FormalPara Key Summary Points

Why carry out this study?

Staphylococcus aureus has been associated with a higher rate of prosthetic joint infection (PJI) and treatment failure compared with other pathogens.

Changes in the management of S. aureus PJI (SA-PJI) in recent decades make it necessary to reassess the incidence and risk factors for treatment failure, including functional loss which has previously been neglected as an outcome.

What was learned from the study?

A considerable proportion of SA-PJIs failed after initial surgical treatments, although a substantial part of them can be rescued by additional procedures.

Significant functional loss must be considered in addition to clinical failure, mainly in patients in whom the prosthesis was removed, with no possibility of recovery.

In patients managed with debridement, antibiotic and prosthesis retention (DAIR), the loss of function was lower even if this procedure fails.

Anaemia and obesity were risk factors for treatment failure that are rarely reported.

The importance of some risk factors for treatment failure was reinforced, including the protective role of rifampicin-based treatment in DAIR.

Digital Features

This article is published with digital features, including [list digital features available e.g. a video abstract and slide deck], to facilitate understanding of the article. To view digital features for this article go to https://doi.org/10.6084/m9.figshare.21163246.

Introduction

Staphylococcus aureus is widely recognized as a key microorganism causing prosthetic joint infection (PJI), and has been associated with higher rates of treatment failure (TF) compared with other aetiologies [1]. However, studies evaluating TF have traditionally focused on outcomes related primarily with control of the infection [2,3,4,5,6,7,8], while functional outcomes have been given little consideration. A definition of TF that integrates clinical and functional aspects could provide a more realistic measurement of the consequences of S. aureus-PJI (SA-PJI).

TF rates may vary according to patient characteristics, infection type (early, delayed or late) and surgical management. Implementation of the recommendations included in the guidelines for the management of PJI [9, 10] may have positively influenced changes in the rates and predictors of TF in general, and particularly in patients managed with debridement, antibiotics and implant retention (DAIR). For SA-PJI, however, there is little published data on the outcomes associated with different therapeutic strategies and their predictors, and the timing and role of rifampicin in combination with other antibiotics, particularly in patients undergoing DAIR, remain controversial.

Assessments of TF tend to be analysed for the first surgical procedure performed to treat the infection, which is important to evaluate the outcomes associated for the decision about which procedure must be performed. However, additional procedures are usually performed if that first procedure fails to cure the infection, which may rescue some patients but could also affect the functional outcome. Evaluations of TF considering all the procedures performed are not usually made. Such an evaluation would provide a more global view of the final consequences of SA-PJI.

The objectives of this study were to provide updated rates of TF in SA-PJI, taking into account clinical and functional aspects, and to investigate the predictors of TF according to different management strategies. The assessment of both objectives will be carried out for both the first surgical procedure and all additional procedures performed on patients in whom the first one failed.

Methods

Study Design, Sites and Period

This study is part of the ARTHR-IS project (registered at clinicaltrials.gov: NCT03826108) and was designed to evaluate the incidence, risk factors and predictors of SA-PJI TF after primary hip and knee arthroplasty. A retrospective cohort study was conducted in 19 European hospitals to identify patients older than 18 years who received a primary arthroplasty between 1 January 2014 and 31 December 2016 and developed post-surgical knee or hip PJI due to S. aureus within the first year after the procedure. The participating sites, located in Spain, Italy, France, Germany, UK and the Netherlands, were selected using the CLIN-NET network (https://www.combacte.com/about/clin-net/), based on their research experience and data collection capability.

The STROBE recommendations were followed for the reporting of the study (Supplementary Table S1).

Participants

Patients with post-surgical hip or knee PJI due to S. aureus diagnosed within the first year after primary arthroplasty were included. The criteria used to define SA-PJI were as follows: presence of at least one sign or symptom of PJI, including joint pain and/or swelling, or a sinus tract communicating with the prosthesis; and the isolation of S. aureus from (a) ≥ one joint aspirate culture, (b) ≥ two periprosthetic tissue samples and (c) blood cultures with no other obvious source of infection. The patients were identified by reviewing medical records from microbiological laboratory databases, local PJI surveillance databases and discharge reports.

Variables and Definitions

The primary outcome variable was TF until month 18 after the first surgical procedure performed, and was analysed separately both for the first surgical procedure performed (mimicking an intention-to-treat analysis for that procedure) and for all procedures performed (including those performed after failure of the first one). TF was defined as a composite variable including SA-PJI-related mortality, clinical failure and functional failure. Clinical failure was defined as persistence or recurrence of signs or symptoms of infection. For the analysis of the first procedure, this also included the need for an additional course of antibiotics beyond the initial one, the need to use long-term suppressive antibiotic therapy and need of prosthesis removal if not performed as the initial surgical procedure. Functional failure was defined as impeded or significantly impaired walking due to prosthetic loosening or the need to perform a Girdlestone procedure or arthrodesis. Finally, TF was also analysed in the subgroup of patients who underwent DAIR as the first procedure.

