Predictive value of antithrombin III and serum C-reactive protein concentration in critically ill patients with suspected sepsis

Objective. To evaluate at admission the performance of serum antithrombin III, serum C-reactive protein, white blood cell and platelet counts, and thromboplastin time values in prediction of hospital mortality rates in critically ill patients with suspected sepsis. Design. Prospective, cohort study. Setting. University hospital medical-surgical intensive care unit. Patients. One hundred eight consecutive critically ill patients with suspected sepsis. Interventions. None. Measurements and Main Results., The outcome measure was hospital mortality rate. Hospital survivors (n = 66) and nonsurvivors (n = 42) differed statistically significantly in admission antithrombin III activity (percentage of normal): survivors' median 66% (interquartile range, 48% to 82%) vs. nonsurvivors' median 46% (37% to 65%, p =.0002 by Mann-Whitney test). Analysis revealed similarly statistically significant differences between survivors and nonsurvivors in admission platelet count, admission thromboplastin time, day 1 Logistic Organ Dysfunction score, and Acute Physiology and Chronic Health Evaluation III score, but not in serum C-reactive protein concentrations or in white blood cells. However, the areas under the receiver operating curves (AUC) showed significantly worse discriminative power for admission antithrombin III concentration (AUC, 0.71; SE, 0.05), platelet count (AUC, 0.67; SE, 0.05), thromboplastin time (AUC, 0.65; SE, 0.05), C-reactive protein concentration (AUC, 0.60; SE, 0.05), and white blood cell count (AUC, 0.53; SE, 0.06) than did the day 1 Logistic Organ Dysfunction score (AUC, 0.82; SE, 0.04) and the Acute Physiology and Chronic Health Evaluation III score (AUC, 0.84; SE, 0.04). Multivariate logistic regression analysis revealed that only the Acute Physiology and Chronic Health Evaluation III score was independently associated with hospital mortality rate. Conclusions. Admission antithrombin III concentrations, but not C-reactive protein concentrations, differ significantly between hospital survivors and nonsurvivors among critically ill patients with septic infection. However, in prediction of hospital mortality rate, the discriminative power of admission antithrombin III concentration is poor, as judged by analysis of areas under the receiver operating curves, and is not independently associated with hospital mortality rate.