Plasma proteome-based integrative prognostic nomogram in metastatic castrate-resistant prostate cancer (mCRPC).

225 Background: Currently non-specific clinical factors including non-tumor biology proteins (Alkaline Phosphatase-ALK, LDH, albumin) are used for identifying high, intermediate or low prognostic survival risk in mCRPC. We integrated these biomarkers with plasma proteins overexpressed exclusively in mCRPC to develop patient-centric nomograms for estimating 1,2 and 3-year (yr) overall survival (OS) probabilities. Methods: Proteomic profiling of plasma for 3072 proteins using Olink Explore NGS platform was performed in 32 local stage prostate cancer, 123 metastatic hormone-sensitive (mHSPC) and 157 mCRPC state samples (Total N=312) in a clinically annotated real-world cohort study that enrolled patients (pts) between 2020 and 2024. Protein assays were measured in normalized protein expression (NPX) units through a normalization process. We compared NPX values of all protein assays across all three cancer states to identify Differentially Expressed Proteins (DEPs) exclusively overexpressed in mCRPC using t-test with multiple corrections. Clinical outcome in the mCRPC was OS. A composite score was generated aggregating DEPs overexpressed in mCRPC with OS, by multiplying each individual DEP’s univariate-level survival Hazard Ratio (HR) coefficient with their corresponding NPX value. The composite score range for the mCRPC cohort (N=121 pts) was dichotomized above and below the cohort median as “High” and “Low” and evaluated for survival using Cox Proportional Hazard Regression along with current prognostic protein biomarkers (ALK, LDH, albumin) at the univariate followed by multi-variable analysis (MVA) for variables observed with univariate significance (P<0.05). Prognostic nomograms integrating MVA significant clinical factors with the DEP composite score was developed and performance of the nomogram with and without the composite score for estimating 1,2 and 3-years survival probabilities was determined using ROC curves and Area Under Curve (AUC). All analyses were conducted in RStudio. Results: Median OS of the mCRPC cohort (N=121) was 28 months (range: 0.3-45). There were 353/3072 DEPs exclusive to mCRPC and 19/353 DEPs were associated with OS (univariate P<0.05) (SCRN1, GP2, NFU1, DPY30, SNRPB2, KRT19, FKBP5, PGD, IL6, L3HYPDH, MYDGF, DH1, IQGAP2, TOMM20, YAP1, CLGN, IMMT, PKD2, PALM2). MVA Hazard Ratios (HR) with 95% CIs for nomogram input included: DEP score HR-3.48 (1.77-6.83; P<0.0003); LDH: HR-1.01 (1-1; P=0.13); ALK: HR-1 (1-1; P=0.000035); Albumin: HR-0.61 (0.24-1.51; P=0.28). AUCs for estimating mCRPC OS probabilities integrating the DEP score with current factors versus clinical factors alone increased to 0.87 from 0.85 (1-yr); to 0.78 from 0.73(2-yr) and to 0.79 from 0.70 (3-yr). Conclusions: A patient-centric approach using nomograms to estimate survival in mCRPC which includes novel protein classifiers can enhance current prognostication models.