Patients with a surgically treated PJI in the hip or knee caused by streptococci or staphylococci were included, whereas patients with PJI caused by other pathogens than these two strains, those with polymicrobial infections, patients with a PJI caused by unidentified pathogens and those with incomplete data were excluded (Figure 1). In case one patient was found to have several infections only the chronological first one was included in this study in order to avoid dependency issues. The criteria for diagnosing PJI have changed over the study time, and for this study, the revised definition of PJI was used, i.e., a combination of clinical signs, blood and synovial fluid biomarkers, synovial fluid cytology, microbiology, histology, and nuclear imaging (Shohat et al., 2019). The presence of a joint communicating sinus tract, more than two samples of congruent intraoperative microbiological findings, more than 80% polymorphonuclear cells and elevated leukocyte cell count in synovial fluid are examples of findings where PJI was considered confirmed (McNally et al., 2021). Medical chart review was performed to assess findings of pre-, intra-, and postoperative cultures in tissue, synovial fluid, abscess drainage, blood, and wound secretion. The following hierarchy regarding the significance of culture findings was applied: (1) tissue, (2) synovial fluid, (3) abscess drainage, (4) blood, and (5) wound secretion, whereby tissue cultures were deemed most and cultures from wound secretion least representative. When multiple cultures were sampled the location with the highest significance rank was considered representative. If preoperative cultures indicated growth of more than one type of bacteria but intraoperative cultures showed only one, the divergent preoperative finding was considered a contamination, and patients were included, given otherwise fulfilled inclusion criteria. The included patients were categorized into three groups according to the findings: PJI caused by S. aureus, CoNS, or Streptococci.

Patients with PJI underwent either DAIR, one- or two-stage revision followed by antibiotic therapy at Uppsala University Hospital. Even though surgical strategies have changed over the years, some basic principles were consistent over the study period: Regardless the type of surgery, the scar tissue as well as any potential fistula is excised and a minimum of 5 tissue samples taken at an early stage during the procedure are collected and sent for analysis.

Excessive debridement is performed to remove all apparently infected tissue around the joint. In cases of DAIR, all modular parts are removed and sent for sonication (during the latest part of the study period). In case of 1- or 2-stage revision surgeries, all implants including all cement are carefully removed and the extracted implants are sent for sonication (during the latest part of the study period). Pulsed lavage of 6 liters 0.9% saline followed by soaking the wound with Prontosan^® (B. Braun, Germany) is performed before the wound is temporarily closed. After re-draping, the second stage advances making use of new, sterile instruments. Repeated pulsatile lavage with a minimum of 3 liters 0.9% saline and – if necessary – further debridement is carried out. Depending on the surgical strategy, new modular components (DAIR), new definite implants (1-stage) or a spacer (2-stage) are implanted followed by thorough closure of the wound.

Index surgery was defined as the first surgery where a monomicrobial PJI caused by S. aureus, CoNS, or Streptococci was treated at Uppsala University Hospital. Previous surgery was defined as the last surgery prior to index surgery. Previous and consecutive surgeries in the affected joint were examined. Information on pre-, intra-, and postoperative antibiotic treatment that had been administered was collected, including types of antibiotics, routes of administration, cause of altered antibiotic therapy and total duration of the antibiotic therapy.

Infection-free survival was the primary outcome and was defined as the absence of all of the following: renewed surgery of the index joint due to PJI, initiation of antibiotic suppression therapy, amputation, or death due to PJI, all due to the same bacteria as found at index surgery. Competing events were defined as re-revision surgery of the index joint, either due to infection caused by other pathogens or due to other reasons than infection, whereas non-informative censoring was defined as emigration (loss to follow-up), end of follow-up, or death due to other reasons than PJI, whichever came first. Outcome events were defined in time by the date of surgery when either a new pathogen or a relapse of the initial pathogen was identified by microbiological cultures, the date of initiation of antibiotic suppression therapy, the date of amputation, the date of death, the date of end of follow-up or the last medical chart entry, whichever came first. To minimize different sources of bias such as unmeasured confounders, selection bias or misclassification, thorough chart reviews were performed to minimize the effect on the mentioned sources of errors on the estimates.

Quantitative variable distributions were described using medians, and interquartile ranges, and categorical variables were summarized in frequency tables. The probability of infection-free survival (equal to the absence of relapse of infection, see previous definition) after index surgery was estimated with 95% confidence intervals (CI) using the Kaplan–Meier method, and Mantel–Haenszel’s log-rank test was applied to investigate differences between groups. Cox regression models were fitted to estimate the adjusted risk of infection relapse, expressed as a hazard ratio (HR) with 95% CI for patients with PJI caused by CoNS or Streptococci compared to those with PJI caused by S. aureus, adjusted for the confounders affected joint, age, gender, type of index surgery, i.e., DAIR, one-stage or two-stage exchange, the onset of the PJI, characterized as early, delayed or late as well as the type of surgery performed prior to the index surgery, i.e., primary arthroplasty, revision arthroplasty due to infection or revision arthroplasty due to any other reason than infection. Schoenfeld-residuals were plotted to confirm that the underlying assumptions of proportionality were met. Furthermore, one analysis stratified for the type of index surgery and one stratified for the type of infection, i.e., “early,” “delayed” or “late,” were performed to estimate differences in unadjusted survival between the three groups. The threshold for statistical significance was defined as p < 0.05.

