1164 - Deep Genomic and Pathologic Profiling Reveals Treatment Modality-Specific Predictors of Biochemical Recurrence In Localized Prostate Cancer

05:25pm - 05:30pm ET

Room 205

Presenter(s)

Ahmed El-Sehemy, MD, PhD, MS, BS - University of Toronto, Toronto, ON

A. El-Sehemy¹, X. Ci², M. Lam³, M. Ramotar⁴, A. Berlin⁵, T. van der Kwast², and H. H. He²; ¹University Health Network (UHN), University of Toronto, Toronto, ON, Canada, ²University Health Network (UHN), Toronto, ON, Canada, ³University of Toronto, Toronto, ON, Canada, ⁴Department of Radiation Oncology, Princess Margaret Cancer Centre, University of Toronto, Toronto, ON, Canada, ⁵Department of Radiation Oncology, University of Toronto, Toronto, ON, Canada

Purpose/Objective(s): Biochemical recurrence (BCR) after definitive therapy reflects heterogeneous tumor biology. We hypothesized that genomic instability, quantified by percent genome altered (PGA), preferentially predicts recurrence timing after image-guided radiation therapy (IGRT), whereas pathologic architectural features preferentially predict recurrence after radical prostatectomy (RadP), and that ERG fusion status and intraductal/cribriform carcinoma (IDC) define biologically distinct instability-driven subtypes.

Materials/Methods: Retrospective cohort of genomically profiled localized prostate cancers (N=424, median follow up ? 10 years) treated with IGRT or RadP. PGA was derived from copy-number profiling. Pathologic variables included IDC, extraprostatic extension (EPE), lymphovascular invasion (LVI), and perineural invasion (PNI). Primary endpoints were BCR occurrence and time to BCR. Secondary endpoint was metastatic progression. Spearman correlation assessed timing associations. Multivariable logistic and linear regression models adjusted for age and Gleason score. Interaction testing evaluated treatment-by-PGA effects.

Results: PGA predicted earlier BCR after IGRT (?=-0.41,p<0.001) but not after RadP (?=-0.22,p=0.07). Quartile analysis demonstrated marked separation: lowest PGA quartile median BCR 7.6 years versus 2.8 years in the highest quartile. Age predicted earlier BCR after IGRT (?=-0.25,p=0.030) but showed no effect in RadP.

IDC independently predicted BCR after RadP (39%vs26%,p=0.020, OR˜2.0,p=0.017) and predicted earlier recurrence (?=-0.26,p=0.036); 68% of early RadP recurrences (<2 years) were IDC+. EPE strongly predicted BCR occurrence (43%vs14%,p=0.0001) but not timing. LVI predicted BCR (67%vs20%,p=0.006) but not timing. PNI was the strongest metastatic predictor (18%vs2%,p=0.013). ERG-/IDC+ tumors exhibited highest genomic instability (mean PGA 10.4%vs3.4% ERG+/IDC-,p<0.001) and highest metastasis rate (22%vs11%). ERG frequency declined with age (67%<60y to 18%=75y), whereas PGA increased (3.5%to6.1%), independent of Gleason grade.

Table 1. Modality-Specific Predictors Of Recurrence

Conclusion: Recurrence biology diverges fundamentally by treatment modality. Genomic instability drives recurrence timing after radiation, whereas pathologic architecture and extent govern recurrence risk after surgery independent of genomic instability. ERG-negative IDC represents a genomically unstable, high-risk subtype. These findings support treatment-specific biological risk stratification and suggest genomic instability may guide treatment approach and intensification.

Variable	IGRT	RadP	Clinical impact
PGA	Timing (?=-0.41,p<0.001)	NS	7.6y vs 2.8y (Q1–Q4)
Age	Timing (?=-0.25,p=0.030)	NS	Earlier BCR onset in younger patients
IDC	NS	OR ˜2.0,p=0.017	39% vs 26% BCR rate
EPE	N/A	p=0.0001	Predictor of BCR
LVI	N/A	p=0.006	Predictor of BCR
PNI	N/A	Mets: 18% vs 2%, p=0.013	9× increased risk of metastasis