Potential predictors of TF were selected according to previous studies [2,3,4,5,6,7,8] and additional hypotheses developed by the project team, and are included in Tables 1 and 2.

Table 1 Characteristics of 128 patients with Staphylococcus aureus prosthetic joint infections (SA-PJI)
Full size table
Table 2 Bivariate analysis of potential predictors of treatment failure among patients with SA-PJI: after the first surgical procedure, after DAIR, and after all surgical procedures performed
Full size table

DAIR as primary treatment procedure was considered appropriate if it was performed < 21 days from onset of PJI symptoms, there was no sinus tract communicating with the joint prosthesis and replacement of polyethylene or mobile components was performed according to IDSA guidelines [9]. The definitions for other variables are included in the Supplementary Table S2.

Data Collection and Ethical Aspects

Data collection was supervised locally by staff with relevant expertise in the field. Data were entered into an anonymized electronic case report form and checked for missing values and inconsistencies. The study was approved by the Ethics Committees at each site (Supplementary Table S5). The need to obtain written informed consent was waived owing to the retrospective nature of the study and anonymized data, except in the case of the French hospitals where a letter of non-opposition was sent to eligible patients. All patients included in these centres therefore gave their authorization to participate.

Statistical Analysis

For bivariate analysis of the association of exposure variables with TF, relative risks with 95% confidence interval (CI) were calculated; p values were calculated by Chi-square or Fisher’s exact test, as appropriate. Continuous variables were categorized after analysing for stratified associations with TF. Multivariable analyses were performed by logistic regression: the effect of study site was controlled for using a generalized linear mixed model in which study sites were considered random effects. Variables with p value < 0.15 in bivariate analysis, and those considered as potentially relevant from clinical judgement, were entered into the models and selected using a manual stepwise backward procedure. Variables with p value < 0.1 were kept in the models. Collinearity and modification effects between variables were studied when clinically sound. The predictive ability of each model was examined by calculating their areas under the receiver operating characteristic (AUROC) curves with 95% CIs. For the effect of rifampicin in the subgroup of patients who underwent DAIR as first surgical treatment, a propensity score (PS) was calculated using a non-parsimonious multivariate logistic regression model, in which the outcome variable was combination therapy with rifampicin. As further sensitivity analyses for the impact of rifampicin therapy, parsimonious multivariate logistic regression models were performed, in which one or two other variables were removed. In addition, to avoid immortal time bias, landmark analysis was used, excluding patients who died or failed treatment in the first 21 days after debridement.

Results

Patient Characteristics and Treatment Failure Rates

A total of 130 cases of SA-PJI were detected, and 128 were included (relevant follow-up data were missing for the other two). The median number of cases per hospital was 7 (interquartile range [IQR] 5–9). The median age of patients was 73 years (IQR 59–81 years); 65 (50.8%) were males; 77 (60.2%) had hip arthroplasty (50 total and 27 partial) and 51 (39.8%) knee arthroplasty (47 total and 4 partial). Patient characteristics are presented in Table 1. Infection-related symptoms started a median of 24 (IQR 15–36) days after the primary arthroplasty (Supplementary Fig. S1), while the first surgical procedure for treatment of infection was performed a median of 4 days (IQR 1–11) after symptom onset. Bacteraemia occurred in 25 cases (19.5%). Overall, out of 128 PJI cases, 28 (21.9%) were due to methicillin-resistant S. aureus (MRSA) strains.

Figure 1 shows patient outcomes according to the first and additional surgical procedures performed. The rate of TF after the first procedure was 32.8% (42 patients; 95% CI 25.2–41.3%). TF was due to clinical failure in 27 cases (21.1%), related deaths in 9 (7%) and loss of function in 6 (4.7%). Median days until failure was 126 (IQR 34–335). Deaths occurred at a median of 21 (IQR 13–48) days after the first surgical procedure performed.