The original register consisted of 572 unique patients and 46 entries representing patients who had PJI present in two or more joints, resulting in a total of 618 entries unique for one joint. After exclusion according to the previously stated criteria, a study cohort of 299 patients was identified; 114 with S. aureus, 121 with CoNS, and 64 with Streptococci (Figure 1). Date of index surgery varied from 2001 to 2021. Median follow-up time was 4.0 years (IQR = 1.0–7.7 years). More men (54.8%, n = 164) than women were identified, and more hips (63.5%, n = 190) than knees were affected. Median age of the patients was 72 years (IQR = 14, Table 1).

The most frequently performed surgery was DAIR, representing 74.9% (n = 224) of all index surgeries (Table 2), yet it was the chosen method in more cases among the PJI caused by S. aureus and Streptococci than in CoNS-caused PJI. Consequently, 2-stage revision surgery was more often performed in PJI caused by CoNS. A summary of antibiotic therapies is provided in Table 2 and intraoperative microbial findings are listed in Appendix 1.

This study had several limitations. First, the retrospective nature of this study limits the validity of our findings. Human sources of error due to retrospective examination of medical charts are apparent. Second, even though the final study population size consisted of a relatively large number of patients (n = 299), subgrouping by type of bacteria and surgical intervention rendered smaller subgroups not equal in size. The number of events compared to the total number of observations in this study was relatively low, resulting in relatively large confidence intervals representing considerable estimation uncertainty. Due to the scarcity of events, Cox regression models were deemed non-meaningful, for the stratified analysis. We chose to summarize all CoNS as one entity, but there is ample evidence that – among others – S. lugdunensis, while taxonomically a CoNS, clinically more resembles S. aureus (Lourtet-Hascoët et al., 2016; Herry et al., 2022). One major limitation to our study is the definition of both the exposure, PJI, and the outcome, relapse of PJI. Despite efforts in standardizing the diagnostic criteria, the diagnosis of PJI is dependent on a high degree of clinical expertise and correct interpretation of clinical and laboratory findings. In the most current attempt at categorizing the diagnosis of PJI the remaining level of uncertainty has been addressed by creating an intermediate level of “likely PJI.” We tried to follow the updated consensus definition (Shohat et al., 2019), but nonetheless a certain amount of uncertainty regarding the reliability of the PJI diagnosis of patients included in this study is present.

Another limitation to our study is the lack of clinical follow-up. To evaluate patient outcomes based only on the review of medical charts induces uncertainty as to whether patients suffered PJI relapses unknown to us. We believe this to be a limited problem since patients with relapses would mostly have been referred to our tertiary referral center if the need for renewed surgery had been apparent. Nonetheless, patients may have been considered too frail to undergo further surgery by their local physicians, or they may have declined further surgery, and opted for antibiotic suppressive treatment. However, we have no reason to believe that the probability of this scenario should be higher in one of the three investigated groups of patients, and we therefore believe that no considerable bias was introduced. Due to a large proportion of deceased or geographically very distant patients less than half of our study cohort would have been expected to attend clinical follow-up in any case. Outcomes were dated by date of renewed surgery due to infection, start of suppression therapy, date of amputation, death, or last medical chart entry from the orthopedic department before emigration. Possible reinfection once emigrated is lost due to loss to follow-up. This risk was minimized by censoring from the last date of medical entry before emigration, and only six patients emigrated, indicating that the uncertainty is small.

Lastly, whether the optimal surgical strategy was applied regarding time since onset of symptoms was not evaluated in this study. Thus, the question of whether patients received optimal surgical strategy, possibly influencing relapse-free survival rate in this study, remains unclear.

Our study also comes with some strengths. Most studies within the field of PJI are as mentioned earlier small in their sample size. This study included 299 PJI patients, and even the smallest sub-group of streptococcal PJI composed of 64 PJI. Thus, this is the largest retrospective cohort study comparing treatment outcomes after PJI caused by S. aureus, CoNS or Streptococci. Another strength is the stringent selection process that was used to delineate our study population: By excluding all polymicrobial infections, all patients who had PJI in another joint and all with incomplete data we attempted at reducing remaining uncertainties and dependency issues, rendering an observational study that to us appears as a valid description of surgically treated patients affected by monomicrobial PJI by the three included pathogens.