3339 - Multicohort Validation of Genetic Variants Associated with Radiation-Induced Hematuria in Prostate Cancer Patients

Purpose/Objective(s):

Late radiation-induced bladder injury can significantly impact quality of life in prostate cancer survivors following radiotherapy (RT). Although previous genome-wide association studies (GWAS) identified single nucleotide polymorphisms (SNPs) linked to hematuria risk, independent validation and the implementation of these markers into predictive modeling are essential for clinical utility. This study validates key SNPs identified by the Radiogenomics Consortium (RGC), using combined data from the REQUITE and URWCI cohorts to refine hematuria risk prediction.

Materials/Methods:

Candidate SNPs, potentially associated with hematuria risk, were selected based on the RGC GWAS summary statistics (N = 3,988) using a nominal significance threshold of p < 5×10^-5. Following linkage disequilibrium (LD) clumping using PLINK v1.9, lead SNPs were then determined. Genome-wide SNP data from the prospective REQUITE and URWCI cohorts, comprising 2,034 prostate cancer patients treated with RT, was generated using Illumina arrays and imputed using the TOPMed server. Lead SNPs identified in the RGC study were extracted from the combined dataset. Hematuria was defined as a patient-reported grade =2 event (gross bleeding) assessed annually up to 5-years after RT. To develop the final prediction model, regularized Cox regression modeling was applied to the selected lead SNPs using 10-fold cross-validation. For biological analysis, lead SNPs were mapped to their proximal genes using the topr R package and gene ontology analysis was conducted utilizing the MetaCore database to identify relevant functional pathways. Unlike our prior analysis (Farazi et al., 2025), which employed a traditional approach to develop a polygenic risk score, the current study utilized a machine learning approach with cross-validation to mitigate overfitting.

Results:

Based on the RGC GWAS summary statistics, 67 lead SNPs were selected following LD clumping with a threshold of p < 5×10^-5 and r2 < 0.3. In total, 1,708 participants had evaluable hematuria grade information, with 155 total gross hematuria events. The regularized Cox regression model, incorporating 67 lead SNPs, yielded a C-index of 0.59 under 10-fold cross-validation, indicating a modest predictive signal. The bladder-V70Gy, heart disease, ACE inhibitor use, previous abdominal surgery, TURP, prostatectomy, and cohort (REQUITE/URWCI) were incorporated into the model, resulting in a C-index of 0.68. Gene ontology analysis highlighted metanephros morphogenesis, smooth muscle contraction, and neutrophil homeostasis, all with a false discovery rate of 6.9E-5.

Conclusion:

This study validated SNPs associated with late radiation-induced hematuria. While the genetic prediction model showed a modest signal, the incorporation of clinical variables yielded reasonable predictive performance. Notably, our biological analysis revealed plausible mechanisms underlying hematuria risk, which warrant further functional characterization.