onc brain

The longer read

The result a radiation oncologist should take from this analysis is narrower than the headline suggests, and more useful. STAR-TREC randomised two radiotherapy schedules head to head inside a response-adapted organ-preservation pathway, which is a comparison the field has largely avoided: most organ-preservation experience has accrued as single-schedule cohorts or as watch-and-wait registries built on long-course chemoradiotherapy, so the assumption that short-course radiotherapy can substitute for it in a preservation intent has been carried more on convenience than on randomised evidence. Here, at 12 months, it does not substitute. 50 Gy in 25 fractions with concurrent capecitabine produced clinical complete response in 64% at the 16-20 week assessment against 36% after 25 Gy in five fractions, and the downstream consequence of that gap is visible in what happened next: transanal excision for near-complete response was needed almost twice as often after short-course, and TME-free survival at 12 months was 78.5% against 60.6%.

The mechanism matters more than the hazard ratio because it tells you where the difference is generated. This is a response-adapted pathway, so the radiotherapy schedule acts by determining how much tumour has regressed at a fixed assessment window, and the schedule that produces deeper regression by 16-20 weeks keeps more patients on the non-operative arm of the decision tree. That framing also flags the trial's main interpretive vulnerability. A fixed assessment interval may not be equally fair to both schedules, since regression after short-course radiotherapy without concurrent chemotherapy can continue past the window where the decision was made, and the authors themselves note that downstaging is time dependent and assessed the interval from treatment initiation to surgery for exactly that reason. Whether the 12-month gap represents a real difference in achievable preservation or a difference in when the two schedules are judged is a question the 30-month endpoint is better placed to answer.

Two other features should move confidence downward rather than up. The phase 2 estimate, HR 3.7 with a confidence interval from 1.7 to 8.0, was much larger than the pooled figure, and the investigators ran exploratory post-hoc work on stratification variables, tumour length, and centre organ-preservation volume to understand why the phases differed. A treatment effect that shrinks substantially as a trial accrues, in a setting where centre experience with transanal excision and with the judgment call at response assessment plausibly varies, is a signal that some of the early magnitude was context rather than schedule. Separately, the primary TME group in phase 3 was self-selected under a patient-preference design, so the favourable toxicity contrast between organ preservation and surgery, gastrointestinal grade 3-4 events of 2% and 4% versus 8%, and one postoperative death after anastomotic leak, is a comparison between groups that differ by choice and by baseline characteristics rather than by randomisation. It is a reasonable descriptive argument for offering preservation; it is not evidence of a randomised toxicity benefit.

The finding that should temper enthusiasm on both arms is nodal. Baseline MRI excluded nodal involvement by protocol, yet 22% of resected TME specimens carried positive nodes. Whatever occult nodal burden exists in that surgical group also exists among those who kept their rectum, and a 12-month TME-free endpoint by construction cannot see it. That is the specific reason this analysis, however clean the randomised schedule comparison is, should not yet be read as establishing organ preservation as equivalent oncologically. It establishes which radiotherapy schedule to use when the preservation attempt is being made, and defers whether the attempt holds.