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. First surveillance colonoscopy at 5 versus 3 years after high-risk adenoma removal: interim analysis (EPoS II, New England Journal of Medicine)
Rodrigo Jover (Hospital General Universitario Dr. Balmis, Alicante, Spain), Michael Bretthauer (Clinical Effectiveness Research Group, University of Oslo), J. Cubiella, S. J. E. Barry, et al.; corresponding authors Rodrigo Jover and Michael Bretthauer
Key points
- Randomized noninferiority trial of 10,799 patients aged 40 to 74 years with high-risk adenomas at 49 endoscopy centers in 8 European countries: 5,398 assigned to a first surveillance colonoscopy at 5 years and 5,401 to a first colonoscopy at 3 years (plus a further one at 5 years).
- At the 5.5-year interim analysis, cumulative colorectal cancer incidence was 0.77% with the 5-year interval versus 0.82% with the 3-year interval (difference -0.05 percentage points; upper bound of the 99.12% CI 0.68, within the 0.7-point noninferiority margin).
- Colorectal cancer deaths were rare: 3 (0.06%) in the 5-year group versus 2 (0.04%) in the 3-year group, with no significant differences in early- or late-stage diagnosis.
- The 5-year strategy required 4,434 follow-up colonoscopies versus 8,614, a 48.5% reduction.
- Weighted detection was lower with the 5-year interval for advanced adenomas (11.07% vs 14.55%) and any adenoma (54.82% vs 71.77%); the clinical value of the extra detection is unknown.
Key limitations
- Interim analysis; the primary endpoint is 10-year cumulative colorectal cancer incidence, which has not yet been reported.
- Noninferiority was met only narrowly against a 0.7-percentage-point margin.
- Only about 70% (5-year group) and 65% (3-year group) completed the planned 5-year colonoscopy.
Future directions
- Final 10-year incidence results to confirm whether the 5-year interval remains noninferior.
- Translation into surveillance guidelines and colonoscopy capacity planning for patients with high-risk adenomas.
- Clarifying whether the additional adenomas found with 3-year surveillance translate into any patient benefit.
Surveillance colonoscopy after polypectomy consumes a large and growing share of endoscopy capacity in countries with organized colorectal cancer screening. Patients whose adenomas are judged high risk, traditionally defined by size of 10 mm or more, high-grade dysplasia or villous histology, or multiple adenomas, have generally been advised to return for a first surveillance examination three years after clearance. The interval has been justified by concern that such patients are at elevated risk of metachronous advanced neoplasia and cancer, yet the strength of the evidence behind a three-year recommendation, rather than a longer one, has long been debated, and each surveillance examination carries cost, discomfort and a small procedural risk.
Much of the evidence base for post-polypectomy surveillance has come from observational cohorts, which cannot readily separate the effect of the surveillance interval from the characteristics of the patients and endoscopists involved. Guideline recommendations have therefore rested on limited evidence, and practice varies between and within countries. A large randomized comparison of surveillance intervals, powered for colorectal cancer incidence rather than adenoma detection, was needed to establish whether colonoscopies could safely be spaced further apart in this population, and the EPoS program was designed to provide that evidence.
EPoS II enrolled 10,799 patients aged 40 to 74 years with high-risk adenomas, defined as at least one adenoma of 10 mm or more, high-grade dysplasia or villous growth, or three to ten adenomas of any kind, at 49 endoscopy centers in Spain, Norway, Denmark, Poland, the Netherlands, Sweden, Portugal and Austria between April 2015 and March 2020. Patients were randomized to a first surveillance colonoscopy at 5 years, followed by a further examination at 10 years (5,398 patients; mean age 62.6 years; 65.1% men), or at 3 years, followed by examinations at 5 and 10 years (5,401 patients; mean age 62.2 years; 65.3% men). The primary endpoint is cumulative colorectal cancer incidence at 10 years, with a noninferiority margin of 0.7 percentage points; the present report is an interim analysis at 5.5 years.
At the interim analysis, the cumulative incidence of colorectal cancer was 0.77% in the 5-year group and 0.82% in the 3-year group, an absolute difference of -0.05 percentage points. The upper boundary of the 99.12% confidence interval for the difference was 0.68 percentage points, which lay within the prespecified margin, so the 5-year strategy met the criterion for noninferiority. More than 99% of patients in both groups remained free of a colorectal cancer diagnosis. Colorectal cancer deaths were few: three (0.06%) in the 5-year group and two (0.04%) in the 3-year group. There were no significant differences between the groups in the proportions of cancers diagnosed at early or late stage.
The resource implications were substantial. Patients in the 5-year group underwent 4,434 colonoscopies during follow-up compared with 8,614 in the 3-year group, a reduction of 48.5%. As expected, the more intensive schedule found more neoplasia: weighted advanced adenoma detection was 11.07% with the 5-year interval and 14.55% with the 3-year interval, and weighted detection of any adenoma was 54.82% and 71.77%, respectively. Completion of the planned 5-year surveillance colonoscopy was about 70% in the 5-year group and 65% in the 3-year group. The investigators noted that whether the higher level of detection is beneficial for patients "remains to be seen."
Several features temper the conclusions. This is an interim look at a trial whose primary endpoint is 10-year incidence, and noninferiority was met narrowly, with the upper confidence bound close to the margin. Adherence to the 5-year examination was incomplete in both arms, which may dilute differences between strategies. Event counts are small, so estimates for colorectal cancer mortality are imprecise. The between-group difference in advanced adenoma detection was small, and the trial does not test directly whether detecting more adenomas lowers cancer risk; the authors cite observational data suggesting that adenoma detection beyond roughly 25% to 30% may add little protection, but that hypothesis lies outside the trial.
