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 August 15, 2026 · covering August 1-14, 2026

1. High-dose vitamin D3 added to first-line chemotherapy plus bevacizumab in metastatic colorectal cancer (SOLARIS/Alliance A021703, JAMA)

Kimmie Ng (Dana-Farber Cancer Institute, Boston), Fang-Shu Ou, Tyler Zemla, Nadine A. Jackson, et al.; senior author Jeffrey A. Meyerhardt (Dana-Farber Cancer Institute)

Key points

  • Double-blind, randomized phase 3 trial (NCT04094688) in the US National Clinical Trials Network; 455 patients with previously untreated metastatic colorectal cancer enrolled October 2019 to December 2022 at 151 academic and community centers.
  • High-dose vitamin D3 (8,000 IU daily for 14 days, then 4,000 IU daily; n=228) versus standard-dose vitamin D3 (400 IU daily; n=227), both with mFOLFOX6 or FOLFIRI plus bevacizumab.
  • Median progression-free survival (PFS), the primary endpoint: 11.8 months (95% CI 10.3-13.3) vs 10.3 months (95% CI 9.4-12.2); HR 0.92; one-sided P=0.25.
  • Objective response rate 51% vs 44% (P=0.12); median overall survival 25.6 vs 27.0 months (P=0.66).
  • Grade 3 or higher neutropenia 32% vs 30% and hypertension 20% vs 23%; vitamin D-associated toxicities were uncommon.

Key limitations

  • Standard-dose arm received 400 IU daily rather than placebo, and vitamin D levels rose in that arm, narrowing the contrast between groups.
  • Baseline vitamin D levels were higher than in the phase 2 SUNSHINE trial, which may have reduced the room for benefit.
  • The left-sided PFS signal comes from an exploratory subgroup analysis and requires independent confirmation.

Future directions

  • Testing vitamin D3 in patients selected for vitamin D deficiency at baseline.
  • Prospective evaluation of the left-sided primary tumor subgroup suggested by the interaction analysis.
  • Mechanistic work on how circulating 25-hydroxyvitamin D relates to tumor biology and chemotherapy response.

Vitamin D has long attracted interest in colorectal cancer. Observational studies have linked higher circulating 25-hydroxyvitamin D concentrations to lower colorectal cancer incidence and to better survival after diagnosis, and laboratory work has described antiproliferative, prodifferentiating and immunomodulatory effects of the active metabolite. Vitamin D3 is inexpensive, orally available and generally well tolerated, which makes it an attractive candidate for addition to standard systemic therapy. Whether supplementation actually changes the course of established metastatic disease, however, can only be answered in an adequately powered randomized trial.

The rationale for SOLARIS came from the randomized phase 2 SUNSHINE trial, in which high-dose vitamin D3 added to first-line chemotherapy and bevacizumab improved progression-free survival compared with standard-dose supplementation. That trial was small, and phase 2 PFS gains frequently fail to replicate. SOLARIS (Alliance A021703) was designed as the confirmatory phase 3 test of the same strategy, led by Kimmie Ng of Dana-Farber Cancer Institute, who also led the earlier work, with Jeffrey Meyerhardt as senior author and statistical support from the Alliance Statistics and Data Management Center at Mayo Clinic.

The trial was double-blind and enrolled 455 patients with previously untreated, unresectable locally advanced or metastatic colorectal adenocarcinoma, measurable disease and ECOG performance status 0 or 1. Patients were randomized 1:1 to high-dose vitamin D3 (a loading dose of 8,000 IU daily for 14 days, then 4,000 IU daily; n=228) or standard-dose vitamin D3 (400 IU daily; n=227), and all received mFOLFOX6 or FOLFIRI with bevacizumab every two weeks. The primary endpoint was PFS, analyzed with an unstratified log-rank test. The median age was 59 years, 40% of participants were women, and the median follow-up was 20 months.

The trial did not meet its primary endpoint. Median PFS was 11.8 months (95% CI 10.3-13.3) with high-dose vitamin D3 and 10.3 months (95% CI 9.4-12.2) with standard dosing, corresponding to a hazard ratio of 0.92 and a one-sided P value of 0.25. The supplementation itself achieved its pharmacological aim: at first restaging, median plasma 25-hydroxyvitamin D was 38.3 ng/mL in the high-dose arm versus 23.5 ng/mL in the standard-dose arm, a difference that persisted through treatment discontinuation. The failure to improve PFS therefore cannot be explained by inadequate exposure.

Secondary endpoints were consistent with the primary result. The objective response rate was 51% (95% CI 44%-58%) with high-dose vitamin D3 and 44% (95% CI 37%-50%) with standard dosing (P=0.12), and median overall survival was 25.6 versus 27.0 months (P=0.66). In a prespecified but exploratory analysis, patients with left-sided primary tumors had a median PFS of 13.8 versus 10.2 months, with a significant treatment-by-tumor-location interaction. Grade 3 or higher toxicity was similar between arms, including neutropenia (32% vs 30%) and hypertension (20% vs 23%), and hypercalcemia, hyperphosphatemia and renal calculi were uncommon.

