Dialysis vintage modifies the effect of adsorption-based therapies on protein-bound toxin clearance

The efficiency of hemodialysis varies substantially across different solute classes, but the determinants of this variability remain poorly characterized, limiting personalized prescription.

In this prospective randomized study, 60 maintenance hemodialysis patients were assigned to three modalities: hemodiafiltration (HDF), hemodialysis with HA130 hemoperfusion, and hemodialysis with KHA130 hemoperfusion. Clearance of p-cresyl sulfate (PCS), β2-microglobulin (β2-MG), advanced glycation end products (AGEs), and indoxyl sulfate (IS) was assessed. Multivariable linear regression, interaction analysis, subgroup analysis stratified by dialysis vintage tertiles, nonlinear quadratic regression, and machine learning (LASSO and Elastic Net) were applied.

HD + KHA130 significantly improved PCS clearance compared with HDF (β = 45.26, 95% CI: 6.26-84.27, P = 0.023), an effect robust to multivariable adjustment. Significant interactions between treatment modality and dialysis vintage were observed for PCS (P = 0.008), β2-MG (P = 0.002), and IS (P = 0.044). Exploratory subgroup analyses suggested that adsorption may benefit β2-MG and IS more in short dialysis vintage, whereas HD + KHA130 appeared more advantageous for PCS in long vintage patients. The relationship between dialysis vintage and PCS clearance followed a U-shaped pattern, with a nadir at approximately 60 months (bootstrap median; 95% CI: 9.6-218.1; P = 0.038). Machine learning identified dialysis vintage, age, treatment method, and CRP as the most stable predictors of PCS clearance.

HD + KHA130 significantly enhances PCS clearance, and may be most beneficial in patients with longer dialysis vintage. These findings support personalized treatment selection based on both solute type and dialysis vintage.