BACKGROUND Mediastinal bleeding is common after pediatric cardiopulmonary bypass (CPB) surgery. Thromboelastography (TEG) may predict bleeding and provide insight into likely mechanisms. We aimed to (a) compare perioperative temporal profiles of TEG and laboratory hemostatic variables between patients with significant hemorrhage (BLEED) and those without (CONTROL), (b) investigate the relationship between TEG variables and routine hemostatic variables, and (c) develop a model for prediction of bleeding. METHODS TEG and laboratory hemostatic variables were measured prospectively at 8 predefined times for 50 children weighing <20 kg undergoing CPB. RESULTS Patients who bled demonstrated different TEG profiles than those who did not. This was most apparent after protamine administration and was partly attributable to inadequate heparin reversal, but was also associated with a significantly lower nadir in mean (sd) fibrinogen for the BLEED group compared with CONTROL group: 0.44 (0.18) and 0.71 (0.40) g/L, respectively (P = 0.01). Significant nonlinear relationships were found between the majority of TEG and laboratory hemostatic variables. The strongest relationship was between the maximal amplitude and the platelet-fibrinogen product (logarithmic r(2) = 0.71). Clot strength decreased rapidly when (a) fibrinogen concentration was <1 g/L, (b) platelets were <120 x 10(9)/L, and (c) platelet-fibrinogen product was <100. A 2-variable model including the activated partial thromboplastin time at induction of anesthesia and TEG mean amplitude postprotamine discriminated well for subsequent bleeding (C statistic 0.859). CONCLUSIONS Hypofibrinogenemia and inadequate heparin reversal are 2 important factors contributing to clot strength and perioperative hemorrhage after pediatric CPB. TEG may be a useful tool for predicting and guiding early treatment of mediastinal bleeding in this group.