Several features of the design bear on interpretation. Because the comparator was standard-dose vitamin D3 rather than placebo, and because 25-hydroxyvitamin D levels also rose in the control arm, the trial tested the incremental value of high-dose over modest supplementation rather than vitamin D versus none. The investigators suggested that the discrepancy with SUNSHINE may reflect higher baseline vitamin D levels in SOLARIS together with the rise in the standard-dose arm. The left-sided subgroup finding is hypothesis-generating; the hazard ratio and confidence interval for that subgroup were not reported in the sources reviewed, and multiplicity is a concern.

The authors concluded that high-dose vitamin D3 added to standard chemotherapy cannot be recommended for unselected patients with previously untreated metastatic colorectal cancer. SOLARIS is a useful reminder that biologically plausible, low-toxicity interventions still require phase 3 confirmation, and that phase 2 PFS signals are fragile. Remaining questions include whether a benefit is confined to patients who are vitamin D deficient at baseline or to those with left-sided primaries, and whether vitamin D has any role in survivorship or prevention settings that this trial did not address.

Source: Ng et al., SOLARIS randomized clinical trial (Alliance A021703), JAMA, published online August 3, 2026. Read the paper

2. Intensified doublet total neoadjuvant therapy versus conventional chemoradiotherapy in high-risk locally advanced rectal cancer (Journal of Clinical Oncology)

Wang et al.; corresponding author Ziqiang Wang (Colorectal Cancer Center, Department of General Surgery, West China Hospital, Sichuan University, Chengdu)

Key points

  • Multicenter, open-label, randomized phase 3 trial at nine Chinese institutions; 458 patients with stage II/III rectal adenocarcinoma and at least one MRI-defined high-risk feature (232 doublet TNT, 226 conventional nCRT), enrolled June 2017 to December 2023.
  • Doublet long-course TNT (induction, concurrent and consolidation capecitabine plus oxaliplatin with long-course radiotherapy) versus capecitabine-based long-course chemoradiotherapy, surgery and adjuvant chemotherapy.
  • 3-year disease-free survival (DFS), the primary endpoint: 74.8% vs 66.0% (HR 0.674; 95% CI 0.489-0.929; P=0.016); 3-year metastasis-free survival 77.7% vs 67.6% (HR 0.655; 95% CI 0.469-0.915; P=0.013).
  • Pathologic complete response among operated patients: 26.37% vs 9.80% (P<0.001); 3-year overall survival 90.2% vs 87.5% (HR 0.727; P=0.167).
  • Grade 3 or higher adverse events during neoadjuvant therapy: 27.59% vs 8.56%; severe adverse events across the whole treatment course were comparable (28.02% vs 24.32%).

Key limitations

  • The conventional chemoradiotherapy control reflects an earlier treatment paradigm, given the long accrual period from 2017 to 2023.
  • Open-label, single-country design; overall survival was not significantly different and follow-up for late outcomes remains limited.
  • No comparison with short-course radiotherapy-based or non-doublet concurrent TNT regimens now widely used.

Future directions

  • Longer follow-up for overall survival, late toxicity and functional outcomes.
  • Head-to-head comparisons with short-course radiotherapy-based TNT and other modern TNT schedules.
  • Exploring whether the higher pCR rate can be leveraged for organ preservation strategies.

Preoperative fluoropyrimidine-based chemoradiotherapy followed by total mesorectal excision and adjuvant chemotherapy was for many years the standard of care for locally advanced rectal cancer. Although this approach achieves good local control, distant metastasis remains the dominant cause of treatment failure, and adjuvant chemotherapy after surgery is often delayed, dose-reduced or never delivered. Patients with MRI-defined high-risk features, such as cT4 tumors, extensive nodal disease, a threatened mesorectal fascia or extramural venous invasion, are at particular risk of systemic relapse.

Total neoadjuvant therapy (TNT), in which systemic chemotherapy is moved before surgery, was developed to address these problems by delivering systemic treatment earlier and more completely. Randomized trials of TNT have used different combinations of short- or long-course radiotherapy with various chemotherapy backbones, and the optimal design remains debated. One unresolved question is whether adding oxaliplatin across the induction, concurrent and consolidation phases of a long-course schedule improves outcomes compared with conventional capecitabine-based chemoradiotherapy in a population selected for high-risk MRI features.

This multicenter, open-label phase 3 trial, led from the Colorectal Cancer Center at West China Hospital, Sichuan University, randomized 458 patients 1:1 between June 2017 and December 2023 at nine Chinese institutions. The experimental doublet long-course TNT arm received capecitabine plus oxaliplatin as induction, concurrently with long-course radiotherapy, and as consolidation before surgery (n=232). The control arm received capecitabine with long-course radiotherapy, followed by surgery and adjuvant chemotherapy (n=226). High-risk features were common: 47.82% had cT4 disease, 75.98% cN2, 70.96% a threatened mesorectal fascia and 54.80% extramural venous invasion. The primary endpoint was disease-free survival.

