Work

Colorectal Cancer Research Roundup

Every two weeks, a critical review of the most significant colorectal cancer research indexed in PubMed, with key findings, methodological appraisal, and clinical implications.

Issue of November 1, 2026 · covering October 15-31, 2026

1. Nivolumab plus ipilimumab versus nivolumab or chemotherapy in MSI-H/dMMR metastatic colorectal cancer: final overall survival (CheckMate 8HW, ESMO Congress 2026)

Presenting author Sara Lonardi (Veneto Institute of Oncology IOV-IRCCS, Padua, Italy), on behalf of the CheckMate 8HW investigators

Key points

  • Final OS analysis of the phase 3 CheckMate 8HW trial (LBA32, ESMO Congress 2026); not yet published in a peer-reviewed journal. Median follow-up 68.1 months across all lines and 63.0 months in the first line.
  • All lines, nivolumab plus ipilimumab versus nivolumab (296 vs 286 patients): OS HR 0.61 (95% CI 0.46-0.82; P = 0.0009); 5-year OS 74% versus 62%.
  • First line, nivolumab plus ipilimumab versus chemotherapy with or without targeted therapy (171 vs 84): OS HR 0.37 (95% CI 0.24-0.57; P < 0.0001); 5-year OS 77% versus 46%, despite 79.8% of the chemotherapy group receiving subsequent immunotherapy.
  • First line, nivolumab plus ipilimumab versus nivolumab: OS HR 0.66 (95% CI 0.44-0.99; P = 0.0432), which did not meet the prespecified significance threshold of P = 0.0058; 5-year OS 77% versus 66%.
  • Grade 3-4 treatment-related adverse events: 22% with nivolumab plus ipilimumab, 14% with nivolumab, and 51% with chemotherapy; no additional treatment-related deaths with longer follow-up.

Key limitations

  • The first-line comparison of the two immunotherapy regimens did not reach formal statistical significance, so superiority of the doublet over PD-1 monotherapy in untreated patients is not established.
  • Open-label design, and the chemotherapy comparison is confounded by extensive crossover to immunotherapy (79.8%).
  • Results are from a conference presentation; the full peer-reviewed report and patient-reported outcomes are awaited.

Future directions

  • Identification of biomarkers that select patients who need the CTLA-4 component versus those adequately served by PD-1 blockade alone.
  • Optimal treatment duration and safe de-escalation for the large proportion of patients with durable complete responses.
  • Extension of dual checkpoint blockade into earlier, curative-intent dMMR settings.

Mismatch repair-deficient or microsatellite instability-high (dMMR/MSI-H) metastatic colorectal cancer is a minority subgroup with a distinctive biology: a high mutational burden and abundant neoantigens make these tumors unusually sensitive to immune checkpoint blockade. Over the past decade this has transformed the outlook for affected patients, moving anti-PD-1 therapy into the first line. The open question has been whether adding CTLA-4 blockade to PD-1 blockade provides a meaningful survival gain over PD-1 monotherapy, given the additional immune-related toxicity the combination carries. CheckMate 8HW was designed to answer both whether dual immunotherapy beats chemotherapy and whether it beats nivolumab alone.

Earlier analyses of CheckMate 8HW had already established a large progression-free survival advantage. In previously reported first-line data, the 24-month PFS rate was 72% with nivolumab plus ipilimumab versus 14% with chemotherapy, and across all lines the combination improved PFS over nivolumab monotherapy (HR 0.62; 95% CI 0.48-0.81; P = 0.0003), with median PFS not reached versus 39.3 months. Those results supported regulatory approvals of the combination, but overall survival, the endpoint most relevant to whether the combination should displace single-agent PD-1 therapy, remained immature until this final analysis.

CheckMate 8HW (NCT04008030) is an international, randomized, open-label phase 3 trial. Patients with MSI-H/dMMR metastatic colorectal cancer were randomized 2:2:1 to nivolumab plus ipilimumab, nivolumab alone, or investigator's choice of chemotherapy with or without targeted therapy; patients who had received two or more prior lines were randomized only between the two immunotherapy arms. The final analysis, with a data cutoff of May 28, 2026, included 296 versus 286 patients in the all-lines comparison of the immunotherapy arms and 171, 170 and 84 patients in the first-line nivolumab plus ipilimumab, nivolumab and chemotherapy groups. Median follow-up was 68.1 months for all lines and 63.0 months in the first line.

Across all lines, nivolumab plus ipilimumab reduced the risk of death compared with nivolumab alone (HR 0.61; 95% CI 0.46-0.82; P = 0.0009); median OS was not reached in either arm, and 5-year OS was 74% versus 62%. In the first line, the combination substantially prolonged survival compared with chemotherapy (HR 0.37; 95% CI 0.24-0.57; P < 0.0001), with median OS not reached versus 50.1 months and 5-year OS of 77% versus 46%. This benefit was observed even though 79.8% (67 of 84) of patients assigned to chemotherapy subsequently received immunotherapy. In the first-line comparison of the two immunotherapy regimens, the HR was 0.66 (95% CI 0.44-0.99; P = 0.0432), with 5-year OS 77% versus 66%, but this did not cross the prespecified significance boundary of P = 0.0058.

