GI Oncology Weekly — Issue 5

This week’s calls — 5 readouts · 2 to act on · ~9 min read. Tap any row to jump.

ACS 2026 colorectal screening guideline Blood-based tests are formally ranked second-line; stool DNA and stool RNA join FIT as preferred options Practice-shaping ASCO Living Guideline v2026.1.2 First-line selection in advanced gastroesophageal cancer now turns on HER2, PD-L1 and MSI status Practice-shaping SimpleScreen CRC (PREEMPT CRC) An approved cfDNA blood test: 81.1% sensitivity for cancer, 13.7% for advanced precancer Worth watching Daraxonrasib resistance mapping Resistance runs through mutant KRAS amplification, not secondary KRAS mutations On the radar γδ T-cell signature (AtezoBev in HCC) A candidate predictive marker for atezolizumab–bevacizumab that does nothing under sorafenib On the radar

This week · 5 readouts

The Verdicts

Every trial that matters this week, distilled to a single call — act on it, watch it, or keep it on the radar.

Practice-shaping Colorectal · screening and early detection

ACS 2026 colorectal screening guideline

The guideline that tells you what to do with the blood test that just got approved. The ACS reaffirmed screening from 45 through 75 and then graded the molecular tests against each other: multitarget stool DNA and stool RNA join high-sensitivity FIT as preferred options, and blood-based tests explicitly do not — on the grounds of lower sensitivity for advanced precancerous lesions and stage I cancers. Offer a blood test to the patient who declines or fails to complete a preferred test. That is a real and useful indication. It is not equivalence.

· Screening ages 45–75 for average-risk adults with life expectancy over 10 years· Preferred: annual high-sensitivity FIT or gFOBT, multitarget stool DNA and stool RNA at 3-year intervals· Blood-based tests recommended only for those who decline or do not complete a preferred test

Go deeper

The American Cancer Society reaffirmed screening from age 45 through 75 for average-risk adults with a life expectancy over 10 years, and then did something more consequential: it graded the new molecular tests against each other. The next-generation multitarget stool DNA test and the multitarget stool RNA test both showed high sensitivity for colorectal cancer and moderate sensitivity for advanced precancerous lesions, and join annual high-sensitivity FIT and gFOBT as preferred stool-based options at three-year intervals. Blood-based tests did not. In the guideline's own words, they "demonstrated lower sensitivity for both advanced precancerous lesions and stage I cancers, with modeling studies predicting less effectiveness in reducing CRC incidence and mortality," and "at this time, blood-based tests should be recommended only to individuals who decline or do not complete preferred screening tests." Full daily entry →

CA: A Cancer Journal for Clinicians — Wolf AMD, Smith RA and the ACS Guideline Development Group Debate on X →

Practice-shaping Gastroesophageal · advanced disease, first line

ASCO Living Guideline v2026.1.2

The living-guideline format working as designed. Version 2026.1.2 updates immunotherapy and targeted therapy for advanced gastroesophageal adenocarcinoma and squamous cell carcinoma, and its organising principle is biomarker-guided selection rather than a single chemotherapy backbone. The practical consequence for clinic is that the first-line conversation now begins with a biomarker panel. The live question is no longer whether to test HER2, PD-L1 and MSI, but in what order — and what you do when two come back positive at once.

· Covers advanced gastroesophageal adenocarcinoma and squamous cell carcinoma· Selection axis: HER2 status, PD-L1 expression, MSI-H/dMMR· Living-guideline format — updates on evidence, not on a republication cycle· ASCO Expert Panel, multi-institutional

Go deeper

ASCO's living-guideline format exists so that a recommendation set can move when the evidence moves rather than on a fixed republication cycle, and version 2026.1.2 is that mechanism working as designed for advanced gastroesophageal cancer. The update addresses immunotherapy and targeted therapy across advanced gastroesophageal adenocarcinoma and squamous cell carcinoma, and its organising principle is biomarker-guided selection: HER2 status, PD-L1 expression, and MSI-H/dMMR status together determine which of several competing first-line platforms a patient should receive. It is worth being clear about what this document is and is not — it is a synthesis authored by an ASCO Expert Panel, so it reports no new endpoint data of its own, and its value lies in adjudicating between trials rather than adding to them. Full daily entry →

Journal of Clinical Oncology — Shah MA, Kennedy EB and the ASCO Expert Panel

Worth watching Colorectal · screening and early detection

SimpleScreen CRC (PREEMPT CRC)

The approval is real and the validation is serious — over 48,000 asymptomatic adults with screening colonoscopy as the reference standard applied to everyone, which is the design a screening test should have to survive. The gap between the two headline numbers is the whole story. A test that finds four in five cancers and roughly one in seven advanced precancers is a detection tool, not a prevention tool, and screening is supposed to prevent. Worth knowing that the label figures are census-adjusted and run about two points above observed.