At a median follow-up of 51 months, the trial met its primary endpoint. Three-year DFS was 74.8% with doublet TNT versus 66.0% with conventional chemoradiotherapy (HR 0.674; 95% CI 0.489-0.929; P=0.016). The benefit appeared to be driven by improved systemic control: three-year metastasis-free survival was 77.7% versus 67.6% (HR 0.655; 95% CI 0.469-0.915; P=0.013), whereas locoregional failure was similar at 6.03% and 6.19%. Three-year overall survival was numerically higher at 90.2% versus 87.5%, but the difference was not statistically significant (HR 0.727; P=0.167).

Pathologic outcomes favored the intensified regimen. Among patients who underwent surgery, the pathologic complete response rate was 26.37% with doublet TNT versus 9.80% with conventional treatment (P<0.001), tumor regression grade 0-1 was achieved in 49.25% versus 28.92%, and ypN0 status in 81.09% versus 69.12%. Toxicity was front-loaded: grade 3 or higher adverse events during neoadjuvant therapy occurred in 27.59% versus 8.56%, with thrombocytopenia the most common, but severe adverse events across the entire course were comparable (28.02% vs 24.32%), as were major postoperative complications.

The investigators acknowledge that the conventional control arm reflects an earlier treatment paradigm, a consequence of an accrual period spanning more than six years during which TNT became widely adopted. The comparison therefore establishes superiority over sequential chemoradiotherapy plus adjuvant chemotherapy, not over contemporary TNT schedules. The trial was open-label and conducted only in China, the overall survival difference was not significant, and longer follow-up is needed for late toxicity and function. The exact online publication date of the article could not be confirmed from the sources reviewed; it was reported by The ASCO Post on August 10, 2026.

Within these limits, the trial provides phase 3 evidence that an oxaliplatin-intensified, long-course TNT regimen improves disease-free and metastasis-free survival in rectal cancer selected for MRI high-risk features, with manageable toxicity. The investigators state that the findings support this doublet-based regimen as a standard option within the modern TNT paradigm. The authors call for comparisons against short-course radiotherapy-based and non-doublet concurrent TNT regimens, and the high pathologic complete response rate raises the question of whether the regimen could also support organ preservation in selected patients.

Source: Wang et al., Journal of Clinical Oncology, 2026 (reported August 10, 2026). Read the paper

3. Anti-DKK1 antibody sirexatamab plus bevacizumab and chemotherapy as second-line therapy for advanced colorectal cancer (DeFianCe, Clinical Cancer Research)

Zev A. Wainberg (UCLA), Kyong W. Lee, Jin G. Kim, et al.

Key points

  • Randomized, open-label, multicenter phase 2 trial (NCT05480306); Part B randomized 188 patients 1:1 after progression on one prior systemic therapy to sirexatamab plus FOLFIRI or mFOLFOX6 and bevacizumab, or chemotherapy plus bevacizumab alone.
  • The primary endpoint was not met: median PFS 9.2 vs 8.3 months (HR 0.84; 95% CI 0.58-1.21); ORR 35.1% vs 26.6%; OS HR 0.83 (95% CI 0.46-1.48).
  • Baseline plasma DKK1 above the median (n=88): median PFS 9.0 vs 7.1 months (HR 0.61; 95% CI 0.37-1.00); OS HR 0.42 (95% CI 0.19-0.91; P=0.0118).
  • Upper-quartile DKK1 (n=44): median PFS 9.4 vs 5.9 months (HR 0.46; 95% CI 0.22-0.96); median OS not reached vs 9.5 months (HR 0.17; 95% CI 0.05-0.53; P<0.001); treatment-by-DKK1 interaction P=0.0129 (PFS) and P=0.0027 (OS).
  • Grade 3 or higher treatment-emergent adverse events: 59.3% vs 67.0%; serious events 19.8% vs 19.3%.

Key limitations

  • Primary ITT endpoint not met; the DKK1-high benefit derives from exploratory biomarker analyses in small subgroups with wide confidence intervals.
  • Fewer PFS events than planned (119 vs 145) and an open-label phase 2 design.
  • Higher DKK1 was also prognostic for worse outcomes on standard therapy, complicating interpretation of the predictive signal.

Future directions

  • A biomarker-selected phase 3 registrational trial in DKK1-high disease, as the authors propose.
  • Prospective validation and standardization of a plasma DKK1 assay and threshold.
  • Mechanistic work on DKK1 and the tumor immune microenvironment to explain the effect.

After progression on first-line therapy, most patients with metastatic colorectal cancer receive a second fluoropyrimidine-based doublet with an antiangiogenic agent such as bevacizumab. Outcomes in this setting are modest, and for patients with microsatellite-stable disease there are few biomarker-directed options beyond RAS, BRAF and HER2 status. New targets that could be added to established second-line backbones, ideally with a blood-based biomarker to select patients, are therefore a clinical priority.

Dickkopf-1 (DKK1) is a secreted modulator of Wnt signaling that has been implicated in tumor growth and in shaping an immunosuppressive microenvironment. Sirexatamab (DKN-01) is a monoclonal antibody directed against DKK1. Interim data from the DeFianCe trial were presented at the ESMO Congress 2025 (abstract LBA34), and in May 2026 the FDA granted fast track designation to sirexatamab with chemotherapy and bevacizumab for DKK1-high metastatic colorectal cancer progressing after one prior systemic therapy. The full Part B results were published in Clinical Cancer Research, as announced on August 12, 2026.

