Continuous monitoring of relative blood volume allows real-time assessment of intradialytic hypotension risk.

BACKGROUND

Intradialytic hypotension (IDH) is a frequent complication of hemodialysis, associated with patient discomfort, end-organ injury, and higher mortality. Prediction remains challenging. The value of relative blood volume (RBV) monitoring is debated. We investigated whether RBV indicates IDH risk and enables real-time prediction.

METHODS

In this prospective, multicenter study, 56 patients across 459 hemodialysis sessions were monitored with RBV sensors. IDH was defined as a systolic blood pressure <90 mmHg with a ≥20 mmHg decrease from baseline. RBV trajectories were analyzed using unsupervised clustering to identify risk patterns, and a time-varying RBV threshold was derived. Associations with IDH were tested using generalized linear mixed models (GLMMs) with patient as random effects. Predictive performance was evaluated using GLMMs and benchmarked against XGBoost, with validation performed session wise and patient wise.

RESULTS

IDH occurred in 29.7% of sessions and was associated with RBV decline. At the time of measurement, each 1% decrease in RBV increased IDH risk (OR = 1.05; P < .001). Falling below the dynamic RBV threshold doubled the odds of IDH (OR = 2.37; P < .001). In predictive analysis assessing IDH occurrence within the subsequent 10-60 minutes, GLMMs showed good discrimination in session-wise validation (AUC = 0.77) but were less accurate in patient-wise validation (AUC = 0.62). Incorporating patient-specific features improved patient-wise performance (AUC = 0.85 for GLMM; 0.86 for XGBoost).

CONCLUSIONS

RBV decline is a robust predictor of IDH. Including RBV in real-time risk models may help personalize ultrafiltration strategies and improve dialysis safety.