PRIMARY2 NCT05154162
ForBiopsy-naive suspected PCa, PI-RADS 2 + red flag or PI-RADS 3, PSA ≤20, ≤cT2
12% vs 16%
–3.7% (95% CI –8.9 to 1.5), p=0.0093; non-inferior at 10% margin
TL;DRBiopsy avoided in 49%; csPCa 12% vs 16% with PSMA-PET vs systematic biopsy, non-inferior in ITT, in biopsy-naive men with equivocal or high-risk non-suspicious MRI.
The csPCa shortfall sits in GS 3+4 disease: GS ≥4+3=7 was found in 14 (4%) PSMA-PET vs 16 (5%) biopsy pts, while clinically insignificant diagnoses fell from 105 (32%) to 46 (14%). Men reaching definitive-treatment consults through this pathway are filtered away from low-grade disease, not from GS ≥4+3; only 7 (1%) have had RT so far.
In a biopsy-naive man with PI-RADS 3, or PI-RADS 2 plus a red flag, and PSA ≤20 ng/mL, this informs replacing systematic biopsy with PSMA-PET triage and PSA surveillance if PET-negative; it does not extend to PI-RADS 4–5 MRI.
Half the PSMA-PET arm (163/331) avoided biopsy, and biopsied pts needed fewer cores (mean 19.6 vs 24.8). The catch for the proceduralist: 35 control pts never got their biopsy vs 7 PSMA-PET, and non-inferiority fails in the worst case once 4 of those 35 are assumed csPCa.
8 details
Investigator-initiated phase 3 non-inferiority RCT at seven Australian hospitals; 1:1 web-based randomisation stratified by site, no masking. N=660 (329 systematic biopsy, 331 PSMA-PET); median follow-up 6.0 mo (IQR 5.1–7.0).
Biopsy-naive men with suspected csPCa and PI-RADS 3, or PI-RADS 2 plus ≥1 red flag (PSA density >0.1 ng/mL/mL, abnormal DRE, strong family history, BRCA mutation, PSA >10 ng/mL, PSA doubling time <36 mo, PSA velocity >0.75 ng/mL/yr); PSA ≤20 ng/mL, ≤cT2. Median age 61, median PSA 5.2 ng/mL, PI-RADS 2 in 335 (51%).
Experimental: pelvic-only [68Ga]Ga-PSMA-11 PET-CT (1.8–2.2 MBq/kg) read on the PRIMARY score; score 3–5 → PSMA-targeted TP biopsy, score 1–2 → no biopsy and PSA review at 6, 12, 18, 24 mo. Control: systematic TP biopsy, ≥12 cores, MRI-targeted cores permitted for PI-RADS 3.
Co-primary: csPCa (GS 3+4 with ≥10% pattern 4, or higher) on biopsy within 3 mo, NI margin 10%; and biopsy avoidance within 6 mo in the PSMA-PET arm vs a 20% threshold. Secondary: clinically insignificant PCa, alternate csPCa definitions, core number, inter-reader agreement, complications and PROs.
Both co-primaries met in ITT: csPCa detection non-inferior (p=0.0093) and biopsy avoided in 163 (49%) of 331. 166 (50%) of 329 scanned were PRIMARY-negative; central vs local agreement κ=0.81 (95% CI 0.75–0.87).
| Endpoint | PSMA-PET | Systematic biopsy | Difference |
|---|---|---|---|
| csPCa (GS 3+4 with ≥10% pattern 4, or higher) | 39 (12%) | 51 (16%) | –3.7% (95% CI –8.9 to 1.5), p=0.0093, NI |
| Clinically insignificant PCa | 46 (14%) | 105 (32%) | –18% (97.5% CI –25 to –11), p<0.0001 |
| No cancer on biopsy | 73 (22%) | 138 (42%) | not reported |
| GS ≥3+4=7 (any pattern 4 fraction) | 56 (17%) | 79 (24%) | –7.1% (95% CI –13.2 to –1.0) |
| GS ≥4+3=7 | 14 (4%) | 16 (5%) | –0.6% (–3.8 to 2.5) |
| Mean cores per biopsied pt | 19.6 (SD 12.1) | 24.8 (SD 9.0) | not reported |
| MRI | PSMA-PET | Systematic biopsy |
|---|---|---|
| PI-RADS 2 | 15/155 (10%) | 22/180 (12%) |
| PI-RADS 3 | 24/176 (14%) | 29/149 (19%) |
| Analysis | PSMA-PET | Systematic biopsy | Difference (95% CI) |
|---|---|---|---|
| Covariate-adjusted (prespecified) | 8% | 13% | –4.8% (–9.5 to –0.1) |
| Per-protocol (prespecified) | 27/236 (11%) | 47/255 (18%) | –7.0% (–13.3 to –0.7) |
| Biopsy per protocol (post-hoc) | 39/324 (12%) | 51/294 (17%) | –5.3% (–11.0 to 0.3) |
| Best case (prespecified) | 14% | 16% | –1.6% (–7.3 to 4.1) |
Among biopsied pts, pain 33 (21%) PSMA-PET vs 62 (21%) control, haematuria 60 (38%) vs 126 (43%), haematospermia 77 (48%) vs 133 (45%). SHIM-assessed erectile dysfunction did not rise after biopsy: 50 (31%) of 160 vs 91 (31%) of 294.
