Every trial that matters this week, distilled to a single call — act on it, watch it, or keep it on the radar.
Practice-shapingGastric / GEJ · perioperative immunotherapy
MATTERHORN
The definitive readout of the week, and the one that settles a question this digest has been tracking since NCCN added durvalumab plus FLOT as category 1 back in April. MATTERHORN randomised 948 patients with resectable gastric or gastro-oesophageal junction adenocarcinoma across 147 centres in 20 countries, and overall survival favoured perioperative durvalumab with a hazard ratio of 0.78 (95% CI 0.63–0.96, p=0.021), clearing the prespecified threshold of p<0.0499. Treatment-related deaths were uncommon and near-balanced, 6 versus 2. The authors conclude this is a new standard treatment option, and on these numbers that is a fair reading. One structural point worth holding onto when you quote it: overall survival here was a key secondary endpoint, with event-free survival the previously reported primary — which does not weaken the result, but does mean the OS figure should be described as what it is.
· Overall survival HR 0.78 (95% CI 0.63–0.96, p=0.021), clearing the prespecified threshold of p<0.0499· 948 patients randomised, 147 centres across 20 countries· Treatment-related deaths 6 (1%) vs 2 (<1%)· Overall survival was a key secondary endpoint; event-free survival was the previously reported primary
Go deeper
MATTERHORN randomised 948 patients with resectable gastric or gastro-oesophageal junction adenocarcinoma to perioperative durvalumab plus FLOT or placebo plus FLOT across 147 centres in 20 countries. Overall survival favoured durvalumab with an HR of 0.78 (95% CI 0.63–0.96, p=0.021), clearing the prespecified significance threshold of p<0.0499. Treatment-related deaths were uncommon and near-balanced at 6 (1%) versus 2 (<1%). The authors conclude that durvalumab plus FLOT is a new standard treatment option. Worth reading the endpoint structure carefully: overall survival was a key secondary endpoint, with event-free survival the previously reported primary.Full daily entry →
The most instructive number in this issue is a comparison the trial itself makes possible. IMHOTEP's oesophagogastric cohort treated 41 patients with localised dMMR/MSI adenocarcinoma with perioperative pembrolizumab; all 27 who reached surgery had an R0 resection, and 4 of 27 (14.8%) achieved the primary endpoint of pathological complete response. Set that against the colorectal cohort of the very same umbrella trial, which this digest led with in June at 52.7%. Same drug, same design, same trial — and a fourfold difference in pathological response by organ of origin. The longer-term signal is reassuring in isolation: 93% were free of progression, recurrence or death at a median 25.8 months, with grade ≥3 treatment-related events in 22%. The authors' own conclusion is deliberately restrained — feasible and safe, with the duration and nature of neoadjuvant immunotherapy still to be optimised — and that restraint looks well judged. dMMR status is not a single phenotype across the GI tract.
· Pathological complete response 4/27 (14.8%), the primary endpoint· R0 resection 27/27 (100%)· Free of progression, recurrence or death at median 25.8 months: 25/27 (93%)· Grade ≥3 treatment-related adverse events 22%· 41 patients enrolled, 36 evaluable for efficacy· For contrast, the IMHOTEP colorectal cohort reported pCR 52.7% (covered in the June 5, 2026 edition)
Pathological complete response, by IMHOTEP cohort · 0–100% scale
52.7%Colorectal cohort (reported June 2026)
14.8%Oesophagogastric cohort (this report)
050100%
Go deeper
The oesophagogastric cohort of the IMHOTEP umbrella trial treated 41 patients with localised dMMR/MSI adenocarcinoma with perioperative pembrolizumab, 36 evaluable for efficacy. All 27 patients who went to surgery achieved an R0 resection, and 4 of 27 (14.8%) reached the primary endpoint of pathological complete response. At a median 25.8 months, 25 of 27 (93%) were free of progression, recurrence or death, and grade ≥3 treatment-related adverse events occurred in 22%. The authors' conclusion is deliberately restrained — feasible and safe, with the duration and nature of neoadjuvant immunotherapy still to be optimised. These are two cohorts of one umbrella trial, not randomised arms — the comparison is across organ of origin, not across treatments.Full daily entry →
Worth watchingAppendiceal · adjuvant therapy and relapse risk
Appendiceal adenocarcinoma: relapse and adjuvant therapy
Appendiceal adenocarcinoma is usually managed by analogy to colorectal cancer, and this cohort is a careful argument that the analogy fails where it matters. Across 439 patients at MD Anderson followed a median 62.6 months, with external validation in 128 stage II patients from MSKCC, relapse occurred in 9.4% — 6% of stage II and 19.5% of stage III — and carried a steep penalty, 5-year OS 95.7% without relapse versus 77.2% with. What predicted relapse was not the familiar list: histology and molecular features did the work, with mucinous HR 5.60 and enteric-type HR 6.60 against goblet-cell, TP53 in goblet-cell HR 6.93, and GNAS in non-goblet HR 17.0. Adjuvant chemotherapy was not associated with improved recurrence-free or overall survival on multivariable analysis (RFS HR 0.98, p=0.90). Read that null carefully: the univariate signal pointed the other way, which is confounding by indication. The honest conclusion is absence of demonstrated benefit, not evidence of harm.
