A confirmed diagnosis of PJI is difficult to achieve due to the limited sensitivity of diagnostic methods. There is little real-life data on the benefits of combining SFC with currently recommended microbiology methods. Our study showed that the combination of sonication and standard tissue/fluid samples can increase the number of confirmed microbiology diagnoses in patients undergoing a complete change of their hip and knee implant because of a suspected infection. Our approach combining the results by patient and by strain is original and allowed us to determine the true contribution of sonication to the microbiology diagnosis in PJI. These results corroborate and complete recent work on the diagnosis of PJI, showing higher sensitivity when the two methods are combined (standard tissue/fluid cultures and implant sonication), without a drop in specificity [8, 10].

Other studies have reported cases similar to those we examined, such as the study by Dudareva et al., in which a confirmed diagnosis for 11 patients could only be established by combining a positive tissue biopsy result with a positive SFC result [8]. However, we do not have details on these cases, as this study included only 44% of sonication of complete hip or knee implants: the remaining cases included, in particular, osteosynthesis hardware, bone sequestra, and cement, for which the utility of the technique is very poorly documented.

Ribeiro et al. also explored the idea of combining sonography and biopsy to diagnose PJI as a whole, but they did not provide details on the specific cases in which this combination was essential for establishing the diagnosis [9].

In our study, the rate of additional confirmed diagnosis enabled by sonication was comparable when patients had been treated with antibiotics before the samples were collected. The published data on this topic are contradictory. The 2016 meta-analysis by Liu et al. concluded that sonication of implants had higher sensitivity in patients who had received antibiotics compared to standard perioperative sample culture However, only 4 of the 16 included studies provided data on patients treated with antibiotics, and these studies reported short incubation periods for tissue sample culture (5 days for aerobic and 7 days for anaerobic culture) compared with current guidelines, particularly in the context of prior antibiotic therapy, which can lead to a delayed bacterial growth [11]. This could partly explain the difference compared with our results, as our protocol used longer incubation times, as suggested in the recommendations following these studies [5]. Another meta-analysis in 2023, which included these same studies, noted in a subgroup analysis that the sensitivity of the SFC remained intact during antibiotic therapy, but did not provide any comparison with tissue samples [12]. Surprisingly, sensitivity even appeared to be increased during antibiotic therapy, which calls for caution regarding this result, which contradicts the literature as a whole (at least a minimal decrease in sensitivity during antibiotic therapy).

Later studies in 2017 and 2018 reported no difference, although the methods used for culturing the tissue samples had been optimized [8, 13]. Our findings, obtained by following the recommendations for culture techniques, suggest that any patient, whether or not they had received antibiotics, could potentially benefit from sonication of their implants post-removal to optimize the microbiology diagnosis.

The most important of our study’s limitations is its retrospective design. Consequently, sample harvesting, whether pre- or intraoperative, was not standardized. This may have modified the influence of standard sample collection (especially intraoperatively) in some of the diagnoses, as the number and type varied depending on the surgeon and operating conditions. Some of the surgeons who performed the surgeries specialized in the management of PJI, which may have improved the quality of the tissue samples collected in certain cases, given their experience in this area. The opposite is also true for patients operated on by non-specialist surgeons.

The practice may also have evolved over the period in which the patients were included, although it was fairly short (32 months). However, this corresponds to real clinical practice, in which the possibility of poor sample collection and inter-surgeon variability is real, as demonstrated by Gandhi et al. [14]. To our knowledge, no study up to now has shown that a specific sample harvesting pattern is superior to another, and Bémer et al. found no difference between the types of samples (interface, synovial, capsule) in terms of positivity rates [4], which suggests that the variation in the nature of the samples likely had a moderate impact on our results. Thus, this lack of standardization likely does not reduce the reproducibility of our findings in current practice. Nevertheless, it surely has led to some heterogeneity between cases, particularly with regard to the number of standard positive samples.

