AREST
ForResected pT1-2N0 OSCC, margins ≥5mm, ≥1 intermediate-risk feature
HR 0.52
95% CI 0.30-0.91, p=0.02 · 3-yr 89.2% vs 80.9% (ITT)
TL;DR3-yr LRFS 89.2% vs 80.9% (HR 0.52, p=0.02) with adjuvant RT vs observation in resected intermediate-risk pT1-2N0 OSCC; no significant DFS/OS difference.
Competing-risk LRF incidence was 10.6% vs 18.9% (HR 0.52) with 60Gy/30fx to tumor bed and at-risk neck. Controls had ≥5mm margins and a level I-III dissection with ≥16 nodes, so this is RT's increment over high-quality surgery alone. It moves adjuvant RT vs observation most for oral tongue, an exploratory subgroup.
In pT1-2N0 OSCC resected with ≥5mm margins and an adequate level I-III neck dissection but carrying PNI, LVE, poor differentiation or DOI ≥5 to ≤10mm, this supports discussing adjuvant RT for loco-regional control; it does not extend to close-margin, limited-neck or node-positive disease.
Competing-risk LRF incidence 10.6% vs 18.9% (HR 0.52) with 60Gy/30fx to tumor bed and at-risk neck, after surgery that already met ≥5mm margins and ≥16 nodes. It moves adjuvant RT vs observation in pT1-2N0 OSCC, strongest for oral tongue in an exploratory subgroup; smaller target volumes untested.
Benefit was measured against surgery with clear margins ≥5mm and an ipsilateral level I-III dissection yielding ≥16 nodes, so RT added loco-regional control even after high-quality resection. Pathology reporting of PNI, LVE, differentiation and DOI now gates the adjuvant referral; the data do not cover close margins or a limited neck.
9 details
Multicenter, open-label, phase III RCT from India, 1:1, stratified by oral cavity subsite, PNI/LVE and differentiation. N=392 (191 RT, 201 observation); median follow-up 47.2 mo (IQR 30-59.4). Powered at 80% to detect HR 0.6256, assuming 3-yr LRFS 70% with observation.
pT1-T2, pN0 OSCC after adequate surgery: clear margins ≥5mm and at least ipsilateral level I-III neck dissection with ≥16 nodes. Required ≥1 intermediate risk factor: DOI ≥5 to ≤10mm, PNI, LVE, or poor differentiation.
60Gy in 30 fractions over 6 weeks to the resected tumor bed and at-risk neck nodal region. Technique and nodal volume definitions not stated in source; no concurrent systemic therapy described.
Primary: loco-regional recurrence-free survival, from randomization to first local and/or regional recurrence. Also reported: cumulative incidence of loco-regional failure with death as competing event, DFS, OS.
LRFS benefit held on PP and competing-risk analyses; DFS and OS not significantly different, with no survival HRs in source.
| Endpoint | Adjuvant RT | Observation | HR (95% CI), p |
|---|---|---|---|
| 3-yr LRFS, ITT | 89.2% (84.3-93.3) | 80.9% (74.6-86.1) | 0.52 (0.30-0.91), p=0.02 |
| 3-yr LRFS, PP | 91.1% | 80.9% | 0.43 (0.23-0.80), p=0.01 |
| LRF cumulative incidence (death competing), ITT | 10.6% (6.1-15.1) | 18.9% (13.3-24.6) | 0.52 (0.30-0.91), p=0.021 |
| LRF cumulative incidence (death competing), PP | 8.7% (4.3-13.1) | 18.9% (13.3-24.6) | 0.43 (0.23-0.79), p=0.007 |
The abstract describes the evidence for adjuvant RT in this setting as largely retrospective; AREST is the randomised test of that practice. The RTOG and EORTC postoperative trials tested adding concurrent chemo to RT in high-risk resected disease and did not randomise RT itself against observation.
Control arm beat its design assumption (3-yr LRFS 80.9% vs 70% assumed), so the absolute gain sits against a lower-risk baseline than planned. Oral tongue vs buccal mucosa is an exploratory subgroup with no HR in source, and no result is given per risk factor. Salvage treatment at recurrence is not described, which bears on the OS read.
