Pulse pressure variation (PPV) has been proposed as a promising resuscitation goal, but its ability to predict fluid responsiveness has been questioned in various conditions. The purpose of this study was to assess the performance of PPV in predicting fluid responsiveness in experimental hemorrhagic shock with endotoxemia, while comparing it with goals determined by a conventional set of guidelines. Twenty-seven pigs were submitted to acute hemorrhagic shock with intravenous infusion of endotoxin and randomized to three groups: (i) control; (ii) conventional treatment with crystalloids to achieve and maintain central venous pressure (CVP) 12 to 15 mmHg, mean arterial pressure of 65 mmHg or greater, and SvO2 (mixed venous oxygen saturation) of 65% or greater; (iii) treatment to achieve and maintain PPV of 13% or less. Parametric data were analyzed by two-way analysis of variance and Tukey test and differences in crystalloid volumes by t test. Predictive values of variables regarding fluid responsiveness were evaluated by receiver operating characteristic curves and multiple logistic regression. Both treatments produced satisfactory hemodynamic recovery, without statistical differences in fluid administration (P = 0.066), but conventional treatment induced higher CVP (P = 0.001). Areas under receiver operating characteristic curves were larger for CVP (0.77; 95% confidence interval, 0.68-0.86) and PPV (0.74; 95% confidence interval, 0.65-0.83), and these variables were further selected by multiple logistic regression as independent predictors of responsiveness. Optimal PPV cutoff was 15%, with false-positive results involving mean pulmonary arterial pressure of 27 mmHg or greater. Acute resuscitation guided by PPV was comparable with the strategy guided by CVP, mean arterial pressure, and SvO2. Central venous pressure and PPV were individually limited but independently predictive of fluid responsiveness.