U-shaped association of time-averaged cumulative diastolic blood pressure slope with mortality in hemodialysis patients

BACKGROUND: Blood pressure (BP) is an important modifiable prognostic factor in maintenance hemodialysis patients. We previously introduced intradialytic time-averaged cumulative blood pressure (TAC-BP), calculated from all BP measurements obtained during each dialysis session using the trapezoidal rule. We hypothesized that the longitudinal slope of session-level TAC-BP values (TAC-BP Slope) is associated with all-cause mortality. METHODS: This single-center retrospective cohort study included 209 maintenance hemodialysis patients followed from August 2018 to March 2025. Four longitudinal BP slope indicators were calculated: TAC-systolic BP (SBP) Slope, TAC-diastolic BP (DBP) Slope, Predialysis SBP Slope, and Predialysis DBP Slope. Multivariable Cox regression was performed. Non-linear associations were examined using tertile analysis. Incremental discrimination was assessed using bootstrap mean C-index comparison. RESULTS: Over a mean follow-up of 2.8 years, 75 patients (35.9%) died. In linear Cox models, all four indicators were significantly associated with mortality, with TAC-DBP Slope showing the strongest linear association (HR 1.767, 95% CI 1.404-2.224). For TAC-DBP Slope, both declining (HR 4.21, 95% CI 2.28-7.75) and increasing (HR 2.81, 95% CI 1.44-5.50) tertiles showed significantly elevated mortality versus stable tertile, indicating a U-shaped association. In sensitivity analyses excluding measurements from the 6 months preceding death or censoring, the association in the declining tertile was attenuated and no longer significant (HR 1.44, 95% CI 0.73-2.84), whereas the increasing tertile remained significant (HR 2.07, 95% CI 1.06-4.06). C-index analysis demonstrated that TAC-DBP Slope deviation provided significantly improved internal discrimination beyond established and predialysis variables (ΔC = +0.048, p=0.010). CONCLUSIONS: Longitudinal TAC-DBP Slope shows a U-shaped association with all-cause mortality, but the two arms differ in origin: the declining arm largely reflects terminal decline preceding death, whereas the increasing arm persists after exclusion of terminal measurements. TAC-DBP Slope may therefore be useful for risk stratification rather than as a validated prediction tool.