Bronchoscopic or Blind Sampling Techniques for Diagnosis of Ventilator-Associated Pneumonia

Abstract

This is a prospective, postmortem study to determine the accuracy of various bronchoscopic and non-bronchoscopic techniques in the diagnosis of ventilator-associated pneumonia. The authors prospectively studied patients who died in their intensive care unit. Within 15 minutes of death, patients underwent first blind bronchial sampling (BBS), then mini-bronchoalveolar lavage, protected specimen brush (PBS), and bronchoalveolar lavage. The specimens obtained from each study were cultured in a similar way. Within 30 minutes of death, a pneumonectomy was performed at the bedside under sterile conditions while maintaining mechanical ventilation. Either the right or left lung was removed, depending on the area of infiltrate on chest film. Six to 10 specimens were obtained aseptically and cultured. Histologic examination was performed on the lung to look for histologic evidence of bronchopneumonia. Ventilator-associated pneumonia was considered definite if histology exhibited signs of bronchopneumonia and lung tissue culture was positive. Ventilator-associated pneumonia was considered histologic if the histology was positive but the lung culture was negative. During the 2-year period of the study, a total of 146 patients died in the intensive care unit, but only 38 were included in the study. The other 118 patients were excluded for a variety of reasons, including autopsy not available within 30 minutes of death, sampling techniques not performed within 15 minutes of death, and immunocompromised hosts. Of the 38 patients in the study, 21 were receiving antibiotics at the time of death. Four had been on antibiotics that had been discontinued for 72 hours or more before death, and 13 were on no antibiotics during the hospitalization or on no antibiotics during the 7 days preceding death. On histology, 18 of the 38 patients had evidence of bronchopneumonia. Half of these 18 patients had moderate pneumonia, and half had evidence of severe pneumonia. Twenty patients had no evidence of bronchopneumonia. Twelve of the 18 patients with evidence of bronchopneumonia had positive cultures from the lung tissue. None of the lung tissue from the 20 patients without histologic evidence of bronchopneumonia yielded bacteria. In comparing the results of the various sampling techniques with diagnoses of definite ventilator-associated pneumonia, BBS had the greatest sensitivity and acceptable specificity at ≥104 organisms per milliliter. PBS had a sensitivity of only 42% at the ≥103 level but a 95% specificity. When the sampling techniques were compared in the 18 patients with definite or histologic evidence of ventilator-associated pneumonia and the 20 patients without ventilator-associated pneumonia, BBS maintained very good sensitivity and specificity at ≥104 organisms. PBS had a sensitivity of only 33% at ≥103 organisms but a 95% specificity. Another interesting result of the study was that clinical characteristics of the patients, including post-surgical status, previous trauma, simplified acute physiologic score, APACHE II score, McCabe score, body temperature, white blood cell count, Po2, Pco2, and radiologic score were not able to distinguish patients with pneumonia from those without pneumonia. Interestingly, the duration of mechanical ventilation did appear to differ significantly; patients with pneumonia were ventilated for 12.7 ± 9 days versus those without pneumonia, who were actually ventilated longer, at 33.1 ± 34.6 days. The only clinical measure which appeared to be discriminatory was the clinical pulmonary infection score (CPIS). Patients with pneumonia had a CPIS significantly higher than did those without pneumonia. At a CPIS of 6, the overall accuracy of the measure for the diagnosis of pneumonia was 79%. The sensitivity of a CPIS ≥6 was 72%, and its specificity was 85%. When the sum of the CPIS and logarithmic concentration of the organisms obtained in culture from BBS samples was >10, the specificity was 95%, and the sensitivity was 67%.