In critically ill patients, fiberoptic bronchoscopy is well tolerated as long as usual contraindications are respected (1). In our study, no severe deleterious effects occurred. Different diagnostic techniques, including nondirected mini-bronchoalveolar lavage, have been tried, with apparently acceptable results. These "intermediate" techniques may reduce costs and simplify the procedure; unfortunately, in the absence of clear confirmation of preliminary findings, they cannot be recommended as the cornerstones of a diagnostic strategy. We agree with Dr. Johnson that semi-quantitative cultures can be considered an acceptable alternative microbiological technique to quantitative cultures for distinguishing colonizing pathogens from infecting pathogens. The approach tested by Singh and associates on a subgroup of patients with a low probability of ventilator-associated pneumonia (2), which was based on periodic systematic clinical reevaluation, merits prospective analysis in a large study sample.

We cannot agree with Ewig and colleagues' comments. First, we compared a strategy that used specimens obtained by bronchoscopic protected specimen brush or bronchoalveolar lavage from affected areas for quantitative cultures with a strategy based on clinical evaluation and qualitative cultures of tracheal aspirates. The latter clinical approach is applied routinely in most intensive care units. The two strategies could be compared only by using strict and widely accepted criteria for defining clinical suspicion of ventilator-associated pneumonia. The criteria used in our study have been repeatedly reported in the literature.

Second, concerning mortality rates, great care must be taken to compare data that are comparable. It is true that mortality for the invasive group was 16.2% at 14 days and 30.9% at 28 days. These results are concordant with epidemiologic data published previously by our group and others (3). Mortality rates observed for both groups actually concerned the entire studied sample, including patients with true pneumonia and those with clinically suspected but not bacteriologically proven pneumonia. These two situations are associated with different mortality rates (3). Third, the frequency (11.5%) of inadequate initial antimicrobial therapy is in fact extremely low in the clinical management group compared with that reported in other studies (4), probably because of the therapeutic regimens recommended in the protocol. Given our definition of inappropriate treatment—"antibiotics to which at least one cultured isolate was resistant in vitro"—and the number of microorganisms identified in both groups, this rate is easily explained. It is undeniably a real challenge for a strategy to be effective against the 312 pathogens isolated from the 179 patients in the noninvasive group compared with only 121 pathogens from the 90 patients in the invasive group.