The authors reply
We excluded patients with chronically impaired laryngeal mobility because they had often received antibiotics for undocumented infections before ICU admission. Including them would have biased the microbiological evaluation of specimens from antibiotic-free patients.
Stroke as an admission diagnosis was not recorded in our electronic case report form. In the recent Preventive Antibiotics in Stroke Study (PASS) trial (3), pneumonia/pneumonitis was diagnosed in 7% of controls. However, the patients were not comatose (median National Institutes of Health Stroke Scale, 5 [3–9]), and pneumonia/pneumonitis was probably overdiagnosed compared with the strict definition used in our study (2) (only 3% of pneumonia/pneumonitis diagnoses in controls were validated by the adjudication committee).
Telescopic-plugged catheter samples were inoculated on blood agar specifically designed for anaerobes, then incubated in anaerobiosis. No specific attention was given to identifying anaerobes, since these are oropharyngeal commensals (e.g., Fusobacterium species and Prevotella species were not identified and may have been present in the seven telescopic-plugged catheter samples with more than one species). However, our probabilistic antibiotic regimen was effective against anaerobes, and noncommensal anaerobes (e.g., Bacteroides species) were identified and tested for susceptibility to antibiotics.
We agree that polymerase chain reaction (PCR) tests can correctly identify bacteria. However, they have several limitations. The high sensitivity of PCR tests can prompt continued antibiotic therapy in patients with small bacterial burdens (e.g., tracheobronchitis as opposed to pneumonia). Clavel et al (4) reported that endotracheal aspirates were inadequate for PCR testing in 14% of patients with suspected ventilator-associated pneumonia. In addition, false-positive PCR tests were most frequent for Haemophilus influenzae and Streptococcus pneumoniae, the second and third most common pathogens, respectively, in our study (2). Finally, PCR tests do not provide information on antibiotic resistance.
We agree fully that further work on antibiotics in the specific context of pneumonia/pneumonitis in comatose patients is needed. Clinical decisions must rely on a careful analysis of the symptoms, laboratory tests, and imaging study findings, and not on a single test, to protect against antibiotic overuse. Eliminating the unnecessary use of antibiotics is of paramount importance. As with all components of ICU management (5), PCR tests for diagnosing pneumonia need to be further assessed in terms of patient outcomes and antibiotic prescription before they can be considered for routine use.