Descriptive PFS results remained consistent: across all lines, median PFS was not reached with the combination versus 39.3 months with nivolumab (HR 0.60; 95% CI 0.47-0.77), and in the first line the PFS HR was 0.21 (95% CI 0.14-0.32) versus chemotherapy and 0.68 (95% CI 0.48-0.96) versus nivolumab. Objective response rates across all lines were 72% versus 58%, with complete responses in 40% versus 31%. Grade 3-4 treatment-related adverse events occurred in 22% of patients treated with nivolumab plus ipilimumab, 14% with nivolumab and 51% with chemotherapy; treatment-related discontinuation rates were 14%, 6% and 34%. There were two treatment-related deaths with the combination and one with nivolumab, and no additional deaths with longer follow-up.

The trial's main interpretive difficulty lies in the first-line immunotherapy comparison. A point estimate of 0.66 and an 11-percentage-point difference in 5-year survival are clinically suggestive, but the result did not meet the alpha allocated to it, and it should not be read as formal proof of superiority. The all-lines comparison is statistically robust but mixes treatment-naive and pretreated patients. The open-label design could influence investigator-assessed endpoints, although OS is not susceptible to that bias. Finally, the data come from a congress presentation, so full subgroup analyses, quality-of-life data and the peer-reviewed report are still awaited.

With more than five years of follow-up, CheckMate 8HW confirms that dual checkpoint blockade produces durable survival in MSI-H/dMMR metastatic colorectal cancer and decisively outperforms chemotherapy in the first line. Whether every patient needs the ipilimumab component remains a matter of judgment: the doublet's higher rate of serious immune toxicity must be weighed against the numerically better survival and response rates. Lonardi's group and others will need to define which patients derive the incremental benefit, how long treatment must continue in patients with complete responses, and how these results translate to earlier-stage disease, where checkpoint inhibitors are already reshaping management.

Source: CheckMate 8HW final overall survival analysis, presented by S. Lonardi at the ESMO Congress 2026 (abstract LBA32), Madrid, October 2026; not yet published in a peer-reviewed journal. Conference coverage (OncoDaily) · Interview (Targeted Oncology)

2. Eicosapentaenoic acid versus placebo after resection of colorectal cancer liver metastases (EMT2, ESMO Congress 2026)

Presenting author Mark A. Hull (Leeds, United Kingdom), on behalf of the EMT2 investigators

Key points

  • Phase 3, multicentre, double-blind, placebo-controlled trial (NCT03428477; LBA30, ESMO Congress 2026) of 4 g/day pure EPA ethyl ester around curative-intent liver resection; not yet published in a peer-reviewed journal.
  • 418 patients randomized 1:1 (209 per arm); median age 64 years; 61.5% male.
  • Primary endpoint not met: median PFS 13.5 months with EPA versus 12.9 months with placebo (HR 0.89; 95% CI 0.70-1.13; P = 0.33).
  • Key secondary endpoint: 5-year OS 62.0% versus 45.1% (HR 0.68; 95% CI 0.48-0.96; P = 0.030), to be interpreted cautiously given the negative primary endpoint.
  • No excess of total adverse events, bleeding or atrial fibrillation with EPA, and quality of life was similar between arms.

Key limitations

  • The primary PFS endpoint was negative, so the OS difference is hypothesis-generating rather than confirmatory.
  • No mechanism explains an OS benefit in the absence of a PFS benefit; post-progression treatment differences have not been reported in available coverage.
  • Results are from a congress presentation; full data on subsequent therapy and causes of death are awaited.

Future directions

  • Peer-reviewed publication with detailed analyses of post-progression therapy and survival after recurrence.
  • Translational work on omega-3 fatty acid biology in micrometastatic disease and the tumour microenvironment.
  • A confirmatory trial with OS as the primary endpoint before EPA can be recommended.

Liver resection offers the best chance of long-term survival for patients with colorectal cancer liver metastases, but most patients relapse after surgery, often within the first two years. Perioperative chemotherapy has had only modest effects on long-term outcomes, and there is no established adjunct that reduces recurrence after hepatectomy. Inexpensive, well-tolerated agents that could act on residual micrometastatic disease are therefore of considerable interest, and nutritional and repurposed compounds such as omega-3 fatty acids have been studied for their anti-inflammatory and anti-neoplastic properties.

Eicosapentaenoic acid (EPA) is an omega-3 polyunsaturated fatty acid available as a purified pharmaceutical-grade ethyl ester. Experimental work has suggested anti-colorectal cancer activity, and the EMT2 investigators had previously conducted a phase II study of EPA given before liver resection. That earlier study did not show a significant difference in tumour proliferation, leaving open whether a longer course of EPA spanning the perioperative period and beyond could alter clinical outcomes. EMT2 was designed as a definitive placebo-controlled phase 3 test of this hypothesis.

