Validation of Modelflow Estimates of Cardiac Output in Hemodialysis Patients.

ABSTRACT

Volume-clamp technology (e.g. Finometer) has become a popular method of collecting continuous, non-invasive, hemodynamic information during hemodialysis. There is minimal data validating the technique in this patient group. A gold standard cardiac output measurement can be obtained using ultrasound dilution in patients with arterio-venous fistulae. Continuous cardiac output was measured in 124 hemodialysis sessions in 27 patients using a volume-clamp device (Finometer PRO). Ultrasound dilution measurement was first taken at baseline (Transonic HD03), then used to calibrate the Finometer. Ultrasound dilution measurement was repeated 2 h into hemodialysis to assess drift following calibration. Pearson's correlation and Bland-Altman statistics, modified for repeated measures, were used to assess agreement between methods. Linear mixed models were constructed to identify factors that could explain session-level and patient-level variation in agreement. For baseline cardiac output before calibration, agreement between volume-clamp and ultrasound dilution measurements was poor, at 25 ± 75% (correlation 0.26, P < 0.001). There was significant variation in agreement between patients, with age, peripheral vascular disease and hemodialysis vintage contributing to poorer agreement. For cardiac output 2 h after calibration, agreement was -5.2 ± 57.5% (correlation 0.6, P < 0.001). Dynamic changes in blood pressure and fluid balance during hemodialysis resulted in greater drift over time after calibration. There was a large error, both random and systematic, in volume-clamp estimates of absolute, pre-calibration cardiac output in this prevalent hemodialysis population. There was minimal bias and reasonable correlation for cardiac output 2 h post-calibration, but limits of agreement remained too wide to meet current clinical standards.