We evaluate the performance of the GenoType® MTBDRsl (Hain Lifescience Nehren, Germany) for the detection of second-line resistant tuberculosis and we correlate the frequency of mutations to different Mycobacterium tuberculosis genotypes. We tested 175 strains and 59 clinical specimens interpreting the results according to the Standards for Reporting of Diagnostic Accuracy recommendations. All the strains were also investigated by spoligotyping and Mycobacterial Interspersed Repetitive Units-Variable Number of Tandem Repeats typing. The performances of the MTBDRsl in detecting resistance to fluoroquinolones (FQ), second-line injectable drugs (SLID), and ethambutol (EMB) on clinical isolates were similar (specificity ∼ 99%, sensitivity ∼ 70%, and positive predictive value (PPV) ∼ 99%). Of the 59 respiratory specimens, three samples were classified as "indeterminate". The specificity in detecting resistances was similar for FQs and EMB 100% (95% CI 92.7-100%) and 100% (95% CI 83.9-100%), respectively with a PPV of 100% (95% CI 64.6-100%) and 100% (95% CI 87.9-100%), respectively. Detection of SLID showed a specificity of 89.1% (95% CI 77.0-95.3%) and a PPV of 58.3% (95% CI 32.0-80.7%). Sensitivity for FQ-resistance detection was 100% (95% CI 64.6-100%), whereas for SLID and EMB it was 89.1% (95% CI 77.0-95.3%) and 86.1% (95% CI 71.3-93.9%), respectively. We detected a significant association between mutations in the rrs gene and Beijing lineage. The MTBDRsl can be used to "rule in" extensively drug-resistant strains of tuberculosis in a high risk group; the low sensitivity and negative predicted value (NPV) make confirmation by conventional drug susceptibility testing mandatory when mutations are not identified. NPV for SLID is higher in Beijing strains, showing that the predictive values of the molecular tests are related to the genetic background.