Fig. 1
figure 1

Outcome of patients with SA-PJI according to type of procedure performed to treat the infection. This figure shows the outcome of patients with SA-PJI according to the first type of procedure performed to treat the infection. For boxes on the left, 99 patients receiving initial debridement, antibiotic and implant retention (DAIR) were included: treatment was successful in 68 patients and 31 failed, including 6 deaths, 23 clinical failures and 2 loss of function. The 25 surviving patients with failure after the first procedure underwent subsequent procedures. After 18 months and considering additional procedures, 15 were still failing (9 clinical failures and 6 loss of function). Similar explanations apply to other surgical procedures. aFailure: a composite variable that included related mortality, clinical failure and functional loss (see Methods section). bIncludes: persistence or relapse of infection, need for extra courses of antibiotics after initial scheduled treatment and need for suppressive antibiotic therapy. cIncludes severe impairment of limb function that impedes or makes walking very difficult, including Girdlestone or arthrodesis. *Two aseptic loosening. **Two arthrodesis (one hip, one knee), four prosthesis removal waiting for a new prosthesis. ***One arthrodesis

Full size image

After further surgical interventions, 11 patients who failed the first procedure (9 with persistent infection, 2 with functional failure due to prosthesis loosening after DAIR) were rescued. After 18 months of follow-up, TF was 24.2% (95% CI 17.5–32.3%). The reasons for failure were 9 related deaths (7.0%), 11 cases of clinical failure and 11 cases of functional loss (8.5%, respectively). Excluding functional loss, the failure rate was 15.5%.

Of the 99 patients who received DAIR as a first intervention to treat the SA-PJI (Fig. 1), 31 (31.3%) failed treatment due to death (n = 6), clinical failure (n = 23) or loss of function (n = 2). Despite further interventions, 15 were still failing at the end of the 18-month follow-up.

Of the 29 patients who received other types of first interventions to treat SA-PJI, 11 (37%) failed treatment, broken down as follows: related death (n = 3), clinical failure (n = 4) and loss of function (n = 4). Notwithstanding further interventions, 7 were still failing at the end of the 18-months follow-up.

A summary of the rates and reasons for treatment failure is provided in Supplementary Table S3, including the rate of TF for prosthesis removal as first procedure.

Analysis of Variables Associated with Treatment Failure

Table 2 presents the bivariate analysis of risk factors for TF after the initial surgery. Variables with p-value < 0.15 were Charlson index ≥ 2, haemoglobin < 10 g/dL, hip fracture as the reason for arthroplasty, type of arthroplasty, radiological signs of infection, sinus tract, MRSA, polymicrobial infection, bacteraemia, need for additional procedures not due to persistent infection, inadequate empirical antimicrobial treatment, and treatment with rifampicin and fluoroquinolones. After controlling for hospital and type of initial surgery performed, multivariable analysis retained Charlson index ≥ 2 [adjusted OR 2.5 (95% CI 1.1–5.9)], haemoglobin < 10 g/dL [aOR 3.6 (95% CI 1.6–8.4)], bacteraemia [aOR 2.7 (95% CI 1.0–7.6)], polymicrobial infection [aOR 3.1 (95% CI 1.2–8.2)] and need for additional procedures not due to persistent infection [aOR 2.7 (95% CI 1.1–6.6)] as variables independently associated with TF following the first surgical procedure (Table 3). The AUROC of the model for the observed data was 0.82 (95% CI 0.74–0.89).

Table 3 Multivariate analysis of associations of potential predictors of treatment failure among patients with SA-PJI after the first surgical procedure, after DAIR, and after all surgical procedures performed
Full size table

In the specific analysis performed to investigate the factors associated with TF of SA-PJI after DAIR as the first surgical procedure, the variables with p-values < 0.15 identified in bivariate analysis were the same as for the whole cohort, except for radiological signs of infection and MRSA, and with the addition of body mass index (BMI) > 30 kg/m2, interval > 21 days from symptom onset to surgery, polyethylene/mobile component replacement and appropriate indication for DAIR (Table 2). No statistically significant differences were found between patients treated with and without combination rifampicin therapy (Supplementary Table S4). As we were unable to obtain a PS sufficiently predictive for the probability of receiving rifampicin (AUROC curves < 0.5), a PS was not useful to provide additional control for confounding for the indication of rifampicin. On multivariable analysis (Table 3), variables independently associated with TF of DAIR were: BMI > 30 kg/m2 [aOR 3.9 (95% CI 1.2–12.6)], haemoglobin < 10 g/dL [aOR 5.2 (95% CI 1.9–14.5)], bacteraemia [aOR 6.4 (95% CI 1.7–24.2)], days from symptom onset to surgery [aOR 1.03 (95% CI 1.01–1.06)], and need for additional debridements after the first procedure not due to persistent infection [aOR 4.4 (95% IC 1.4–13.9)]; use of rifampicin was a protective factor [aOR 0.2 (95% CI 0.1–0.7)]. The AUROC of the model for the observed data was 0.84 (95% CI 0.76–0.93). The model prediction was similar when rifampicin was considered to be used for more than 21 or 28 days. For the different models made in which one or two variables other than rifampicin were removed, the estimates of the OR for rifampicin ranked from 0.15 (95% CI 0.04–0.5) to 0.2 (95% CI 0.1–0.7). In the landmark analysis excluding patients who died or failed in the first 21 days after debridement, use of rifampicin remained protective for TF [adjusted ORs (95%CI), 0.22 (0.06–0.80); AUROC curve of the model, 0.82 (95% CI 0.72–0.91)].

