PRIMARY2 NCT05154162
ForBiopsy-naive men, PI-RADS 2-3 MRI with a clinical red flag, PSA ≤20, ≤cT2
TL;DRcsPCa 12% vs 16% (difference -3.7%, 95% CI -8.9 to 1.5, p=0.0093 non-inferiority), and PSMA-PET avoided biopsy in 49%.
The RT-relevant read is downstream case-mix, not the PET itself: clinically insignificant cancer fell from 32% to 14% and no-cancer biopsies from 42% to 22%, so referrals arriving from this pathway are enriched for GS 3+4 with ≥10% pattern 4 or higher. GS 4+3 or higher was near-identical (4% vs 5%), so the high-risk pool a radiation oncologist actually treats does not shrink.
In a biopsy-naive man with PI-RADS 2 or 3 MRI and one clinical red flag (PSA density >0.1, family history, abnormal DRE), this supports PSMA-PET as a triage step before transperineal biopsy; it does not extend to PI-RADS 4-5, PSA above 20, or previously biopsied men.
The downstream referral mix shifts, not the treatment: clinically insignificant cancer fell from 32% to 14% and no-cancer biopsies from 42% to 22%, while GS 4+3 or higher was 4% vs 5%. Fewer men enter the treatment discussion with disease that never needed it, and the high-risk pool is unchanged.
This directly changes the biopsy decision: a PRIMARY score of 1-2 sent 159 of 331 men to PSA surveillance instead of transperineal biopsy, and among those still biopsied mean cores fell from 24.8 to 19.6. The trade is 35 (11%) missing control biopsies and a worst-case analysis that crosses the margin at four true positives among them.
11 details 5 trials watching
Investigator-initiated, multicentre, non-inferiority phase 3 RCT at seven Australian hospitals, randomised 1:1 (block sizes 2 or 4, stratified by site) via a central web system. No masking of participants or investigators. Enrolment March 2, 2022 to Aug 24, 2025; data cutoff Dec 5, 2025; median follow-up 6.0 months (IQR 5.1-7.0).
Biopsy-naive men with clinical suspicion of significant prostate cancer and PI-RADS 2 or 3 mpMRI plus at least one red flag (PSA density >0.1 ng/mL/mL, abnormal DRE, strong family history, BRCA mutation, PSA >10, PSA doubling time <36 months, or PSA velocity >0.75 ng/mL per year), PSA ≤20 ng/mL, clinical T2 or less. Median age 61 (IQR 56-66), median PSA 5.2 ng/mL, median PSA density 0.13; PI-RADS 2 in 335 (51%) and PI-RADS 3 in 325 (49%).
Experimental: pelvic-only [68Ga]Ga-PSMA-11 PET-CT within 28 days of randomisation, 1.8-2.2 MBq/kg, furosemide 20 mg recommended, acquisition 60-70 min post-injection, low-dose CT without contrast. PRIMARY score 3-5 → PSMA-targeted (plus MRI-targeted) transperineal biopsy; score 1-2 → no biopsy, PSA surveillance at 6, 12, 18 and 24 months. Control: systematic transperineal biopsy, minimum 12 cores.
Co-primary: proportion with clinically significant cancer (Gleason 3+4 with ≥10% pattern 4 or higher) on biopsy within 3 months, non-inferiority margin 10%; and proportion in the PET arm avoiding biopsy within 6 months, threshold 20%. Secondary: clinically insignificant cancer, alternate csPCa definitions, core number, interobserver variability, and patient-reported outcomes.
Both co-primaries met. Covariate-adjusted sensitivity analysis gave 8% vs 13%, difference -4.8% (95% CI -9.5 to -0.1); per-protocol gave 27 (11%) of 236 vs 47 (18%) of 255, difference -7.0% (-13.3 to -0.7). In the PET arm 166 (50%) of 329 scanned were PRIMARY-negative and 163 (50%) positive.
Post-biopsy symptoms were similar between arms: pain 33 (21%) experimental vs 62 (21%) control, haematuria 60 (38%) vs 126 (43%), haematospermia 77 (48%) vs 133 (45%). SHIM-assessed erectile dysfunction did not increase after biopsy (50 [31%] of 160 vs 91 [31%] of 294). The real safety argument is exposure avoided rather than toxicity reduced: 159 men had no biopsy at all.
The only prior evidence was the non-randomised PRIMARY cohort (2021), which reported improved NPV for csPCa in an MRI-triaged population going to systematic biopsy. A literature search to Dec 11, 2025 found no randomised trials of PSMA-PET in the diagnostic setting, so this is the first randomised test of the question rather than a confirmation of one.
Missing protocol biopsies were differential (35 [11%] control vs 7 [2%] experimental), and the prespecified worst-case scenario crosses the 10% margin once four of the 35 unbiopsied controls truly had csPCa, so the non-inferiority claim rests on how those men are counted. The 159 men who avoided biopsy carry no confirmatory histology at a median 6.0 months, and the trial reports 51 (8%) prostatectomies and 7 (1%) radiotherapy with definitive management deferred to the final analysis.
