Objective: To determine the use of plasma C-reactive protein (CRP) concentrations, body temperature (BT) and white blood cell count (WBC) in the detection of sepsis in critically ill patients. Design: All patients admitted for more than 24 h in the intensive care unit (ICU) were prospectively included. Patients were followed up to ICU discharge and each patient-day was classified in one of four categories according to the infectious status: 1) Negative, patient-day without systemic inflammatory response syndrome (SIRS); 2) Definite, patient-day with SIRS and a positive culture; 3) SIRS, patient-day with SIRS and negative or no cultures. The last group was subdivided according to the following criteria: a) new, or persistence of, pulmonary infiltrates, b) the presence of pus in a place known to be sterile. Patient-days without these criteria were classified as SIRS with improbable sepsis (Unlikely), and with one criteria or more as SIRS with probable sepsis (Probable). Setting: Medical/surgical intensive care unit. Patients: Twenty-three patients were followed. Measurements and results: A total of 306 patient-days were analysed: 20 Negative, 15 Definite, 63 Unlikely and 208 Probable. The median (range) CRP values for Negative, Unlikely, Probable and Definite groups were as follows: 24.5 (7–86), 34 (5–107), 143 (39–544), and 148 (52–320) mg/l. The plasma CRP levels were significantly related to the infectious status (Negative, Unlikely, Probable or Definite) of the patient-day classification (p < 0.05). Concentrations of CRP in the Negative and Unlikely groups were significantly lower than in the Probable and Definite ones (p < 0.05). A plasma CRP of 50 mg/l or more was highly suggestive of sepsis (sensitivity 98.5 %, specificity 75 %). Conclusions: Daily measurement of CRP is useful in the detection of sepsis and it is more sensitive than the currently used markers, such as BT and WBC.