DeFianCe is a two-part, randomized, open-label, multicenter phase 2 trial. In Part B, 188 patients with histologically confirmed colorectal adenocarcinoma, measurable disease and ECOG performance status 0 or 1, who had progressed on one prior line of systemic therapy, were randomized 1:1 to sirexatamab 400 mg intravenously on day 1 of each 14-day cycle (with an additional dose on day 8 of cycle 1) plus investigator-choice FOLFIRI or mFOLFOX6 and bevacizumab, or to chemotherapy and bevacizumab alone. The primary endpoint was investigator-assessed PFS in the intention-to-treat population; baseline plasma DKK1 was a prespecified exploratory biomarker.

In the intention-to-treat population the primary endpoint was not met. Median PFS was 9.2 months (95% CI 7.3-10.9) with sirexatamab versus 8.3 months (95% CI 7.2-9.5) with control (HR 0.84; 95% CI 0.58-1.21), the objective response rate was 35.1% versus 26.6%, and median overall survival had not been reached in either arm (HR 0.83; 95% CI 0.46-1.48). The final analysis included 119 PFS events, fewer than the 145 planned, which reduced the statistical power of the comparison.

The biomarker analyses told a different story. Plasma DKK1 was detectable in all patients across about an eightfold range. Among patients with DKK1 above the median (n=88), median PFS was 9.0 versus 7.1 months (HR 0.61; 95% CI 0.37-1.00) and overall survival favored sirexatamab (HR 0.42; 95% CI 0.19-0.91; P=0.0118). In the upper quartile (n=44), response rates were 44.0% versus 15.8%, median PFS 9.4 versus 5.9 months (HR 0.46; 95% CI 0.22-0.96) and median overall survival not reached versus 9.5 months (HR 0.17; 95% CI 0.05-0.53; P<0.001). Treatment-by-DKK1 interaction tests were significant for PFS (P=0.0129) and OS (P=0.0027). Grade 3 or higher adverse events occurred in 59.3% versus 67.0%.

These results must be read cautiously. The trial failed its primary endpoint, and the DKK1-high findings come from exploratory analyses in small subgroups, with an upper confidence bound of 1.00 for PFS in the above-median group. Event accrual fell short of plan, the design was open-label, and higher DKK1 was itself associated with worse outcomes on standard therapy, so part of the apparent effect could reflect differences in prognosis. Tumoral DKK1 mRNA was low in most tissue samples, leading the authors to favor plasma DKK1 for selection, but the assay and threshold have not been prospectively validated.

DeFianCe nonetheless identifies a plausible, blood-based biomarker that could enrich a future trial for patients most likely to benefit from DKK1 blockade, and the consistency of the signal across continuous, threshold-based and quartile analyses is notable. The authors present the results as support for a biomarker-selected phase 3 registrational trial in DKK1-high disease. Confirmation in such a trial, with a standardized plasma assay and a prespecified threshold, will be essential before sirexatamab can be considered for clinical use.

Source: Wainberg et al., "Sirexatamab in Combination with Bevacizumab and Chemotherapy as Second-Line Therapy for Advanced Colorectal Adenocarcinoma: the Phase II DeFianCe Trial," Clinical Cancer Research, announced August 12, 2026. Read the paper

4. Colorectal cancer mortality over 21 years after an organized FIT-based screening program in the Basque Country (JAMA Network Open)

Luis Bujanda (Biogipuzkoa Health Research Institute-Donostia University Hospital, University of the Basque Country), Beatriz Val, Isabel Idigoras-Rubio, Jesus M. Banales, Maria A. Gutierrez-Stampa, Joaquín Cubiella, Laura Izquierdo-Sanchez, et al.; last author Isabel Portillo Villares (Basque Country Colorectal Cancer Screening Programme, Osakidetza)

Key points

  • Retrospective population-based cohort covering all residents of the Basque Country, Spain (about 2.2 million people), 2004-2024; biennial single-sample FIT screening for ages 50-69 began in 2009, with people under 50 as a nonscreened reference group.
  • Of 438,134 deaths, 16,298 (3.7%) were from colorectal cancer; the overall age-standardized CRC mortality rate fell from 32.8 to 23.1 per 100,000, a 29.6% reduction (P=0.002).
  • In the screened age group (50-69), CRC mortality fell 50.1%, from 39.7 to 19.8 per 100,000; counterfactual analysis estimated a 46.2% relative reduction versus projected trends.
  • Segmented regression identified a trend breakpoint in 2012 (95% CI 2011-2014; P<0.001), three years after the program started.
  • Of 4,892 screen-detected cancers with staging, 3,464 (70.8%) were stage I or II; mortality also fell 19.6% at age 70 or older.

