Timely diagnosis of Pulmonary Tuberculosis (PTB) is associated with good prognosis, but remains difficult in primary healthcare facilities and particularly in children and patients living with HIV. The aim of this study was to compare the GeneXpert ® MTB/RIF assay (Xpert) performed using a stool sample (3-5 g) and using the first Respiratory Tract Sample (RTS; i.e., sputum, bronchoalveolar or gastric aspirate; as normally done) concomitantly collected from 119 patients with suspected PTB to improve PTB diagnosis in Burkina Faso, a high tuberculosis burden country with limited resources. Overall, microbiological, microscopic and molecular analysis of the 119 first RTS and 119 stool specimens led to Mycobacterium tuberculosis complex detection in 28 patients (23 positive RTS cultures and 5 negative RTS cultures-RTS Xpert positive). When using the 28 clinical confirmed cases as reference standard, the sensitivities of the stool-based and RTS-based Xpert assays were not different (24/28, 85.7%, versus 26/28, 92.86%; p > 0.30), and 22 results were fully concordant. Considering the first RTS culture as the gold standard, the sensitivities of the stool-based and RTS-based Xpert assays to detect PTB in patients with positive RTS culture were 100% (23/23) and 91.3% (21/23), respectively (p >0.05). The stool-based Xpert assay specificity for excluding PTB was 99% (95/96) (compared with 95%, 91/96, when using RTS) and its negative and positive predictive values were 100% (95/95) and 96% (23/24), respectively. Compared with the 23 positive RTS cultures, the incremental yield rates of the RTS-based and stool-based Xpert assays were 4.2% (5/119) and 0.84% (1/119), respectively. Overall, our findings support using the stool-based Xpert assay as an alternative method for earlier PTB diagnosis, when RTS are difficult to obtain.< Réduire