Loco-regional control improved without DFS or OS separation; one explanation is effective salvage of isolated recurrences in observed pts, which the abstract cannot confirm. The decision it moves is adjuvant RT vs observation after adequate resection, weighed against RT morbidity the abstract does not report.
CONSORT flow
Randomised phase III hit LRFS primary at 47.2 mo median f/u; DFS/OS not significantly different and no toxicity data in abstract, so supports rather than mandates adjuvant RT.
- Which intermediate-risk feature (PNI, LVE, DOI, grade) drives the benefit?
- Does the oral tongue benefit hold as a prespecified subgroup?
- Late RT toxicity and salvage outcomes in the observation arm
📚 Sources · 🐦 1 tweet · 📄 1 paper
#ASCO26
— Dr Rishabh Jain (@DrRishabhOnco) May 27, 2026
🗣️ The AREST trial tackles one of the biggest gray zones in oral cavity cancer.
After adequate surgery in pT1-2N0 OSCC with intermediate-risk features:
✅ Adjuvant RT improved loco-regional control
❌ No OS benefit observed
3-year LRFS:
🔹 89.2% vs 80.9%
🔹 HR 0.52… https://t.co/qsALPFX032 pic.twitter.com/F4XzTMETih
Abstract
The longer read
AREST gives a randomised answer to a question head and neck teams have mostly settled on retrospective series: whether a cleanly resected, pathologically node-negative early oral cavity primary earns adjuvant RT because it carries perineural invasion, lymphovascular emboli, poor differentiation, or an intermediate depth of invasion. On the prespecified endpoint the answer is yes. The ITT hazard ratio of 0.52 is echoed by the per-protocol estimate of 0.43 and by the competing-risk analysis, which answers the usual objection that a Kaplan-Meier recurrence endpoint can flatter an arm when deaths without recurrence are censored. The observed effect was also larger than the hazard ratio of 0.6256 the trial was powered to detect, although an upper confidence bound of 0.91 still admits a modest benefit.
The design feature that matters most for transfer is the surgical floor. Every pt had margins of at least 5mm and at least an ipsilateral level I-III neck dissection with 16 or more nodes, so the pN0 label is reliable and the observation arm was not a proxy for undertreatment. What RT added is an increment over good surgery, not a repair of inadequate surgery. The same feature limits extrapolation: a pt with a close margin, a limited neck dissection, or a low nodal yield sits outside the trial, and the tempting inference that such pts would benefit more is plausible but untested here. The observation arm also did better than planned, with 3-yr LRFS of 80.9% against an assumed 70%, so the absolute gain a reader weighs against RT morbidity is set against a lower-risk baseline than the protocol anticipated.
DFS and OS were not significantly different, and two readings compete. The trial was powered for loco-regional control, not survival, so a smaller survival effect could be missed. Alternatively, isolated loco-regional recurrence in the observed arm may often be salvaged, which turns the choice into a trade between up-front RT for everyone and salvage treatment for the minority who recur. The abstract reports neither salvage rates nor salvage morbidity, so these cannot be separated, and that gap is where the decision actually lives. Six weeks of RT to the oral cavity and neck carries late effects this abstract does not quantify, so the balance of benefit and harm cannot yet be read from the report.
Three further points temper application. The oral tongue versus buccal mucosa difference is labelled exploratory, and no subgroup hazard ratio or interaction test appears in the source; stratification by subsite makes the comparison fairer but it remains hypothesis-generating. No result is given by individual risk factor, so whether intermediate depth of invasion alone justifies RT as much as perineural invasion does is unanswered. The target was the tumor bed plus the at-risk neck, so the trial cannot say whether a primary-bed-only volume would capture most of the gain.
Taken together this is a confirmatory randomised read. It moves adjuvant RT for intermediate-risk pT1-2N0 OSCC from retrospective justification to randomised support for loco-regional control, without showing that pts live longer. The full publication needs to supply toxicity, salvage outcomes and per-factor results before the benefit-harm balance can be stated for an individual pt.