Key limitations

  • Ecological, descriptive design; declines cannot be attributed to screening alone, given concurrent changes in treatment and other factors.
  • No individual-level comparison of screened versus unscreened people, so effect sizes are not hazard ratios.
  • A single regional program in Spain; results may not generalize to settings with different participation or colonoscopy follow-up.

Future directions

  • Individual-level linkage studies to estimate the effect of participation, FIT positivity and colonoscopy completion on mortality.
  • Evaluation of extending screening below age 50 given rising early-onset incidence.
  • Cost-effectiveness and equity analyses of long-running organized FIT programs.

Organized screening programs based on the fecal immunochemical test (FIT) are now the backbone of population colorectal cancer control in much of Europe. Randomized trials of guaiac-based fecal occult blood testing established that stool-based screening can reduce colorectal cancer mortality, and FIT has since replaced guaiac testing because of its higher sensitivity and ease of use. Because FIT programs were introduced as public health policy rather than within randomized trials, however, evidence about their long-term population impact must largely come from well-conducted observational analyses of complete regional populations.

The Basque Country in northern Spain offers an unusually well-documented setting. Its organized program, targeting asymptomatic adults aged 50 to 69 with biennial single-sample FIT, began in 2009 and has now been running for more than a decade, long enough for effects on mortality, which lag behind effects on incidence and stage, to become visible. Earlier evaluations of FIT programs have reported early stage shifts and declines in mortality, but long follow-up with formal trend modeling and counterfactual projections has been less common.

Bujanda and colleagues, from Biogipuzkoa Health Research Institute and Donostia University Hospital, the Basque screening program, the Basque Department of Health and Galician hospitals, conducted a retrospective population-based cohort study of all residents, about 2.2 million people, from 2004 to 2024. Age-standardized colorectal cancer mortality rates per 100,000 (European Standard Population 2013) and annual trends were estimated, and segmented regression and counterfactual analyses were used to quantify changes after screening began. People younger than 50 served as a nonscreened reference group. Participation, FIT positivity, colonoscopy adherence, detection and stage distribution were also examined.

Among 438,134 deaths over 21 years, 16,298 (3.7%) were attributed to colorectal cancer. The overall age-standardized mortality rate fell from 32.8 to 23.1 per 100,000, a 29.6% reduction (P=0.002). The largest change occurred in the screened age group: among people aged 50 to 69, colorectal cancer mortality fell by 50.1%, from 39.7 to 19.8 per 100,000. Segmented regression located a significant inflection in the mortality trend in 2012 (95% CI 2011-2014; P<0.001), three years after the program began, and counterfactual analysis estimated a 46.2% relative reduction compared with the trend projected without screening in this age group.

Mortality also declined among people aged 70 or older, by 19.6% (from 165.4 to 132.9 per 100,000), with a 45.4% relative reduction against projected trends, consistent with the aging of previously screened cohorts. Across all ages, the counterfactual relative reduction was estimated at 41.7%. Screening appeared to shift diagnosis toward earlier disease: of 4,892 screen-detected cancers with staging information, 3,464 (70.8%) were stage I or II, a pattern consistent with the mechanism through which stool-based screening is expected to reduce mortality.

The study is descriptive and ecological, and its findings are associations rather than causal estimates. Mortality declines over two decades also reflect improvements in surgery, systemic therapy and multidisciplinary care, and the authors did not compare screened with unscreened individuals, so the reported relative reductions are not individual-level hazard ratios. Counterfactual projections depend on modeling assumptions, and the program operated within a single, well-resourced regional health system with its own participation and follow-up colonoscopy rates. Some of the reported confidence intervals relate only to breakpoint years and projected rates.

Despite these caveats, a 50% fall in colorectal cancer mortality in the screened age range, with a trend change three years after implementation, is strong real-world support for organized biennial FIT screening as a long-term public health strategy. The findings are relevant to countries considering or expanding stool-based programs and to debates about screening modality. Future work linking individual screening histories to outcomes, and assessing the case for extending screening to younger adults given rising early-onset incidence, would strengthen causal inference and inform program design.

Source: Bujanda et al., "Colorectal Cancer Mortality Following an Organized Fecal Immunochemical Test-Based Screening Program," JAMA Network Open, published online August 5, 2026. Read the paper

5. Genetic ancestry, colorectal cancer burden and age-specific risk in the All of Us Research Program (JAMA Network Open)

Odysseas P. Chatzipanagiotou (Department of Surgery, The Ohio State University Wexner Medical Center and James Comprehensive Cancer Center, Columbus), Charalampos M. Charalampous, Adam Cordle, Abdulaziz Elemosho, et al.; senior author Timothy M. Pawlik (The Ohio State University Wexner Medical Center)

Key points

  • Cohort study of 316,624 All of Us participants with whole-genome sequencing and linked health records (data July 1986 to October 2023; median follow-up 133.1 months); 2,914 (0.9%) developed colorectal cancer.
  • European genetic ancestry was associated with greater overall CRC burden versus all other ancestries combined (OR 1.50; 95% CI 1.39-1.62).
  • Median age at diagnosis was 63.4 years in European ancestry and 8.4 years earlier in American admixed-Latino and 7.9 years earlier in East Asian ancestry.
  • Age-specific hazard versus European ancestry: American admixed-Latino HR 1.30 (95% CI 1.14-1.47); East Asian HR 1.43 (95% CI 1.06-1.94).
  • A multiethnic XGBoost risk model achieved an ROC AUC of 0.898 (95% CI 0.882-0.912) and a precision-recall AUC of 0.338 (95% CI 0.296-0.379).

