Optimal threshold for diagnosis of ventilator-associated pneumonia using bronchoalveolar lavage

BACKGROUND Identification of ventilator-associated pneumonia (VAP) with invasive methods such as bronchoalveolar lavage (BAL) paired with treatment is associated with improved mortality. Inappropriate antibiotic use, however, is known to increase bacterial resistance, making future treatment problematic. Thus, the diagnostic threshold for VAP in BAL must yield adequate sensitivity while limiting exposure of patients to unnecessary antibiotics. Our institution uses a cutoff of > or = 10(5) colony-forming units (CFUs)/mL, but the optimal cutoff remains an area of debate. In this project, the effects of lower diagnostic cutoffs on VAP diagnosis and unnecessary antibiotic use are examined.

Records of all patients admitted to the trauma intensive care unit over a 2-year period requiring > 48 hours of mechanical ventilation were reviewed. Number of BALs, quantity of organism on each BAL, and presence of VAP (> or = 10(5) CFUs/mL) were noted. Indication for BAL was pulmonary infiltrate, sepsis syndrome, and C-reactive protein > 17 microg/dL at > or = 48 hours after admission.

From January 1, 2000, to December 31, 2001, 563 patients were admitted to the trauma intensive care unit. Two hundred fifty-seven required > 48 hours of mechanical ventilation, and 257 BALs were performed in 168 (65%) of these patients. One hundred thirty-nine episodes of VAP occurred in 109 (42%) patients. Subdiagnostic quantities of bacteria (> or = 10(2) but < 10(5) CFUs/mL) were seen in 98 BALs. Of these, only 16 (16%) episodes of VAP with the same organism were seen later during hospitalization. At a threshold of > or = 10(4) CFUs/mL, 4 of 28 (14%) patients went on to develop pneumonia. A similar pattern was seen at diagnostic thresholds of > or = 10(3) CFUs/mL (10 of 72 [14%]) and > or = 10(2) CFUs/mL (16 of 98 [16%]).

A threshold of > or = 10(5) CFUs/mL for VAP diagnosis carries a low false-negative rate. Over 80% of additional patients who would have been treated had a threshold of > or = 10(4) CFUs/mL been used recovered without treatment and thus would have undergone unnecessary antibiotic exposure. A similar pattern is seen at all lower thresholds. Lower diagnostic thresholds would lead to marginal increase in sensitivity, and many would receive unnecessary VAP treatment with potential for increasing bacterial resistance.