Résumé : The line probe assay (LiPA), a rapid molecular method for detecting rifampicin resistance (RMPr) in Mycobacterium tuberculosis, correctly identified all 145 rifampicin-sensitive (RMPs) and 262 (98.5%) of 266 RMPr strains among 411 isolates collected from diverse countries. If used as a marker of multidrug-resistant tuberculosis (MDR-TB), detection of RMPr by LiPA would have detected 236 of the 240 MDR strains in this study but would have wrongly suggested the presence of MDR in 26 RMP-monoresistant isolates (sensitivity 98.3%, specificity 84.8%). Hence, the reliability of using LiPA (or any other rapid RMPr-detection method) as a surrogate marker of MDR-TB largely depends on the prevalence of RMP-monoresistance in the study population. This approach must therefore be validated in each local situation.