· Relapse 9.4% overall — 6% of stage II, 19.5% of stage III· 5-year OS 95.7% without relapse vs 77.2% with relapse (HR 5.50, 95% CI 3.07–9.83, p<0.001)· Mucinous HR 5.60 and enteric-type HR 6.60 versus goblet-cell· TP53 in goblet-cell HR 6.93· GNAS in non-goblet HR 17.0· Adjuvant chemotherapy not associated with improved RFS on multivariable analysis (HR 0.98, p=0.90)· 439 patients, median follow-up 62.6 months, external validation in 128 stage II patients from MSKCC
Relapse rate by stage · 0–100% scale
19.5%Stage III
6%Stage II
050100%
Go deeper
This retrospective cohort followed 439 patients with stage I–III appendiceal adenocarcinoma at MD Anderson for a median 62.6 months, with external validation in 128 stage II patients from MSKCC. Relapse occurred in 9.4% of surgical patients — 6% of stage II and 19.5% of stage III — and carried a steep survival penalty (5-year OS 95.7% without relapse versus 77.2% with, HR 5.50, 95% CI 3.07–9.83, p<0.001). Histology and molecular features drove risk: mucinous HR 5.60 and enteric-type HR 6.60 versus goblet-cell, TP53 in goblet-cell HR 6.93, GNAS in non-goblet HR 17.0. Adjuvant chemotherapy was not associated with improved recurrence-free or overall survival on multivariable analysis (RFS HR 0.98, p=0.90). Note the univariate signal pointed the other way, which is confounding by indication — the multivariable null is the honest reading, not evidence of harm.Full daily entry →
On the radarPancreatic · neoadjuvant chemotherapy duration
PACT-21/CASSANDRA
A negative phase 3 that earns more attention than most positive ones, because of the specific way it is negative. The second randomisation of the Italian PACT-21/CASSANDRA trial gave 86 patients two extra months of preoperative chemotherapy and 79 the shorter schedule, in resectable and borderline resectable pancreatic ductal adenocarcinoma. Event-free survival was flat — unadjusted HR 0.98, 95% CI 0.68–1.41, p=0.90. Everything the surgeon and the pathologist look at improved: CA19-9 response 97% versus 84%, pathological complete response 5% versus 0%, N0 resection 45% versus 27%. None of it changed when events happened. This is about as clean a dissociation between pathological response and clinical outcome as the pancreatic literature has produced, and it is a caution for any trial design that treats pathological response as a surrogate primary endpoint. Overall survival data are not yet mature, so the final word is still pending — but the EFS result is unambiguous as it stands.
· Primary endpoint event-free survival NOT met — unadjusted HR 0.98 (95% CI 0.68–1.41), p=0.90· CA19-9 response 58/60 (97%) vs 41/49 (84%), p=0.02· Pathological complete response 4/86 (5%) vs 0/79, p=0.05· N0 resection 39/86 (45%) vs 21/79 (27%), p<0.01· 86 assigned to long vs 79 to short preoperative chemotherapy (PAXG or mFOLFIRINOX), 17 Italian academic hospitals· Overall survival data not yet mature
Go deeper
The second randomisation of the Italian PACT-21/CASSANDRA phase 3 trial asked whether two additional months of preoperative chemotherapy improve outcomes in resectable and borderline resectable pancreatic ductal adenocarcinoma. Between February 2021 and September 2024, 86 patients were assigned to long and 79 to short preoperative chemotherapy with either PAXG or mFOLFIRINOX. The primary endpoint was not met — event-free survival was effectively identical between arms (unadjusted HR 0.98, 95% CI 0.68–1.41, p=0.90). Every pathological and biochemical secondary endpoint nevertheless favoured the longer schedule: CA19-9 response 58/60 (97%) versus 41/49 (84%), p=0.02; pathological complete response 4/86 (5%) versus 0/79, p=0.05; and N0 resection 39/86 (45%) versus 21/79 (27%), p<0.01. Overall survival data are not yet mature.Full daily entry →
Until now only bi-allelic MSH3 and MLH3 carriers were clearly linked to colorectal cancer. Studying roughly 12,000 CRC and multiple-polyp cases against 460,000 controls, with cancer genome sequencing in 2,023 patients, this work shows simple heterozygotes are also at risk — 2.2-fold for MSH3 and 1.6-fold for MLH3. The mechanism is the satisfying part, because it is classical two-hit tumour suppression rather than anything exotic: somatic second hits inactivate the wild-type allele, and a single second hit sometimes knocks out both MSH3 and its neighbour. The resulting tumours are microsatellite-stable yet hypermutant, with ≥2 bp deletions enriched roughly 12-fold and mutational signature ID4 usually present — and tumours from heterozygotes without a second hit showed no hypermutation at all, which is the internal control that makes the causal claim stick. This is the first causal link between specific gene defects and ID4 in human tumours. Practically, it gives these carriers a phenotype resembling MSH6-mutant Lynch syndrome with incomplete penetrance, and it is a reminder that an MSS result does not rule out a mismatch-repair story.