In contrast, the explanted implant analyzed by SFC represents a single, standardized microbiological sample. Its contribution must be interpreted differently from that of additional periprosthetic samples, whose number, type, and anatomical location may vary from one patient and surgeon to another. In our study, SFC provided the second concordant microbiological result for 32 strains that were identified by only a single positive standard culture. This number was of the same order of magnitude as that of strains confirmed by two concordant standard periprosthetic samples despite a negative SFC. This suggests that, when interpreted in combination with standard cultures, the SFC constitute an additional standardized microbiological source, less dependent on the variability of intraoperative samples.

We chose to use the microbiology diagnosis made by the team treating a patient when the major microbiology criterion was not met (i.e., only one positive culture instead of two positive, matching cultures), thus false positives may have been considered as being infected. This is true for pathogens said to be “virulent” (which are always treated to be on the safe side, but whose contaminating nature cannot be ruled out) and for those more likely to be considered as contaminants. Our analysis focused solely on cases meeting the major criterion, with no consequence on our primary objective.

This descriptive approach, which did not include a calculation of sensitivity and specificity and theoretically provides a moderate level of evidence, was justified in our opinion given the difficulty of not having a simple reference method for the diagnosis of PJI but only a composite gold standard (i.e. guidelines criteria [5]). Studying the combination of tissue sample cultures and SFC allowed us to avoid the incorporation bias, which happens when the results provided by the studied method form part of the reference test in a diagnostic study, which would be the case here. The challenge posed by incorporation bias has been highlighted by the authors of other studies on sonication, such as Dudareva et al. who, in their work, have attempted to make multiple comparisons with various existing diagnostic definitions and create their own composite score to address these limitations. [8].

We wanted to have a fixed outcome by setting an expected rate of at least 5% of diagnoses made by fulfilling the major microbiology criterion when sonication is done. This number is arguable because it comes from an approximation of the increased sensitivity achieved by the latest advance in laboratory techniques, namely inoculation of paediatric blood culture bottles [4]. This reference may be controversial, in part because of the unknown specificity associated with results of blood culture bottles and in part because the incremental gain expected by adding a new technique to the existing arsenal is likely small. This reasoning may have led us to setting a lower expected rate, but we wanted to ensure a true clinical benefit by aiming for a level of 5%.

To place our findings in the context of the latest recommendations, the EBJIS diagnostic criterion of positive sonication at > 50 CFU/ml [5] was not used because our study was designed to add sonication to existing techniques to achieve two matching positive samples, not evaluate sonication by itself. The culture cutoff point for SFC is still controversial. Alvarez Otero et al. [15]. proposed lowering it to ≥ 20 CFU/10 ml, which is 2 CFU/ml. They proposed this value subject to centrifugation of the sonication fluid, and their analysis consisted solely of monomicrobial infections. They admit this cut-off point could depend on which sonication technique is used. Furthermore, in the meta-analysis conducted by Peng et al., the different thresholds used in the studies do not appear to significantly alter the sensitivity of sonication for the diagnosis of PJI (76%, 75%, and 73% for 50, 20, and 2 CFU/mL, respectively) [12]. However, the heterogeneity among the studies and the fact that this result is based on subgroup analyses cast doubt on the weight of this data.

In our study, among the 32 strains for which sonication contributed to confirmation of the microbiological diagnosis, 17 strains grew only in paediatric blood culture bottles. However, all were concordant with one standard positive culture. These findings illustrate the limitation of applying a quantitative threshold. Although it may strengthen the specificity, it may also classify concordant results as negative. These findings, in line with guidelines, further justify our concordance-based diagnostic strategy rather than using SFC as a standalone test.

We did not include cases where SFC was the only positive result and at a significant threshold (either 2 or 50 CFU/mL), as this was not the aim of our study.

Nor did we report cases where SFC was the only positive result and identified a virulent pathogen, which in clinical practice often leads to the patient being treated. However, taking these two scenarios into account, the contribution of sonication could be greater than what has been reported here.