EMT2 (NCT03428477) randomized adults undergoing curative-intent resection of colorectal cancer liver metastases 1:1 to 4 g/day pure EPA ethyl ester (Vascepa) or mineral oil placebo capsules. Treatment began at least one day before surgery and was planned for 2 to 4 years, continuing after progression. Randomization used minimisation by neoadjuvant chemotherapy, extrahepatic metastases, randomization-to-surgery interval, aspirin use and centre. The primary endpoint was progression-free survival, the key secondary endpoint was overall survival, and efficacy and quality of life were analysed in the intention-to-treat population. Of 418 patients randomized (209 per arm), 411 were included in the safety analysis; median age was 64 years and 61.5% were male.

The trial did not meet its primary endpoint. Median PFS was 13.5 months (95% CI 11.8-17.9) with EPA versus 12.9 months (95% CI 10.4-16.3) with placebo (HR 0.89; 95% CI 0.70-1.13; P = 0.33). The similarity of the PFS estimates indicates that EPA did not meaningfully delay recurrence after liver resection in this population, in line with the negative proliferation result of the earlier phase II study. The confidence interval for the PFS hazard ratio spans both a modest benefit and a modest harm, so the trial effectively excludes a large effect of EPA on recurrence.

The key secondary endpoint showed a different picture: 5-year OS was 62.0% (95% CI 52.7-70.0) with EPA versus 45.1% (95% CI 33.7-55.9) with placebo (HR 0.68; 95% CI 0.48-0.96; P = 0.030), corresponding to a roughly 30% lower hazard of death. EPA was well tolerated, with no excess of total adverse events, bleeding or atrial fibrillation compared with placebo, which is relevant given the antiplatelet effects attributed to omega-3 fatty acids and the perioperative setting. Quality of life was similar between arms, with no detrimental effect of EPA reported.

The central methodological problem is the discordance between a null primary endpoint and a statistically significant secondary endpoint. In a hierarchical framework, an OS benefit following a negative PFS result cannot be claimed as confirmatory, and coverage of the presentation explicitly cautions against over-interpretation. A survival difference without a recurrence difference could reflect effects on post-recurrence disease, imbalances in subsequent treatment, chance, or non-cancer mortality, and none of these explanations can be evaluated from the data available so far. The double-blind, placebo-controlled design is nevertheless a strength, reducing the risk of bias in outcome assessment.

EMT2 does not support EPA as a standard adjunct after resection of colorectal liver metastases, but the survival signal, combined with low cost and good tolerability, justifies further investigation. Priorities include full peer-reviewed publication with analyses of subsequent therapy and causes of death, translational studies of how EPA might influence the biology of residual or recurrent disease, and, if the signal holds, a confirmatory trial powered for overall survival. Until then, the trial is best regarded as a well-conducted negative study with an intriguing secondary finding.

Source: EMT2 trial, presented by M.A. Hull at the ESMO Congress 2026 (abstract LBA30), October 2026; not yet published in a peer-reviewed journal. Conference coverage (OncoDaily)

3. Multiomic characterization of microsatellite-stable colorectal cancer reveals targetable vulnerabilities (Cancer Cell)

Joint lead investigators Anguraj Sadanandam (The Institute of Cancer Research, London) and Annette Byrne (RCSI University of Medicine and Health Sciences, Dublin), et al.; co-senior author Rodrigo Dienstmann (Vall d'Hebron Institute of Oncology, Barcelona)

Key points

  • Multiomic analysis of about 2,000 patient tumour samples of microsatellite-stable (MSS) colorectal cancer, integrating DNA, RNA, epigenetic, protein, metabolic, single-cell and clinical data.
  • MSS disease accounts for 80-85% of colorectal cancers, and is largely resistant to the checkpoint inhibitors that have transformed MSI-H disease.
  • A new AI-based approach, iPhenMap, was developed to build a biological map of each tumour linking genetic, metabolic and clinical features.
  • Previously unrecognised tumour vulnerabilities were identified, along with drug combinations that could target them, including drugs already approved for other cancers.
  • In laboratory models, trametinib combined with other drugs significantly reduced tumour growth and extended survival; the authors stress that the findings are preclinical.

Key limitations

  • All therapeutic findings are preclinical; no patient has yet been treated according to the proposed vulnerabilities.
  • Predictive biological features to select patients for the combinations still need to be identified.
  • Public summaries do not report effect sizes for the model experiments, and partner drugs for trametinib were not specified in available coverage.

Future directions

  • Definition and validation of biomarkers that identify the MSS subgroups most likely to respond.
  • Early-phase clinical trials of trametinib-based combinations, particularly in MSS RAS-mutant metastatic disease.
  • Wider application of the iPhenMap framework to other heterogeneous tumour types.

Microsatellite-stable (MSS) colorectal cancer accounts for 80-85% of all colorectal cancers and remains the principal clinical challenge in the disease. Unlike mismatch repair-deficient tumours, MSS tumours are generally resistant to immune checkpoint inhibitors, and patients with metastatic MSS disease, particularly those with RAS mutations, have limited options beyond sequential chemotherapy and a small number of targeted agents. MSS colorectal cancer is also biologically heterogeneous, and existing classification schemes have had only limited success in guiding therapy. New, biology-based ways of selecting treatment for this majority of patients are therefore a priority.

