onc brain

The longer read

The interesting thing about ARTO is not that metastasis-directed therapy improved survival, it is where it did so. STOMP and ORIOLE both tested MDT in hormone-sensitive oligometastatic disease against observation, so the counterfactual there was no treatment at all and the endpoints were ADT-free survival and progression. SABR-COMET carried the OS signal but pooled histologies and enrolled up to five lesions across tumor types. ARTO puts every patient on abiraterone plus ADT and asks only whether ablating all visible disease adds anything on top of a modern androgen-receptor backbone in the castration-resistant state. That is a harder question, and the fact that HR 0.55 survives it is what should move a reader's confidence more than the p-value.

The methodological caveat that matters most is not the phase 2 label. It is that overall survival was never the endpoint this trial was designed to answer. The primary was a 6-month PSA decline of at least 50%, the power calculation for survival was rewritten after that endpoint was met, and this analysis is explicitly unplanned. The authors handled it about as well as it can be handled, finalising the revised calculation before unmasking, but the reader should still treat 0.33-0.92 as a confidence interval whose lower bound is doing a lot of work in a 157-patient trial where the experimental median has not been reached. A single-institution reader deciding whether to change practice on this alone is extrapolating from an interval that comfortably contains a much smaller benefit than the point estimate implies.

The radiotherapy specification deserves more weight than it usually gets in the coverage of these trials. One to five fractions at a biologically effective dose of at least 100 Gy, delivered to every site of disease, is a genuinely ablative prescription and a demanding one. It is not the same intervention as treating an index lesion for symptom control or PSA effect, and a patient whose disease is anatomically positioned such that one of three lesions cannot receive that dose has not received the intervention ARTO tested. This is the parameter that gates transfer to a reader's own clinic, and it is also the most plausible mechanistic account of the survival difference: if the benefit came from delaying systemic progression by eliminating the clones that seed it, then incomplete or subablative coverage should not be expected to reproduce it.

The alternative account is selection. Disease that remains confined to three or fewer sites at the point of castration resistance is biologically unusual, and a trial enriched for that phenotype will show large treatment effects for reasons that have little to do with the treatment. Randomisation protects the comparison internally, but it does not tell you whether the effect size travels to a patient whose three lesions are the leading edge of something more diffuse that PSMA imaging has not yet resolved. The toxicity picture, at least, is reassuring on its own terms: the grade 3-4 events reported skewed toward the control arm, including all five infectious complications and the single treatment-related death, so nothing here argues that comprehensive ablation is being paid for in acute harm.

What would have to be true for this to be wrong is a chance survival split in a small trial whose OS comparison was not prespecified. That is not a remote possibility at this sample size. A prespecified phase 3 in the same setting is the thing that would settle it, and until one reports, ARTO is best read as a strong reason to offer comprehensive ablation in a discussion, not as a result that has already redefined the standard.