Why This Study Matters
Pancreatic adenocarcinoma is projected to be the third and fourth leading cause of cancer death among U.S. women and men, respectively, despite ranking far lower in incidence — a stark reflection of the disease’s poor prognosis. Curative-intent management depends on complete surgical resection, but surgery alone is insufficient, and effective adjuvant therapy remains essential.
Radiotherapy was historically part of early adjuvant regimens, but its use diminished in the U.S. after the ESPAC-1 trial suggested adjuvant radiotherapy might be associated with inferior outcomes when added to 5-fluorouracil-based chemotherapy. Even so, concerns about ESPAC-1’s radiotherapy technique, combined with local-regional recurrence rates of 30–40% after chemotherapy alone, kept adjuvant chemoradiotherapy (CXRT) in use at many U.S. centers. RTOG 0848 was designed to test, in a modern randomized framework, whether adding CXRT to adjuvant chemotherapy improves survival after resection of pancreatic head adenocarcinoma.
Study Design
RTOG 0848 (NRG Oncology) was a multicenter, randomized, phase III, two-step trial. Step 1, reported previously, randomized patients to adjuvant gemcitabine versus gemcitabine plus erlotinib. Step 2, the subject of this analysis, randomized patients who completed five cycles of Step 1 chemotherapy without disease progression to a sixth chemotherapy cycle with or without chemoradiotherapy.
The trial was powered assuming a 17-month median overall survival with chemotherapy alone, targeting 354 patients and 316 observed OS events (HR 0.76, 80% power, one-sided α=0.05). Step 2 accrued from 2009 to 2018. At the December 10, 2023 data cutoff, the trial had accumulated 270 OS events — short of the 316-event target — with a median follow-up of 2.2 years across all patients and 7.4 years among survivors.
Primary Endpoint: Overall Survival
The primary endpoint, overall survival, was not significantly improved by adding CXRT. Median OS was 31.1 months (90% CI: 25.3–37.6) with chemotherapy alone versus 27.3 months (90% CI: 24.5–31.4) with chemotherapy plus CXRT (HR 0.96, 90% CI: 0.79–1.18; two-sided p=0.77). Five-year OS rates were 23.1% with chemotherapy alone and 27.9% with chemotherapy plus CXRT.
Because 14% of patients randomized to CXRT ultimately refused it, the investigators also ran an as-treated sensitivity analysis, which produced results consistent with the intention-to-treat analysis (HR 0.98, 90% CI: 0.80–1.20, two-sided p=0.85). Disease-free survival showed a nonsignificant trend favoring CXRT: HR 0.82 (95% CI: 0.65–1.03, two-sided p=0.089), with median DFS of 12.4 months versus 15.6 months.
Subgroup Analyses: The Nodal Status Signal
On multivariable analysis — adjusting for race, for which non-white/unknown patients had significantly worse OS and DFS than white patients — the treatment-by-nodal-status interaction was the only prospectively designated variable showing a significant interaction with treatment, for both OS (p=0.0063) and DFS (p=0.014). Patients with negative nodes (n=91, 26% of the cohort) had better OS and DFS when treated with chemotherapy plus CXRT, while patients with positive nodes (n=263, 74%) did not derive the same benefit.
The published article reports the significance of this treatment-by-nodal-status interaction and describes the direction of effect qualitatively, but it does not provide a standalone hazard ratio or confidence interval for the node-negative subgroup in the article text — the underlying multivariable Cox regression results are referenced to a data supplement not reviewed for this article. That number is reported here as absent from the source rather than estimated.
Safety Profile
Adding CXRT did not significantly increase the most severe adverse events: grade 4 or 5 events occurred at similar rates between arms (6% vs. 5%, p=0.68). Grade 3 events, however, were significantly more common with chemotherapy plus CXRT (38% vs. 19%, p<0.001), driven mainly by gastrointestinal toxicity — particularly diarrhea — and decreased lymphocyte counts.
| System Organ Class | Chemo Alone (n=174) | Chemo+CXRT (n=180) |
|---|---|---|
| Gastrointestinal disorders | 4 (2%) | 19 (11%) |
| – Diarrhea | 1 (1%) | 10 (6%) |
| Investigations (incl. lymphocyte count decreased) | 25 (14%) | 46 (26%) |
| General disorders/administration site conditions | 5 (3%) | 11 (6%) |
| Overall Highest Grade 3 | 33 (19%) | 68 (38%) |
| Overall Highest Grade 4 | 7 (4%) | 9 (5%) |
| Overall Highest Grade 5 | 1 (1%) | 1 (1%) |
| Nodal Status | Chemo Alone (n=174) | Chemo+CXRT (n=180) | Total (n=354) |
|---|---|---|---|
| N0 (node-negative) | 42 (24%) | 49 (27%) | 91 (26%) |
| N1 (node-positive) | 132 (76%) | 131 (73%) | 263 (74%) |
The addition of adjuvant CXRT to adjuvant gemcitabine chemotherapy did not statistically significantly improve OS or DFS or increase grade 4 or 5 toxicity. CXRT improved DFS and OS among node negative patients.
— Study Conclusions, Abrams et al., Journal of Clinical Oncology, 2026
Interpretation and Broader Context
Taken as a whole, RTOG 0848’s Step 2 randomization is a negative trial for its primary endpoint: adding CXRT to adjuvant chemotherapy did not significantly extend overall or disease-free survival across the full resected pancreatic head adenocarcinoma population, and it came with a meaningfully higher rate of grade 3 toxicity. That result is broadly consistent with the cautious trajectory of adjuvant radiotherapy use in pancreatic cancer since ESPAC-1.
The treatment-by-nodal-status interaction is the study’s most clinically provocative finding: node-negative patients — about one in four in this cohort — appeared to benefit from CXRT on both OS and DFS, while node-positive patients did not. The authors are explicit that this observation, while statistically significant on multivariable analysis, will need confirmation in future studies, ideally incorporating current or more effective systemic therapy regimens, before it can change practice broadly.
Limitations
The node-negative benefit is a subgroup finding from a treatment-by-covariate interaction test on multivariable analysis, not a prespecified, independently powered comparison — it is hypothesis-generating rather than confirmatory, and the published article does not report a standalone hazard ratio or confidence interval for that subgroup. The trial also closed with 270 of its targeted 316 overall-survival events, an interim shortfall relative to its planned power, which widens the uncertainty around the primary null result. Median follow-up was only 2.2 years across all randomized patients (versus 7.4 years among survivors), limiting how confidently the study can speak to long-term outcomes. Finally, 14% of patients assigned to CXRT did not receive it, and although the as-treated sensitivity analysis was consistent with the intention-to-treat result, nonadherence of this size can still dilute a true treatment effect in either direction.
Funding & Trial Registration
Trial registration: ClinicalTrials.gov NCT01013649. Funding: National Cancer Institute/NIH award numbers U10CA180868 (NRG Oncology Operations), U10CA180822 (NRG Oncology SDMC), UG1CA189867 (NCORP), and U24CA180803 (IROC).
In this randomized phase III trial of 354 patients with resected pancreatic head adenocarcinoma, adding chemoradiotherapy to adjuvant chemotherapy did not significantly improve overall or disease-free survival overall, and came with more grade 3 toxicity. A statistically significant interaction suggested a survival benefit from chemoradiotherapy in the roughly one-quarter of patients who were node-negative, but that signal comes from a subgroup interaction test without its own reported hazard ratio, reached only 270 of a targeted 316 survival events, and the authors say it needs confirmation before it can guide practice.