Biomarkers in the Post-operative Setting for Prostate Cancer (review)
TL;DRSpratt ASTRO 2026 review: PAM50 luminal B predicts apalutamide benefit with salvage RT in NRG GU006 (5y MFS 95% vs 82%, HR 0.27); non-luminal B no benefit.
Reported via UroToday →
POSEIDONMARCAPRTOG 9601RTOG 0534NRG GU002NRG GU006BALANCESYMPHONY
The ADT decision with salvage RT now has three selectors: pre-RT PSA (OS benefit only above 0.50 ng/mL in POSEIDON), PAM50 luminal B (GU006 5y MFS 95% vs 82%), and MMAI-high. Adverse pathology count does not select (interaction p=0.40). SYMPHONY will test 5-fraction prostate-bed SBRT, dose de-escalation, and elective nodal RT.
In a post-RP pt with PSA ≤0.50 ng/mL and non-luminal B PAM50, these data question adding ADT to salvage RT; the GU006 benefit applies to luminal B tumors only.
Pre-RT PSA gates ADT benefit with salvage RT: POSEIDON OS benefit appears only above 0.50 ng/mL, so early salvage is where ADT omission is most defensible. SYMPHONY will test 5-fraction prostate-bed SBRT, dose de-escalation and elective nodal RT.
PAM50 luminal B identified apalutamide benefit with salvage RT in GU006 (5y MFS HR 0.27); non-luminal B showed none (HR 1.06). Longer ADT duration did not improve OS (HR 0.96), and adverse pathology count did not predict HT benefit.
10 details 3 trials watching
Conference review talk by Dr. Spratt at ASTRO 2026, summarized by UroToday. It pulls together IPD meta-analyses, the prospective biomarker-stratified NRG GU006, and post-hoc biomarker analyses of RTOG 9601/0534.
POSEIDON: adding HT to post-op RT gives OS HR 0.87 (0.76-1.01), p=0.06 overall, with benefit confined to PSA >0.50 ng/mL (HR 0.72 and 0.69). In GU006, apalutamide benefit is restricted to luminal B (MFS HR 0.27). Non-luminal B shows no benefit (MFS HR 1.06).
| Pre-RT PSA (ng/mL) | OS HR (95% CI) | p |
|---|---|---|
| Overall | 0.87 (0.76-1.01) | 0.06 |
| ≤0.20 | 1.14 (0.83-1.57) | 0.42 |
| 0.21-0.50 | 0.94 (0.74-1.19) | 0.70 |
| 0.51-1.0 | 0.72 (0.54-0.96) | 0.02 |
| >1.0 | 0.69 (0.48-0.98) | 0.03 |
| Endpoint | Luminal B: apa vs pbo | Luminal B HR | Non-luminal B: apa vs pbo | Non-luminal B HR |
|---|---|---|---|---|
| 5y bPFS | 72% vs 54% | 0.45 (95% CI 0.24-0.86), p=0.0062 | 70% vs 71% | 0.95 (80% CI 0.65-1.41), p=0.44 |
| 5y MFS | 95% vs 82% | 0.27 (95% CI 0.07-0.95), p=0.029 | 90% vs 89% | 1.06 (95% CI 0.41-2.78), p=0.90 |
Upcoming SYMPHONY trials: PRESTO (prostate bed 20-33 fx vs 5 fx SBRT, N=948), ADAGIO (normal vs reduced dose, N=948), CRESCENDO (± elective nodal RT, N=870), and FULL-SCORE (MDT ± prostate-bed/nodal RT in N+/M+, N=480).
PAM50 prediction is prospective in NRG GU006 and post-hoc in RTOG 0534, so the two lines of evidence agree. Decipher has been validated prospective-retrospectively across RTOG 9601, RTOG 0534, NRG GU002 and NRG GU006.
Clinicopathologic adverse features predict risk but not differential HT benefit. Benefit comes from PSA level and molecular or pathology biomarkers. The speaker frames PAM50 as a level 1A predictive biomarker for systemic intensification with salvage RT.
GU006 luminal B estimates rest on a subgroup with a wide MFS CI. MMAI benefit separation is reported only as approximate curve readings, and RTOG 0534 PAM50 data are retrospective.
- Does PAM50 luminal B predict ADT benefit at PSA ≤0.20 ng/mL?
- Can 5-fraction prostate-bed SBRT replace conventional salvage fractionation? recruiting Post-prostatectomy Radiation Therapy--Moderate Versus Ultra-hypofractionated (Also Known as Stereotactic Body Radiation Therapy [SBRT]) Phase 2n=136 · primary completion 2026-11 · randomised ph 2, ultra vs moderate hypofx, 2y QOLrecruiting Salvage Stereotactic Body Radiotherapy of the Prostate Bed for Biochemical Recurrence After Radical Prostatectomy. Phase NAn=284 · primary completion 2029-02 · randomised ph II/III salvage SBRT vs SOC RT post-RPn=206 · primary completion 2031-09 · phase 3, 5-fx SBRT vs mild hypofx RT, adjuvant/salvage
- Does elective nodal RT add benefit in salvage setting?
📚 Sources · 📄 1 paper
Abstract
The longer read
The argument of this talk is that the salvage setting is moving from asking whether hormone therapy helps everyone to asking who it helps. POSEIDON frames the starting point well. Across 6,057 pts from six randomized trials, adding hormone therapy to post-operative RT produced an OS HR of 0.87 with a confidence interval crossing 1, and lengthening ADT from 4-6 to 24 months did nothing for OS. Read on its own, that would push toward omitting ADT. The PSA-stratified rows say otherwise: the point estimate exceeds 1 at PSA ≤0.20 ng/mL, and benefit appears only above 0.50 ng/mL. The practical consequence for an RT reader is that early salvage, which most of us already favor on local-control grounds, is also the setting where the case for adding ADT weakens.
The adverse-feature analysis is the useful negative result here. Seminal vesicle invasion, high grade group, ECE and positive margins are what most clinicians use to justify ADT with salvage RT. They predict worse OS and MFS, but the interaction with hormone therapy was null for OS (p=0.40) and borderline for MFS (p=0.08). Pathology tells you the risk, not whether the drug changes it. That separation of prognostic from predictive is the conceptual core of the talk.
NRG GU006 carries the most weight because it is the only prospective, double-blind, biomarker-stratified randomized test of a predictive classifier in this setting. The luminal B versus non-luminal B split is large on both bPFS and MFS, and the non-luminal B curves are essentially superimposed. Two cautions apply. First, the luminal B MFS HR of 0.27 comes with an interval running from 0.07 to 0.95, which reflects few metastatic events, so the magnitude is uncertain even if the direction holds. Second, the non-luminal B bPFS estimate is reported with an 80% CI rather than a 95% CI, which a reader should not compare directly with the luminal B interval. The retrospective RTOG 0534 PAM50 analysis points the same way, which strengthens the case without making it independent proof.
The MMAI digital pathology data are the least mature line of evidence. The speaker himself noted that the model went from training straight into randomized-trial validation without a standalone independent validation step. The benefit separation in MMAI-high pts is described only as approximate 12-year distant metastasis readings, with no HR in the source. It is a plausible way to scale biomarker selection, since it needs only routine H&E, but it is not yet at GU006's level.
What the talk does not settle is how the selectors interact: whether a luminal B pt with PSA ≤0.20 ng/mL benefits, or whether MMAI and PAM50 identify the same pts. The SYMPHONY program shifts the RT questions to fractionation, dose de-escalation and elective nodal coverage, which suggests the next decade of salvage trials will tailor the radiation itself, not only the ADT added to it.