Key limitations

  • All of Us is not population-based; restriction to participants with genomes and linked records may introduce selection bias.
  • Locus-specific ancestry, tumor molecular features and stage were unavailable.
  • The prediction model has not been externally validated, and its precision-recall performance is modest.

Future directions

  • External validation of the multiethnic risk model in population-based cohorts.
  • Investigation of the environmental, socioeconomic and genetic factors behind earlier diagnosis in Latino and East Asian ancestry groups.
  • Prospective testing of ancestry-informed risk enrichment as a complement to age-based screening.

Colorectal cancer incidence and mortality differ substantially across racial and ethnic groups in the United States, and the rising incidence of early-onset disease has intensified interest in who should start screening earlier. Most studies of these disparities rely on self-reported race and ethnicity, which combine social, environmental and biological influences and may poorly capture underlying genetic ancestry. Large biobanks that couple genome sequencing with electronic health records allow ancestry to be measured directly and related to disease risk across the lifespan.

The National Institutes of Health All of Us Research Program was established to build a diverse cohort in which groups historically underrepresented in biomedical research are well represented. Prior work had not systematically examined, within one dataset, how genetically inferred ancestry relates to both the overall burden of colorectal cancer and the age at which it occurs, nor whether ancestry and clinical variables could be combined into a multiethnic prediction model that might support risk-stratified screening.

Chatzipanagiotou, Pawlik and colleagues in the Department of Surgery at The Ohio State University Wexner Medical Center analyzed Controlled Tier data from 316,624 participants with whole-genome sequencing and linked records, collected between July 1986 and October 2023 (median follow-up 133.1 months). Ancestry groups were European (54.4%), African (19.6%), American admixed-Latino (16.7%), East Asian, South Asian and Middle Eastern combined (3.1%), and other (6.2%). Logistic regression, cause-specific hazard models of attained age accounting for the competing risk of death, and penalized LASSO and XGBoost prediction models were used.

Overall, 2,914 participants (0.9%) developed colorectal cancer. European ancestry was associated with greater crude odds of colorectal cancer than all other ancestries combined (OR 1.50; 95% CI 1.39-1.62), whereas odds were lower for African (OR 0.77; 95% CI 0.70-0.85) and American admixed-Latino ancestry (OR 0.63; 95% CI 0.56-0.71). Age at diagnosis told a different story. The median age at diagnosis was 63.4 years among people of European ancestry, but 8.4 years earlier among American admixed-Latino, 7.9 years earlier among East Asian and 5.0 years earlier among African ancestry participants.

Modeling risk by attained age showed higher age-specific hazards for American admixed-Latino (HR 1.30; 95% CI 1.14-1.47) and East Asian ancestry (HR 1.43; 95% CI 1.06-1.94) compared with European ancestry. In adjusted analyses, family history of colorectal cancer (aOR 2.82; 95% CI 2.42-3.29), type 2 diabetes (aOR 1.95; 95% CI 1.79-2.13), prior cancer (aOR 1.61) and family history of polyps (aOR 1.52) were associated with colorectal cancer. A multiethnic XGBoost model reached an ROC AUC of 0.898 (95% CI 0.882-0.912), with similar performance in European (0.900) and non-European (0.869) participants, compared with 0.733 for the LASSO model.

The authors note that All of Us is not a population-based sample and that limiting the analysis to participants with genomes and linked records may introduce selection bias and limit generalizability. Locus-specific ancestry, tumor molecular characteristics and stage were not available. The higher crude burden among European-ancestry participants partly reflects their older age, a reminder that crude and age-specific comparisons answer different questions. The high ROC AUC should also be read alongside a modest precision-recall AUC of 0.338, reflecting the low prevalence of the outcome, and the model has not yet been validated externally.

The study suggests that genetic ancestry is associated with meaningful differences not only in how often colorectal cancer occurs but also in when it occurs, with earlier diagnosis in Latino and East Asian ancestry groups. The authors propose that a multiethnic model of this kind may complement screening as a risk-enrichment tool, helping identify people who might benefit from earlier or more intensive screening. External validation, integration of social and environmental exposures, and prospective evaluation within screening programs will be needed before such tools can inform practice.