Earlier efforts to stratify colorectal cancer, including transcriptomic subtyping, demonstrated that tumours with similar mutational profiles can behave very differently, yet these schemes have rarely translated into treatment selection. Single data layers such as DNA mutations or gene expression capture only part of the biology, and integrating multiple omic layers at scale has been technically difficult. The investigators reasoned that combining genomic, transcriptomic, epigenetic, proteomic, metabolic, single-cell and clinical information across a large cohort could reveal functional vulnerabilities in MSS disease that single-layer analyses miss.

The study, co-led by Anguraj Sadanandam at The Institute of Cancer Research, London, and Annette Byrne at RCSI University of Medicine and Health Sciences, Dublin, with co-senior author Rodrigo Dienstmann of the Vall d'Hebron Institute of Oncology, analysed about 2,000 patient tumour samples. The team developed an AI-based method, iPhenMap, which builds a biological map of a tumour by linking genetic, metabolic and clinical data. Candidate vulnerabilities emerging from this map were then tested in laboratory models. The work was funded by the European Union's Horizon 2020 programme through the COLOSSUS project and by a Research Ireland Project Award.

The analysis identified previously unrecognised vulnerabilities within MSS colorectal cancer and drug combinations predicted to exploit them, including drugs already approved for other malignancies. The most prominent example was trametinib, a MEK inhibitor approved for melanoma, lung cancer and thyroid cancer with specific mutations. In laboratory models, trametinib combined with other drugs significantly reduced tumour growth and extended survival. The investigators highlighted MSS RAS-mutant metastatic disease as a setting where these findings could be particularly relevant.

A major strength emphasised by the authors is the scale and diversity of the dataset, which spans multiple molecular layers and clinical annotation rather than a single omic modality. Repurposing approved drugs may also shorten the path to clinical testing, because their toxicity profiles are already known. However, the publicly available summaries do not report the magnitude of the effects in individual models, nor the specific partner drugs in the trametinib combinations, and readers should consult the full paper for these details.

The principal limitation is that all therapeutic findings remain preclinical. Effects seen in laboratory models frequently fail to translate to patients, and MEK inhibition has a mixed history in colorectal cancer, where feedback signalling and toxicity have limited single-agent activity. The authors themselves note that predictive biological features must still be identified before the combinations can be evaluated clinically. In addition, large integrative datasets are susceptible to batch effects and to selection of tumours from specialist centres, which may limit generalisability.

This work offers a framework for moving beyond broad MSS-versus-MSI classification toward biology-driven treatment selection in the majority of patients with colorectal cancer. Its immediate value is in generating prioritised, testable hypotheses: biomarker-defined subsets, rational drug combinations built around agents with known safety profiles, and a computational tool that could be applied to other cancers. The critical next steps are biomarker validation and carefully designed early-phase trials, particularly in RAS-mutant metastatic MSS disease, where new options are most needed.

Source: "Multiomic characterization of microsatellite-stable colorectal cancer reveals distinct biology and targetable vulnerabilities," Cancer Cell, October 23, 2026 (doi:10.1016/j.ccell.2026.09.014). Read the paper · RCSI press release

4. Colorectal cancer incidence among U.S. adults, 2000-2023 (NEJM Evidence)

Caitlin Murphy (University of Chicago), et al.

Key points

  • Analysis of U.S. National Cancer Institute incidence data for colorectal cancer from 2000 to 2023.
  • Age-adjusted incidence fell from 55.2 per 100,000 in 2000 to 36.7 in 2022, then rose to 38.3 in 2023, the first overall increase in more than 20 years.
  • The sharpest rise was in adults aged 45-49, at 12.76% per year from 2019 to 2023, likely partly reflecting the lowering of the screening start age to 45.
  • Incidence rose across age groups up to 65-69, despite about 80% screening uptake in that group, and plateaued at 70-74.
  • The increase persisted when 2020-2021 data were excluded, arguing against a COVID-19 artefact; the authors describe a birth cohort effect.

Key limitations

  • Registry-based trend analysis cannot identify the risk factors driving the increase.
  • The 2023 uptick is a single data year and requires confirmation in subsequent releases.
  • Screening changes complicate interpretation, especially in the 45-49 age group.

Future directions

  • Screening of environmental chemical exposures in blood samples and testing their effects on colon organoids, as planned by the investigators.
  • Continued surveillance with annual registry updates to confirm whether overall incidence keeps rising.
  • Reassessment of screening strategies in older age groups where incidence is rising despite high uptake.

For several decades, colorectal cancer incidence in the United States fell steadily, a success largely attributed to screening and removal of precancerous polyps in older adults. That decline has coexisted with a well-documented rise in early-onset disease: American Cancer Society data cited alongside this study indicate that incidence among adults aged 20 to 49 rose about 3% per year from 2013 to 2022. A key question has been whether the rise in younger adults is an isolated phenomenon or the leading edge of a broader shift that will eventually affect older age groups as well.

The birth cohort hypothesis holds that people born from around the 1960s onward carry a higher lifetime risk of colorectal cancer than earlier generations, presumably because of changes in early-life exposures. If this is correct, rising risk should progressively appear in older age groups as these cohorts age, and could begin to offset the gains achieved by screening. Testing this requires up-to-date, population-level incidence data analysed across the full adult age range, rather than focusing only on adults younger than 50.

