onc brain

The longer read

The value of this readout is not that 40 Gy in 15 fractions works, which has not been a live question in most of Europe for a decade, but that the late-toxicity advantage survives a decade of follow-up in a population enrolled and treated under modern planning standards. The persistent worry about moderate hypofractionation has always been the late-reacting-tissue argument: a higher dose per fraction should, on radiobiological grounds, extract a fibrosis penalty that only becomes visible at long follow-up, and the fear was that the equivalence seen at three and five years would invert by ten. It does not. The induration curve goes the other way, and the hazard ratio of 0.76 with a confidence interval excluding one means the shorter schedule is superior on the trial's own primary endpoint rather than merely acceptable.

That matters more than a non-inferiority result would have. Non-inferiority on toxicity plus equivalence on control is an argument from convenience: fewer visits, less machine time, lower cost. Superiority on the toxicity endpoint changes the register of the conversation with the patient. The residual defence of a 25-fraction course in this population, that the longer schedule buys some late-tissue insurance, has no data behind it in these results, and a clinician still prescribing it for a node-negative patient should be able to name what they think they are purchasing.

The caution is in what the slides do not report. Locoregional recurrence, distant failure and breast cancer mortality are given only as an absence of significant difference, with no event counts, no absolute rates and no confidence intervals. In a node-negative and DCIS population followed for a median of 12.8 years, the number of locoregional events will be small, and small event counts mean wide intervals; an absence of a significant difference in that setting is weak evidence of equivalence, not strong evidence. It is almost certainly true here, because it is consistent with everything START-B and its successors showed, but the strength of the claim in this dataset alone should not be overstated from these numbers, because these numbers were not shown.

The second caveat is endpoint measurement. Breast induration graded by a clinician is a subjective assessment, and the two arms are not concealable: anyone assessing the breast knows, or can trivially learn, whether the patient came fifteen times or twenty-five. Whether assessors were blinded is not stated in the source. The requirement for the finding at two consecutive visits or at final follow-up is a real methodological guard against single-visit noise, and it should raise confidence in the direction of the result even if it cannot correct for an expectation effect.

What this does not settle is the next question the field has already moved to. The comparator here is 50 Gy in 25 fractions, and the live argument in 2026 is 40 Gy in 15 fractions against five-fraction schedules, where the late-tissue question reopens with an even larger dose per fraction and a shorter interval of mature follow-up. DBCG HYPO gives the moderate-hypofractionation arm of that comparison a ten-year toxicity benchmark to be measured against, which is arguably its most durable contribution. It says nothing about node-positive disease, regional nodal irradiation, or the post-mastectomy chest wall, all of which were outside the enrolled population.