onc brain

The longer read

Discussant sessions are worth reading precisely because they arbitrate, and the arbitration here lands on selection rather than escalation. Three separate threads (bladder preservation with checkpoint blockade added to chemoradiation, adjuvant IO in non-clear cell RCC, and ctDNA-stratified adjuvant risk) all resolve to the same complaint: the field can deliver the treatment, and cannot yet name the patient who needs it.

For a radiation oncologist the most immediately usable fact is the delivery table. A hypofractionated 55 Gy in 20 fractions schedule with mitomycin C and 5-FU was given in full to every one of the 54 pts, and 47 of them had no extension or delay, with durvalumab layered on top. That matters because the standard worry when a checkpoint inhibitor joins chemoradiation is not efficacy but attrition: treatment breaks, reduced dose intensity, and a bladder-preservation course that quietly becomes a lower-dose course. It did not happen. The pressure instead fell on the systemic components, with chemotherapy discontinued early in 12 (22) and durvalumab discontinued early in 21 (39). That asymmetry is the useful read, and it argues that RT is the robust element of the combination rather than the fragile one. The discussant is careful to say what this does not establish, which is anything about long-term bladder preservation, bladder function, symptom burden or QoL, all of which need longer follow-up. Delivery feasibility is a necessary condition for a preservation strategy, not evidence that the bladder is worth preserving on these terms.

The structural point sits in the EV trials. EV-209 asks whether enfortumab vedotin plus pembrolizumab can produce a clinical CR durable enough to hold in cystectomy-eligible cT2-4a N0 disease, with non-CR patients falling through to surgery or radiotherapy. EV-309 is the more consequential design for an RT reader, because it randomises cystectomy-ineligible or refusing patients against chemoradiotherapy on BIEFS and OS. Trimodality therapy has spent two decades arguing for parity with cystectomy; it is now the comparator that a drug-only strategy has to beat. Whether that comparison is fair will depend on what chemoradiation regimen the control arm actually uses, and on whether local salvage is counted as an event or a feature, since a bladder-intact endpoint treats cystectomy and death as equivalent failures in a way that flatters whichever arm defers surgery longer.

The RCC material should move confidence in the opposite direction. The RAMPART non-clear cell subgroups have few patients and few events, the point estimates swing widely, the confidence intervals are wide, and central pathology review has not finished. The discussant's conclusion, that adjuvant IO has no proven benefit in non-clear cell RCC, is the honest reading of an underpowered subgroup rather than a positive claim of no effect. Against that, the ctDNA data are the striking contrast: separation by ctDNA status (24-mo DFS 79.9% and 72.9% in ctDNA-negative pembrolizumab and placebo pts, against 38.8% and 18.3% in ctDNA-positive pts) dwarfs anything pathologic staging delivers. The obvious caution is that only 30 patients per arm were ctDNA-positive, so this is prognostic signal with thin numbers, and it says nothing yet about whether treating on the basis of ctDNA changes an outcome.