Even as an interim result, EPoS II provides strong randomized evidence that the first surveillance colonoscopy after high-risk adenoma removal can be deferred from 3 to 5 years without an apparent increase in colorectal cancer incidence, while nearly halving the number of surveillance procedures. If the 10-year results confirm these findings, guideline committees will have grounds to lengthen recommended intervals, freeing endoscopy capacity for screening and diagnostic work. In the meantime, the data support shared decision-making with patients who would prefer fewer procedures, and they reinforce the importance of high-quality baseline colonoscopy and complete polyp clearance.
Source: Jover et al. (EPoS II), "Colonoscopy Intervals and Colorectal Cancer Incidence after Adenoma Removal," New England Journal of Medicine 2026;395:1051-1061 (September 2026). Read the paper · PubMed
2. Covalent Werner helicase inhibitor RO7589831 (VVD-214) in advanced MSI/dMMR solid tumors, including colorectal cancer (first-in-human phase 1, Nature Medicine)
Timothy A. Yap (The University of Texas MD Anderson Cancer Center), Elisa Fontana, Natalie Cook, Marwan Fakih, John H. Strickler, et al.; senior author Sophie Postel-Vinay (Gustave Roussy and UCL Cancer Institute)
Key points
- First-in-human phase 1 trial (NCT06004245, part 1) of an oral covalent WRN inhibitor in 88 patients with MSI and/or dMMR advanced solid tumors at 17 centers in eight countries; 64.8% had colorectal cancer and 95.5% had received a prior checkpoint inhibitor.
- Among 66 MSI efficacy-evaluable patients, the disease control rate was 74.2% (95% CI 63.9-82.9) and the objective response rate 10.6% (7 confirmed partial responses), with median duration of response 10.2 months.
- Median progression-free survival was 6.7 months (95% CI 4.1-8.5) and median overall survival 17.6 months (95% CI 17.6-not estimable).
- In 41 evaluable patients with colorectal cancer, the disease control rate was 80.5%, three patients responded (7.3%), median PFS was 7.3 months and estimated 12-month OS 78.3%.
- No maximum tolerated dose was identified; treatment-related grade 3-4 events occurred in 15.9%, discontinuation for adverse events in 3.4%, and there were no grade 5 events. Recommended phase 2 doses: 150 mg and 600 mg twice daily.
Key limitations
- Single-arm, dose-escalation study without a comparator; response rates were low and disease control is hard to interpret without a control arm.
- Direct intratumoral target engagement was not demonstrated: paired biopsies from 26 patients yielded no interpretable pharmacodynamic results.
- Small subgroups outside colorectal cancer, and a change of sponsor from Roche to Vividion during the study.
Future directions
- Formal dose optimization between the two recommended phase 2 doses.
- Combination studies with bevacizumab (phase 1b) and with pembrolizumab in untreated MSI-H/dMMR colorectal cancer (phase 2, NCT07811193).
- Development of pharmacodynamic and selection biomarkers, and of endpoints suited to an agent that may mainly stabilize disease.
Microsatellite instability (MSI) and mismatch repair deficiency (dMMR) define a subset of colorectal and other solid tumors that are highly sensitive to immune checkpoint blockade. Yet some patients with MSI/dMMR metastatic colorectal cancer are primarily resistant to immunotherapy or progress after an initial response, and once checkpoint inhibitors fail, the remaining options are largely the conventional chemotherapy and late-line agents used in microsatellite-stable disease. New strategies that exploit the biology of MSI tumors themselves, rather than the immune response to them, are therefore of considerable interest.
Werner helicase (WRN) emerged from functional genomic screens as a synthetic lethal dependency of MSI cancer cells: MSI tumors accumulate expanded DNA repeats that appear to require WRN for their resolution, whereas microsatellite-stable cells do not depend on it. This selective vulnerability makes WRN an attractive drug target with, in principle, a wide therapeutic window. Initial data with this covalent WRN inhibitor were presented at the 2025 AACR Annual Meeting, and the present paper reports the full first-in-human experience, in a population that was overwhelmingly checkpoint-inhibitor pretreated.
The trial (NCT06004245, part 1) used dose escalation with randomized backfill, enrolling 88 patients at 17 centers in eight countries between January 2024 and June 2025, with data cutoff in September 2025. RO7589831 (VVD-133214), an oral agent that binds WRN covalently at Cys727, was tested at 150 mg once daily, 150 to 1,000 mg twice daily and 200 to 600 mg three times daily. Patients had a median age of 60 years and a median of three prior lines of therapy; 64.8% had colorectal cancer, 9.1% endometrial cancer, and 95.5% had received a checkpoint inhibitor. Efficacy was assessed by RECIST 1.1 in 66 MSI efficacy-evaluable patients.
Among the 66 evaluable patients, disease control was achieved in 49 (74.2%; 95% CI 63.9-82.9), with seven confirmed partial responses (objective response rate 10.6%): three in colorectal cancer, three in endometrial cancer and one in ovarian cancer. Median duration of response was 10.2 months, and five of seven responders remained in response at cutoff. Median progression-free survival was 6.7 months (95% CI 4.1-8.5) and median overall survival 17.6 months. In the 41 evaluable patients with colorectal cancer, disease control was 80.5%, three patients (7.3%) responded, median PFS was 7.3 months and estimated 12-month overall survival was 78.3%. No responses occurred among seven patients found to be microsatellite stable on central testing or nine with unknown MSI status.
