Background: Pulmonary tuberculosis (TB) is the most frequent site of TB and the one leading its spread worldwide. Multiple specimens are commonly collected for TB diagnosis including those requiring invasive procedures. This study aimed to review the sampling strategy for the microbiological diagnosis of pulmonary TB.

Methods: A retrospective analysis of collected samples from September 1^st 2014 to May 1^st 2016 in the Bacteriology laboratory of Pitié-Salpêtrière Hospital (Paris, France) was performed. All the samples collected in patients aged over 18 years for the bacteriological diagnosis of pulmonary TB were included.

Results: A total of 6,267 samples were collected in 2,187 patients. One hundred and twenty-six patients (6%) had a culture confirmed pulmonary TB. Among them, multiple sputum collections were sufficient for TB diagnosis in 63.5%, gastric lavages permitted to avoid bronchoscopy in only 7.1%, and bronchoscopy was necessary in 29.4%. The culture positivity of sputa (8.6%) was higher than that of bronchial aspirations (3.1%), bronchiolo-alveolar lavages (BAL) (2.3%) or gastric lavages (4.8%) (p<0.001). From its 70.0% theoretical PPV value, the 46.1% selection in bronchial aspirations allocated to molecular test increased PPV up to 88.9%.

Conclusions: Based on our data, we suggest to collect sputum consistently. If smear negative a bronchoscopy should be performed and molecular diagnosis be performed on a subset of bronchial aspirations based on expertise of the bronchoscopist.