Predicting long-term metastatic risk after radical prostatectomy: Adding Genomic Prostate Score, grade group, and cribriform size to unfavorable histology burden translates into minimal incremental discrimination.

INTRODUCTION

Current risk stratification after radical prostatectomy (RP) relies on Grade Group (GG) and genomics, which may not fully capture metastatic potential. Unfavorable histology (UH), defined by adverse architectural patterns such as large cribriform carcinoma and intraductal carcinoma, has emerged as a strong predictor. We evaluated whether integrating genomic and other clinicopathologic variables improves prediction beyond quantitative UH burden.

METHODS

We analyzed 418 men from an event-enriched RP cohort (1987-2004) with centralized pathology review, Genomic Prostate Score (GPS), and long-term follow-up. We recorded the percentage of tumor with UH (UH%), cribriform size, GG, stage, margin status, prostate-specific antigen, and GPS. Multivariable Cox models and a sequential modeling framework evaluated independent associations and incremental predictive performance for metastasis at 15 years.

RESULTS

During median 15.5-year follow-up, 102 men (24%) developed metastases, all within UH-positive tumors. UH ≥10% was strongly associated with metastasis (hazard ratios [HR] 50.95, 95% confidence intervals [CI] 10.4-249). GPS (HR 1.03 per unit, 95% CI 1.01-1.05) and cribriform size (HR 1.0007 per μm, 95% CI 1.0003-1.0010) were independently associated but added only modest discrimination. UH% alone achieved high 15-year discrimination (c-index ∼0.85); inclusion of GPS and cribriform size increased this to 0.901. Limitations include retrospective design and single-institution pathology review.

CONCLUSION

UH% is the dominant determinant of metastatic risk after RP, offering greater prognostic discrimination than GG, stage, and other clinicopathologic variables. Genomic classifiers and cribriform size provide secondary refinement but do not materially alter risk once UH burden is established. These findings support pathology-anchored risk models.