Caitlin Murphy and colleagues at the University of Chicago analysed national incidence data from the U.S. National Cancer Institute covering 2000 through 2023, the most recent year available. They examined age-adjusted incidence for the whole adult population and trends within age groups, with particular attention to recent years. To address the possibility that pandemic-related disruption of diagnosis distorted recent trends, they repeated analyses excluding 2020 and 2021. The study was published in NEJM Evidence in late October 2026.

Age-adjusted colorectal cancer incidence fell from 55.2 per 100,000 in 2000 to 36.7 per 100,000 in 2022, then rose to 38.3 per 100,000 in 2023. According to Murphy, this is the first increase in the overall rate in more than 20 years. The steepest increase was among adults aged 45 to 49, in whom incidence rose 12.76% per year from 2019 to 2023, a change the authors consider likely to reflect, at least in part, the extension of screening to age 45.

Beyond the 45-49 age group, incidence rose across age groups up to 65 to 69 years and plateaued in those aged 70 to 74. The rise among people aged 65 to 69, a group with about 80% screening uptake, is particularly notable because screening should be driving rates down; the authors suggest that unidentified risk factors may be offsetting screening gains. The increase persisted when 2020-2021 data were excluded, making it unlikely that COVID-19-related disruption explains the trend. The authors interpret the pattern as a birth cohort effect, with people born in the 1960s and later carrying higher risk as they age.

As a descriptive analysis of registry data, the study cannot identify why incidence is rising, and the 2023 increase represents a single data year that will need confirmation as registries are updated. The effect of changing screening recommendations on recorded incidence is difficult to disentangle from true changes in disease occurrence, particularly in adults aged 45 to 49, where earlier detection of prevalent cancers is expected to produce a transient rise. Details of the full statistical methods and confidence intervals were not available in the coverage reviewed here.

If confirmed, these findings would mark an inflection point: colorectal cancer would no longer be a disease whose incidence is falling overall in the United States, and the drivers of early-onset disease would be acting across the age spectrum. The investigators plan to screen environmental chemical exposures in blood samples and to study their effects on colon organoids, in collaboration with toxicologists. The findings also argue for continued surveillance, aggressive efforts to improve screening uptake and quality, and research into exposures that could explain a cohort-wide rise in risk.

Source: Murphy et al., "Colorectal cancer among U.S. adults, 2000-2023," NEJM Evidence, October 2026. Read the paper · Coverage (ecancer)

5. National trends in watch-and-wait organ preservation for rectal cancer, 2006-2023 (JAMA Network Open)

Stephanie S. Keeling (MD Anderson Cancer Center and Baylor College of Medicine, Houston), et al.; senior author George J. Chang (MD Anderson Cancer Center)

Key points

  • Retrospective National Cancer Database cohort of 101,001 adults with stage II-III rectal adenocarcinoma treated with neoadjuvant chemoradiation, diagnosed 2006-2023 (84,324 underwent surgery; 16,677 did not).
  • Nonsurgical management rose from 14.02% in 2006 to 44.64% in 2023, with annual percentage change accelerating to 21.71% in 2019-2023 (P < 0.001).
  • Watch and wait, estimated as nonsurgical management above an age-specific baseline, rose from 2.30% in 2015 to 32.94% in 2023.
  • Nonsurgical management was more likely in 2023 than 2006 (OR 6.23; 95% CI 5.48-7.09) and in patients aged 80 or older versus younger than 50 (OR 4.25; 95% CI 3.98-4.54).
  • Older patients adopted watch and wait earlier, but rates increased in all age groups; insurance-related differences disappeared by 2020-2023.

Key limitations

  • Watch and wait is inferred indirectly; the NCDB does not record clinical response or the intent behind nonsurgical management.
  • The baseline method assumes that non-guideline nonsurgical care stayed constant after 2014.
  • No survival or recurrence outcomes were reported, so the safety of this practice at population level cannot be assessed.

Future directions

  • Linkage to outcome data to evaluate oncologic safety, regrowth and salvage surgery in routine practice.
  • Standardised response assessment and surveillance protocols outside specialist centres.
  • Monitoring of equity in access as organ preservation becomes routine.

For decades, total mesorectal excision after neoadjuvant chemoradiation was the standard of care for locally advanced rectal cancer, at the cost of substantial bowel, urinary and sexual dysfunction and, for many patients, a permanent stoma. Watch and wait, in which patients with a clinical complete response after neoadjuvant therapy are managed with intensive surveillance rather than immediate surgery, has emerged as a way to preserve the rectum. The approach has moved from specialist centres into guidelines, but how widely and how evenly it has been adopted across the United States has been unclear.

Support for watch and wait has accumulated from cohort studies and prospective trials of total neoadjuvant therapy showing that a substantial proportion of patients achieve a clinical complete response and can avoid surgery, with salvage surgery available for most regrowths. Organ preservation is also highly relevant to the growing number of younger patients with rectal cancer. Yet registry data have not routinely captured watch and wait as a treatment category, making it difficult to separate deliberate organ preservation from nonsurgical management driven by frailty, comorbidity or patient refusal, especially among older adults.

