onc brain

The longer read

The question of whether radiotherapy belongs in the perioperative management of gastric cancer has been asked and answered negatively twice in the modern era, and the interesting thing about this trial is not that it disagrees but where it disagrees. CRITICS delivered radiotherapy after surgery, in a population that had already been selected by having survived resection, and against a surgical standard that was not uniformly D2. TOPGEAR delivered it preoperatively but enrolled broadly and used a systemic backbone that changed mid-trial. Neo-CRAG holds surgery constant at standardised D2, restricts entry to node-heavy cT3N2-3 or cT4 disease, and gives the radiotherapy before the operation. Those three choices are the trial's entire argument, and the result should be read as a claim about that specific configuration rather than about radiotherapy in gastric cancer generally.

What makes the efficacy claim more credible than a single hazard ratio would be is the internal consistency of the intermediate endpoints. pCR roughly doubles, ypN0 rises from 36.4% to 56.1%, downstaging to ypT0-2 from 23.6% to 42.6%, and locoregional recurrence among R0-resected pts falls from 18.3% to 9.4%. That is a dose-response-shaped chain running from local cytoreduction to local failure to disease-free survival, and it is the mechanistic story the negative trials never produced. A reader sceptical of the DFS curve has to explain away four concordant surrogates, which is harder than dismissing one endpoint.

The weakness is the comparator, and it is not a small one. XELOX was a defensible perioperative standard when accrual opened in 2013; it was no longer the strongest available option when accrual closed in 2022. The control arm's median overall survival of 37.6 months is the number to sit with, because FLOT's benefit over a doublet is itself substantially locoregional, mediated through deeper pathologic response. If a stronger systemic regimen captures part of the same downstaging effect that radiotherapy is capturing here, the two interventions are partly competing for the same mechanism rather than stacking, and the increment radiotherapy adds on top of FLOT would be smaller than the increment shown here. Nothing in this trial can distinguish those possibilities.

The survival figures also deserve a sceptical eye on their own terms. A median overall survival separation of 67.5 versus 37.6 months is very large for a hazard ratio of 0.781 with a confidence interval reaching 0.970, which tells you the medians sit at a steep and unstable part of the curve and that the difference in medians overstates the average treatment effect. The 5-year rates, 50.1% against 44.2%, are the honest summary of magnitude. Enrolment at 13 Chinese referral centres with uniformly high-quality nodal dissection is a further boundary: the value of adding locoregional therapy is generally largest where surgery is weakest, so a Western population with more variable D2 quality might see a different, plausibly larger, effect, while the absolute event rates would not transfer.

For practice, the trial reopens a question most centres considered closed rather than settling it. It makes the omission of radiotherapy in bulky node-positive gastric and junctional disease a position that now needs defending, and it does so without a toxicity argument against it: G3+ postoperative complications were 9.0% versus 7.6%, so the old worry that preoperative radiotherapy compromises a D2 gastrectomy is not supported here. The confirmatory question is specific and answerable: the same design on a FLOT backbone.