GI Upper
2026-06-10
RTOG 0848 NCT01013649
ForResected pancreatic head adenocarcinoma, post 6 cycles adjuvant gemcitabine-based chemo
HR 0.96
90% CI 0.79-1.18, 1-sided P=.38; did not meet OS
TL;DRAdjuvant CXRT after 6 cycles gemcitabine-based chemo missed OS (HR 0.96), but node-negative pts gained: 5yr OS 48.1% vs 28.6%.
The whole RT read sits in 91 node-negative pts: 5yr OS 48.1% vs 28.6%, median OS 3.9 vs 3.0 yr, with interaction P=.022. Delivery was standard and QA'd (50.4 Gy/28 fx, tumor bed plus pancreatojejunostomy plus regional nodes, 81% IMRT), so the technique transfers directly. G4-5 toxicity was unchanged (6% v 5%).
In a resected pancreatic head adenocarcinoma pt who is node-negative and progression-free after adjuvant chemotherapy, this supports discussing CXRT; it does not extend to node-positive disease, where no treatment effect was seen, or to margin-positive pts (only 10 node-negative/margin-positive across both arms).
The RT signal is confined to 91 node-negative pts: 5yr OS 48.1% vs 28.6%, median OS 3.9 vs 3.0 yr, interaction P=.022. Technique transfers unchanged (50.4 Gy/28 fx, tumor bed plus pancreatojejunostomy plus regional nodes, 81% IMRT, real-time central review), and G4-5 toxicity was flat at 6% v 5%.
The systemic backbone dates the result: 67% got single-agent gemcitabine, 28% gemcitabine plus erlotinib, and no pt received FOLFIRINOX. Unselected CXRT after six cycles adds nothing (OS HR 0.96), so referral for adjuvant RT stays off the default path, though the grade 3 excess (38% v 19%) is GI and lymphopenic, not prohibitive.
10 details 2 trials watching
Prospective randomized phase IIR/III, two-step, multicenter (RTOG then NRG). Step 2 opened November 2009, closed October 2018, analyzed December 2023. Randomization 1:1, unmasked, stratified by nodal status, CA19-9, margins and adjuvant systemic treatment.
Curative-intent gross total resection of pancreatic head adenocarcinoma with documented microscopic margin status; body/tail excluded. Only pts without recurrence after five cycles of chemotherapy entered step 2: 546 entered step 1, 354 randomized (174 chemo, 180 chemo+CXRT). 74% node-positive, 17% margin-positive, 81% T3.
Step 1 was gemcitabine (67%), gemcitabine plus erlotinib (28%) or non-oxaliplatin gemcitabine combinations (5%); no pt received FOLFIRINOX. Both arms then took a sixth cycle; the experimental arm followed with CXRT within 21 days, sensitized by capecitabine (83%) or infusional 5-FU (15%).
50.4 Gy in 28 fractions to the postoperative tumor bed, pancreatojejunostomy and regional nodes; median delivered dose 50.4 Gy over 28 fractions and 38 days. 81% IMRT, 19% 3D conformal, 79% with no interruptions, 88% received ≥50 Gy. Centers were IROC-credentialed and every plan underwent real-time central review before start.
Primary: overall survival from second-step randomization. Secondary: DFS and adverse events (CTCAE v4.0). Final analysis triggered at the earlier of 316 OS events or 5 years of potential follow-up for all pts; it fired on the latter with 270 events, cutting power to 72%.
Grade 4 or 5 AEs were not increased (6% v 5%, P=.68), with one grade 5 in each arm (hepatic failure, sepsis). Grade 3 rose to 38% v 19% (P<.001), driven by GI events and decreased lymphocyte count; grade 3 nonhematologic 22% v 11% (P=.004).
ESPAC-1 reported adjuvant CXRT as detrimental, but accrued 1994-2000 with split-course lower-dose 2D planning, no postoperative imaging to exclude incomplete resection or early metastases, and no QA. RTOG 9704's post hoc analysis linked RT protocol deviations to worse survival, the specific failure mode 0848 designed its real-time review to prevent.