The safety profile was dominated by gastrointestinal effects. Grade 3-4 treatment-emergent events occurred in 46.6% of the 88 patients and treatment-related grade 3-4 events in 15.9%, most commonly transaminase elevations; there were no grade 5 events and only 3.4% discontinued for an adverse event. The most frequent events were nausea (54.5%), diarrhea (43.2%), fatigue (39.8%), anemia (37.5%) and vomiting (30.7%). No maximum tolerated dose was found; the only dose-limiting toxicity was grade 2 nausea. Higher peak concentrations tracked with gastrointestinal toxicity, and a short half-life of about 4.6 hours led to twice-daily dosing. Recommended phase 2 doses of 150 mg and 600 mg twice daily were selected for formal optimization.
As a first-in-human study, the trial was not designed to estimate efficacy precisely, and the absence of a comparator makes the high disease control rate difficult to interpret in a population with variable disease kinetics. Objective responses were infrequent. Importantly, target engagement in tumor tissue could not be shown: paired biopsies from 26 patients were uninterpretable, so the degree of WRN inhibition rests on pharmacokinetic modeling and on supportive but indirect circulating tumor DNA and FDG-PET changes. Candidate selection biomarkers such as TA-repeat expansions could not be characterized. The authors argue that RECIST may undervalue an agent that stabilizes rather than shrinks tumors, but that remains a hypothesis.
These data provide the first clinical evidence that pharmacologic WRN inhibition is tolerable and has activity in checkpoint-refractory MSI tumors, with colorectal cancer the main population studied. Responses were few, so the immediate path forward lies in combinations and in earlier lines of therapy. A phase 1b study is combining the drug with bevacizumab in previously treated MSI-H/dMMR colorectal cancer, and a phase 2 trial is testing it with pembrolizumab in untreated disease. Robust pharmacodynamic markers and appropriate endpoints will be essential to judge whether WRN inhibition can become a standard option after or alongside immunotherapy.
Source: Yap et al., "Werner helicase inhibitor for advanced microsatellite instability solid tumors: a phase 1 trial," Nature Medicine, September 29, 2026. Read the paper · PubMed
3. Pretreatment gut microbiome signatures of cancer type, early-onset colorectal cancer and fluorouracil-associated diarrhea in a real-world cohort (Mayo Clinic, Cell)
Mayo Clinic investigators (Mayo Clinic, Arizona, Florida and Minnesota), et al.; co-corresponding authors Ruben Mars and Purna Kashyap (Mayo Clinic)
Key points
- Real-world cohort of 1,364 patients with cancer from 40 U.S. states, with pretreatment stool samples linked to clinical data on treatment side effects and outcomes, compared with 287 people without cancer.
- 341 bacterial species were associated with five cancer groups (neuroendocrine tumors, liver and intrahepatic bile duct cancers, esophageal cancer, lymphoid leukemia, and multiple myeloma and related plasma cell cancers).
- Patients with colorectal cancer diagnosed at age 50 or younger had higher lactate levels and more Veillonella parvula, a lactate-consuming bacterium, than those diagnosed later in life.
- Patients receiving 5-fluorouracil who later developed diarrhea had fewer bacterial genes capable of breaking down the drug, much of this function coming from Anaerostipes hadrus; the signal was not seen with carboplatin.
- The associations are observational and do not establish that microbial or metabolic differences cause early-onset disease or toxicity.
Key limitations
- Observational design: associations cannot establish causality, and the direction of the lactate-Veillonella link in early-onset colorectal cancer is unknown.
- Single health system (Mayo Clinic) cohort; findings need external validation in other populations.
- The toxicity signal for 5-fluorouracil requires confirmation in larger patient groups before any clinical use.
Future directions
- Studies of whether cancer-specific microbial signals, including V. parvula in early-onset colorectal cancer, play a causal role.
- Validation of microbiome-based prediction of fluoropyrimidine toxicity in larger patient groups.
- Exploration of whether microbial drug-metabolizing capacity could be used to personalize or protect against chemotherapy side effects.
The gut microbiome has been implicated in colorectal carcinogenesis through inflammatory, genotoxic and metabolic mechanisms, and it is increasingly recognized as a possible modifier of response and toxicity with systemic cancer therapy. Interest has intensified with the rise in colorectal cancer among younger adults: the release accompanying this study cites American Cancer Society figures of an increase in incidence of about 3% a year among adults aged 20 to 49. Whether microbial differences are specific to particular cancers, or simply reflect illness, treatment and lifestyle, has been difficult to establish.
Many microbiome-cancer studies have compared one tumor type with healthy controls, often after treatment has begun, which makes it hard to separate disease-specific signals from the effects of therapy or hospitalization, or from features shared by many cancers. Comparing several cancer types against one another, and sampling stool before treatment begins, offers a way to narrow the search to microbial changes most specific to individual cancers. As co-corresponding author Ruben Mars put it, those are "the signals we need to understand first" to determine whether the microbiome plays a causal role. Pretreatment sampling also allows microbial features to be related to subsequent toxicity.
Investigators at Mayo Clinic assembled a real-world cohort of 1,364 patients with cancer recruited at sites in Arizona, Florida and Minnesota and drawn from 40 U.S. states, together with 287 people without cancer as comparators. Stool samples were collected before treatment and linked to clinical records describing subsequent side effects and outcomes. Analyses compared patients with cancer to those without, and compared cancer types with one another, to identify taxa and microbial functions associated with specific cancers, with early-onset presentation, and with treatment-related adverse events.