The trial answers a triage question, not a detection-accuracy question: it shows a PSMA-negative result can stand in for a negative biopsy over 6 months in a low-yield population, and that the biopsies still done return less indolent disease. What it does not settle is whether the avoided biopsies were truly negative, or whether the same result holds with an 18F-labelled tracer, at a non-accredited site, or in a health system where a PET costs more than the biopsy it replaces.
CONSORT flow
Randomised phase 3, prespecified co-primaries both met, so a triage step that halves biopsies is now tested evidence against a biopsy-everyone pathway. Short follow-up limits durability, not internal validity.
- Do the 159 men who avoided biopsy stay negative beyond 6 months
- Does the result hold with 18F-labelled PSMA tracers recruiting Fully Hybrid 18F-PSMA PET/MRI as One-stop Approach for the Diagnosis of Clinically Significant Prostate Cancer. Phase 2n=167 · primary completion 2026-03 · 18F-PSMA PET/MRI to cut unnecessary biopsiesn=250 · primary completion 2027-12 · RCT: MRI vs MRI+18F-PSMA biopsy in equivocal MRIrecruiting An Investigational Scan (18F-rhPSMA-7.3 PET-mpMRI) for Targeted Prostate Biopsy Using TRUS-MR Fusion Technique Phase 2n=90 · primary completion 2028-12 · 18F-rhPSMA-7.3 PET-mpMRI targeted biopsy
- Cost-effectiveness of PET triage vs systematic biopsy recruiting Dutch National Randomized Study: PSMA-PET/CT As a Triage Tool for Pelvic Lymph Node Dissection in Prostatectomy Patients Phase NAn=706 · primary completion 2025-07 · randomised PSMA-PET triage, endpoint incl. costsactive PSMA PET/CT Guided Intensification of Therapy in Patients at Risk of Advanced Prostate Cancer Phase 3n=800 · primary completion 2029-01 · phase 3 PSMA-PET guidance, cost-effectiveness EP
📚 Sources · 📄 1 paper
The longer read
The question PRIMARY2 asks is narrower than "does PSMA-PET diagnose prostate cancer", and that narrowness is what makes the result usable. The enrolled population is defined by a negative or equivocal MRI plus a clinical reason not to trust it, exactly the group where the MRI-first pathway established by PRECISION leaves the clinician without a decision rule. Median PSA was 5.2 ng/mL and median PSA density 0.13, and the control arm's yield of 16% csPCa against 32% clinically insignificant cancer describes the problem precisely: biopsy this group and you find twice as much disease you would rather not have found. Against that baseline, a triage test does not need to detect more cancer. It needs to detect the same cancer in fewer men, which is what the -3.7% difference and the 49% biopsy-avoidance rate together claim.
The methodological question is whether the non-inferiority claim survives its own missing data, and here the paper is more candid than most. Protocol biopsies were missing differentially, 35 (11%) in the control arm against 7 (2%) in the experimental arm, and the trial's prespecified worst-case analysis states that the lower bound crosses the 10% margin once four of those 35 unbiopsied controls truly had csPCa. Four of 35 is not an implausible number in a population with a 16% underlying rate. Two things push back. The covariate-adjusted analysis (8% vs 13%, difference -4.8%, 95% CI -9.5 to -0.1) and the per-protocol analysis (11% vs 18%, difference -7.0%, -13.3 to -0.7) both move the point estimate further in favour of the experimental arm rather than toward the margin, and both exclude it. Consistency across analyses that handle the missingness differently is the strongest internal evidence on offer, and it is reassuring without being conclusive.
The more durable uncertainty is the follow-up, not the analysis set. 159 men left the trial without histology on the strength of a PRIMARY score of 1 or 2, and the median follow-up is 6.0 months (IQR 5.1-7.0). A negative triage test is only as good as the surveillance that follows it, and at this maturity the trial cannot distinguish a true negative from a deferred diagnosis. That is not a design flaw, it is the reading order: the co-primary endpoints were always going to report first and the confirmatory PSA-surveillance data later. But a reader deciding whether to adopt this pathway is implicitly betting on data the trial has not yet published.
Transferability is the other constraint, and it is technical rather than statistical. Sites were accredited for camera performance and onsite tracer production by a national trials network before activation, scans were pelvic-only at 60-70 min post-injection, and the PRIMARY score was read by nuclear medicine physicians with dedicated prostate experience, with central-versus-local agreement of κ=0.81. That number is the honest ceiling on how well this travels: strong, from experienced readers, on one tracer. The trial's own interpretation calls for validation with other PSMA radiopharmaceuticals, which is the correct caveat given how much of the effect depends on the negative score meaning what the trial says it means.
For a radiation oncologist the trial changes the referral population rather than any treatment decision. Insignificant cancer diagnoses fell from 32% to 14% and no-cancer biopsies from 42% to 22%, while the GS 4+3 or higher yield was essentially unchanged (4% vs 5%). The men who arrive for a treatment discussion out of this pathway are therefore fewer and more concentrated toward disease worth treating, with no evidence that the higher-grade pool is being thinned.