Keeling and colleagues at MD Anderson Cancer Center used the National Cancer Database, which covers Commission on Cancer-accredited facilities, to study 101,001 adults with stage II-III rectal adenocarcinoma who received neoadjuvant chemoradiation between 2006 and 2023; 84,324 underwent surgery and 16,677 did not. Trends were analysed with joinpoint regression. To isolate watch and wait, the investigators estimated an age-specific baseline rate of nonsurgical management from 2006 to 2014 and treated nonsurgical management above that baseline as excess attributable to watch and wait. Multivariable logistic regression examined associated factors across three periods.

The proportion of patients managed without surgery rose from 14.02% in 2006 to 44.64% in 2023. Joinpoints were identified in 2014 and 2019: the annual percentage change was -0.45% before 2014 (P = 0.42), 9.83% in the middle period (P = 0.001) and 21.71% from 2019 to 2023 (P < 0.001). The estimated watch-and-wait rate rose from 2.30% in 2015 to 32.94% in 2023. Nonsurgical management was strongly associated with calendar year (2023 vs 2006: OR 6.23; 95% CI 5.48-7.09).

Baseline nonsurgical management during 2006-2014 ranged from 6.73% among patients aged 18-49 to 28.15% among those aged 80 or older. Older patients adopted watch and wait earlier, with excess nonsurgical management above 5 percentage points by 2015 and above 10 points by 2017; in patients younger than 50, excess reached 6.15 points in 2019 and 14.02 points in 2021. Age 80 or older versus younger than 50 was associated with an OR of 4.25 (95% CI 3.98-4.54). Black patients were more likely than White patients to be managed nonsurgically in all periods, though the association weakened over time, and the lower odds among privately insured patients seen early on had disappeared by 2020-2023 (OR 0.97; P = 0.75).

The key methodological limitation is that watch and wait cannot be directly observed in the NCDB, which records neither clinical response nor the reason surgery was not performed. The baseline-subtraction approach assumes that non-guideline nonsurgical care remained stable after 2014, an assumption that cannot be verified. As a retrospective registry study, the analysis is subject to missing data and unmeasured confounding, and minor inconsistencies between the abstract and results (for example, in the period attached to the 9.83% APC) were noted. Importantly, the study reports no survival or recurrence outcomes.

These data show that organ preservation has moved rapidly into mainstream U.S. practice, accelerating sharply after 2019, in parallel with publication of total neoadjuvant therapy and organ-preservation trial data. Nearly half of patients treated with neoadjuvant chemoradiation in 2023 did not undergo surgery. The authors interpret this as largely evidence-concordant adoption but stress careful patient selection, accurate response assessment and rigorous surveillance. The next priority is to link these trends to oncologic outcomes, regrowth rates and salvage surgery in community practice, and to ensure that the expansion of organ preservation is accompanied by the infrastructure to deliver it safely.

Source: Keeling et al., "Organ Preservation With Watch and Wait in Patients With Rectal Cancer," JAMA Network Open, October 28, 2026 (doi:10.1001/jamanetworkopen.2026.40267). Read the paper

6. Centralized mailed FIT outreach for colorectal cancer screening in community health centers (cluster randomized trial, JAMA Network Open)

Maria Elena Martinez (UC San Diego Moores Cancer Center), Scott Roesch, Sheila F. Castañeda, Jesse N. Nodora, Borsika A. Rabin, et al.; senior author Samir Gupta (UC San Diego)

Key points

  • Pragmatic cluster randomized trial (NCT04941300) in two federally qualified health center systems in San Diego County: 16 clinics, 20,250 patients aged 50-75 not up to date with screening.
  • Intervention: three annual rounds of centralized mailed FIT outreach in English and Spanish, with a mailed primer, FIT kit and automated phone and text reminders.
  • Primary outcome not met: up to date with screening at year 3 was 39.4% versus 36.5% (difference 2.9 percentage points; 95% CI -3.2 to 9.0).
  • Screening was higher with outreach at year 1 (39.0% vs 27.8%; difference 11.2 points; 95% CI 5.9-16.5) and year 2 (37.8% vs 27.0%; difference 10.8 points; 95% CI 4.9-16.7).
  • Implementation was disrupted by COVID-19 waves, withdrawal of two health center systems and a protocol change that moved the year 3 assessment earlier.

Key limitations

  • Two FQHC systems discontinued participation, and modest baseline differences existed between arms.
  • The year 3 result is undermined by a protocol change that queried some patients before they were due for repeat FIT.
  • Usual-care clinics may have improved through their own quality-improvement work and post-pandemic rebound.

Future directions

  • Sustained implementation support and fidelity monitoring to maintain gains beyond the first two years.
  • Closer links between centralized outreach hubs and patients' primary care clinicians.
  • Strategies to improve follow-up colonoscopy after abnormal FIT in safety-net settings.

