NRG-GU005 NCT03367702
ForLocalized favorable intermediate-risk prostate, T1-T2b, GG1-2, PSA <20
88.6% vs 92.1%
SBRT not superior; adjusted HR 1.40 (0.91-2.13), P=.12
TL;DR3yr DFS 88.6% SBRT vs 92.1% MH-IMRT, superiority rejected (adjusted HR 1.40, 0.91-2.13); bowel and GU toxicity favored SBRT.
The number that should move practice is biochemical failure: 7.8% vs 4.2% at 3yr (adj HR 1.82, 1.01-3.27), while local failure was flat at 1.2% vs 1.0%. The SBRT prescription was deliberately modest, 36.25 Gy/5 fx with dose uniformity prioritized and urethral max held to 38.78 Gy, so this reads as a dose and margin question, not a verdict on 5 fractions.
In favorable intermediate-risk localized prostate cancer choosing between 5-fraction SBRT and moderate hypofractionation, this supports the toxicity and convenience case for SBRT while flagging a 3-year PSA-failure gap; it does not speak to unfavorable intermediate or high-risk disease, or to dose-escalated SBRT regimens.
Biochemical failure, not DFS, carries the read: 7.8% vs 4.2% at 3yr (adj HR 1.82, 1.01-3.27) with local failure flat at 1.2% vs 1.0%. The prescription was deliberately uniform, 36.25 Gy/5 fx with urethral max 38.78 Gy and no focal boost, so this argues about dose and margin rather than about 5 fractions.
Also covered Jul 8
13 details 5 trials watching
Phase 3, international, open-label, 1:1 randomized superiority trial across 136 centers, accruing November 2017 to June 2022 with last follow-up October 2024. N=698 randomized (353 SBRT, 345 MH-IMRT), median follow-up 3.2 years. Two coprimary end points, patient-reported QoL and DFS, each tested at 2-sided alpha .05 with no study-wise correction, so both had to be positive for a positive trial.
Localized cT1-T2b, either Gleason 3+4 (GG2) with PSA <20 ng/mL or Gleason 3+3 (GG1) with PSA 10-20 ng/mL. Median age 68 (range 42-84); 80% White, 13% Black, 3% Asian; 81.7% T1c and 88.1% Zubrod 0. Stratified by risk group and by rectal manipulation, of whom 55.6% used spacer alone and 40.3% no device.
SBRT 36.25 Gy in 5 fractions (7.25 Gy per fraction), delivered as prescribed in 96.6%; MH-IMRT 70 Gy/28 fx in 70.6% or 60 Gy/20 fx in 26.6%. Rectal constraints were max 38 Gy to 0.03 mL and 18 Gy to 50%; bladder 39 Gy and 15 Gy. Where PTV max exceeded 38.78 Gy the urethra had to be contoured and capped at 38.78 Gy, a deliberately uniform, non-escalated prescription. Protocol-compliant or acceptable variation in 97.7% and 97.2% of arms.
Coprimary: MCID frequency in EPIC-26 urinary irritative/obstructive and bowel domains at 24 months, and DFS at 3 years (powered for HR 0.62). Secondary: the other EPIC-26 domains at 12 and 24 months, overall survival, biochemical DFS, regional and distant failure. EPIC-26 adherence was 82% at 1yr and 84% at 2yr.
Urinary irritative/obstructive MCID was flat (35.4% vs 33.7%, P=.68); bowel MCID favored SBRT (34.9% vs 43.8%, P=.03). DFS superiority was rejected at a 91-event interim (HR 1.38, 0.91-2.09), with 3yr rates 88.6% vs 92.1% and no adjusted difference (HR 1.40, P=.12).
| Endpoint | SBRT | MH-IMRT | Effect |
|---|---|---|---|
| Biochemical failure | 7.8% | 4.2% | adj HR 1.82 (1.01-3.27), P=.046 |
| Local failure | 1.2% | 1.0% | P=.97 |
| Overall survival | n/a | n/a | P=.65; adj HR 1.15 (0.55-2.41), P=.70 |
Grade 3+4 GU adverse events were lower with SBRT (0.6% vs 2.5%, P=.04), as were any-grade rectal hemorrhage (10.5% vs 17.3%, P=.01) and fatigue (39.2% vs 50.8%, P=.002). Longitudinal bowel scores favored SBRT (LS mean 2.68 [1.02-4.34], P=.002) and urinary incontinence scores likewise (LS mean 2.91 [0.85-4.97], P=.006).
PACE-B and HYPO-RT-PC established that 5-fraction prostate SBRT is tolerable and non-inferior on biochemical control; this trial asked the harder superiority question and lost it, and adds a rectal-manipulation stratification neither predecessor used, which balanced spacer use across arms rather than leaving it a center-level confounder.
