soriano, Tomlinson and colleagues hypothesised instead that loss-of-function variants in these genes behave like other mismatch repair genes: causing disease in a constitutional mismatch repair deficiency-like pattern when biallelic, but also raising cancer risk in heterozygotes through a somatic second hit, as in Lynch syndrome. Testing this required very large case-control resources combined with tumor sequencing.
The team, led from the Department of Oncology, University of Oxford, analysed germline data from about 12,000 patients with colorectal cancer or multiple polyps and about 460,000 controls. Cancer genome sequencing was available for 2,023 patients, allowing the investigators to look for second hits on the wild-type allele and to characterise the mutational consequences of MSH3 and MLH3 deficiency, including mutational signatures and patterns of insertion-deletion mutations.
Among cases, one patient carried biallelic MSH3 mutations and another biallelic MLH3 mutations. More importantly, heterozygous carriers of loss-of-function variants had increased colorectal cancer risk, about 2.2-fold for MSH3 (P=6.6×10^-5) and 1.6-fold for MLH3 (P=0.028). Tumor sequencing showed that this risk operated through somatic second hits that inactivated the remaining wild-type allele, the classic two-hit tumor suppressor mechanism. In some tumors a single second hit disabled both MSH3 and the nearby APC gene, providing an efficient route to tumor initiation.
All colorectal cancers with MSH3 or MLH3 deficiency were microsatellite-stable yet hypermutant. Deletions of 2 bp or more were roughly 12 times more frequent than in comparison tumors, and mutational signature ID4 was usually present (P<0.0001). In contrast, cancers arising in heterozygotes without a second hit were not hypermutant, indicating that the germline variant was not always the driver of the tumor. The authors note that ID4 likely has several causes, including MSH3 or MLH3 deficiency.
Penetrance of heterozygous variants appears incomplete and the risk increases are modest, so absolute lifetime risk for individual carriers cannot yet be estimated precisely. The abstract does not provide confidence intervals for the risk estimates, and the biallelic observations rest on single patients. Variant classification for these less-studied genes is challenging, and the ancestry composition of the case and control resources may limit generalisability. Because these tumors are microsatellite-stable, they would be missed by standard Lynch screening approaches based on microsatellite instability or immunohistochemistry for canonical mismatch repair proteins.
The study reframes MSH3 and MLH3 as classical tumor suppressor predisposition genes whose heterozygous loss raises colorectal cancer risk through hypermutation, a pattern the authors compare to PMS2-associated Lynch syndrome, while biallelic carriers resemble some patients with constitutional mismatch repair deficiency. This has implications for germline panel design and interpretation, for the use of tumor signatures such as ID4 to flag possible carriers, and potentially for treatment, since MSS hypermutant tumors may have distinct immunotherapy sensitivity, a question that remains to be tested.
Source: Soriano et al., Gut, published online September 11, 2026. PubMed
5. Global burden of colorectal cancer potentially attributable to Helicobacter pylori exposure: risk attribution modelling study (eGastroenterology)
Yunhao Li (LKS Faculty of Medicine, The University of Hong Kong), et al., with co-authors from Erasmus MC and Brown University; joint senior authors Wai-keung Leung (The University of Hong Kong) and Shailja Shah (University of California San Diego)
Key points
- Pooled analysis of 43 studies including 48,694,403 people found that H. pylori exposure was associated with about 1.6-fold (1.59-fold) higher colorectal cancer risk.
- Applied to GLOBOCAN 2022 case counts, about 22% of colorectal cancers worldwide were estimated to be potentially attributable to H. pylori; the estimate fell to 12% when restricted to 14 population-based and cohort studies.
- Attributable fractions were highest in the Western Pacific region (25%) and Mongolia (36%); China had the largest absolute number of attributable cases (139,619).
- In exploratory birth-cohort analyses, the attributable proportion rose from 9.5% (born before 1947) to about 10% (1948-62) and 15% (1963-77).
- In 4 observational studies, eradication therapy was associated with reduced colorectal cancer risk only after 10 years; no significant differences were seen by sex, age, diagnostic method or presence of a screening programme.