Colorectal cancer screening rates in federally qualified health centers (FQHCs), which serve low-income, uninsured and underinsured populations, lag behind national averages, contributing to disparities in stage at diagnosis and mortality. Mailed fecal immunochemical test (FIT) outreach, in which kits are sent directly to patients' homes, is one of the best-studied strategies for raising screening participation, and has been promoted as a scalable population-health intervention for safety-net systems. This trial, conducted in two FQHC systems in San Diego County, California, tested whether a centralized model could deliver such outreach over several years.

Randomized trials have shown that mailed FIT outreach increases screening completion, typically over a single round. Programmatic screening with FIT, however, depends on repeated annual testing, and whether centralized outreach can sustain higher screening rates over several years in real-world FQHC settings, with their staffing constraints and patient turnover, has been less certain. This trial was designed to test a centralized, multiyear model in which a hub delivers outreach on behalf of participating clinics. The study was funded by the National Cancer Institute.

Martinez and colleagues at UC San Diego conducted a pragmatic cluster randomized trial from June 2021 to May 2024 in two FQHC systems in San Diego County. Sixteen clinics were randomized, stratified by health system: 6 intervention clinics with 8,334 patients and 10 usual-care clinics with 11,916 patients aged 50 to 75 years who were not up to date with screening. The intervention comprised three annual rounds of centralized mailed outreach in English and Spanish, including an educational primer, a FIT kit with invitation letter, instructions and prepaid return envelope, and automated phone and text reminders at 2 and 3 weeks for nonresponders. The primary outcome was the proportion up to date with screening at year 3, analysed by intention to screen.

At year 3, 39.4% (95% CI 36.3%-42.6%) of patients in intervention clinics were up to date with screening compared with 36.5% (95% CI 31.2%-42.2%) in usual-care clinics, a difference of 2.9 percentage points (95% CI -3.2 to 9.0) that was not statistically significant. The primary hypothesis that centralized mailed outreach would sustain a higher screening rate at three years was therefore not supported. The paper reports confidence intervals rather than P values, and the interval for the year 3 difference is wide, ranging from a small decrease to a meaningful increase in screening.

Earlier time points told a different story. At baseline, screening rates were similar (38.7% vs 39.9%). At year 1, 39.0% of intervention patients versus 27.8% of usual-care patients were up to date (difference 11.2 points; 95% CI 5.9-16.5), and at year 2 the figures were 37.8% versus 27.0% (difference 10.8 points; 95% CI 4.9-16.7). From baseline to year 2, screening in intervention clinics was essentially stable (-0.9 points) while it fell by 12.9 points in usual-care clinics, suggesting that outreach chiefly prevented the decline in screening seen during the pandemic period.

The authors acknowledge substantial implementation challenges. Two FQHC systems withdrew, one before implementation and one after the first year; the trial ran through COVID-19 waves; and a resource-driven protocol change moved the year 3 query earlier, so that some patients screened in year 2 were not yet due for repeat FIT, reducing confidence in the year 3 result. Usual-care clinics may have improved through their own quality-improvement activities and post-pandemic rebounds in care, and the central hub had no direct link to patients' primary clinicians. Cluster randomization with relatively few clinics also left modest baseline imbalances.

The trial shows that centralized mailed FIT outreach can protect and raise screening rates in safety-net clinics over two years, but that the effect was not sustained at three years under real-world conditions. As an accompanying commentary framed it, the challenge has shifted from efficacy to implementation. The authors conclude that sustained implementation support and intervention fidelity are needed for lasting impact. For health systems, the message is that mailed outreach is a program, not a one-off intervention: it requires ongoing resources, integration with primary care and attention to follow-up colonoscopy to translate into reduced colorectal cancer mortality.

Source: Martinez et al., "Fecal Immunochemical Test Outreach for Colorectal Cancer Screening in Community Health Centers: A Randomized Clinical Trial," JAMA Network Open, October 21, 2026 (doi:10.1001/jamanetworkopen.2026.39013). Read the paper

7. Food processing, dietary risk factors and colorectal cancer incidence in the NIH-AARP Diet and Health Study (PLOS Medicine)

Leila Abar (Metabolic Epidemiology Branch, National Cancer Institute, Rockville), Caitlin P. O'Connell, Hyokyoung G. Hong, Kirsten A. Herrick, Rashmi Sinha, et al.; senior author Erikka Loftfield (National Cancer Institute)

Key points

  • Prospective cohort of 461,682 cancer-free adults aged 50-71 at enrolment (1995-1996), followed to 2018 (median 18.8 years), with 10,075 incident colorectal cancers.
  • Total ultra-processed food (UPF) intake was not associated with colorectal cancer: highest versus lowest quintile HR 0.97 (95% CI 0.91-1.03; P-trend = 0.55).
  • Within UPF, bread (HR 0.91; 95% CI 0.86-0.97) and breakfast cereals (HR 0.84; 95% CI 0.79-0.90) were inversely associated, whereas reconstituted meat and fish products were positively associated (HR 1.10; 95% CI 1.04-1.18).
  • Fiber from UPF sources was inversely associated (HR 0.72 per 10 g/day; 95% CI 0.64-0.80), and meat was positively associated from both UPF (HR 1.08) and non-UPF sources (HR 1.04).
  • Replacing 10% of UPF grams with minimally processed food was associated with slightly lower risk (HR 0.98; 95% CI 0.96-0.99).

