The overmortality induced by nosocomial infections, especially pneumonia in ventilated patients (VNP), is still a matter of controversy because it is difficult to know precisely the respective effects of VNP per se and both the underlying illness and the severity of the disease that indicates ICU stay. During a 3-yr period, for each patient mechanically ventilated for more than 48 h we recorded underlying illness, reason for mechanical ventilation, clinical and therapeutic data collected during the first 48 h of ventilation, and death in the ICU. Patients with suspicion of VNP (S-VNP) according to clinical, radiologic, and biologic criteria underwent bronchoscopy with protected specimen brush (PSB) and bronchoalveolar lavage culture (BAL-C). VNP was confirmed (C-VNP) if PSB > or = 10(3) cfu/ml and/or BAL-C > or = 10(4) cfu/ml. Prognostic multivariate analysis was performed introducing S-VNP and C-VNP as time-dependent covariates. Of the 387 studied patients, 112 S-VNP and 56 C-VNP were observed with overall mortality of 43% (168 patients). MacCabe, APACHE II score, shock, use of sedatives and absence of enteral nutrition were additively associated with an increased mortality as well as C-VNP (relative risk [RR]: 1.8, p = 0.007). Nevertheless, when S-VNP and C-VNP were simultaneously introduced in the Cox model, only S-VNP remained associated with increased mortality. In patients suspected of VNP, confirmation of VNP using PSB and/or BAL-C adds no prognostic information. Whether this could be explained by the lack of sensitivity of protected distal samples or the severity of underlying conditions of S-VNP patients is still an open issue. A multivariate analysis based on follow-up data during the ICU course of ventilated patients will be initiated in the near future.