onc brain

The longer read

The first randomised answer to a question neuro-oncology has argued over for two decades comes out in favour of adjuvant radiotherapy, but it arrives underpowered. The trial planned for 46 events and stopped at 41 when funding ran out. The hazard ratio of 0.51 has an upper confidence bound of 0.97 and the absolute 5-year DFS difference has a lower bound of 0.6 points. A reader should hold two things together: the point estimate matches what the investigators designed for, a halving of recurrence that pt representatives had said would justify six weeks of treatment, and the precision around it is thin enough that a modestly smaller true effect is compatible with the data. The as-treated estimate of 0.54 has an interval crossing 1, but that analysis was built from post-randomisation groups and does not undercut the ITT result so much as show how little room the sample leaves.

Several design choices push the estimate toward the null rather than away from it, which matters when judging whether the benefit is real. Twelve of 78 pts allocated to RT never started it. DFS counted any recurrence and any death, so out-of-field recurrences and new meningiomas at distant sites, events no adjuvant field could prevent, sit inside the endpoint. Deaths were balanced and none were attributed to treatment. Against those, the main bias toward exaggerating benefit is the unblinded local MRI read, though recurrence of a solid enhancing extra-axial mass is less open to interpretation than progression in a glioma. On balance the direction of the effect looks trustworthy, and the magnitude is uncertain.

The prior evidence was retrospective series that disagreed with each other and two non-randomised phase 2 cohorts reporting high 3-year PFS with 54 to 60 Gy. Those cohorts could not say what observation alone would have achieved. ROAM supplies that control arm, and its 5-year DFS of 64.3% under observation shows that most completely resected grade 2 tumours do not recur within that window. That is the counterweight to the headline. Adjuvant RT here is a way of preventing roughly one recurrence for every six or seven pts treated, and whether that trade suits a given pt depends on age, comorbidity, location, re-resectability, and the size of the field. NRG BN003, testing the same question, will show whether the effect reproduces.

For the radiation oncologist the more consequential questions are technical. The protocol margins, a 5 to 10 mm CTV followed by a 3 to 5 mm PTV, came from earlier EORTC trials that predate current contouring guidance, and the investigators acknowledge that modern IMRT with image guidance allows tighter volumes. The one optic neuritis happened with the optic apparatus dose at constraint limits despite central review. The planned patterns-of-failure analysis, relating recurrences to dose distributions and dural coverage, is what will tell us whether smaller volumes cost control.

The post-hoc methylation analysis is hypothesis-generating. Close to two-thirds of histologically atypical tumours fell into the methylation class-benign group, and adjusting for methylation class or integrated score did not remove the RT effect. That argues against using molecular class, at present, to withhold RT from an individual pt, but the subset is too small and unplanned to say which molecular groups can safely be observed.