onc brain

The longer read

The argument here is about substitution, and it is the argument that matters for how a radiation oncologist actually builds a plan. Two intensification levers have been available for three decades, and the field has often treated them as fungible: push the dose and you can spare the patient hormones. MARCAP's answer is that they act on different endpoints and do not trade against each other. ADT delivered MFS HR 0.83 and OS HR 0.86 across 10,853 patients with 11.4 years of median follow-up, and critically the benefit was identical above and below 74 Gy (HR 0.83 in both strata). That last comparison is the one that closes the dogma, because it is the direct test: if escalation substituted for hormonal therapy, ADT's effect should have shrunk in the high-dose stratum. It did not move at all.

The companion HEAT network analysis supplies the other half. Escalation up to 79.2 Gy improved biochemical RFS without improving MFS or OS, which is a familiar pattern and one worth being precise about rather than dismissive of. The explanation offered is the risk-group mismatch between where escalation helps biochemically and where metastasis actually occurs: a 21% five-year biochemical benefit in low-risk disease (PROG 05/09), where the competing risk of metastasis is small enough that the surrogate cannot cash out, against 3% in high-risk disease (GETUG-18), where metastasis is common but the biochemical gain is thin. That is a coherent account of why the surrogate has repeatedly failed to translate, and it should make a reader more skeptical of biochemical endpoints in escalation trials specifically, not of biochemical endpoints in general.

The sequencing finding deserves more weight than it usually gets. Neoadjuvant prolongation from 3-4 to 6-9 months produced HR 0.95 for both MFS and OS, confidence intervals crossing one in both directions; adjuvant prolongation from 4-6 to 18-36 months produced HR 0.84 and HR 0.85. Those two comparisons are not the same duration contrast, so the cleanest reading is not that adjuvant months are more potent per month than neoadjuvant ones, but that the trials which extended on the front end found nothing and the trials which extended on the back end found survival benefit. For a practice that still routinely front-loads ADT before radiotherapy, the actionable implication is that the extra neoadjuvant months are being spent without demonstrated oncologic return, and the toxicity of those months is real.

Two cautions belong on the read. First, the risk-group invariance of the relative benefit is the source of the most interesting downstream claim, that some intermediate-risk patients warrant long-term ADT and some high-risk patients are adequately served by less. That is an inference from constant relative effect plus differing baseline risk, not a randomised finding, and the NNT split (18.0 vs 8.4 for ADT addition; 16.1 vs 10.4 for adjuvant prolongation) is where the decision actually lives. Second, the specific duration guidance is drawn from DHARMA, described as accepted but not yet published, so the reader is being asked to act on a number they cannot yet check. The rest of the analysis is auditable; that part is not.

What would have to be true for this to be wrong: that the pooled trials' dose strata are confounded by era and technique in a way that masks a real dose-ADT interaction, or that a modern escalation approach (brachytherapy boost, ultrahypofractionation) behaves differently from the external-beam escalation these trials tested. The presentation acknowledges the toxicity axis, contrasting ASCENDE-RT against FLAME, which is the right place to look for whether escalation earns its cost when it is not buying survival.