onc brain

The longer read

Hypofractionation for breast cancer has been settled for the breast and chest wall for roughly two decades, and unsettled for the nodes for exactly as long. The reason was never efficacy: the START trials and their successors made 15-fraction radiotherapy the default where the target was the breast, and nobody expected the nodal volume to behave differently biologically. The reason was morbidity, and specifically arm lymphedema, which sits at the intersection of surgery, nodal dose, and fraction size, and which patients live with for decades. Skagen 1 is valuable because it attacks that objection directly rather than adding a fourth or fifth demonstration that a shorter course controls disease. Making lymphedema the primary endpoint, with a prespecified 5-percentage-point noninferiority margin against an assumed 10% baseline, is the design choice that gives the result its weight.

The answer is not a narrow pass. The point estimate favors the short course (8.0% versus 9.4%, odds ratio 0.84) and the upper confidence bound of 1.14 on the odds scale sits far from the margin the trial set for itself. A reader inclined to skepticism should note that the trial would have been declared successful with a result meaningfully worse than what it observed, and the observed result does not lean on that permissiveness. That matters for how the finding travels: a marginal pass inside a generous margin is a weaker argument than a numerically favorable point estimate, and this is the latter.

The cancer outcomes are the part that deserves the more careful reading. Locoregional recurrence, the endpoint most directly tied to the dose reduction, gives a hazard ratio of 0.96 with a confidence interval from 0.62 to 1.51. That interval is wide because locoregional recurrence is now an uncommon event in adequately treated high-risk breast cancer, and a trial powered for a morbidity endpoint at this sample size cannot narrow it. The honest read is that a moderate relative difference in locoregional control has not been excluded, but that the point estimate sits on the null and there is no signal to explain away. Breast cancer mortality at 1.25 (0.93 to 1.66) is the number that will draw attention and the one most likely to be over-interpreted. It is a secondary endpoint, in a trial not sized for it, with a confidence interval crossing one, and it is discordant with the locoregional and distant recurrence estimates that would have to mediate any true mortality effect. A dose effect that raises breast cancer death without raising locoregional or distant recurrence is not a coherent mechanism. It should be watched in longer follow-up rather than treated as a finding.

What this changes in practice depends on where a reader already stands. Centers that extrapolated the UK data to nodal volumes years ago now have prospective randomized support for what they were already doing, which is worth more than it sounds: it converts a defensible extrapolation into an evidence-based default. Centers holding 25 fractions for comprehensive nodal irradiation lose their principal justification. What remains genuinely open is the interaction with the things this trial did not isolate, chiefly a simultaneous integrated boost, immediate reconstruction, and the internal mammary chain, along with the shoulder and plexus endpoints that a 4.1-year median follow-up cannot yet characterize. Lymphedema was the right endpoint to pick first, because it was the specific fear. It was not the only one.