Key limitations
- Exploratory modelling built on observational associations; causality is not established.
- The attributable fraction roughly halves (22% to 12%) when restricted to higher-quality population-based and cohort studies.
- Evidence on eradication rests on only 4 observational studies, and the birth-cohort pattern was not formally tested.
Future directions
- Randomized trials and prospective cohorts testing whether H. pylori eradication lowers colorectal cancer incidence.
- Mechanistic studies of how gastric H. pylori infection might influence colorectal carcinogenesis.
- Weighing antibiotic resistance and side effects before any population-level eradication strategy targeting colorectal cancer.
Colorectal cancer accounted for nearly 2 million new cases and about 900,000 deaths worldwide in 2022, and its incidence is rising in many regions and in younger birth cohorts. Most of that burden is attributed to non-modifiable factors such as age and genetics and to modifiable lifestyle and dietary exposures. Infectious and microbial contributors have attracted growing interest, and Helicobacter pylori, the leading infectious cause of gastric cancer and one of the most common chronic bacterial infections globally, has been repeatedly linked to colorectal neoplasia in epidemiologic studies.
Population-level H. pylori screening and eradication is already considered for gastric cancer prevention in high-prevalence regions. If H. pylori also contributes to colorectal cancer, the population benefits of eradication could be larger than currently assumed. However, previous reports on the H. pylori-colorectal cancer association have been heterogeneous in design, diagnostic method and region, and there had been no global estimate of how much colorectal cancer might be attributable to the infection, or of how that fraction varies by region and birth cohort.
Li and colleagues, led by Wai-keung Leung at the University of Hong Kong and Shailja Shah at the University of California San Diego, with collaborators at Erasmus MC and Brown University, performed an exploratory risk attribution modelling study. They drew on systematic reviews and pooled analyses of the H. pylori-colorectal cancer association, stratified by region, diagnostic method, sex, access to a national screening programme and use of eradication therapy, and combined the pooled estimates with H. pylori prevalence and GLOBOCAN 2022 colorectal cancer case counts to estimate population attributable fractions globally, regionally and by country.
Across 43 studies including 48,694,403 people, H. pylori exposure was associated with a 1.59-fold increase in colorectal cancer risk. On that basis, about 22% of colorectal cancer cases worldwide, roughly one in five, were estimated to be potentially attributable to H. pylori. The proportion was highest in the Western Pacific region (25%) and in Mongolia (36%), and China had the largest absolute number of potentially attributable cases, estimated at 139,619. When the analysis was restricted to the 14 population-based and cohort studies, which are less prone to selection bias than case-control designs, the attributable fraction fell to 12%.
Exploratory birth-cohort analyses suggested an increasing attributable share in more recent cohorts: 9.5% among people born before 1947, about 10% for those born in 1948-62, and 15% for those born in 1963-77. In 4 observational studies, eradication therapy was associated with reduced colorectal cancer risk, but only after 10 years. The association did not differ significantly by sex, by age above or below 60 years, by diagnostic method, or by the presence of a national screening programme. Confidence intervals for the attributable fractions were not reported in the press materials reviewed for this summary.
The authors are explicit that the work is exploratory and does not establish causality. Attributable-fraction modelling assumes a causal relationship and is sensitive to the input relative risk; the halving of the estimate when only population-based and cohort studies were used shows how much study design drives the result. Residual confounding by socioeconomic status, diet and other infections that cluster with H. pylori is likely. The birth-cohort trend was descriptive rather than formally tested, and evidence on eradication rests on only four observational studies.
If the association proves causal, H. pylori control programs designed to prevent gastric cancer could also reduce colorectal cancer burden in high-prevalence regions, although the authors caution that benefits will depend on local factors and that large-scale eradication raises concerns about antibiotic resistance and side effects. They call for well-designed randomized trials and prospective studies to test whether H. pylori and its eradication affect colorectal cancer risk, which would also clarify the plausibility of a gastric-colonic link.
