Predicting Fluid Responsiveness in Acute Liver Failure: A Prospective Study.

Background: The profound hemodynamic changes seen in acute liver failure (ALF) resemble the hyperdynamic state found in the later stages of septic shock. Vasopressor support frequently is required after initial volume therapy. Markers of preload dependency have not been studied in this patient group. Dynamic maneuvers such as passive leg raising or end-expiratory hold, which have shown good predictive accuracy in a general intensive care unit population, cannot be considered safe in this cohort because of the concerns of intracranial hypertension.

Methods: Mechanically ventilated patients with ALF admitted to a tertiary specialist intensive care unit in shock and multiorgan failure were enrolled. Markers of fluid responsiveness derived from transpulmonary thermodilution, pulse contour analysis, and echocardiography were compared between responders (cardiac index ≥15%) and nonresponders to a colloid fluid challenge (5 mL/kg predicted body weight). The ability to predict fluid responsiveness of stroke volume variation, pulse pressure variation (PPV), and respiratory change in peak (delta V peak) left ventricular outflow tract velocity for preload dependency were analyzed.

Results: Thirty-five patients (mean ± SD age, 38 [14] years, 13 male, 22 female]) were assessed after a single fluid challenge. Ten patients (29%) were fluid responders. Changes in cardiac index and stroke volume index in the cohort of 35 patients were correlated (R = 0.726 [99% confidence interval, 0.401-0.910]; P < .001). PPV predicted fluid responsiveness (area under the receiver operating characteristic curve [AUROC], 0.752 [95% confidence interval, 0.565-0.889]; P = .005; cutoff >9%). The AUROC for stroke volume variation was 0.678 ([95% confidence interval, 0.499-0.825]; P = .084; cutoff >11%). The AUROC for [delta] V peak before fluid bolus was 0.637 (95% confidence interval, 0.413-0.825; P = .322).

Conclusions: PPV based on pulse contour analysis predicted fluid responsiveness in ALF.