onc brain

The longer read

The value of this report is not that SBRT works for low-risk prostate cancer, which 5-yr data settled some time ago, but that it puts a number on when failures actually arrive. Two-thirds of the relapses in this cohort occurred between years 5 and 10, and the authors report the same pattern in Fuller's CyberKnife series. That single observation does more to shape how a reader should hold the ultrahypofractionation literature than any of the survival estimates here. PACE-B's 5-yr non-inferiority result, which is the evidence most clinicians are actually leaning on, is measured at a point where a minority of the eventual events have occurred. It is not wrong, but it is provisional in a way that a 5-yr readout in a faster-failing disease would not be. The authors say as much when they write that they await longer PACE-B follow-up to determine whether SBRT improves long-term relapse rates.

The risk-group results argue against treating intermediate-risk as one population. Pooled IR relapse-free survival of 86% reads as a success, and against the 74-78% reported with dose-escalated EBRT in RTOG 0126 it looks like more than that. But the MSK favorable and unfavorable subgroups separate at 92% and 77% with p = 0.002, which is the only comparison in the paper that reaches significance, while low-risk versus intermediate-risk does not (p = 0.19). The conventional risk trichotomy is doing less work here than the favorable/unfavorable split within IR. The authors' finding that grade group 3 independently predicted 10-yr failure points the same direction, and their inference, that unfavorable IR patients merit consideration of androgen deprivation, is the practical read. It is a hypothesis this trial cannot test, since androgen suppression was prohibited by protocol. That prohibition is what makes the efficacy numbers cleanly RT-attributable and simultaneously what prevents the trial from answering the question its own subgroup analysis raises.

The toxicity comparison deserves more scepticism than the efficacy comparison. Grade 3 rates of 1.4% and 1.5% with no grade 4-5 events are genuinely low, and the absence of new late grade 3+ events after year 5 is worth knowing, since the fear with ultrahypofractionation has always been that late GU injury accrues on a long clock. But the argument the discussion builds, that intrafractional tracking explains why toxicity here is lower than in the PACE-B SBRT arm and PATRIOT, is a between-trial comparison of physician-reported CTCAE across different eras, different case mixes, and different platforms, funded by the manufacturer of the platform that comes out ahead. The MIRAGE result gives that argument real support from a randomised design, so the mechanism is plausible. The attribution of this particular toxicity gap to that mechanism is not established by these data. A reader should also note that the outcome is physician-graded, which reliably reads lower than patient-reported instruments; grade 2+ GU of 14% against grade 2+ GI of 2.1% is the more informative pairing, and it says the bladder, not the rectum, is where the residual burden of five-fraction treatment sits.

What would have to be true for this to mislead? Chiefly that a 2008-2010 cohort selected across 21 sites willing to adopt robotic SBRT early differs from a contemporary population in ways that flatter the result. Central pathology review argues against gross stage migration, but there is no way to exclude selection. The honest conclusion is the authors' own: these results support the durability of SBRT in low-risk and favorable intermediate-risk disease, and a randomised comparison remains required before the cross-trial toxicity claims carry weight.