Accrual took almost 9 years and the OS event rate ran below projection, so the trial read out on a follow-up trigger at 270 of 316 planned events (power 72% for the hypothesized HR 0.76). The node-negative arms hold 42 and 49 pts, and the subgroup rests on an interaction P=.022 across nine tested subgroups.
The negative primary settles that unselected adjuvant CXRT adds nothing after gemcitabine-based chemotherapy, while the safety profile removes the ESPAC-1-era objection that RT here is harmful. What stays open is whether the node-negative signal reflects a real locoregional benefit in pts with lower competing distant risk, and whether it survives a modern FOLFIRINOX backbone.
| Endpoint | Chemo (n=42) | Chemo+CXRT (n=49) |
|---|---|---|
| 5yr OS (95% CI) | 28.6 (14.9-42.2) | 48.1 (33.3-62.9) |
| Median OS, yr (95% CI) | 3.0 (2.2-4.0) | 3.9 (2.5-NR) |
| Median DFS, yr (95% CI) | 1.5 (0.8-2.7) | 2.3 (1.4-NR) |
CONSORT flow
Prespecified stratified phase III, primary OS endpoint negative overall. Node-negative benefit rests on a subgroup interaction (P=.022) in 91 pts, hypothesis-generating not definitive.
- Does the node-negative CXRT benefit hold on a FOLFIRINOX backbone?
- Can SBRT substitute for conventional CXRT in this setting? active mFOLFIRINOX With or Without Stereotactic Body Radiotherapy in Locally Advanced Pancreatic Adenocarcinoma Phase 2n=92 · primary completion 2025-01 · candidate match
- How to select for RT benefit when nodal status follows neoadjuvant therapy? active Carboplatin, Nab-Paclitaxel, Durvalumab Before Surgery and Adjuvant Therapy in Head and Neck Squamous Cell Carcinoma Phase 2n=39 · primary completion 2022-02 · candidate match
📚 Sources · 📄 1 paper
2026-05-31
RASolute 302
ForPreviously treated metastatic pancreatic cancer, RAS G12-mutant
TL;DRMedian OS 13.2 vs 6.6 mo, HR 0.40 (95% CI 0.30-0.54), P<0.001, daraxonrasib over chemo in previously treated RAS G12 metastatic pancreatic cancer.
In previously treated metastatic pancreatic cancer with a RAS G12 mutation, this supports daraxonrasib over investigator's choice chemotherapy on OS; it does not extend to treatment-naive or non-metastatic disease.
| Population | Arm | N | Events, n (%) | mOS, mo (95% CI) | HR for death (95% CI) |
|---|---|---|---|---|---|
| RAS G12 | Daraxonrasib | 228 | 72 (32) | 13.2 (10.0-NR) | 0.40 (0.30-0.54), P<0.001 |
| RAS G12 | Chemotherapy | 231 | 127 (55) | 6.6 (5.4-8.2) | ref |
| Overall | Daraxonrasib | 248 | 79 (32) | 13.2 (10.0-NR) | 0.40 (0.30-0.53), P<0.001 |
| Overall | Chemotherapy | 252 | 141 (56) | 6.7 (5.8-8.0) | ref |
8 details 2 trials watching
Phase, randomization and primary endpoint not stated in source. HRs came from a stratified Cox model and P values from a stratified log-rank test. Presented at ASCO as abstract LBA5.
Previously treated metastatic pancreatic cancer. OS reported in a RAS G12 population (228 vs 231) and an overall population that also includes G13 or Q61 mutations or no RAS mutation identified (248 vs 252).
Daraxonrasib vs investigator's choice chemotherapy. Dose, schedule and the chemo options allowed are not stated in source.
OS favored daraxonrasib in both populations with HR 0.40, P<0.001; per-arm values are in the figure table.