Source: Li et al., "Global and regional burden of colorectal cancer potentially related to Helicobacter pylori exposure: a risk attribution modelling study," eGastroenterology, September 8, 2026. Read the paper
6. Relapse risk and adjuvant chemotherapy in localized appendiceal adenocarcinoma (MD Anderson and MSK cohorts, JAMA Surgery)
Sacha El Khoury (Department of Gastrointestinal Medical Oncology, University of Texas MD Anderson Cancer Center, Houston), et al., with co-authors from Memorial Sloan Kettering Cancer Center; senior author John Paul Shen (University of Texas MD Anderson Cancer Center)
Key points
- Retrospective cohort of 439 patients with stage I-III appendiceal adenocarcinoma at MD Anderson (2000-2024; 202 resected there), with a validation cohort of 128 stage II patients from Memorial Sloan Kettering; median follow-up 62.6 months.
- Relapse after resection was uncommon: 19 of 202 surgical patients (9.4%), including 6% with stage II and 19.5% with stage III disease.
- Relapse strongly predicted death: 5-year overall survival 95.7% without relapse versus 77.2% with relapse (HR 5.50; 95% CI 3.07-9.83; P<.001).
- Compared with goblet cell tumors, mucinous (HR 5.60; 95% CI 2.1-15) and enteric-type histology (HR 6.60; 95% CI 2.9-15) and pathologic T4 (HR 3.30; 95% CI 1.9-5.7) were independently associated with relapse.
- Adjuvant chemotherapy was not associated with benefit after adjustment: relapse-free survival HR 0.98 (95% CI 0.43-2.28; P=.90) and overall survival HR 0.71 (95% CI 0.24-2.1; P=.53).
Key limitations
- Retrospective design with a single-institution primary cohort; treatment allocation was not randomized and confounding by indication is likely.
- Few relapse events (19) limit the precision of multivariable and molecular subgroup estimates.
- Wide confidence intervals around the adjuvant chemotherapy estimates cannot exclude a modest benefit.
Future directions
- Prospective, histology-specific trials of adjuvant therapy in appendiceal adenocarcinoma rather than extrapolation from colon cancer.
- Validation of TP53 and GNAS as relapse biomarkers in independent cohorts.
- Risk-adapted surveillance and treatment strategies that distinguish goblet cell, mucinous and enteric-type tumors.
Appendiceal adenocarcinoma is a rare malignancy that is frequently managed with protocols borrowed from colon cancer, including adjuvant fluoropyrimidine-based chemotherapy after resection of node-positive or high-risk disease. Yet appendiceal tumors comprise distinct histologic subtypes, notably goblet cell, mucinous and enteric-type (colonic-type) adenocarcinomas, with different biology and natural history, and their genomic profiles differ from those of colorectal cancer. Because the disease is uncommon, randomized evidence to guide treatment of localized appendiceal adenocarcinoma is essentially absent.
Without disease-specific trials, clinicians extrapolate from colon cancer data, where adjuvant chemotherapy provides a well-established survival benefit in stage III disease. Whether that benefit transfers to appendiceal adenocarcinoma, and which patients are at sufficient risk of relapse to justify chemotherapy, has remained unclear. Large single-institution series with long follow-up and molecular annotation are therefore one of the few available sources of guidance for estimating relapse risk by histology and stage and for exploring the association between adjuvant treatment and outcomes.
El Khoury, Shen and colleagues at the University of Texas MD Anderson Cancer Center performed a retrospective cohort study of 439 patients with stage I-III appendiceal adenocarcinoma seen between January 2000 and February 2024, of whom 202 underwent surgery at MD Anderson. A validation cohort of 128 patients with stage II disease came from Memorial Sloan Kettering Cancer Center. The exposure was surgical resection with or without adjuvant chemotherapy, and relapse-free and overall survival were analyzed with Kaplan-Meier methods and Cox proportional hazards models. Median follow-up was 62.6 months.
