NRG/RTOG 9601 (transcriptomic analysis)
ForProstate cancer, salvage RT ± high-dose bicalutamide, GC score available
TL;DRSecondary analysis, n=352, 18.2y median f/u: GC per 0.1 prognostic for PCSM, sHR 1.34 (1.17-1.54); AAT benefit by GC showed no significant interaction.
The sRT decision in play is hormone omission at low pre-sRT PSA: at PSA <0.7, 15y PCSM benefit from AAT was -2% with GC <0.6 vs 16% with GC ≥0.6. At PSA ≥0.7 the gradient reverses (12% vs 6%), so GC here ranks risk more reliably than it selects for hormones.
In sRT candidates with pre-sRT PSA <0.7 and GC <0.6, this informs the discussion of sRT without hormone therapy as hypothesis-level evidence (15y PCSM benefit with AAT -2%); it does not extend to contemporary ADT, which this trial did not test.
The sRT decision in play is hormone omission at low pre-sRT PSA: at PSA <0.7, 15y PCSM benefit from AAT was -2% with GC <0.6 vs 16% with GC ≥0.6. At PSA ≥0.7 the gradient reverses (12% vs 6%), so GC ranks risk more reliably than it selects for hormones. RT dose and volume not reported.
11 details
Secondary analysis of NRG/RTOG 9601, a phase 3 trial of sRT ± high-dose bicalutamide, using GC scores (Veracyte) from a prior translational project. GC-available cohort n=352, median follow-up 18.2 years. Multivariable Cox or Fine-Gray models, adjusted (s)HR with 95% CI.
Pts treated with sRT on NRG/RTOG 9601 who had a GC score. Median GC 0.59 (Q1-Q3 0.42-0.76); 29%, 23%, 38% and 11% were low, intermediate, high and very high risk. Other baseline features not reported in source.
sRT with or without high-dose bicalutamide (AAT). Bicalutamide dose and duration not reported in source.
Salvage RT in both arms. Dose, fractionation and target volume not reported in source.
Primary objective: prognostic impact of GC (per 0.1) on PCSM. Secondary: OS, MFS, DM. Additional: PCSM and OS by pre-defined GC strata, and PCSM benefit of AAT by GC (<0.6 vs ≥0.6) within pre-sRT PSA subgroup (<0.7 vs ≥0.7).
GC was independently prognostic for every profiled endpoint (tables above). The AAT-by-GC estimates are absolute 15-year PCSM differences with no significant interaction in either PSA stratum.
| Endpoint | Adjusted (s)HR | 95% CI |
|---|---|---|
| PCSM (primary objective) | sHR 1.34 | 1.17-1.54 |
| OS | HR 1.18 | 1.09-1.27 |
| MFS | HR 1.16 | 1.08-1.25 |
| DM | sHR 1.20 | 1.07-1.35 |
| Endpoint at 15 years | Low GC | Very high GC | Difference |
|---|---|---|---|
| PCSM | 4% | 52% | 48% |
| OS | 75% | 25% | 50% |
| Pre-sRT PSA | GC <0.6 | GC ≥0.6 |
|---|---|---|
| <0.7 | -2% | 16% |
| ≥0.7 | 12% | 6% |
Extends the earlier GC validation in this same trial with final outcomes, and layers GC onto the prior 9601 finding that pre-sRT PSA predicts AAT benefit. GETUG-AFU 16, NRG/RTOG 0534 (SPPORT) and RADICALS-HD tested ADT with post-prostatectomy RT; none is addressed in this source.
AAT benefit by GC is given as point estimates only: no CIs, no cell sizes. PCSM was chosen as primary for this secondary analysis because of suspected other-cause mortality with bicalutamide, which sets aside the endpoint where AAT harm would show. Whether the GC-available subset matches the full trial is not reported.
The prognostic claim is well supported at long follow-up. The predictive claim is not: the GC gradient runs in opposite directions in the two PSA strata. The authors frame GC-guided selection for modern ADT in early sRT as a hypothesis, which is the right ceiling.
Secondary biomarker analysis of a GC-available subset; predictive claim rests on non-significant interactions, and the hormone agent was high-dose bicalutamide, not contemporary ADT.
- Does GC predict benefit from contemporary ADT with early sRT?
- Do intermediate and high GC strata separate on 15-year PCSM and OS?
- Does the AAT-by-GC pattern hold for OS, not just PCSM?
📚 Sources · 📄 1 paper
Abstract
The longer read
Two claims sit in this abstract and they deserve different levels of confidence. The first is prognostic: GC score, modelled continuously per 0.1 and adjusted for clinical covariates, tracks with PCSM, OS, MFS and DM at a median follow-up of 18.2 years. This is an update of a validation already performed in the same trial, now with final outcomes, and the direction and consistency across four endpoints make it hard to argue with. The 15-year contrast between the extremes is stark, PCSM 4% vs 52% and OS 75% vs 25%, though it compares the 29% of the cohort with low scores against the 11% with very high scores. The abstract does not report the two middle strata, which is where most counseling decisions actually fall.
The second claim is predictive: that GC might identify who needs hormone therapy with sRT. Here the data are much thinner. At pre-sRT PSA <0.7, the 15-year PCSM benefit from AAT was -2% with GC <0.6 and 16% with GC ≥0.6, which reads as a clean selection story. At PSA ≥0.7 the same split gives 12% and 6%, with the larger benefit in the lower GC group. If GC were a general predictor of hormone benefit, one would expect the gradient to point the same way in both strata. It does not. The authors attribute the non-significant interaction tests to limited power, and that is plausible, but the reversal is equally compatible with noise from carving 352 pts into four cells. No confidence intervals or cell sizes are given, so the reader cannot judge how far any one of these estimates could move.
The endpoint choice also matters. PCSM was selected as the primary endpoint of this secondary analysis because high-dose bicalutamide may raise other-cause mortality, an effect the authors suggest may not occur with contemporary ADT. That reasoning is defensible, yet it removes from view the endpoint on which the hormone arm could look worse. An OS version of the same four-cell table would be the natural companion and is not in the source.
Transferability is the last gap. The hormone agent tested was high-dose bicalutamide, and the conclusion reaches toward modern ADT in early sRT. Trials such as GETUG-AFU 16, NRG/RTOG 0534 (SPPORT) and RADICALS-HD studied ADT with post-prostatectomy RT, and the question of whether GC modifies benefit in those settings is not answered here. RT dose, fractionation and target volume are also absent from the abstract, so the radiotherapy backdrop against which hormone benefit was measured cannot be compared to current practice from this source alone.
The practical read: GC is a credible long-term risk ranker for pts headed to sRT, and that alone can inform how a clinician frames prognosis. The suggestion that a low GC score at low PSA marks pts who gain nothing from hormones is a single unreplicated cell with a negative point estimate and no interval. It would be wrong if the GC-available subset differs systematically from the full trial, or if the low-PSA pattern fails to appear in a cohort treated with contemporary ADT. A prospective GC-stratified comparison is what would settle it.