Across the cohort, 341 bacterial species were associated with five cancer groups. Neuroendocrine tumors showed a broad loss of common health-associated gut bacteria; liver and intrahepatic bile duct cancers had higher levels of several bacteria, including Enterococcus faecalis; and esophageal cancer had higher levels of six species, including Streptococcus. Distinct associations were also found for lymphoid leukemia and for multiple myeloma and related plasma cell cancers. For colorectal cancer, the key finding concerned age: patients diagnosed at 50 or younger had higher lactate levels and more Veillonella parvula, a bacterium that feeds on lactate, than patients diagnosed later in life. Tumors can produce high levels of lactate, but whether this relationship contributes to early-onset disease is not known.
The treatment analyses produced a drug-specific toxicity signal. Patients treated with 5-fluorouracil who went on to develop diarrhea had lower levels of bacterial genes capable of breaking down the drug, with much of that capacity attributable to Anaerostipes hadrus. The same signal was not linked to diarrhea in patients receiving carboplatin, suggesting that it reflects fluoropyrimidine metabolism rather than a nonspecific marker of gastrointestinal vulnerability. The study also identified 64 bacterial species that differed in early-onset breast cancer. Co-corresponding author Purna Kashyap described the toxicity result as "a proof of concept" for understanding why some patients experience particular side effects.
The design is observational, and the investigators state explicitly that the colorectal and breast cancer findings do not establish that microbial or metabolic differences cause early-onset disease; they identify links that warrant further study. The higher lactate and V. parvula levels could be a consequence of the tumor as easily as a contributor to it. The cohort comes from a single, albeit multistate, health system, and real-world sampling introduces variation in diet, medication and comorbidity that may not be fully captured. The fluorouracil toxicity association, while biologically plausible, will need replication before it could inform dosing or supportive care.
By sampling stool before treatment and contrasting many cancer types at once, this study offers a more specific map of cancer-associated microbial features than single-disease comparisons, and it nominates concrete hypotheses for colorectal oncology. The investigators plan to test whether cancer-specific microbial signals play a causal role in disease and to validate the microbiome's ability to predict treatment side effects in larger patient groups. If microbial drug-metabolizing capacity reliably predicts fluoropyrimidine toxicity, it could complement existing approaches to toxicity risk, and the lactate-Veillonella axis merits mechanistic work in early-onset colorectal cancer.
Source: Mayo Clinic investigators (Mars, Kashyap et al.), "Microbiome signatures linked to cancer and treatment adverse events in a real-world cohort," Cell, September 29, 2026. Read the paper · Mayo Clinic release
4. Medicaid expansion and overall mortality among adults aged 40 to 64 with colorectal cancer (National Cancer Database, JAMA Network Open)
Oluwasegun Akinyemi (Clive O. Callender Outcomes Research Center, Howard University College of Medicine), Oladayo Oyebanji, Babawale Oluborode, Kakra Hughes, Terrence Fullum, Quyen Chu, et al.; last author Craig Reickert (Henry Ford Hospital, Detroit)
Key points
- Retrospective difference-in-differences cohort of 658,593 adults aged 40-64 with colorectal cancer in the National Cancer Database, 2006-2021 (316,097 pre-ACA, 342,496 post-ACA).
- Residence in a Medicaid expansion state was associated with lower overall mortality: DID HR 0.94 (95% CI 0.93-0.95), a 6.3% relative reduction (95% CI 5.1%-7.5%).
- The largest association was among Hispanic patients (HR 0.77; 95% CI 0.75-0.80); HRs for Black and White patients were 0.96 (95% CI 0.93-0.99 and 0.94-0.97, respectively).
- Benefits were seen at every stage and were most pronounced in stage IV (HR 0.88; 95% CI 0.83-0.94).
- Expansion was associated with a 2.13-percentage-point increase in early-stage diagnosis (95% CI 1.60-2.65) and a 1.54-point increase in surgical treatment (95% CI 1.19-1.90).
Key limitations
- Observational design; associations, not causal effects, despite difference-in-differences methods.
- The NCDB does not record cause of death, so colorectal cancer-specific mortality could not be assessed.
- Hospital-based registry covering about 70% of new U.S. cancers; states with late or partial expansion were excluded.
Future directions
- Understanding why the survival benefit differs across racial and ethnic groups.
- Directly testing earlier detection and access to definitive treatment as mechanisms.
- Informing policy debates on Medicaid eligibility and funding using cancer outcome data.
Insurance coverage shapes nearly every step of the colorectal cancer pathway, from access to screening and timely diagnostic colonoscopy to surgery, systemic therapy and survivorship care. The Affordable Care Act (ACA) allowed states to expand Medicaid to adults with incomes up to 138% of the federal poverty level; 40 states and the District of Columbia have adopted expansion, but it remains politically contested. Colorectal cancer is the second leading cause of cancer death in the United States, and the 40- to 64-year-old adults most affected by expansion span patients with early-onset disease through the years just before Medicare eligibility.
Prior work has linked Medicaid expansion to increased coverage and, for several cancers, improved outcomes; this group had previously examined breast, lung and cervical cancers. For colorectal cancer, however, the magnitude of any survival association, its variation by race, ethnicity and stage, and the mechanisms through which it might operate have been less well defined. Large registry datasets spanning periods well before and after expansion allow a difference-in-differences design, which compares changes over time in expansion states with those in nonexpansion states and so accounts for secular trends in outcomes.