Relapse after resection of localized appendiceal adenocarcinoma was uncommon, occurring in 19 of 202 surgical patients (9.4%); the abstract reports relapse in 6% of patients with stage II and 19.5% with stage III disease. When relapse occurred, it carried substantial prognostic weight: 5-year overall survival was 95.7% among patients without relapse versus 77.2% among those who relapsed (HR 5.50; 95% CI 3.07-9.83; P<.001). Histologic subtype was a major determinant of relapse. Compared with goblet cell tumors, which had the lowest relapse risk, mucinous (HR 5.60; 95% CI 2.1-15; P<.001) and enteric-type tumors (HR 6.60; 95% CI 2.9-15; P<.001) were independently associated with relapse, as was pathologic T4 stage (HR 3.30; 95% CI 1.9-5.7; P<.001).
Several features often used to justify adjuvant therapy, including poor differentiation, perforation, lymphovascular invasion and perineural invasion, were not significantly associated with relapse. In the full localized cohort, adjuvant chemotherapy appeared harmful on univariate analysis (relapse-free survival HR 2.06; 95% CI 1.36-3.13; overall survival HR 1.80; 95% CI 1.0-3.2), reflecting its use in higher-risk patients, but after multivariable adjustment there was no association with relapse-free survival (HR 0.98; 95% CI 0.43-2.28; P=.90) or overall survival (HR 0.71; 95% CI 0.24-2.1; P=.53). Molecular factors also mattered: TP53 mutation in goblet cell tumors (HR 6.93; 95% CI 1.50-31.00) and GNAS mutation in nongoblet tumors (HR 17.0; 95% CI 3.09-93.3) were associated with relapse.
The retrospective design and single-institution primary cohort mean that the decision to give chemotherapy was not random, and the reversal between univariate and adjusted estimates illustrates strong confounding by indication that adjustment may not fully remove. With only 19 relapse events among resected patients, multivariable and molecular subgroup estimates are imprecise, as the wide confidence intervals for GNAS and TP53 show. The adjusted chemotherapy estimates are compatible with both meaningful benefit and harm, so the null finding should be read as absence of evidence rather than proof of no effect. The external validation cohort was restricted to stage II disease.
The study suggests that relapse after resection of localized appendiceal adenocarcinoma is uncommon and varies by histologic subtype and molecular features, and that routine extrapolation of colon cancer adjuvant chemotherapy to all patients may not be justified. Goblet cell tumors appear to carry the lowest relapse risk, while mucinous and enteric-type histology and T4 stage identify patients for whom treatment decisions and surveillance deserve more attention. The findings argue for histology-specific prospective trials and for validating TP53 and GNAS as biomarkers to guide risk-adapted management.
Source: El Khoury et al., "Risk of Relapse and Efficacy of Adjuvant Chemotherapy in Localized Appendiceal Adenocarcinoma," JAMA Surgery, published online September 9, 2026. Read the paper
7. Socioeconomic inequalities in colon cancer diagnosis and treatment in England: national population-based cohort (Cancer Epidemiology)
Benjamin A. Pickwell-Smith (Hull York Medical School, University of Hull, and Queen's Centre for Oncology and Haematology, Hull University Teaching Hospitals), et al.; senior author Una Macleod (Hull York Medical School, University of Hull)
Key points
- Population-based cohort of people diagnosed with colon cancer in England in 2016-2017, using National Cancer Registry data linked to Hospital Episode Statistics, the Cancer Pathway, the Systematic Anti-Cancer Dataset and the Diagnostic Imaging Dataset (NCT05185388).
- Patients in the most deprived quintile had lower odds of surgery than the least deprived (OR 0.62; 95% CI 0.56-0.69).
- Among stage II-III patients, the most deprived had lower odds of adjuvant chemotherapy (OR 0.73; 95% CI 0.65-0.83).
- Among stage IV patients, the most deprived had lower odds of receiving chemotherapy (OR 0.63; 95% CI 0.54-0.74).
- Diagnostic and treatment intervals were modestly longer in more deprived areas, by adjusted differences of about 2-5 days at the median and upper quartile.
