To determine the value of procalcitonin (PCT) as a marker of postoperative infection after cardiac surgery. A prospective single institution three phase study. University cardiac surgical intensive care unit (31 beds). Phase 1: To determine the normal perioperative kinetics of PCT, 20 consecutive patients undergoing elective cardiac surgery with cardiopulmonary bypass were included. Phase 2: To determine whether PCT may be useful for diagnosis of postoperative infection, 97 consecutive patients with suspected infection were included. Phase 3: To determine the ability of PCT to differentiate patients with septic shock from those with cardiogenic shock, 26 patients with postoperative circulatory failure were compared. Phase 1: Serum samples were drawn for PCT determination after induction of anesthesia (baseline), at the end of surgery, and daily until postoperative day (POD) 8. Baseline serum PCT concentration was 0.17 +/- 0.08 ng/mL (mean +/- SD). Serum PCT increased after cardiac surgery with a peak on POD 1 (1.08 +/- 1.36). Serum PCT returned to normal range on POD 3 and remained stable thereafter. Phase 2: In patients with suspected infection, serum PCT was measured at the same time of C-reactive protein (CRP) and bacteriologic samples. Among the 97 included patients, 54 were infected with pneumonia (n = 17), bacteremia (n = 16), mediastinitis (n = 9), or septic shock (n = 12). In the 43 remaining patients, infection was excluded by microbiological examinations. In noninfected patients, serum PCT concentration was 0.41 +/- 0.36 ng/mL (range, 0.08-1.67 ng/mL). Serum PCT concentration was markedly higher in patients with septic shock (96.98 +/- 119.61 ng/mL). Moderate increase in serum PCT concentration occurred during pneumonia (4.85 +/-3.31 ng/mL) and bacteremia (3.57 +/- 2.98 ng/mL). Serum PCT concentration remained low during mediastinitis (0.80 +/- 0.58 ng/mL). Five patients with mediastinitis, two patients with bacteremia, and one patient with pneumonia had serum PCT concentrations of <1 ng/mL. These eight patients were administered antibiotics previously and serum PCT was measured during a therapeutic antibiotic window. For prediction of infection by PCT, the best cutoff value was 1 ng/mL, with sensitivity 85%, specificity 95%, positive predictive value 96%, and negative predictive value 84%. Serum CRP was high in all patients without intergroup difference. For prediction of infection by CRP, a value of 50 mg/L was sensitive (84%) but poorly specific (40%). Comparing the area under the receiver operating characteristic curves, PCT was better than CRP for diagnosis of postoperative sepsis (0.82 for PCT vs. 0.68 for CRP). Phase 3: Serum PCT concentration was significantly higher in patients with septic shock than in those with cardiogenic shock (96.98 +/- 119.61 ng/mL vs. 11.30 +/- 12.3 ng/mL). For discrimination between septic and cardiogenic shock, the best cutoff value was 10 ng/mL, with sensitivity of 100% and specificity of 62%. Cardiac surgery with cardiopulmonary bypass influences serum PCT concentration with a peak on POD 1. In the presence of fever, PCT is a reliable marker for diagnosis of infection after cardiac surgery, except in patients who previously received antibiotics. PCT was more relevant than CRP for diagnosis of postoperative infection. During a postoperative circulatory failure, a serum PCT concentration >10 ng/mL is highly indicative of a septic shock.