Akinyemi and colleagues analyzed the National Cancer Database (NCDB), a hospital-based registry capturing roughly 70% of new U.S. cancers, for adults aged 40 to 64 with histologically confirmed colorectal cancer diagnosed between 2006 and 2021. Patients older than 64 were excluded to limit confounding by Medicare, and states with late or partial expansion were excluded. Pre-ACA (2006-2013; n=316,097) and post-ACA (2014-2021; n=342,496) periods were compared between states that expanded Medicaid in 2010-2014 and those that did not. Multivariable Cox models included an expansion-by-period interaction, adjusted for demographic, clinical, facility and neighborhood factors, with event-study checks of parallel trends.
The cohort comprised 658,593 adults (mean age 54.9 years; 55.7% male). Post-ACA, 2.4% of patients in expansion states were uninsured compared with 9.7% in nonexpansion states, and Medicaid coverage was 19% versus 9.5%. Expansion was associated with lower overall mortality, with a difference-in-differences hazard ratio of 0.94 (95% CI 0.93-0.95), corresponding to a 6.3% relative reduction (95% CI 5.1%-7.5%). The association was present across racial and ethnic groups and was largest among Hispanic patients (HR 0.77; 95% CI 0.75-0.80), with smaller but significant reductions among Black (HR 0.96; 95% CI 0.93-0.99) and White (HR 0.96; 95% CI 0.94-0.97) patients.
Benefits appeared at all stages: HRs were 0.95 for stage I, 0.97 for stage II, 0.95 for stage III and 0.88 (95% CI 0.83-0.94) for stage IV, the last representing an 11.6% lower risk of death. In mechanism analyses, expansion was associated with a 2.13-percentage-point increase in early-stage diagnosis (95% CI 1.60-2.65) and a 1.54-point increase in surgical treatment (95% CI 1.19-1.90), along with small changes in other treatment use, including decreases in recorded chemotherapy and immunotherapy that were not explained in coverage of the study. The lead author told Healio that the policy "also works for individuals who have advanced stage disease."
The design is observational, and even a difference-in-differences analysis can be biased if expansion and nonexpansion states diverged for reasons unrelated to Medicaid. Because the NCDB lacks cause of death, overall rather than colorectal cancer-specific mortality was the primary outcome. The NCDB is hospital-based rather than population-based, and excluding late or partial expansion states narrows the comparison. The Cox interaction estimates a multiplicative effect, which is why additive sensitivity analyses were performed. The authors acknowledge that the observed shifts in stage and treatment are consistent with, but do not establish, earlier detection and improved access as the mechanisms.
This is one of the largest analyses of Medicaid expansion and colorectal cancer outcomes, and its findings align with the group's earlier work in other cancers: broader coverage is associated with earlier diagnosis, more definitive treatment and longer survival, including for patients with metastatic disease who depend on continuous, costly care. The especially large association among Hispanic patients warrants further investigation, as the authors note. For clinicians and policymakers, the data add outcome-based evidence to debates about Medicaid eligibility and funding, and they underline insurance status as a modifiable determinant of colorectal cancer survival.
Source: Akinyemi et al., "Medicaid Expansion and Mortality Among Patients With Colorectal Cancer," JAMA Network Open, September 18, 2026. Read the paper
5. Statin initiation after colorectal cancer treatment and mortality: a target trial emulation (Korean nationwide cohort, JAMA Network Open)
Hyeong-taek Woo (Department of Preventive Medicine, Keimyung University School of Medicine, Daegu, Korea) and Aesun Shin (Department of Preventive Medicine, Seoul National University College of Medicine); corresponding author Hyeong-taek Woo
Key points
- Target trial emulation in 88,516 Korean adults with newly diagnosed colorectal cancer and no recent statin use (National Health Insurance Service data, 2005-2015): 2,071 initiated a statin within 6 months of first treatment and 86,445 did not.
- Statin initiation was associated with a modestly lower risk of colorectal cancer-specific death: HR 0.95 (95% CI 0.92-0.99); 60-month risk difference 1.1 percentage points (95% CI -0.1 to 2.3).
- All-cause mortality: HR 0.96 (95% CI 0.93-1.00), not statistically significant; 60-month risk difference 2.4 percentage points (95% CI 1.0-3.8).
- Deliberately introducing immortal time bias produced spurious HRs of 0.17 (cancer-specific) and 0.19 (all-cause); prevalent-user bias gave 0.81 and 0.82, and both biases combined 0.25 and 0.27.
- Conclusion: any survival benefit from post-treatment statin initiation is modest and far smaller than conventional observational estimates suggested.
Key limitations
- Cancer stage was unavailable; treatment modality served as a proxy, and residual confounding is possible.
- A 6-month washout may have misclassified former users as new users; statins were analyzed as a class.
- Few initiators limited assessment of effect modification, and the positive-control analysis gave only partial reassurance.
Future directions
- Randomized trials of statins in colorectal cancer survivors to settle the residual question of a small benefit.
- Application of target trial emulation to other repurposed drugs (for example, aspirin or metformin) where observational signals may be biased.
- Analyses with stage and molecular data to test whether any benefit is concentrated in particular subgroups.