· Colorectal cancer risk 2.2-fold for MSH3 heterozygotes and 1.6-fold for MLH3 (p=0.028)· Roughly 12,000 CRC and multiple-polyp cases against 460,000 controls· Cancer genome sequencing in 2,023 patients· Tumours microsatellite-stable yet hypermutant: deletions ≥2 bp enriched roughly 12-fold, signature ID4 usually present (p<0.0001)· Tumours from heterozygotes without a somatic second hit showed no hypermutation at all
Colorectal cancer risk in heterozygous carriers · 0–6 mo scale
2.2-foldMSH3 heterozygotes
1.6-foldMLH3 heterozygotes
06 mo
Δ +0.6 mo median
Go deeper
MSH3 and MLH3 are non-canonical mismatch repair genes handling insertion-deletion errors, and until now only bi-allelic carriers were clearly linked to colorectal cancer. Studying roughly 12,000 CRC and multiple-polyp cases against 460,000 controls, with cancer genome sequencing in 2,023 patients, the authors show that simple heterozygotes are also at risk: 2.2-fold for MSH3 and 1.6-fold for MLH3 (p=0.028). The mechanism is classical two-hit tumour suppression — somatic second hits inactivate the wild-type allele, and a single second hit sometimes knocks out both MSH3 and the neighbouring gene. The resulting cancers are microsatellite-stable yet hypermutant, with deletions of at least 2 bp enriched roughly 12-fold and mutational signature ID4 usually present (p<0.0001); tumours from heterozygotes without a second hit showed no hypermutation at all. This is the first causal link between specific gene defects and ID4 in human tumours, and it gives these carriers a phenotype that resembles MSH6-mutant Lynch syndrome with incomplete penetrance. Full daily entry →
Three perioperative trials landed in the same week and they do not agree with one another — which is the most useful thing about them. MATTERHORN is the definitive one: 948 patients, overall survival HR 0.78, and a new standard treatment option in resectable gastric and gastro-oesophageal junction adenocarcinoma. The other two are about the limits of the same idea.
Read together, a boundary appears:
MATTERHORN — adding perioperative immunotherapy to FLOT in unselected resectable disease improved survival.
IMHOTEP — the same strategy in dMMR/MSI oesophagogastric disease produced pCR in 14.8%, against 52.7% in the colorectal cohort of the very same umbrella trial. Same drug, same design, different organ, fourfold difference.
PACT-21/CASSANDRA — simply giving more of the same preoperative chemotherapy in pancreatic cancer improved every pathological endpoint and changed event-free survival not at all.
That last one deserves more attention than a negative trial usually gets. CA19-9 response, pathological complete response and N0 resection all moved in the right direction, and the events did not follow. It is the cleanest recent demonstration that downstaging the specimen and changing the disease are separate achievements — and a standing caution for any trial that leans on pathological response as a surrogate primary endpoint.
Away from the operating room, two molecular papers quietly widened who we should be testing and how we should read the answer. Heterozygous MSH3 and MLH3 carriers turn out to carry real colorectal risk through ordinary two-hit biology, producing tumours that are microsatellite-stable yet hypermutant — a reminder that an MSS result does not close the mismatch-repair question. And in appendiceal adenocarcinoma, where the colorectal playbook is usually borrowed wholesale, histology and GNAS/TP53 predicted relapse while the familiar risk factors predicted nothing, with no demonstrated benefit from adjuvant chemotherapy.
A thin Tuesday, sitting at the end of a week that was anything but.
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.