Usefulness of a strategy based on bronchoscopy with direct examination of bronchoalveolar lavage fluid in the initial antibiotic therapy of suspected ventilator-associated pneumonia

Abstract

Objectives: To evaluate (a) the routine accuracy of bronchoalveolar lavage by direct examination (BAL-D) in diagnosing ventilator-associated pneumonia (VAP), and (b) the impact of a diagnostic strategy including clinical judgment, bronchoscopy, and BAL-D on the initial diagnosis and appropriateness of treatment when VAP is suspected.

Design and setting: Prospective cohort study in two academic ICUs in Paris, France.

Patients and participants: Mechanically ventilated patients with suspected VAP underwent bronchoscopy with BAL and protected specimen brush (PSB). BAL-D results were available within 2 h, BAL on culture and PSB results after 24 h, and antibiotic susceptibility after 48 h. At each step in the strategy the senior and the resident in charge of the patient were asked their diagnosis and their therapeutic plan on the basis of presently available data. Definite diagnosis of suspected VAP was based on histology, appearance of cavitation, positive pleural fluid culture, results of PSB and BAL culture, and follow-up.

Measurement and results: A total of 110 episodes of suspected VAP were studied; 94 definite diagnoses were made (47 VAP, 47 no VAP). Using a threshold 1% of infected cells, BAL-D discriminated well between patients with and those without VAP (sensitivity 93.6%, specificity 91.5%, area under the receiver-operating characteristic curve 0.953). The senior clinical judgment was correct in 71% cases. It was correct in 78% and 94% of cases after airway visualization and BAL-D findings, respectively. After BAL-D the positive and negative predictive values in diagnosing VAP were 90% and 98%, respectively. However, the therapeutic plan was correct in only 65% using clinical judgment (15 untreated patients, 3 ineffective treatment, 15 useless treatment), 66% using airway visualization (14 untreated VAP, 4 ineffective treatment, 14 useless treatment), and 88% using BAL-D results (1 untreated patients, 6 ineffective, 4 useless), according to definite diagnosis and final antibiotic susceptibility testings.

Conclusions: A strategy based on bronchoscopy and BAL-D generally leads to a rapid and appropriate treatment of nosocomial pneumonia in ventilated patients.