Repurposing inexpensive, widely used drugs to improve outcomes after colorectal cancer treatment is an appealing prospect, and statins have been among the most studied candidates. Observational studies have repeatedly reported better survival among patients with colorectal cancer who use statins after diagnosis, and biologically plausible anticancer mechanisms have been proposed. On that basis, some have argued for broader statin use in survivors. Yet randomized evidence is lacking, and the apparent benefits in observational data have been strikingly large for a drug whose primary indication is cardiovascular prevention.
The authors argue that much of the earlier evidence is vulnerable to time-related biases. Immortal time bias arises when follow-up time before a patient starts a drug is wrongly credited to the exposed group, and prevalent-user bias arises when patients who have tolerated a drug for years are compared with nonusers. Informative censoring can further distort estimates. Target trial emulation addresses these problems by explicitly specifying the protocol of a hypothetical randomized trial, including eligibility, time zero, treatment strategies and follow-up, and then emulating it in observational data with methods such as cloning, censoring and weighting.
Woo and Shin used the Korean National Health Insurance Service database, which covers nearly the entire population, to identify adults aged 30 years or older newly diagnosed with colorectal cancer between 2005 and 2015 who had no statin prescriptions in the prior 6 months. Time zero was the first colorectal cancer treatment. The strategies compared were initiation of any statin within a 6-month grace period versus no initiation. A clone-censor approach with stabilized inverse probability of censoring weights, updated every 3 months for time-varying comorbidity, estimated hazard ratios, 60-month survival and risk differences over 60 months of follow-up, adjusted for age, sex, insurance status, comorbidity and treatment modality as a proxy for stage.
Of 88,516 eligible patients (mean age 62.7 years; 63.7% male), 2,071 initiated a statin within the grace period and 86,445 did not. Statin initiation was associated with a small reduction in colorectal cancer-specific mortality (HR 0.95; 95% CI 0.92-0.99), corresponding to a 60-month absolute difference of 1.1 percentage points whose confidence interval crossed zero (-0.1 to 2.3). For all-cause mortality the hazard ratio was 0.96 (95% CI 0.93-1.00), not statistically significant, although the 60-month risk difference of 2.4 percentage points (95% CI 1.0-3.8) did exclude zero. The authors characterized the benefit as modest.
The most instructive results came from analyses that deliberately reintroduced common design errors. Allowing immortal time bias produced hazard ratios of 0.17 for cancer-specific and 0.19 for all-cause mortality, an apparent 80% reduction in risk. Prevalent-user bias, even with clone-censor weighting, produced hazard ratios of 0.81 and 0.82, and combining both biases yielded 0.25 and 0.27. As a positive control, the same design applied to 8,212 patients with ischemic heart disease showed lower all-cause mortality with statin initiation (HR 0.77; 95% CI 0.65-0.92) but no clear association with cardiovascular death (HR 1.23; 95% CI 0.65-2.32).
The authors acknowledge several limitations. Cancer stage was not available, and treatment modality is an imperfect proxy for severity that may also reflect age, comorbidity, performance status or preference; residual confounding is possible. The 6-month washout may have misclassified remote former users as new users, which could inflate the apparent benefit. Statins were analyzed as a class, only one grace period was examined, and the small number of initiators limited analyses of effect modification. The positive-control results were only partly reassuring, because the cardiovascular-specific estimate was imprecise and did not move in the expected direction.
This study is a useful corrective. It suggests that the large survival benefits attributed to statins in earlier observational studies of colorectal cancer were largely artifacts of study design, and that any true effect of starting a statin after treatment is small. Statins should continue to be prescribed to survivors for their established cardiovascular indications, but the data do not support initiating them for cancer control. More broadly, the paper is a clear demonstration, using real data, of how immortal time and prevalent-user biases can manufacture apparent benefit, a lesson relevant to other repurposed drugs studied in colorectal cancer.
Source: Woo and Shin, "Statin Initiation and Mortality After Colorectal Cancer Treatment," JAMA Network Open, September 15, 2026. Read the paper
6. Longitudinal quality of life after short-course radiotherapy and FOLFOXIRI total neoadjuvant therapy for rectal cancer (SHORT-FOX, JAMA Network Open)
Katherine D. Arnow (S-SPIRE Center, Department of Surgery, Stanford University School of Medicine), Eleanor Brown, Erqi L. Pollom, Thomas Holden, Jennifer Shearer; corresponding author Arden M. Morris (Stanford University School of Medicine)
Key points
- Patient-reported outcomes from the single-group phase 2 SHORT-FOX trial: 37 adults with rectal adenocarcinoma received short-course radiotherapy followed by FOLFOXIRI (2020-2023); median age 52 years, 70.3% clinical T3.
- 23 patients (62.2%) had a clinical complete response at the post-treatment assessment and 18 (48.6%) pursued organ preservation.
- Overall self-rated health (EQ-5D) improved above baseline by 18 months (+6.50; 95% CI 0.48-12.51) and 24 months (+7.48; 95% CI 0.72-14.24).
- Erectile function in men declined from 12 months onward (IIEF change at 24 months -4.58; 95% CI -7.33 to -1.82); sexual function in women fell at 6 months (FSFI -4.15; 95% CI -8.19 to -0.10).
- Bowel function scores did not differ significantly from baseline by 24 months; LARS scores (no stoma) improved immediately after treatment (-6.55; 95% CI -11.70 to -1.40).
Key limitations
- Small single-group cohort (37 patients), particularly few women.
- Possible bias from attrition over 24 months of questionnaires.
- Comparison of surgical and nonsurgical patients is limited because the need for surgery is unknown before TNT.
