RESUMEN
SETTING: International multicentric study at nine tertiary care centres. OBJECTIVE: The World Health Organization (WHO) currently does not recommend chest radiographs (CXRs) for routine management of pneumonia. We evaluated the use of CXR for the prediction of treatment failure in children with severe pneumonia. DESIGN: We used WHO vaccine trials radiographic assessment, clinical and nasopharyngeal microbiological data from 1121 3-59-month-old children recruited using the WHO definition of severe pneumonia in the Amoxicillin Penicillin Pneumonia International Study (APPIS). Using Poisson regression, we estimated the relative risk of developing clinical treatment failure and predictive preventive benefit of the CXR and examined the concordance of the CXR findings with the nasopharyngeal microbiological data. RESULTS: A CXR with 'significant pathology' (defined by the WHO algorithm as end-point consolidation, pleural fluid and other infiltrates) was associated with a high risk of treatment failure, especially in children who received penicillin as compared to oral amoxicillin. Significant pathology was also associated with nasopharyngeal isolation of penicillin-resistant Streptococcus pneumoniae. Children with a normal CXR had a reduced risk of clinical treatment failure. CONCLUSIONS: CXR with significant pathology independently and additively predicts clinical treatment failure. If CXR and the WHO tool are available, they can be used in the management of severe pneumonia.