Second-line metastatic pancreatic cancer has been a chemotherapy space, with NAPOLI-1 establishing a nal-IRI based option. An OS benefit for a RAS-directed agent over chemo is a different class of result; no cross-trial numbers are reported in source.
Safety and follow-up duration not reported in source; the daraxonrasib median OS upper CI is not reached. Composition of the investigator's-choice control is unknown, which bears on comparator strength.
The identical HR 0.40 in both populations reflects that the overall population is almost entirely RAS G12, not a demonstrated benefit beyond G12. Events of 32% vs 55% and 56% mean the daraxonrasib median of 13.2 mo rests on fewer events and may move with maturity.
OS HR 0.40 vs investigator's-choice chemo, P<0.001, consistent in RAS G12 and overall populations; source frames it as historic. Tweet omits phase and design.
- Benefit in RAS G13, Q61, or no-RAS-mutation tumors?
- Durability of OS benefit with longer follow-up
- Role in first-line metastatic pancreatic cancer recruiting Study of Daraxonrasib and Daraxonrasib + GnP as First-line Treatment in Patients With Metastatic Pancreatic Adenocarcinoma Phase 3n=900 · primary completion 2028-06 · phase 3 daraxonrasib ± GnP vs chemo, 1L metastatic PDACn=400 · primary completion 2029-03 · phase 3 zoldonrasib + daraxonrasib vs GnP, 1L KRAS G12D
📚 Sources · 🐦 1 tweet
#ASCO26
— Nicholas Hornstein (@GIMedOnc) May 31, 2026
This one is special.
This is the hottest paper of 2026 and potentially in the history of pancreatic cancer.
Let’s dive in.
RASolute 302: Daraxonrasib vs investigator’s choice chemotherapy in previously treated metastatic pancreatic cancer
Abstract LBA5 (soon!)… pic.twitter.com/Lq7PEjOWAo
2026-05-30 ASCO Annual Meeting 2026
Neo-CRAG
ForcT3N2-3M0 / cT4 gastric or Siewert II-III EGJ adenocarcinoma, D2-resectable
HR 0.750
95% CI 0.607-0.928, P=0.008; 3yr DFS 55.6% vs 42.4%
TL;DRAdding preop 45Gy/25fx to periop XELOX raised 3yr DFS 55.6% vs 42.4%, HR 0.750; mOS 67.5 vs 37.6mo.
The RT case here rests on locoregional control: LRR after R0 fell to 9.4% from 18.3%, tracking the ypN0 (56.1% vs 36.4%) and pCR gains, which is the mechanistic chain CRITICS and TOPGEAR never produced. Dose was conventional 45Gy/25fx preoperatively, and G3+ postop complications stayed flat (9.0% vs 7.6%), so the deliverability objection to preop RT before D2 loses force.
In node-heavy cT3N2+ or cT4 gastric/Siewert II-III disease heading to D2 resection, this supports revisiting preoperative chemoRT rather than chemo alone; it does not speak to cT2N0-N1 disease or to pts already committed to a FLOT backbone.
Conventional 45Gy/25fx delivered preoperatively, after cycle 1 of chemo, halved locoregional recurrence after R0 resection (9.4% vs 18.3%) with no excess G3+ postoperative complications (9.0% vs 7.6%). That combination, a local-control gain without a surgical-morbidity cost, is what the omission decision in node-heavy cT3N2+ disease has been waiting for.
The systemic backbone was XELOX in both arms, so the DFS gain is attributable to radiotherapy rather than to the regimen, but the control arm's 37.6 mo median OS is the caveat: this does not tell you whether radiotherapy still adds on top of FLOT, whose own advantage is partly locoregional. Sequencing of RT after cycle 1 with attenuated dosing was deliverable.