Future directions
- Larger, comparative patient-reported outcome studies of intensified TNT regimens.
- Proactive assessment and management of sexual dysfunction after short-course radiotherapy.
- Integration of functional outcomes into counseling about organ preservation versus surgery.
Total neoadjuvant therapy (TNT), in which all planned chemotherapy and radiotherapy are delivered before any decision about surgery, has become a standard approach for locally advanced rectal cancer. By increasing the proportion of patients with a clinical complete response, TNT has also made watch-and-wait organ preservation a realistic option for many. As cure rates rise and more patients avoid total mesorectal excision, the long-term functional consequences of treatment, including bowel, urinary and sexual function and overall health, have become central to treatment choice and counseling.
Short-course radiotherapy followed by consolidation chemotherapy is attractive because it shortens the radiation phase and allows early delivery of systemic therapy. Intensifying the chemotherapy component with the triplet FOLFOXIRI may increase response rates further, but adds toxicity, and the patient-reported trajectory after such a regimen has been incompletely described. Function after TNT is also difficult to interpret because patients diverge after restaging: some enter surveillance with an intact rectum while others proceed to surgery. Prospective, repeated measurement from a defined trial cohort is needed to describe what patients actually experience.
SHORT-FOX was a single-group phase 2 trial at Stanford of FOLFOXIRI following short-course radiotherapy in adults with rectal adenocarcinoma. All 37 enrolled patients completed treatment between 2020 and 2023; their median age was 52 years, 35.1% were women, 27.0% were non-White and 70.3% had clinical T3 tumors. Validated instruments were administered at baseline, 8 weeks after treatment and at 6, 12, 18 and 24 months, including the EQ-5D for overall health, the International Index of Erectile Function (IIEF) for men, the Female Sexual Function Index (FSFI) for women, the International Prostate Symptom Score (IPSS) for urinary symptoms, and the MSKCC Bowel Function Instrument and LARS score for bowel function.
At the primary post-TNT assessment, 23 patients (62.2%) had a clinical complete response and 18 (48.6%) pursued organ preservation. Overall self-rated health was essentially unchanged at 6 and 12 months but rose above baseline at 18 months (mean change 6.50; 95% CI 0.48-12.51) and 24 months (7.48; 95% CI 0.72-14.24). Bowel function, measured by the MSKCC instrument, showed small, nonsignificant declines through 24 months (-2.56 at 24 months; 95% CI -6.35 to 1.23). Among patients without a stoma, LARS scores improved immediately after treatment (-6.55; 95% CI -11.70 to -1.40) and did not differ from baseline at later time points.
Sexual function was the domain most affected. In men, IIEF scores declined significantly from 12 months onward, with mean changes of -3.26 at 12 months, -3.85 at 18 months and -4.58 at 24 months (95% CI -7.33 to -1.82). In women, FSFI scores fell significantly at 6 months (-4.15; 95% CI -8.19 to -0.10) and remained lower at 24 months, although that estimate was imprecise (-3.93; 95% CI -8.38 to 0.52). Urinary symptoms on the IPSS worsened transiently at 12 months (2.64; 95% CI 0.66-4.61) but were not significantly different from baseline at 24 months.
The authors identify the main limitations: a small cohort, particularly of women, and possible bias from attrition across 24 months of follow-up. Estimates of sexual dysfunction were imprecise, especially at later time points. Because it is not known before TNT who will need surgery, planned comparisons between surgical and nonsurgical patients are of limited value. As a single-group study at one academic center, SHORT-FOX lacks a comparator arm, so the functional profile of short-course radiotherapy plus FOLFOXIRI cannot be directly contrasted with long-course chemoradiotherapy or doublet consolidation.
These data provide reassurance that an intensified TNT regimen can be followed by recovery and even improvement in self-rated health over two years, with bowel and urinary function returning close to baseline in this cohort, nearly half of whom preserved the rectum. They also highlight persistent sexual dysfunction, particularly erectile dysfunction in men, as a consequence that warrants explicit discussion before treatment and active management afterward. Larger comparative studies with standardized patient-reported outcomes will be needed to weigh the oncologic advantages of intensified TNT against its functional costs.
Source: Arnow et al., "Longitudinal Quality of Life After Total Neoadjuvant Therapy for Rectal Cancer in the SHORT-FOX Trial," JAMA Network Open, September 25, 2026. Read the paper
7. Tumor vascular phenotypes and colorectal cancer-specific mortality (Nurses' Health Study and Health Professionals Follow-up Study, JAMA Network Open)
Kosuke Matsuda (Program in Molecular Pathological Epidemiology, Brigham and Women's Hospital and Harvard Medical School), Satoshi Miyahara, Satoko Ugai, Mingyang Song, Jeffrey A. Meyerhardt, Marios Giannakis, Jonathan A. Nowak, et al.; corresponding author Shuji Ogino
Key points
- Prospective cohort incident-tumor biobank study within the Nurses' Health Study and Health Professionals Follow-up Study; vessels analyzed in 837 patients with resected colorectal cancer (median follow-up 11.8 years; 263 colorectal cancer-specific deaths).
- Multispectral immunofluorescence (ACKR1, CD34, CD36, KDR, LAMB1, MADCAM1, KRT) plus machine learning classified more than 61,000 tumor vessels by molecular and morphologic features.
- Higher overall CD34+ vessel density was associated with lower colorectal cancer-specific mortality (quartile 4 vs 1, HR 0.39; 95% CI 0.25-0.60; P for trend <.001), as was a higher proportion of microvessels (HR 0.48; 95% CI 0.31-0.74).
