onc brain

The longer read

The useful way to read this document is by its disagreements. Twenty-four of thirty-one statements cleared 85%, and the ones that did not are not scattered: they sit on dose prescription, fractionation, and cumulative organ-at-risk limits. Where the panel agreed completely (review the prior plan, avoid reirradiation inside six months, skip elective nodal volumes), the statements describe things most experienced centres already do. Where it fell to 70% (a cumulative cord D0.1cc of 50 Gy or less, brainstem 55 Gy or less), the statement is precisely the number a planner most wants handed to them. That inversion is the honest signal in the paper, and it should temper how firmly any single threshold here is quoted back in a tumour board.

The elective nodal irradiation statement is the one that carries real practice weight, and it is worth noting how thin its footing is relative to its 100% agreement. It rests largely on Caudell's 505-patient multi-institution cohort, where 186 patients received elective coverage without a reduction in locoregional failure or a gain in overall survival, but with more acute toxicity. That is retrospective, and elective coverage in a reirradiated neck is exactly the kind of decision confounded by why the treating physician chose it. The panel's unanimity probably reflects that the toxicity argument is compelling on its own: in a field where pooled grade 3 or higher late toxicity runs at 29% and carotid blowout mortality is reported at 76% among those with data, an intervention with no demonstrated efficacy signal and a demonstrated toxicity cost does not need randomised refutation to be dropped.

The technique guidance is more defensible because it is volume-conditioned rather than modality-preferential. The Vargo comparison showed IMRT beating SBRT on 2-year overall survival in RPA class II, 35.4% versus 18.6%, but that gap closed when SBRT was restricted to 35 Gy or more in five fractions for targets of 25 cc or less. So the 25 cc line the panel draws is not an arbitrary cut point, it is the reconstruction of a subgroup in which the two techniques stopped differing. A reader should treat it as a statement about target selection and SBRT dose adequacy, not about SBRT being inferior. Underdosed SBRT to a large infiltrative recurrence is the failure mode the data actually identify.

What should move confidence downward is the voting mechanics. No-opinion responses stayed in the denominator, which means a statement can land at 76% because four panellists declined to vote rather than because four disagreed, and the paper does not separate those. There was a single round with no re-vote after discussion, so the moderate statements record an unresolved split rather than a considered one. Combined with evidence levels that are mostly 3 and 4, and none level 1, the appropriate posture is to use the high-consensus statements as defaults and the moderate ones as prompts for individual discussion.

The gap the document names but cannot fill is sequencing against immune checkpoint inhibitors. With first-line IO now routine in recurrent or metastatic disease, the practical question is whether to reirradiate a symptomatic local recurrence before starting systemic therapy or after. The panel states plainly that no randomised data define this, and that the available combination data show no unexpected safety signal. That is a fair reading of the evidence and also an admission that the single most common decision point in this population is currently outside the reach of consensus.