onc brain

The longer read

The central claim here is less that hormone therapy fails with postoperative radiotherapy than that the benefit the field has inferred from biochemical and metastasis endpoints does not carry through to survival for most men now being treated. GETUG-AFU 16 and RTOG 0534 made short-course HT with salvage RT a common default on the strength of biochemical control. POSEIDON pools those trials with RADICALS-HD and RTOG 9601 at the patient level, and the pooled MFS benefit (HR 0.79) is real but falls short of the ICECaP surrogate threshold of 0.81 at every pre-PORT PSA. That is the methodological hinge: the paper is not just reporting a null OS primary, it is arguing, with a prespecified surrogacy framework, that the prior positive endpoints were never expected to become survival gains in this population.

The PSA story deserves more caution than the trade-press headline gives it. The binary split at 0.5 ng/mL produced an interaction p of 0.02, but the four-strata interaction was 0.08, and PSA as a continuous variable was non-significant in the one-stage model (0.19) and significant only in the two-stage model (0.04). No multiplicity correction was applied, and the authors themselves label secondary subgroup findings hypothesis-generating. The direction is biologically coherent, since higher pre-PORT PSA plausibly marks occult disease beyond the RT field where systemic therapy should matter, and it matches the prior suggestion that PSA is predictive for long-term HT benefit. But the threshold itself should be read as a gradient, not a validated cut point.

The duration question is weaker still as a comparison. Long-term HT looked better on OS (HR 0.79, p=0.049) than short-term (HR 0.93), yet the duration interaction was 0.17, and the long-term trials enrolled men with higher PSA and more adverse pathology. Directly randomised prolongation from short to long course in the RADICALS cohorts improved MFS but not OS (HR 0.89). The spline analysis puts any long-term OS benefit at PSA of 1.6 ng/mL or higher, a range rarely seen in contemporary early salvage. The practical read is that when HT is used, 4 to 6 months carries most of what the data support, with 24 months reserved for the high-PSA tail. The RTOG 9601 bicalutamide arm also means part of the long-term signal comes from a regimen few would use now.

The largest threat to transportability is era. These trials enrolled before PSMA PET and genomic classifiers; the discussant restricts the findings to PET-negative biochemical recurrence. A modern man at PSA 0.3 with a negative PSMA scan is probably well represented, and arguably lower risk than the pooled population. A man whose PSA is 0.3 but whose PSMA PET shows pelvic nodes is not represented at all, and the result should not be stretched to him. For the RT decision itself, POSEIDON does not inform field design, dose, or nodal coverage; the source excerpt reports no RT parameters.

What would make it wrong: if pre-PORT PSA is mostly a proxy for disease biology that a classifier captures better, the PSA threshold could misallocate HT in both directions. NRG GU006, testing PAM50 with short-term apalutamide, is the trial positioned to show whether biology beats PSA as the selector.