- Poster presentation
- Open Access
Bronchoalveolar lavage cytology in the diagnosis of ventilator-associated pneumonia
© BioMed Central Ltd 2006
- Published: 21 March 2006
- Bronchoalveolar Lavage
- Nosocomial Infection
- Fluid Culture
- Intensive Care Patient
- Cytology Analysis
Ventilator-associated pneumonia (VAP) is a common nosocomial infection in intensive care patients. The most accurate diagnosis relies on the quantitative culture results of broncho-alveolar lavage (BAL) fluid. In our hospital, BAL fluid quantitative culture and cytology analysis are performed routinely in patients suspected of VAP. In this study, we assessed the value of the combination of various BAL fluid cytology findings for the diagnosis of VAP.
All BAL fluid samples obtained from patients suspected of VAP were collected. VAP was defined as BAL fluid cultures with ≥ 104 colony forming units/ml BAL fluid. BAL fluid cytology included the total cell count/ml, the differential cell count on 500 nucleated cells including the percentage of polymorpho-nuclear neutrophils and the percentage of infected cells containing phagocytised organisms (intracellular organisms [ICOs]). Areas under the curve (AUCs) of receiver operating characteristic curves were calculated and plotted for various cytologic parameters and their combinations.
Over a 61-month period (January 1999-February 2004), a total of 335 BAL fluid samples were included from 287 patients. In 126 (37.6%) patients VAP was confirmed by positive culture. Highest AUCs were as follows: total cell count: 0.65, percentage polymorphonuclear neutrophils: 0.71 and percentage ICO: 0.90. The combination of percentage ICO with any other cytological parameter did not increase the AUC. At a threshold of 2%, the percentage ICO had an 86.2% positive predictive value and a 88.1% negative predictive value for VAP.
Among all cytologic parameters analysed in BAL fluid obtained under the suspicion of VAP, the percentage of ICOs was the parameter most predictive of VAP. Combination with other cytological parameters did not improve its predictive value.