Diagnosis
Nearing the close of the 20th century, nosocomial pneumonia, now more often referred to as VAP, was a clinical diagnosis suspected by the onset of a new or changing pulmonary infiltrate on X-ray, fever, leukocytosis, and purulent tracheobronchial secretions. Although the development of this clinical picture in a patient previously free of pulmonary disease indicates bacterial pneumonia with a high likeliness, many existing conditions in a mechanically ventilated patient present a similar clinical picture, including chemical pneumonitis, acute respiratory distress syndrome, systemic inflammatory response syndrome (SIRS), pulmonary contusion, atelectasis, pulmonary edema, and pleural effusion. In addition, intubated patients often have proximal airway colonization by potentially pathogenic organisms and often have purulent secretions because of either tracheobronchitis or oropharyngeal secretions that escape the barrier of the endotracheal tube cuff. As a result, patients were often assumed to have pneumonia and treated for such with empiric antibiotics. Investigations performed in the 1980s began to illuminate this deficiency in diagnostic accuracy.3–5 Taken together, these studies suggested that no combination of clinical variables was accurate in the prediction of nosocomial pneumonia and the misdiagnosis of nosocomial pneumonia was a common occurrence.
Stimulated by their medical colleagues at the University of Tennessee Health Sciences Center who were among the early proponents of fiberoptic bronchoscopy diagnosis of pneumonia,6 Dr Fabian and his surgical colleagues embarked on the first of many studies that would subsequently confirm and refine the efficacy of a lower-respiratory culture-driven approach to the diagnosis of pneumonia in trauma patients.7 In 107 trauma patients with clinical suspicion of pneumonia, respiratory cultures were obtained in triplicate—the first by routine sputum collection, the second by fiberoptic bronchoscopy-guided protected specimen brushing, and the third by fiberoptic bronchoscopy-guided bronchoalveolar lavage (BAL). They observed that the incidence of pneumonia according to culture positivity was 73% by sputum culture, 34% by protected brush specimen, and 25% by BAL. Importantly, this article included the details of a provocative pilot study to determine if empiric therapy could be stopped safely should the lower respiratory tract cultures be negative. It is important to note that it was common practice at that time to continue antibiotics irrespective of culture results should the clinical picture resemble pneumonia, and cessation of antibiotics would have been perceived as putting the patient at risk. Ten patients were studied and seven did not meet the prespecified lavage criteria for diagnosis of pneumonia of ≥105 colony-forming units (cfu)/mL and antibiotics were discontinued. One of these patients subsequently died of their head injury, but the remaining six patients clinically improved without continuation of antibiotic therapy. From this observation, the Memphis group proposed that bronchoscopy with BAL was able to differentiate trauma patients with de facto lower respiratory tract infection from those with SIRS, paving the way for their next study.
The proof-of-concept study was completed in 1994 and presented at the Eastern Association of Trauma Annual Meeting in 1995.8 This study sought to answer two questions with respect to suspected pneumonia in mechanically ventilated trauma patients: (1) can quantitative BAL culture differentiate pneumonia from SIRS, and (2) can antibiotic therapy be based solely on quantitative BAL cultures? In this prospective study, patients with clinically suspected VAP (based on the presence of abnormal body temperature, leukocytosis, grossly purulent sputum, and new or changing infiltrate on chest X-ray) underwent fiberoptic bronchoscopy with BAL, and only those with significant bacterial colony counts (≥105 cfu/mL) were treated with a full course of therapeutic antibiotics. Forty-three patients underwent bronchoscopy 55 times, and 20 were identified to have pneumonia according to quantitative culture result (the remaining 23 were designated as having SIRS). For the patients with SIRS, antibiotics were stopped after culture result (average 3.3 days). Sixty-five percent of these patients clinically improved after antibiotic cessation. The remaining 35% continued to demonstrate clinical suspicion of pneumonia and underwent repeat bronchoscopy with BAL. Among these eight patients, three were identified to have developed pneumonia per quantitative culture result on repeat bronchoscopy; the remaining five ultimately had clinical improvement without continuation of antibiotics. From this study, it was concluded that bronchoscopy with quantitative culture from BAL could in fact distinguish pneumonia from SIRS and dictate the appropriateness of the continuation of antibiotic therapy. This approach to the diagnosis of pneumonia then became standard practice in the Memphis Trauma ICU.
A follow-up prospective study involving 232 patients during a 2-year period was presented at the annual meeting of the Southern Surgical Association in 1997 and served to establish a false negative rate of 7% when using a cut-off point of ≥105 cfu/mL to establish the diagnosis of pneumonia.1 In this study, empiric antibiotic therapy was instituted on all patients after BAL. A third-generation antipseudomonal cephalosporin, a quinolone, or a carbapenem was administered at the discretion of attending physician, and vancomycin was added if the gram stain demonstrated gram-positive organisms. This study revealed two important findings. Organisms identified by quantitative culture ≥105 cfu/mL were compared with the gram stain, and it was observed that the gram stain of the BAL effluent correlated poorly with the ultimate culture result and would therefore not be useful in guiding specific antibiotic empiric therapy. However, the investigators identified that duration of hospital stay could guide empiric therapy with respect to antibiotic choice. In the first week of ICU stay, BAL primarily identified Haemophilus influenzae and gram-positive organisms, whereas Acinetobacter and Pseudomonas were more common after the first week (figure 1).
The third study from Memphis related to the diagnosis of pneumonia among ventilated trauma patients was presented at the annual meeting of the American Association for the Surgery of Trauma (AAST) in 2003.9 The purpose of that study was to determine the optimal diagnostic threshold for pneumonia with respect to quantitative cultures. Based on their own mortality data, the Memphis group had chosen a threshold of 105 cfu/mL to distinguish pneumonia from SIRS. As derived from their first study, the mortality rate of all patients with 105 cfu/mL was significantly higher than patients with less than 105 cfu/mL (29% vs. 14%, p<0.04). Nonetheless, there was not a consensus at the time as to what diagnostic threshold was optimal and, in fact, a lower diagnostic threshold was generally recommended in the medical ICU setting. In this study of 526 patients who underwent 1372 bronchoscopies with BAL, the sensitivities, specificities, positive and negative predictive values were determined with respect to the diagnostic thresholds of 105 cfu/mL and 104 cfu/mL. As demonstrated in table, test performance was better at 105 cfu/mL with improved specificity and positive predictive value without any large reduction in either sensitivity or negative predictive value. This study solidified the diagnostic threshold of 105 cfu/mL for trauma patients and remains the diagnostic threshold in the Memphis trauma ICU today. In accordance with the theme of continuous performance improvement, the diagnostic threshold was re-evaluated a decade later, presented at the annual meeting of the AAST in 2014.10 This study comprised 1679 patients who underwent 3202 bronchoscopies with BAL, during a 9-year period. Using 105 cfu/mL as a diagnostic threshold, a low false negative rate continued to be achieved (2.3%), the false negative rate among those with 104 cfu/mL remained low (7.5%) as well. The current clinical pathway for initial diagnosis is demonstrated in figure 2.