- Higher proportions of CD34+CD36+ (HR 1.58; 95% CI 1.07-2.32) and CD34+LAMB1+ vessels (HR 1.78; 95% CI 1.22-2.61) were associated with higher cancer-specific mortality.
- Observational associations; the authors call for validation before vascular features are used for risk stratification.
Key limitations
- Observational study; associations do not establish that vessel phenotypes drive outcomes.
- Tissue was available for 837 of 4,476 incident cases, requiring inverse probability weighting for selection.
- Cohort participants are predominantly health professionals, and several subgroup findings (for example, by MSI status) rest on few events.
Future directions
- External validation of vascular metrics as prognostic biomarkers.
- Study of how vessel phenotypes relate to known and novel colorectal cancer risk factors, including in early-onset disease.
- Testing whether specific vessel subtypes are therapeutic targets, for example combining vessel-targeting therapy with immunotherapy.
Angiogenesis is a hallmark of cancer, and antiangiogenic therapy with agents such as bevacizumab is a standard component of treatment for metastatic colorectal cancer. Yet the tumor vasculature is far from uniform: vessels differ in size, shape and patency, and in their molecular identity, from immature angiogenic sprouts to more organized microvessels. Whether these differences carry prognostic information in colorectal cancer, as corresponding author Shuji Ogino put it, remains "relatively understudied," despite the fact that blood vessels extend deep into tumors and shape oxygenation, immune cell trafficking and drug delivery.
Previous studies of tumor vasculature in colorectal cancer have typically relied on single-marker microvessel density counts in modest cohorts, with inconsistent results. Advances in multiplexed tissue imaging and machine learning now allow vessels to be classified by several markers and morphologic features simultaneously at scale. The Program in Molecular Pathological Epidemiology at Brigham and Women's Hospital, which maintains tumor tissue from incident cancers in two long-running U.S. cohorts, had previously reported that certain vessel features characterize early-onset colorectal cancer, providing a platform to test vascular phenotypes against long-term survival.
The investigators used the prospective cohort incident-tumor biobank method within the Nurses' Health Study and the Health Professionals Follow-up Study, which together contributed 4,476 incident colorectal cancer cases with resected specimens. They developed an in situ multispectral immunofluorescence assay targeting ACKR1, CD34, CD36, KDR, LAMB1, MADCAM1 and keratin, and a pathologist-supervised machine learning model (92% accuracy in validation) classified vessels as micro, collapsed, patent or irregular. More than 61,000 vessels were characterized in 837 patients. Multivariable Cox models with inverse probability weighting for tissue availability assessed colorectal cancer-specific mortality, using a stringent significance threshold of 0.005 to account for nine vessel variables.
The 837 analyzed patients had a median age at diagnosis of 69 years, and 56% were women. Over a median follow-up of 11.8 years among censored cases, there were 628 deaths, 263 of them from colorectal cancer. Higher overall CD34+ vessel density was associated with markedly lower colorectal cancer-specific mortality (highest vs lowest quartile, HR 0.39; 95% CI 0.25-0.60; P for trend <.001), and a higher proportion of microvessels was likewise favorable (HR 0.48; 95% CI 0.31-0.74; P for trend = .004). Conversely, higher proportions of vessels co-expressing CD36 (HR 1.58; 95% CI 1.07-2.32) or LAMB1 (HR 1.78; 95% CI 1.22-2.61), subtypes the authors link to angiogenesis, were associated with worse prognosis.
Results for overall mortality were directionally consistent: HRs were 0.53 (95% CI 0.38-0.74) for the highest quartile of CD34+ density, 0.61 (0.44-0.84) for microvessel proportion, 1.45 (1.07-1.98) for CD34+CD36+ and 1.36 (1.00-1.86) for CD34+LAMB1+ vessels. A higher proportion of patent vessels was also associated with higher overall mortality (HR 1.49; 95% CI 1.06-2.08). In an exploratory analysis, the association of CD34+MADCAM1+ vessels differed by microsatellite instability status, with lower cancer-specific mortality in non-MSI-high tumors (HR 0.67) but a reversed association in MSI-high tumors that rested on only five events.
The study's strengths include its prospective design, long follow-up, ascertainment of deaths through the National Death Index, multiplexed vessel phenotyping and adjustment for selection bias. Nonetheless, it is observational and cannot show that vessel phenotypes influence outcome rather than mark other tumor characteristics. Only 837 of 4,476 incident cases contributed vessel data, and although weighting addresses measured differences, unmeasured selection could remain. The cohorts consist largely of health professionals, which may limit generalizability, and subgroup findings such as those by MSI status are based on few events. The authors state that additional studies are needed to validate the findings.
If validated, multiplexed vascular profiling could add prognostic information to stage and molecular markers, helping to stratify risk after resection irrespective of treatment. The finding that dense, small, rounder vessels accompany better outcomes while larger, abnormal and angiogenesis-associated vessels accompany worse ones also raises therapeutic questions. Ogino has suggested that particular vessel subtypes might be targetable and that combining immunotherapy with vessel-directed therapy could outperform immunotherapy alone, a hypothesis that remains untested. Future work will relate vessel features to known and novel risk factors, including in early-onset disease.
Source: Matsuda et al., "Tumor Vascular Features and Colorectal Cancer-Specific Mortality," JAMA Network Open, September 21, 2026. Read the paper