Key limitations
- Area-level deprivation (Index of Multiple Deprivation) may misclassify individual socioeconomic position.
- Registry data cannot fully capture patient preference, frailty or fitness for treatment, so some differences may reflect appropriate clinical choice.
- Diagnoses from 2016-2017 predate pandemic-era and later service changes.
Future directions
- Identifying the structural and system-level barriers that drive lower treatment receipt in deprived groups.
- Targeted interventions, such as navigation or prehabilitation, tested for their effect on treatment equity.
- Monitoring treatment inequalities in more recent cohorts and their effect on survival.
Socioeconomic inequalities in cancer survival persist in high-income countries, including those with universal health coverage such as the United Kingdom. In colon cancer, survival depends heavily on receipt of surgery and, for selected patients, adjuvant or palliative chemotherapy, as well as on timely diagnosis. Understanding where along the care pathway inequalities arise is essential for designing effective interventions, because differences in survival could stem from later diagnosis, from differences in treatment, or from both.
Previous work in England and elsewhere has documented poorer outcomes for patients from more deprived areas, but many studies lacked linked data on both diagnostic intervals and the full range of treatments, or examined these in isolation. National linkage of cancer registration with hospital, chemotherapy, imaging and cancer pathway data now allows the entire pathway to be examined within one population-based cohort. This makes it possible to ask whether deprivation is associated with receipt of each major treatment and with the timing of diagnosis and treatment.
Pickwell-Smith, Macleod and colleagues at Hull York Medical School and Hull University Teaching Hospitals used National Cancer Registry data for people diagnosed with colon cancer in England in 2016-2017, linked to Hospital Episode Statistics, the Cancer Pathway data, the Systematic Anti-Cancer Dataset and the Diagnostic Imaging Dataset. Deprivation was measured in quintiles of the Index of Multiple Deprivation. Multivariable logistic regression assessed receipt of surgery, adjuvant chemotherapy for stage II-III disease and chemotherapy for stage IV disease, and quantile regression assessed diagnostic and treatment intervals. The study was registered as NCT05185388.
Clear treatment gradients emerged. Patients living in the most deprived quintile had 38% lower odds of undergoing surgery than those in the least deprived quintile (OR 0.62; 95% CI 0.56-0.69). Similar inequalities were seen for adjuvant chemotherapy among patients with stage II-III disease (OR 0.73; 95% CI 0.65-0.83) and for chemotherapy among those with stage IV disease (OR 0.63; 95% CI 0.54-0.74). These differences persisted after multivariable adjustment, indicating that deprivation was associated with lower receipt of potentially beneficial treatment at every stage examined.
Diagnostic and treatment intervals were also modestly longer for patients in more deprived areas, with adjusted differences of roughly 2 to 5 days at the median and upper quartiles of the interval distributions. Confidence intervals for these timing differences were not reported in the abstract. The authors interpret the overall pattern, lower treatment receipt combined with slightly longer intervals, as reflecting structural and systemic barriers rather than isolated delays at a single point in the pathway.
Because deprivation was measured at area level, individual socioeconomic circumstances may be misclassified, which would tend to dilute rather than exaggerate associations. Registry data cannot fully capture comorbidity severity, frailty, performance status or patient preference, so part of the treatment gap may reflect appropriate, individualized decisions rather than inequitable care. The cohort was diagnosed in 2016-2017, before pandemic-era disruption and subsequent service reforms, and the abstract does not report how much of the survival gap these treatment differences explain.
The study shows that, even within a universal health system, people with colon cancer from more deprived areas are less likely to receive surgery and chemotherapy and wait slightly longer along the pathway. The authors call for targeted interventions to address the structural barriers involved. Next steps include identifying the specific drivers, such as transport, comorbidity management or referral patterns, testing interventions like patient navigation or prehabilitation, and tracking whether treatment equity improves in more recent cohorts.
Source: Pickwell-Smith et al., "Are there inequalities in colon cancer diagnosis and treatment in England? A population-based study," Cancer Epidemiology, published online September 12, 2026. PubMed