Early diagnosis of ventilator-associated pneumonia. Is it possible to define a cutoff value of infected cells in BAL fluid?

To assess the usefulness of quantification of infected cells (ICs) in BAL fluid for the diagnosis of ventilator-associated pneumonia (VAP). A prospective study. A medico-surgical ICU in a tertiary health-care institution. One hundred thirty-two patients (mean age, 52 +/- 19 years). The suspicion of nosocomial pneumonia was strong in these patients: all had fever (> or = 38.5 degrees C), purulent tracheal aspirates, leukocytosis (> or = 10,000 cells per cubic millimeter), and new or persistent radiographic lung infiltrates. One hundred sixty-three samples (BAL and protected specimen brushes [PSB]) were obtained. VAP was present in 56 cases. The diagnosis was excluded in the remaining 107 cases. The IC count was performed on 100 cells in BAL fluid. The percentage of IC was significantly higher (12.6 +/- 12.4 vs 1.14 +/- 3.39; p < 0.0001) in patients with pneumonia: the area under the receiver operating characteristic (ROC) curve was 0.888 and a threshold of 2% of IC corresponded to a sensitivity of 84%, a specificity of 80%, a positive predictive value of 69%, and a negative predictive value of 90%. It is possible to define a threshold of IC in BAL fluid with a good reliability by using an ROC curve. This technique is useful for the early diagnosis (< 2 h) of nosocomial bacterial pneumonia in mechanically ventilated patients and allows a rapid and appropriate treatment of most of the patients with suspected VAP.