Key limitations

  • Diet was measured once at baseline with a questionnaire not designed to capture food processing.
  • The cohort was older, mostly White and recruited in the mid-1990s, limiting generalisability to today's food supply and younger adults.
  • Observational design with possible residual confounding; the proportional hazards assumption was violated.

Future directions

  • Studies with repeated, processing-specific dietary assessment and ingredient and additive data.
  • Evaluation in younger and more diverse cohorts, particularly in relation to early-onset colorectal cancer.
  • Dietary guidance that targets specific foods and nutrients rather than processing level alone.

Ultra-processed foods (UPFs), industrially formulated products typically classified using the Nova system, make up a large share of energy intake in high-income countries and have been linked to obesity, cardiometabolic disease and all-cause mortality. Because diet is an established determinant of colorectal cancer risk, and because the rise of UPF consumption has paralleled the increase in early-onset disease, UPFs have attracted intense attention as a possible colorectal carcinogen. Whether UPFs as a category raise risk, or whether specific foods and nutrients within that category drive any association, has major implications for dietary advice.

Previous cohort studies have reported inconsistent associations between total UPF intake and colorectal cancer, with some suggesting increased risk in men and others finding no relationship. The Nova category groups together foods with very different nutritional profiles, from wholegrain breads and yogurts to processed meats, and established colorectal cancer risk factors such as red and processed meat and protective factors such as fiber, whole grains, dairy and calcium span both UPF and non-UPF foods. Disentangling processing from composition requires a large cohort with long follow-up and detailed dietary data.

Abar and colleagues at the National Cancer Institute analysed the NIH-AARP Diet and Health Study, which enrolled adults aged 50 to 71 in 1995-1996. Of 461,682 cancer-free participants (274,269 men and 187,413 women), dietary intake was assessed by food frequency questionnaire at baseline and foods were classified by Nova processing level. Participants were followed through 2018, a median of 18.8 years. Cox models estimated hazard ratios for sex-specific quintiles of UPF intake (grams per 1,000 kcal per day), for UPF subgroups and for nutrients and food groups from UPF and non-UPF sources, with substitution analyses.

During follow-up, 10,075 colorectal cancers were diagnosed. Total UPF intake was not associated with colorectal cancer incidence: the HR comparing the highest with the lowest quintile was 0.97 (95% CI 0.91-1.03; P-trend = 0.55), and estimates were similar after additional adjustment for diet quality and nutrients (HR 0.94; 95% CI 0.88-1.00) or for body mass index (HR 0.94; 95% CI 0.89-1.01). There was no effect modification by sex or BMI. By subsite, HRs were 1.02 for proximal colon, 0.94 for distal colon and 0.91 for rectal cancer, none statistically significant in the primary analysis.

Associations differed markedly by UPF subgroup. Bread (HR 0.91; 95% CI 0.86-0.97), frozen desserts and milk drinks (HR 0.82; 95% CI 0.77-0.88) and breakfast cereals (HR 0.84; 95% CI 0.79-0.90) were inversely associated with colorectal cancer, whereas reconstituted meat and fish products were positively associated (HR 1.10; 95% CI 1.04-1.18). In mutually adjusted models, calcium, dairy, whole grains and fiber from UPF sources were inversely associated, with fiber showing an HR of 0.72 per 10 g/day (95% CI 0.64-0.80), and meat was positively associated whether from UPF (HR 1.08; 95% CI 1.02-1.14) or non-UPF sources (HR 1.04; 95% CI 1.01-1.06). Substituting 10% of UPF grams with minimally processed food was associated with an HR of 0.98 (95% CI 0.96-0.99). Among participants with poor diet quality, higher UPF intake was associated with elevated risk.

The authors list several limitations. Diet was measured once, by a questionnaire and databases not designed to capture food processing, and population-level Nova assignments may blur differences between individuals. Participants were older, mostly White and recruited in the mid-1990s, so the findings may not reflect today's food supply or apply to younger adults, in whom early-onset disease is rising. Variation in UPF intake across quintiles was limited, the proportional hazards assumption was violated, no ingredient or additive data were available, and the nutrient and food-group analyses were exploratory with multiple testing. Residual confounding remains possible, although an unmeasured confounder would need a more than 2.1-fold association to explain the fiber finding.

In this large cohort, the degree of processing per se was not associated with colorectal cancer; what mattered was the nutritional composition of the foods. These findings support existing dietary guidance for colorectal cancer prevention, more whole grains, fiber, dairy and calcium and less processed meat, rather than blanket avoidance of all UPFs. They also caution against treating UPFs as a single exposure in research and policy. Future work should use repeated, processing-specific dietary assessments, include ingredient and additive data and examine younger, more diverse populations, particularly in the context of early-onset colorectal cancer.

Source: Abar, O'Connell et al., "Associations of food processing and dietary risk factors with colorectal cancer incidence in the NIH-AARP Diet and Health Study: A cohort study," PLOS Medicine, October 15, 2026. Read the paper