Source: Chatzipanagiotou et al., "Genetic Ancestry and Colorectal Cancer in the All of Us Dataset," JAMA Network Open, published online August 12, 2026. Read the paper

6. Body mass index at diagnosis, prediagnostic weight loss and the obesity paradox in colorectal cancer survival (DACHS, JAMA Network Open)

Marko Mandic (Cancer Prevention Graduate School, German Cancer Research Center [DKFZ], Heidelberg), Durgesh Wankhede, Fatemeh Safizadeh, Michael Hoffmeister, et al.; corresponding author Hermann Brenner (DKFZ, Heidelberg)

Key points

  • Population-based case-control study with prospective follow-up of 5,521 patients with colorectal cancer recruited from more than 20 hospitals in southwest Germany between 2001 and 2016; median follow-up 5.2 years, with 2,429 deaths (1,381 CRC-related).
  • Compared with normal weight, overweight at diagnosis was associated with modestly better overall survival (HR 0.89; 95% CI 0.81-0.98); obesity showed no significant association.
  • Weight loss of more than 10% in the 5-14 years before diagnosis was associated with worse overall survival (HR 1.55; 95% CI 1.39-1.72).
  • After mutual adjustment for BMI at diagnosis and prediagnostic weight change, the apparent benefit of overweight and obesity became null.
  • The authors conclude the obesity paradox is largely explained by cancer-associated prediagnostic weight loss and should not prompt weight gain or changes to survivorship advice.

Key limitations

  • Height and weight were self-reported, including recalled weight 5-14 years before diagnosis, introducing measurement and recall error.
  • Observational design; residual confounding by comorbidity, frailty and tumor characteristics is possible.
  • BMI does not capture body composition such as muscle mass or visceral fat.

Future directions

  • Studies using measured body composition (for example CT-derived muscle and fat) to refine the relationship.
  • Routine incorporation of prediagnostic weight trajectories into prognostic analyses.
  • Investigation of unintentional weight loss as an early warning sign that could prompt earlier diagnostic workup.

Excess adiposity is an established risk factor for developing colorectal cancer, yet several observational studies have reported that patients who are overweight or obese at diagnosis survive longer than those of normal weight. This counterintuitive observation, often called the obesity paradox, has fueled debate about whether some excess weight might be protective after diagnosis, for example by providing metabolic reserve during treatment, or whether it reflects bias in how weight is measured and when.

One candidate explanation is reverse causation. Colorectal cancer can cause weight loss for months or years before it is diagnosed, so some patients classified as normal weight at diagnosis may have been heavier before their illness and may be sicker, with more advanced or aggressive disease. If so, BMI measured at diagnosis would partly mark disease severity rather than adiposity. Testing this requires information on weight trajectories well before diagnosis, which few cohorts have collected systematically.

Mandic, Brenner and colleagues at the German Cancer Research Center (DKFZ) in Heidelberg used a multicenter, population-based case-control study with prospective follow-up of patients aged 30 or older with histologically confirmed colorectal cancer, recruited from more than 20 hospitals in southwest Germany between 2001 and 2016 and followed through September 2018. Exposures were BMI at diagnosis and percentage weight change over the 5 to 14 years before diagnosis, both based on self-report. Cox models for overall, colorectal cancer-specific and recurrence-free survival were adjusted for demographic, lifestyle, tumor and comorbidity factors.

The cohort included 5,521 patients (39% female; median age 69 years). Over a median follow-up of 5.2 years there were 2,429 deaths, 1,381 of them related to colorectal cancer. In models that considered BMI at diagnosis alone, overweight was associated with modestly better overall survival than normal weight (HR 0.89; 95% CI 0.81-0.98), while no significant association was seen for obesity. Prediagnostic weight loss of more than 10% was strongly associated with worse overall survival (HR 1.55; 95% CI 1.39-1.72).

The key analysis adjusted BMI at diagnosis and prediagnostic weight change for each other. Once weight loss before diagnosis was taken into account, the associations between overweight or obesity at diagnosis and survival became null, indicating that much of the apparent survival advantage was attributable to the poorer prognosis of patients who had lost substantial weight before diagnosis and had thereby moved into lower BMI categories. The study also analyzed colorectal cancer-specific and recurrence-free survival; the abstract reports the overall survival estimates summarized here.

The findings rest on self-reported height and weight, including recalled weight from 5 to 14 years earlier, which is vulnerable to recall error and misclassification. As an observational study, residual confounding by comorbidity, frailty, treatment and tumor biology cannot be excluded, and BMI is a crude measure that does not distinguish muscle from fat or capture visceral adiposity. The study population was drawn from one region of Germany with recruitment ending in 2016, and treatment has evolved since, although the methodological point about reverse causation is unlikely to depend on era.

The authors conclude that the obesity paradox in colorectal cancer appears to be largely explained by cancer-related weight loss before diagnosis, and that the survival advantage seen with higher BMI should not be read as a reason for people of healthy weight to gain weight or as grounds to revise survivorship recommendations. The work argues for routinely accounting for prediagnostic weight trajectories in prognostic research and supports further study of measured body composition. Unintentional weight loss may also deserve attention as a clinical signal of undiagnosed disease.