Every pt was intended for standardized D2 resection, and preoperative chemoRT did not degrade it: G3+ postoperative complications were 9.0% vs 7.6%, and 448 of 620 randomised pts reached gastrectomy. ypN0 rose to 56.1% from 36.4%, which changes what the surgeon can expect to find in the nodal basin rather than the extent of dissection required.
| Endpoint | CRT | CT | Effect size |
|---|---|---|---|
| 3yr DFS | 55.6% (50.1-61.1) | 42.4% (36.9-47.9) | HR 0.750 (0.607-0.928), P=0.008 |
| Median DFS | 52.7 mo | 24.4 mo | n/a |
| 5yr OS | 50.1% | 44.2% | HR 0.781 (0.628-0.970), P=0.025 |
| Median OS | 67.5 mo | 37.6 mo | n/a |
9 details
Randomised, multicenter, phase 3 trial, N=620 (310 per group), 13 referral hospitals in China, accrual 2013-2022. Follow-up on the DFS curve extends past 120 months.
High-risk locally advanced gastric or EGJ adenocarcinoma: cT3N2-3M0, cT4aN+M0, or cT4bNanyM0, Siewert II/III permitted. 225 (36.3%) had an EGJ primary, and every pt was intended for standardized D2 resection.
Both arms: 3 cycles preoperative XELOX (oxaliplatin 130 mg/m² D1, capecitabine 1000 mg/m² BID D1-14, Q3W) then D2 gastrectomy then 3 adjuvant XELOX cycles.
CRT arm added concurrent 45 Gy / 25 fractions after cycle 1 of preoperative chemo, with attenuated concurrent dosing (oxaliplatin 100 mg/m², capecitabine 825 mg/m²). Target volume not specified in source.
Primary: disease-free survival, from randomization to progression, relapse, or death. Secondary: overall survival, pCR, R0 resection, safety.
Primary endpoint met. The locoregional read is the one an RT reader needs: LRR 9.4% vs 18.3% among R0-resected pts.
G3+ haematologic toxicity 14.6% vs 10.3%, the expected cost of concurrent chemoRT. G3+ postoperative complications were comparable, 9.0% vs 7.6%, so preoperative RT did not measurably worsen D2 surgical morbidity.
CRITICS (postoperative chemoRT after D1+ surgery) and TOPGEAR (preoperative chemoRT with ECF/FLOT) both failed to show a survival gain for radiotherapy in this disease. Neo-CRAG differs on three axes at once: RT given preoperatively, D2 resection mandated, and enrolment restricted to node-heavy cT3N2+ or cT4 disease.
The 2013-2022 accrual window means the control arm reflects a doublet era; a 37.6 mo median OS in the control group is short against contemporary FLOT-treated series. Single-country enrolment with uniformly high-quality D2 surgery also caps generalisability to centres with variable nodal dissection.
The pCR, downstaging, ypN0 and LRR gradient forms a coherent mechanistic chain from RT to the DFS separation, which is what earlier negative trials lacked. What it does not settle is whether that chain survives a stronger systemic backbone, since much of FLOT's advantage over a doublet is itself locoregional.
CONSORT flow
Randomised ph3, prespecified DFS met with OS support, but the chemo backbone is XELOX not FLOT, so it contests RT omission without settling it.
- Does preoperative chemoRT still add benefit on a FLOT backbone?
- Generalizability outside high-volume D2 centers
- Optimal target volume and elective nodal coverage not reported in source
📚 Sources · 🐦 1 tweet
Neo-CRAG: Ph3 RCT (n=620, gastric/GEJ) - adding neoadj chemoRT to peri-op XELOX improved OS (68 v 38 mo).
— Dr. Nina Niu Sanford (@NiuSanford) May 30, 2026
Limitation=non-FLOT, BUT still relevant IMO b/c:
1) DFS/OS benefit substantial.
2) Improvements in pCR, downstg, LRR supports plausible RT effect on OS. #ASCO26 @OncoAlert pic.twitter.com/eeM0Z8p20M