The biochemical failure signal, the one result that argues against SBRT, sits on 3-year rates with a CI whose lower bound touches unity (adj HR 1.82, 1.01-3.27) and cannot be separated from the higher benign PSA bounce rate after SBRT, which the trial was not designed to distinguish from true failure. Local, regional and distant failure events were too few to analyze (8 vs 7, 4 vs 2, 4 vs 4), so the mechanism behind the PSA gap is unobserved.
The authors attribute the PSA-control gap to a possible lower biologically effective dose and smaller SBRT margins, the same two choices that plausibly produced the bowel benefit. If that trade is real, it is tunable: focal boost to the dominant intraprostatic lesion was explicitly excluded here and is where the next version of this question belongs.
CONSORT flow
Randomised phase 3, prespecified coprimary endpoints, ITT: the DFS superiority hypothesis was rejected and biochemical failure ran higher with SBRT, contesting the assumption 5 fractions cost nothing.
- Whether benign PSA bounce explains the biochemical failure gap
- Does focal boost to the dominant intraprostatic lesion close it recruiting Image-guided Focal Dose Escalation- Primary pc Treated With Primary External Beam Hypofract.Stereotactic rt Phase NAn=374 · primary completion 2025-08 · randomised focal dose escalation, SBRT vs IMRT/IGRTn=186 · primary completion 2032-08 · randomised DIL boost vs whole-gland boost in SBRTrecruiting Phase III Adaptive Adaptive Stereostactic Body Radiotherapy (SBRT) With Dose Escalation for High-Risk Prostate Cancer Phase NAn=390 · primary completion 2033-04 · DIL dose-escalated SBRT vs mod hypofx, high risk
- Late GI/GU toxicity and QoL beyond 2 years n=68 · primary completion 2025-10 · 5y cumulative GI/GU/sexual toxicity, MR-linac SBRTactive Germline DNA-Based Radiosensitivity Biomarker Influence on Toxicity Following Prostate Radiotherapy, GARUDA Trial Phase NAn=208 · primary completion 2027-12 · long-term physician-scored GU toxicity after SBRT
📚 Sources · 📄 1 paper
Abstract
The longer read
The framing that matters is what NRG-GU005 asked and what it did not. PACE-B and HYPO-RT-PC put 5-fraction prostate SBRT into practice on non-inferiority footing; this trial went for superiority on both quality of life and disease-free survival, and the DFS half was rejected at interim rather than merely missed. That distinction is doing more work than the headline suggests. A rejected superiority hypothesis is not a demonstration of equivalence, and the authors say so plainly. Nothing here establishes that 36.25 Gy in 5 fractions is non-inferior for cancer control, which means a reader who was already using SBRT on the strength of PACE-B has not had that evidence withdrawn, and a reader who wanted this trial to close the question is left where they started.
The uncomfortable result is biochemical: 7.8% versus 4.2% failure at 3 years, adjusted HR 1.82 with a lower bound of 1.01. Two readings compete and the trial cannot separate them. The first is real undertreatment. The protocol prioritized dose uniformity and capped urethral dose at 38.78 Gy, so the biologically effective dose in this SBRT arm sits below what several institutional 5-fraction series deliver, and the treatment volumes were smaller. The authors raise marginal miss of disease abutting the target as a candidate mechanism, and the MRI substudy collected here, comparing dominant lesion location against site of recurrence, is precisely the analysis that would test it. It is not in this report. The second reading is measurement: benign PSA bounce is more common after SBRT, the trial was not designed to distinguish bounce from failure, and DFS counts a biochemical event the same either way. With 3.2 years of median follow-up, a bounce-inflated failure curve and a genuinely inferior one look similar. The tell will be whether local and distant failure eventually track the PSA signal, and right now they do not: 8 local events versus 7, 4 distant versus 4, too few to analyze at all.
Against that sits a toxicity profile that is not marginal. Grade 3-4 GU events were 0.6% versus 2.5%, rectal hemorrhage 10.5% versus 17.3%, fatigue 39.2% versus 50.8%, and the bowel MCID difference of 8.9% cleared its endpoint. Notably the urinary irritative domain, the one most people expect 5 fractions to worsen, was flat. So the trade on offer is a measurable near-term morbidity and convenience gain against a biochemical signal of uncertain meaning, in a population whose prognosis is favorable enough that a 3.5-point DFS difference at 3 years may never translate into an event that matters.
What would make the negative read wrong is longer follow-up separating true failure from bounce, and what would make it right is the same follow-up showing local failure climbing behind the PSA. Either way the design choice most worth revisiting is not fractionation but dose distribution. This trial excluded focal escalation to the dominant intraprostatic lesion by protocol, which makes it a clean test of uniform-dose SBRT and a poor test of SBRT as it is increasingly practiced. The rectal-manipulation stratification is the other durable contribution: spacer use was balanced rather than confounded, and its independent effect on bowel scores was significant, which is a cleaner estimate of what a spacer buys than the prior trials could give.