Source: Mandic et al., "Body Mass Index at Diagnosis, Prediagnostic Weight Loss, and Survival in Colorectal Cancer," JAMA Network Open, published online August 14, 2026. Read the paper

7. A previously missed late-arising mutational signature in microsatellite-stable colorectal cancer (Mutographs, Nature Communications)

Mariya Kazachkova, B. Otlu, M. Díaz-Gay, A. Abbasi, S. Moody, Z. Jiang, S. Perdomo, D. C. Wedge, P. Brennan, M. R. Stratton, et al.; senior author Ludmil B. Alexandrov (University of California San Diego and Moores Cancer Center)

Key points

  • Whole-genome analysis of 802 treatment-naive, DNA repair-proficient, microsatellite-stable colorectal tumors identified a single-base substitution signature (SBS_D) previously absorbed into known signatures; it is now provisionally designated SBS111 in COSMIC.
  • The signature was independently re-extracted in three validation cohorts from the United States, Sweden and the United Kingdom; UC San Diego reported 2,616 tumors across three independent colorectal cancer datasets.
  • About 10% of colorectal tumors without obvious DNA repair defects carried mutations resembling the pattern, which accounted for roughly one-sixth of mutations where active.
  • SBS_D mutations accumulated late in tumor evolution, were depleted in nucleosome-wrapped DNA, and were associated with a modestly higher burden of small insertions and deletions.
  • Tumors with the signature lacked the high mutation burden and genomic instability of repair-deficient cancers; its molecular cause remains unresolved.

Key limitations

  • Purely computational analysis; the underlying mutational process has not been demonstrated experimentally.
  • Gene expression differences linked to the signature were modest, and its clinical relevance is unknown.
  • Validation cohort sizes and several secondary analyses were not reported in the sources reviewed.

Future directions

  • Experimental work to identify the mechanism, which Cancer Grand Challenges Team CAUSE, led by Alexandrov, plans to pursue.
  • Testing associations of the signature with exposures, early-onset disease and clinical outcomes.
  • Incorporating SBS111 into reference catalogs and tumor sequencing pipelines.

Mutational signatures are characteristic patterns of somatic mutations left in the genome by specific DNA-damaging exposures and by defects in replication or repair. Cataloging them has helped link cancers to causes such as tobacco smoke, ultraviolet light, aristolochic acid and mismatch repair deficiency, and has revealed exposures such as the bacterial genotoxin colibactin in colorectal cancer. Because signature extraction is statistical, however, a genuine but less common process can be mistaken for a combination of better-known signatures and remain hidden.

Most colorectal cancers are microsatellite stable and proficient in DNA repair, and in these tumors the processes that generate somatic mutations are only partly understood. The Mutographs team, funded by Cancer Grand Challenges, has been sequencing cancers from populations with differing incidence to identify exposures that might explain geographic variation and the rising incidence of colorectal cancer in younger adults. Re-examining large whole-genome datasets with careful extraction methods offers a way to detect signatures that earlier analyses had folded into others.

Kazachkova, Alexandrov and collaborators, including investigators associated with Mutographs, analyzed whole-genome sequences from 802 treatment-naive, DNA repair-proficient, microsatellite-stable colorectal tumors from a large international sequencing project. De novo extraction across 288 mutation categories was followed by independent extraction in three validation cohorts from the United States, Sweden and the United Kingdom, restricted to comparable tumors. UC San Diego reported 2,616 tumors across three independent colorectal cancer datasets. Analyses examined mutation timing, nucleosome positioning, associated indel patterns and, where RNA sequencing was available, gene expression.

The investigators identified a signature, initially called SBS_D and now provisionally designated SBS111 in the COSMIC catalog, that had previously been attributed to a known signature closely resembling one linked to reactive oxygen species damage. On closer examination it was not associated with the processes to which it had been assigned. About 10% of colorectal tumors without obvious DNA repair defects carried mutations resembling the pattern, and where it was active it accounted for roughly one-sixth of all mutations. The finding was reproduced in the independent cohorts.

Several features distinguish the signature. Its mutations tend to accumulate late in tumor evolution, whereas the related known signature appears from early stages. SBS_D mutations are notably depleted in nucleosome-wrapped DNA, which the authors describe as unusual, and the signature is associated with a modestly higher burden of small insertions and deletions, correlating with an indel pattern previously linked to replication slippage. Tumors carrying it lack the elevated mutation burden and genomic instability typical of repair-deficient cancers, and associated gene expression differences were modest. The authors link it to DNA repair infidelity rather than outright repair failure.

The work is computational, and the molecular origin of the signature remains unresolved; experimental studies will be required to establish its cause. Signature extraction depends on methodological choices, and although validation across independent cohorts mitigates this concern, the sizes of the individual validation cohorts and the results of some secondary analyses, such as associations with stage and driver mutations, were not reported in the sources reviewed. The clinical significance of the signature, including any association with prognosis, treatment response or patient age, has not been established.

By recovering a recurrent mutational process that had been hidden within existing catalogs, the study refines the map of how microsatellite-stable colorectal cancers acquire mutations and identifies a late-acting process worth explaining. The findings provide a foundation for Team CAUSE, a Cancer Grand Challenges team led by Alexandrov that will investigate the mechanisms behind this and other unexplained signatures. Linking SBS111 to specific exposures, host factors or clinical features, including early-onset disease, will determine whether it has etiological or clinical relevance.

Source: Kazachkova et al., "Identification and validation of a previously missed mutational signature in colorectal cancer," Nature Communications, August 12, 2026. Read the paper