· CRC sensitivity 81.1% census-adjusted (71.3–88.1) vs 79.2% observed (68.4–86.9)· Advanced precancerous lesion sensitivity 13.7% adjusted (12.4–15.0) vs 12.5% observed (11.3–13.8)· PREEMPT CRC n>48,000 asymptomatic adults 45–85, 200+ sites

Sensitivity by lesion type, census-adjusted · 0–100% scale

81.1%Colorectal cancer

13.7%Advanced precancerous lesion

050100%

Go deeper

The FDA approved SimpleScreen CRC on July 27, 2026, a cell-free DNA methylation blood test for average-risk adults 45 and older, with Abbott commercialising it in the US. The validation behind it is genuinely strong: PREEMPT CRC enrolled more than 48,000 asymptomatic average-risk adults aged 45–85 across over 200 sites, all of whom were scheduled for screening colonoscopy — the reference standard applied to everyone, which is the design a screening test should have to survive. The label numbers come from a prespecified post-stratification analysis, and reading it carefully matters. Adjusting the cohort to the US Census age and sex distribution raised CRC sensitivity from an observed 79.2% (95% CI 68.4–86.9) to 81.1% (71.3–88.1), and raised advanced precancerous lesion sensitivity from 12.5% (11.3–13.8) to 13.7% (12.4–15.0). Full daily entry →

Journal of the National Cancer Institute + FDA approval (Freenome / Abbott) — Shaukat A, Levin TR and the PREEMPT CRC investigators (NCT04369053) Debate on X →

On the radar Pancreas · RAS-mutant metastatic adenocarcinoma

Daraxonrasib resistance mapping

A resistance map with an unusually actionable shape. Paired ctDNA from 44 patients on daraxonrasib monotherapy shows that when pancreatic cancer escapes, it does so by amplifying the target rather than mutating it — the opposite of the pattern that defines G12C(OFF) inhibitors. That distinction matters, because you cannot re-engineer binding around a tumour that simply makes more target. It points at combinations rather than at a next-generation single agent, and the preclinical work names candidates.

· Treatment-emergent RAS-pathway alterations 26/44 (59%)· Mutant KRAS amplification 16/44 (36%)· MAPK 11/44 (25%)· RTK 4/44 (9%)· PI3K 4/44 (9%)· Acquired secondary KRAS mutations: none

Treatment-emergent alterations at progression (n=44) · 0–100% scale

59%Any RAS-pathway alteration

0%Acquired secondary KRAS mutation

050100%

Go deeper

Daraxonrasib is the oral RAS(ON) multi-selective tri-complex inhibitor whose phase 1/2 activity in previously treated RAS-mutant metastatic pancreatic adenocarcinoma justified the randomised phase III RASolute 302 trial. This analysis sequenced over 800 genes in paired pretreatment and end-of-treatment circulating tumour DNA from 44 patients in that phase 1/2 study. Treatment-emergent alterations in the RAS signalling pathway appeared in 26 of 44 patients (59%), most notably mutant KRAS amplification in 16 of 44 (36%), alongside receptor tyrosine kinase (4/44, 9%), MAPK (11/44, 25%) and PI3K (4/44, 9%) pathway alterations. Notably, no acquired secondary KRAS mutations were observed — a resistance profile distinct from that of mutant-selective KRAS G12C(OFF) inhibitors. Concordant mechanisms were found or mechanistically established in human and murine preclinical models, including mutant KRAS and MYC amplification and RTK upregulation; combining daraxonrasib with agents targeting DNA damage response, RTKs, or the mutant-selective RAS(ON) G12D inhibitor zoldonrasib averted resistance in those models. Full daily entry →

Nature Medicine — Aronchik I, Kar S, … Aguirre AJ, Singh M (Revolution Medicines, Dana-Farber, Washington University, MD Anderson, MSKCC, NYU Langone, Huntsman — multicentre) Debate on X →

On the radar Liver · advanced hepatocellular carcinoma, first line

γδ T-cell signature (AtezoBev in HCC)

First-line atezolizumab–bevacizumab has no established way to say in advance who benefits. This proposes one, and does the thing that separates a predictive marker from a prognostic one: the signal holds in AtezoBev-treated patients and disappears in sorafenib-treated patients from the same trial dataset. It then replicates in MSI-high colorectal cancer. The cohort is small and the validation retrospective, so this is not a selection tool yet — but the design asks the right question, which is rarer than it should be.

