onc brain

The longer read

The question NRG 1005 answers is narrow and useful: not whether to boost, which EORTC 22881-10882 settled decades ago, but whether the boost has to be tacked onto the end. Everything downstream of that framing matters, because the trial spent its statistical budget establishing that a 15-fraction total course is not meaningfully worse than a 21 to 32 fraction one in exactly the population where clinicians are least comfortable de-intensifying: 52.7% grade 3, 29.6% ER-negative, 16.7% with close or focally positive margins, 16.3% node-positive.

A reader should sit with the direction of the hazard ratio before accepting the headline. The estimate is 1.31, favoring sequential, and noninferiority was declared because the upper 90% bound landed at 2.04 against a margin of 2.12. That is a narrow clearance, and it was achieved on a redesigned margin. The honest read is that concurrent delivery is probably slightly worse for local control and almost certainly not worse enough to matter, since 7-year IBR was 2.2% versus 2.6% and the trial saw only 56 events across 2,255 patients. Whether that trade is acceptable is a judgment about the value of six fractions, and for most patients in most practices it plainly is. What should not happen is the finding being reported as though the two schedules performed identically.

The cosmesis result deserves more scrutiny than the recurrence result, because it is where the mechanistic worry lived. A simultaneous integrated boost concentrates dose in the cavity throughout treatment rather than after it, and the concern has always been late fibrosis and retraction. The trial tested this three ways and all three cleared noninferiority. The problem is attrition: blinded central photo review captured 79.7% at baseline and only 47.1% at three years, and physician rating fell from 90.3% to 50.5%. Patient-reported BCTOS held up better, and imputation analyses reportedly agreed, but a cosmetic conclusion resting on roughly half a cohort in an unblinded trial is softer than the recurrence conclusion resting on adjudicated events. Three years is also early for the fibrosis this design would produce; the EORTC boost data showed fibrosis accruing well past that point.

The practical constraint on transferring this result is the backbone, not the boost. The concurrent arm is 40 Gy in 15 fractions with 8 Gy integrated, and a centre that still delivers 50 Gy in 25 fractions cannot simply graft the integrated boost onto its own schedule and claim this evidence. The comparison also lets the sequential arm use either a conventional or a moderately hypofractionated backbone, which strengthens the generalizability of the control side. Against IMPORT HIGH, which examined integrated boost dose on the same 40 Gy in 15 fraction platform and found its higher 53 Gy arm produced more induration, NRG 1005's chosen 48 Gy cavity dose sits in the range that trial found tolerable, which is reassuring for the plan as specified rather than for integrated boosting generally.

What would have to be true for this to be wrong is a late divergence in either endpoint. Recurrence curves in high-risk breast cancer continue to separate past seven years, and cosmetic outcomes were only measured to three. Neither risk is a reason to withhold the shorter course now, but both are reasons to want the ten-year report.