ROAM/EORTC-1308
ForNewly diagnosed WHO grade 2 atypical meningioma after Simpson I–III resection
HR 0.51
95% CI 0.27–0.97, p=0.0396
TL;DRAdjuvant IMRT 60 Gy/30 fx vs observation after complete resection of WHO grade 2 atypical meningioma: DFS HR 0.51 (0.27–0.97); 5-yr DFS 79.9% vs 64.3%.
The RT effect is probably understated: 12 of 78 assigned pts never started RT, and DFS counts out-of-field and new meningiomas that a 5–10 mm CTV was never aimed at. The planned patterns-of-failure analysis will decide whether dural margins can tighten under modern IGRT, which is the next dosimetric question once adjuvant RT is adopted.
For a newly diagnosed WHO grade 2 atypical meningioma after surgeon-assessed Simpson I–III resection, this supports adjuvant fractionated RT over observation when recurrence prevention is the priority; it does not address subtotal resection, NF2, multiple or radiation-induced meningiomas.
IMRT 60 Gy/30 fx to a 5–10 mm CTV halved the DFS hazard (HR 0.51) with no G3+ AEs among RT-related events and one G2 necrosis. The pending patterns-of-failure analysis decides whether dural margins can tighten.
Surgeon-assessed Simpson I–III resection without early postop MRI still left 30% recurrence under observation at 64 mo. GTR alone is not definitive for atypical meningioma; authors suggest observation mainly where re-resection is easy, such as convexity tumours.
10 details 1 trial watching
International open-label phase 3 RCT, 1:1, 58 hospitals in 11 countries, stratified UK vs non-UK. N=157 randomised from 990 screened; median follow-up 64 mo.
Age ≥16, newly diagnosed atypical meningioma after surgeon-assessed GTR (Simpson I–III), WHO PS ≤2, able to start RT within 12 weeks. Excluded NF2, multiple or radiation-induced meningiomas, optic nerve sheath tumours.
IMRT 60 Gy in 30 fx over 6 weeks, median 2.50 mo from surgery. CTV 5–10 mm isotropic from GTV plus 3–5 mm PTV, with prospective central plan review before treatment.
Primary: DFS, surgery to MRI-confirmed recurrence or death. Secondary: HRQoL, neurocognition (UK/Ireland), time to second-line treatment, OS, cost-effectiveness.
Early RT-related AEs in 30 (45%), all G1-2; late in 17 (26%). G2 or 3 RT-related SAEs 5 (8%), including one G2 radiation necrosis and one G2 optic neuritis; no treatment-related deaths.
Before this, evidence was retrospective series and two non-randomised phase 2 cohorts reporting 3-yr PFS 88.7% (60 Gy/30 fx) and 93.8% (54 Gy/30 fx). NRG BN003 asks the same randomised question and will be the replication test.
No central pathology review (5 pts would now be grade 3) and no mandated early postop MRI. HRQoL data fell to 38% at 5 years and neurocognition was UK/Ireland-only with ~45% missing by 24 mo, so the null QoL and cognitive findings are weak.
With 70% of observed pts recurrence-free at 5 years, the result supports offering RT rather than mandating it. Methylation class stratifies risk but did not erase the RT effect, so it cannot yet select who to observe.
| Endpoint | Radiotherapy (n=78) | Observation (n=79) | Effect |
|---|---|---|---|
| Meningioma recurrence | 11 (14%) | 24 (30%) | n/a |
| 5-yr DFS | 79.9% (67.6–87.9) | 64.3% (51.9–74.2) | Diff 15.6 pp (0.6–30.5) |
| DFS, ITT | n/a | n/a | HR 0.51 (0.27–0.97), p=0.0396 |
| DFS, as-treated | 10/66 (15%) recurred | 25/91 (27%) recurred | HR 0.54 (0.28–1.04), p=0.0642 |
| Deaths | 7 (9%) | 6 (8%) | 5-yr OS OR 0.91 (0.27–3.05) |
| Second-line treatment | 8 (10%) | 16 (20%) | n/a |
CONSORT flow
First randomised evidence and primary DFS met, but accrual closed at 41 of 46 planned events, CI upper bound 0.97, and recurrence was read locally without masking.
- Do patterns of failure support tighter dural margins?
- Can methylation class identify pts safe to observe? n=210 · primary completion 2026-03 · DNA methylation signature for recurrence after GTR
- Is adjuvant RT cost-effective vs active monitoring?
📚 Sources · 📄 1 paper
The longer read
The question this trial answers has been argued from retrospective data for two decades, and the retrospective literature never settled it because the pts who got adjuvant RT were the ones surgeons worried about. ROAM removes that selection problem, and the direction it finds is the one the non-randomised series and the two recent systematic reviews pointed to. What it adds is not a new direction but the first estimate of magnitude that is not confounded by indication: roughly a halving of the DFS hazard, with 5-year DFS 79.9% vs 64.3%.
The confidence a reader should place in that magnitude is moderate, and the reasons pull in opposite directions. Against it: the trial closed at 41 of 46 planned events, the upper CI bound sits at 0.97, and the as-treated estimate (HR 0.54, 0.28–1.04) crosses 1. Recurrence was called by unmasked local investigators without central MRI review, which matters less for a new enhancing dural nodule than for glioma but is still an unmasked subjective read in an open-label trial. For it: 12 of 78 assigned pts never started RT, which dilutes the ITT effect toward the null, and DFS is a blunt endpoint that counts out-of-field recurrence and new meningiomas elsewhere as failures of a treatment never aimed there. The authors plausibly argue the true local effect is larger. Both sets of concerns are real, and the honest summary is that the point estimate is likely right in direction and uncertain within a wide band.
OS was not different (OR 0.91, 0.27–3.05) and the trial could never have shown otherwise with 13 deaths. That makes the decision a trade between recurrence and toxicity, not survival. The toxicity side looks acceptable at 60 Gy/30 fx with central QA: late RT-related events in 26%, mostly alopecia and fatigue, but one G2 necrosis and one optic neuritis with optic dose at the constraint limit. Neurocognitive and QoL data are too incomplete to reassure or alarm.
For RT practice, the more durable questions are dosimetric. The CTV of 5–10 mm from GTV predates modern contouring consensus; the planned patterns-of-failure analysis will show whether failures were in-field, marginal along dura, or distant. If most are distant, tighter margins become defensible and the therapeutic ratio improves. The post-hoc methylation work shows the grade 2 label hides mostly methylation-class benign tumours, yet adjustment did not remove the RT effect, so biology refines prognosis without yet identifying who can safely be observed.
The result would be wrong if the local-read recurrence calls systematically favored RT, or if the truncated event count produced an early overestimate. NRG BN003 asks a closely parallel randomised question and is the obvious replication test. Until then this supports offering adjuvant RT as a default discussion after GTR, with observation a reasonable choice for older pts, easily re-resected convexity tumours, or large treatment volumes.