Association of the ZJU index with PSA-defined high-risk prostate cancer status and mortality: a population-based cohort study

Abstract

Prostate cancer (PCa) ranks among leading causes of male cancer mortality, but PSA screening lacks specificity. Metabolic dysregulation drives PCa pathogenesis; however, composite metabolic indices remain underexplored. The ZJU index—integrating BMI, fasting glucose, triglycerides, and liver enzymes—may enable novel risk stratification. We evaluated the ZJU index in the NHANES population cohort ( N = 3,482) and a biopsy-confirmed clinical PCa cohort ( N = 246) from The Second Affiliated Hospital of Nanchang University. Multivariable logistic regression, restricted cubic splines, Boruta feature selection, and Cox proportional hazards models assessed associations with PSA-defined high-risk status for PCa, PSA levels, and all-cause mortality. Elevated ZJU index independently predicted reduced PSA-defined High-risk PCa. In NHANES, each unit increment lowered high-risk PCa odds by 3% (OR 0.97, 95% CI 0.94–1.00, P = 0.049); however, Q4 versus Q1 showed no statistically significant association in the fully adjusted model (OR 0.63, 95% CI 0.35–1.14, P = 0.13). In the clinical cohort of PSA-selected biopsy patients, Q4 was associated with 93% lower odds of biopsy-confirmed PCa (OR 0.07, 95% CI 0.02–0.19, P < 0.001). ZJU inversely correlated with PSA (linear in NHANES; nonlinear clinically) and ranked highly in feature importance. Survival analysis revealed a nonlinear association with all-cause mortality (P-nonlinear < 0.001), with restricted cubic spline analysis suggesting a potential U-shaped pattern, though categorical analysis showed a significantly increased hazard only in Q4. The ZJU index is associated with PSA-defined high-risk prostate cancer status and biopsy-confirmed PCa among PSA-selected patients, and shows a nonlinear association with all-cause mortality, underscoring metabolic health in oncogenesis. Further validation is needed before incorporating it into screening frameworks.