· Prospective cohort n=31, first-line AtezoBev· High baseline circulating γδ T cells: disease control p=0.006, PFS p=0.02· High intratumoral γδ T signature: response p<0.01, PFS p<0.001, OS p=0.003 — AtezoBev only, not sorafenib· Validation GO30140/IMbrave150 (209 AtezoBev, 58 sorafenib) + 163 MSI-high CRC (PFS p=0.034)· Rising CD8⁺TIGIT⁺ shorter PFS p=0.04, rising CD8⁺PD-1⁺ longer PFS p=0.03

Go deeper

Atezolizumab plus bevacizumab is the first-line standard in advanced hepatocellular carcinoma, and the field still has no reliable way to say in advance who will benefit. This study takes a prospective cohort of 31 patients starting first-line AtezoBev — 87% male, median age 65, 65% BCLC-C — and layers sequential PBMC immunophenotyping, cytokine profiling and whole-blood RNA sequencing across treatment. High baseline circulating γδ T cells were associated with disease control (p=0.006) and longer progression-free survival (p=0.02). In tumour RNA-sequencing, a high intratumoral γδ T signature tracked with higher response (p<0.01), longer progression-free survival (p<0.001) and longer overall survival (p=0.003) — and the critical control is that this held only in AtezoBev-treated patients, not in those given sorafenib, which is what separates a predictive marker from a merely prognostic one. Validation drew on GO30140/IMbrave150 (209 AtezoBev, 58 sorafenib) plus 163 MSI-high colorectal cancers treated with immunotherapy, where a high γδ T-cell signature again tracked with longer progression-free survival (p=0.034). On-treatment dynamics added a second axis: rising circulating CD8⁺TIGIT⁺ cells marked shorter progression-free survival (p=0.04) while rising CD8⁺PD-1⁺ cells marked longer (p=0.03). Full daily entry →

Clinical Cancer Research — Galy-Fauroux I, … Zucman-Rossi J, Nault JC, Allaire M (Centre de Recherche des Cordeliers, multicentre France/Italy)

The take

The take — what a test is for

Four of these five papers are about the same question wearing different clothes: what is this test actually for? Not whether it works — whether the thing it does is the thing you needed done.

The colorectal screening pair is the clearest case, and it is worth reading them in the order they landed. A cell-free DNA blood test earns FDA approval on a genuinely serious validation: 48,000 asymptomatic adults, colonoscopy as the reference standard for everyone, 81.1% sensitivity for cancer. Days later the ACS declines to make it a preferred option, because the same study found 13.7% sensitivity for advanced precancerous lesions. Both bodies are reading the same data and neither is wrong. The FDA asked whether the test performs as claimed. The ACS asked whether the claim is the one that matters for a programme whose purpose is to prevent cancer, not to find it.

That distinction has a clinical shape. Roughly a third of eligible US adults are not screened at all, and for the patient who has declined colonoscopy twice and will not post a stool sample either, a blood draw is unambiguously better than the nothing they are currently getting. The failure mode is not offering it — it is offering it to the patient who would have accepted a colonoscopy, and calling that equivalent. Last issue's DENEB reported 81% advanced-adenoma sensitivity from a case-control design; this issue's approved assay reports 13.7% from a screening population. The gap between those two numbers is mostly study design, and it is the reason the screening-population cohort is the only evidence that settles anything.

The other two run the same logic through treatment rather than screening. The γδ T-cell signature earns attention specifically because it fails where it should — present under atezolizumab–bevacizumab, absent under sorafenib in the same dataset. A marker that predicts outcome regardless of what you give is a prognostic curiosity; a marker that only works under the drug you are choosing is a selection tool. And the daraxonrasib resistance map is valuable because of a negative finding: no acquired secondary KRAS mutations at all. Escape by amplification rather than mutation is not a problem you solve with a better binder, which is why the paper points at combinations instead of a next-generation single agent.

The common thread:

A test that detects is not automatically a test that prevents; say which one you are ordering.
A biomarker earns the word predictive only by failing in the comparator arm.
Approval answers whether a claim is true, not whether it is the right claim.
When two guidelines disagree, check whether they are answering the same question before deciding one is wrong.
A resistance mechanism is useful in proportion to how clearly it names the next combination.

GI Oncology Weekly · curated by Dr. Allan Pereira, Moffitt Cancer Center. Educational summaries for practising oncologists — not medical advice; verify against primary sources before acting.

All weekly issues