Background: The thresholds of the diagnostic procedures performed to d
iagnose ICU-acquired pneumonia (ZAP) are either speculated or incomple
tely tested. Purpose: To evaluate the best threshold of protected spec
imen brush (PSB), plugged telescoping catheter (PTC), BAL culture (BAL
C), and direct examination of cytocentrifugated lavage fluid (BAL D)
to diagnose ZAP. Each mechanically ventilated patient with suspected I
AP underwent bronchoscopy successively with PSB, PTC, and BAL in the l
ung segment identified radiographically. Population: One hundred twent
y-two episodes of suspected IAP (occurring in 26% of all mechanically
ventilated patients) were studied. Forty-five patients had definite ZA
P, and 58 had no ZAP. Diagnosis was uncertain in 19 cases. Results: Us
ing the classic thresholds, sensitivity was 67% for PSB, 54% for PTC,
59% for BAL D, and 77% for BAL C. Specificity was 88% for PSB, 77% for
PTC, 98% for BAL D, and 77% for BAL C. We used receiver operating cha
racteristics methods to reappraise thresholds. Decreasing the threshol
ds to 500 cfu/mZ for PSB, 10(2) cfu/mL for PTC, 2% cells containing ba
cteria for BAL D, 4x10(3) cfu/mL for BAL C increased the sensitivities
(plus 14%, 23%, 25%, 10%, respectively) and moderately decreased the
specificities (minus 4%, 9%, 2%, 4%, respectively) of the four examina
tions. The association of PSB with a 500 cfu/mL threshold and BAL D wi
th a 2% threshold recovered all but one episode of pneumonia (SE 96+/-
4%) with a 84+/-10% specificity. For a similar ICU population, these '
'best'' thresholds increased negative predictive value with a minimal
decrease of positive predictive value. They need to be confirmed in mu
ltiple ICU settings in prospective fashion.