The MRI-first pathway (PROMIS, PRECISION) defines who with a suspicious scan is biopsied; PRIMARY2 targets the equivocal or non-suspicious MRI group it leaves behind. The prior evidence was the non-randomised PRIMARY study, which showed improved NPV for csPCa in an MRI-triaged population, and the authors found no prior RCT of PSMA-PET in the diagnostic setting.
Per-protocol analysis excluded 169 (26%) of 660 for out-of-window imaging or biopsy, or off-protocol biopsy type. Baseline MRI mix was imbalanced (PI-RADS 2 in 55% control vs 47% PSMA-PET), and the csPCa gap was wider in PI-RADS 3 (14% vs 19%). Single-country, accredited-camera Ga-68 protocol; ethnicity not collected.
The biopsy-avoidance benefit is clear; the detection claim rests on the ITT and covariate-adjusted analyses, since per-protocol and post-hoc bounds fall past the margin. The shortfall concentrates below GS 4+3 (GS ≥4+3=7: –0.6%), so the open question is whether PSA surveillance recaptures missed GS 3+4 disease in the 159 unbiopsied PET-negative men.
CONSORT flow
ITT met the 10% NI margin, but per-protocol (lower CI –13.3) and post-hoc (–11.0) bounds cross it; 35 vs 7 unbiopsied pts, worst case fails at 4.
- Missed csPCa in PET-negative men on PSA surveillance beyond 6 months
- Transferability to other PSMA radiopharmaceuticals and centres
- Cost-effectiveness of PSMA-PET triage vs systematic biopsy
📚 Sources · 📄 1 paper
The longer read
The MRI-first pathway (PROMIS, PRECISION) settled who with a suspicious scan goes to biopsy, but it left a gray zone that is still biopsied by default: equivocal MRI, or a non-suspicious MRI with a clinical red flag. Yield there is low and the dominant harm is overdiagnosis of low-grade disease. The non-randomised PRIMARY study suggested PSMA-PET adds negative predictive value in exactly this MRI-triaged group. PRIMARY2 asks the more useful question, not whether the scan is accurate but what happens to a population when the scan decides who is biopsied, and pairing detection non-inferiority with biopsy avoidance is the right structure for a pathway trial.
The avoidance result needs little defence: its interval sits well clear of the 20% threshold, and clinically insignificant diagnoses dropped from 32% to 14% of the arm. The detection result needs a closer read. Every reported analysis puts PSMA-PET detection below control. The ITT lower bound of –8.9 clears the 10% margin and the covariate-adjusted –9.5 only just does, while the per-protocol bound of –13.3 and the post-hoc bound of –11.0 fall past it. A 10-point margin is also wide against a control detection rate of 16%.
Differential missing outcomes compound this. 35 control pts never had their allocated biopsy against 7 in the PSMA-PET arm, and an unbiopsied control cannot contribute a csPCa diagnosis, which pulls control detection down and the difference toward non-inferiority. The prespecified worst case loses non-inferiority once 4 of those 35 are assumed to harbor csPCa. That is a thin cushion for the half of the trial carrying the safety claim.
Where the shortfall sits matters more than its size. On the alternate definitions, GS ≥4+3=7 was found in 14 vs 16 pts (–0.6%), while GS ≥3+4=7 of any pattern-4 fraction differed by –7.1% with an interval excluding zero. The cancers PSMA triage leaves undiagnosed at 6 months look concentrated in GS 3+4 disease, the least consequential place for a triage test to lose sensitivity, provided PSA surveillance recovers them. The gap was also wider in PI-RADS 3 (24/176 vs 29/149) than in PI-RADS 2 (15/155 vs 22/180), and PI-RADS 3 controls could add MRI-targeted cores, so the comparator was strongest where PSMA-PET lost the most ground.
What the trial does not settle is the fate of the 159 PET-negative men who skipped biopsy. With median follow-up of 6.0 months, missed csPCa that declares itself on PSA surveillance has not had time to appear; the final analysis carries that answer. Transfer is a second question: a pelvic-only Ga-68 PSMA-11 protocol on accredited cameras across seven Australian sites is not the same test as another tracer read elsewhere, though central versus local agreement of κ=0.81 is reassuring on the reading side. The result would be wrong in the way that matters if PET-negative men later declare GS ≥4+3 disease that systematic biopsy would have caught up front.