Real-time Fluorescence Microscopy for Intraoperative Assessment of Prostate Biopsy Cores (FLASH)

FLASH is an investigator-initiated, international, multicentre, prospective diagnostic-accuracy and clinical-utility study with a within-patient paired design, conducted and reported in accordance with STARD 2015. Biopsy-naïve men with suspected prostate cancer (serum PSA >4 ng/mL and/or abnormal digital rectal examination) and at least one MRI-visible lesion (PI-RADS v2.1 score 3-5) will undergo standard-of-care combined targeted and systematic prostate biopsy.

Immediately after extraction, every core will undergo ex vivo dual staining (fluorescein sodium for 30 seconds followed by methylene blue for 30 seconds) and contact-mode scanning with the EndoScell real-time fluorescence microscope (model MES-1000P), completed within 5 minutes per core by certified technicians. Each core is assigned to a three-tier category (0 = negative/suspected low-grade; 1 = suspected clinically significant prostate cancer; 2 = equivocal). The interval from core extraction to formalin fixation will not exceed 15 minutes. All cores then undergo routine formalin-fixed paraffin-embedded (FFPE) histopathology with haematoxylin and eosin staining, with immunohistochemistry as clinically indicated, reported by senior uropathologists.

Bidirectional blinding is enforced throughout: technicians are blinded to all clinical data, MRI findings, and core origin; device results are withheld from the treating team and never enter the medical record; uropathologists are blinded to all device data; the statistician works with coded variables until database lock. Central pathology review is performed for discordant and equivocal cases.

The primary outcome is patient-level diagnostic accuracy (sensitivity, specificity, positive and negative predictive values with exact 95% confidence intervals) of the device for clinically significant prostate cancer (ISUP grade group >=2) against FFPE histopathology in the intention-to-diagnose population. Secondary outcomes include core-level concordance with cluster adjustment, inter-reader reproducibility, tissue-integrity assessment, operational metrics, and a prespecified decision-analytic simulation of a "Positive-then-Stop" biopsy-sparing pathway, which is hypothesis-generating and does not alter patient care. The accuracy and utility estimates are intended to inform the design of a future randomised de-implementation trial.

• Ex vivo real-time fluorescence microscopy of prostate biopsy cores (EndoScell system, model MES-1000P) (Diagnostic test): Fresh biopsy cores are stained with fluorescein sodium (30 s) and methylene blue (30 s), scanned in contact mode within 5 minutes per core, and categorised on a three-tier scale; the tissue then proceeds to routine FFPE histopathology unchanged. The test is performed on excised tissue only and does not influence clinical management.

Primary outcomes

• Patient-level sensitivity of the EndoScell system for clinically significant prostate cancer [Index test on Day 0; reference standard completed Day 7-14]: Proportion of reference-positive participants (any core with ISUP grade group >=2 on FFPE histopathology) with at least one core read as Category 1 (suspected clinically significant prostate cancer) on the index test; reported with exact (Clopper-Pearson) 95% confidence interval in the intention-to-diagnose population.
• Patient-level negative predictive value (NPV) of the EndoScell system [Index test on Day 0; reference standard completed Day 7-14]: Proportion of participants with no Category 1 core who are reference-negative; exact 95% CI.
• Patient-level positive predictive value (PPV) of the EndoScell system [Index test on Day 0; reference standard completed Day 7-14]: Proportion of participants with at least one Category 1 core who are reference-positive (ISUP grade group >=2 on FFPE); exact 95% CI.
• Patient-level specificity of the EndoScell system for clinically significant prostate cancer [Index test on Day 0; reference standard completed Day 7-14]: Proportion of reference-negative participants with no core read as Category 1; equivocal (Category 2) readings are grouped with test-negative in the primary analysis, with worst-case/best-case/exclusion sensitivity analyses prespecified; exact 95% CI.

Secondary outcomes

• Core-level diagnostic concordance [Index test on Day 0; reference standard completed Day 7-14]: Per-core sensitivity, specificity, PPV and NPV of the index test against FFPE histopathology, overall and stratified by core origin (targeted vs systematic), with cluster-robust 95% CIs from a generalised estimating equation (logit link, exchangeable correlation) accounting for within-patient clustering.
• Biopsy-sparing rate of the simulated "Positive-then-Stop" pathway [Simulation performed after database lock (estimated December 2028)]: Proportion of participants in whom systematic biopsy would have been omitted (any targeted core read as Category 1), and the corresponding proportion of systematic cores avoided; prespecified deterministic simulation, exact 95% CIs. Hypothesis-generating; does not alter patient care.
• Overdiagnosis reduction in the simulated pathway [Simulation performed after database lock (estimated December 2028)]: Number and proportion of ISUP grade group 1 diagnoses that would have been averted by omitting systematic biopsy after a positive intraoperative reading; exact 95% CI.
• Clinically significant prostate cancer miss rate in the simulated pathway [Simulation performed after database lock (estimated December 2028)]: Number and proportion of reference-positive participants who would have been missed under the simulated pathway (targeted cores negative on both index test and FFPE, with csPCa found only in systematic cores); exact 95% CI.
• Net benefit of the simulated pathway (decision-curve analysis) [Simulation performed after database lock (estimated December 2028)]: Decision-curve analysis with patient-level bootstrap (1000 resamples) comparing three strategies (combined biopsy for all; targeted biopsy alone; Positive-then-Stop) across threshold probabilities.
• Inter-reader agreement of index-test readings [Video re-reading after completion of enrolment (estimated 2028)]: Cohen's kappa (three-tier scale and dichotomised Category 1 vs 0/2) and percent agreement between two blinded technicians re-interpreting a random 20% subset of archived scan videos (\~900-1200 cores).
• Tissue-integrity (tissue-compromise) rate [Day 7-14 (pathology reporting)]: Proportion of cores flagged by uropathologists as degraded, unreadable, or inadequate for definitive H\&E/IHC evaluation attributable to the staining/scanning procedure; one-sided upper 95% confidence bound compared against the 1% acceptability benchmark.
• Operational performance of the index test [Day 0]: Per-core scanning time, extraction-to-fixation interval, equivocal (Category 2) rate, and device/scan failure rate, summarised descriptively by centre and calendar quarter.

Sex: Male, ages: 40 Years to any

Inclusion Criteria:

1. Age >= 40 years;
2. Serum total PSA > 4.0 ng/mL and/or abnormal digital rectal examination;
3. Biopsy-naïve;
4. Pre-biopsy 3.0-T multiparametric MRI demonstrating at least one suspicious lesion (PI-RADS v2.1 score 3-5);
5. Fit for transperineal or transrectal prostate biopsy under local, regional, or general anaesthesia;
6. Written informed consent.

Exclusion Criteria:

1. Prior treatment for prostate cancer (surgery, radiotherapy, focal therapy, or androgen-deprivation therapy), or use of 5-alpha-reductase inhibitors within the preceding 6 months;
2. Acute prostatitis or active urinary tract infection;
3. Contraindications to MRI (e.g., incompatible implants, severe claustrophobia);
4. Severe coagulation disorders or uncorrectable bleeding tendency;
5. History of major anorectal surgery precluding safe insertion of the endorectal ultrasound probe (where transrectal guidance is used);
6. Any other condition that, in the investigator's judgement, makes the participant unsuitable for the trial.

• haifeng wang, Study chair, Changhai Hospital