To improve the microbiological diagnosis of device-related osteoarticular infections, we have developed a protocol based on the sonication of device samples, followed by concentration and inoculation of the sonicate in a broad variety of media in a quantitative manner. Sixty-six samples from 31 patients were included in the study (17 of them with clinical diagnosis of infection). The sonication procedure had a sensitivity of 94.1%, which is better than that of conventional cultures (88.2%). One case of contamination and six cases of unexpected positive cultures were detected (specificity of 42.8%): two of these were considered to represent true infection, while the other four were considered to be nonsignificant (corrected specificity of 50%), although the clinical importance of these isolates is questionable. When we analyzed the number of CFU, no breakpoint between significant and nonsignificant isolates could be established. Based on our results, the procedure of sonication of retrieved implants is better than conventional cultures for the diagnosis of device-related infections. The significance of some isolates in patients without clinical infection remains uncertain. However, they may become pathogens and cannot be routinely considered to be contamination.