Blood pressure variability and serum sodium for long-term mortality risk stratification in maintenance hemodialysis patients: a real-world retrospective cohort study

Background Patients receiving maintenance hemodialysis have a high risk of long-term mortality, but practical risk stratification using routinely available dialysis-unit markers remains challenging. Blood pressure variability (BPV) and serum sodium may reflect complementary dimensions of dialysis-related risk, including hemodynamic instability and electrolyte-volume vulnerability. This study evaluated the prognostic value of BPV and serum sodium for long-term mortality risk stratification in maintenance hemodialysis patients. Methods This real-world retrospective cohort study included 595 maintenance hemodialysis patients. BPV was calculated as the coefficient of variation of predialysis systolic blood pressure measurements during a 3-month baseline assessment period preceding mortality follow-up. The primary outcome was all-cause mortality occurring after completion of the baseline assessment. Cox proportional hazards regression was used to identify prognostic factors. Three nested Cox models were compared to evaluate the incremental prognostic value of BPV and serum sodium. A Cox-based nomogram was constructed and internally validated using 1,000 bootstrap resamples. Model performance was assessed using Harrell's C-index, time-dependent area under the curve (AUC), calibration curves, prediction error curves, and decision curve analysis. Restricted cubic spline analysis was used to assess the functional form of the association between serum sodium and all-cause mortality. Results During follow-up, 120 all-cause deaths occurred. After multivariable adjustment, older age, shorter dialysis vintage, higher BPV, and lower serum sodium were associated with all-cause mortality. Restricted cubic spline analysis showed a significant association between serum sodium and mortality ( P for overall association <0.001), with evidence of nonlinearity ( P for nonlinearity = 0.042). Incorporating BPV and RCS-modeled serum sodium progressively improved model performance, with Harrell's C-index increasing from 0.717 in Model 1 to 0.748 in Model 2 and 0.849 in the final Model 3. The final model yielded time-dependent AUCs of 0.869 and 0.853 for 12- and 36-month mortality prediction, respectively. Conclusions BPV and serum sodium provided complementary and incremental prognostic information beyond conventional clinical factors in maintenance hemodialysis patients. The internally validated Cox-based model incorporating these routinely available markers showed promising performance for long-term mortality risk stratification. External validation in independent cohorts and prospective evaluation are required before clinical implementation.