Table 2 also presents the bivariate analysis of risk factors for patients who failed after all procedures performed. In addition to the variables identified for the first procedure, TF was also associated with age > 80 years, prosthesis removal as first procedure, need to perform additional joint surgery not due to persistent infection and non-use of rifampicin and fluoroquinolones in combination. On multivariate analysis, hip fracture [aOR 4.6 (95% CI 1.6–12.9)], haemoglobin level < 10 g/dL [aOR 2.5 (95% CI 1.0–6.6)] and need to perform additional joint surgery not due to persistent infection [aOR 3.2 (95% IC 1.1–8.9)] were independently associated with TF at 18 months (Table 3). The AUROC of the model for the observed data was 0.80 (95% CI 0.71–0.90).

Discussion

In this study, we found that nearly a third of initial surgical procedures resulted in TF. The TF rate decreased when further surgical procedures were performed. Importantly, we estimated the impact of significant functional loss. When DAIR was used as the first procedure, even though it is a less aggressive strategy, additional procedures rescued a significant proportion of initial failures without increasing loss of function. In addition, the predictors of TF for SA-PJIs were identified, and the role of rifampicin in patients undergoing DAIR was confirmed.

A review of the literature on SA-PJI studies focusing on treatment outcomes highlighted the difficulties of comparing different study results owing to heterogeneity in study design, case definitions adopted, length of follow-up and types of analyses used (Table 4). Overall, previously reported TF rates ranged from 0% to 16.6% for prosthesis removal [4, 5, 8, 11] and 13.6% to 63% for DAIR [2, 4,5,6,7,8, 12,13,14,15]. It should also be noted that, to the best of the authors’ knowledge, functional outcome was not considered at all in previous studies, despite it being critical to the quality of life of patients. Functional loss is significantly influenced by the surgical procedures performed and this information is therefore relevant to the decision-making process.

Table 4 Summary of published studies on the outcome and management of PJI including ≥ 20 cases focusing on S. aureus
Full size table

When DAIR was analysed as the initial procedure, the TF rate was 31.3% (29.2% without considering functional loss), which is slightly higher than reported in more recent observations [7, 12] but lower than in older publications [2,3,4, 6, 13] (Table 4). However, the TF rate decreased to 21.2% after additional procedures, and to 15% if functional loss was not taken into account (which is more consistent with definitions in previous reports). The lower TF rates for DAIR reported by the latest studies (and by this one) could be attributed to better patient selection for this treatment strategy and to the involvement of multidisciplinary teams in the management of PJI. Indeed, Bouaziz et al. [13] found an overall TF rate of 42%, which decreased to 30% when DAIR was performed according to the latest guidelines [9]. Our data further suggest that appropriate patient selection favours more positive outcomes, and that an initial TF can be rescued without significant functional loss in a considerable number of patients.

The overall TF rates, including clinical and functional updates, for prosthesis removal in our cohort (34.4%) may seem relatively high when compared with other studies, but when only clinical cure was considered: the rate of TF for SA-PJI was 17.2%, which is similar to that found in other series [4, 5, 11]. The high proportion of TF after prosthesis removal related to functional loss is noteworthy and reinforces the importance of early diagnosis of SA-PJI to increase the likelihood of being treated with DAIR.

Since the decision to perform DAIR or remove the prosthesis as first procedure is strongly influenced by patient and infection characteristics, we did not try to compare the outcomes of the two procedures as they are not comparable. Instead, we focused our analysis on identifying potential predictors for TF. Regarding the variables identified, the Charlson index is a predictor of survival and also of prognosis of many infections; similarly, other studies have used the ASA index [5] or ≥ 2 comorbidities for similar reasons [11]. Anaemia was previously identified as a predictor in one study [16], but was not assessed at all in most of the others. Anaemia is a potential marker of nutritional status that can increase tissue hypoxia or even indicate a systemic inflammatory process. As in our study, bacteraemia, polymicrobial infection and the need for additional debridement were also predictors of TF in other studies [6, 14].

Since DAIR is the most frequent initial procedure, we also analysed predictors of TF in this subgroup. Apart from the above variables, obesity and delay in performing DAIR were also identified as risk factors, while the use of rifampicin had a protective effect. Obesity is a risk factor for PJI and was also found to be associated with TF in hip PJI undergoing two-stage replacement [17], but was not even considered in most studies. Obesity could be associated with wound complications, additional debridement, impaired innate immune response and changes in the pharmacokinetics of some antimicrobial drugs [18]. Delayed DAIR following onset of PJI symptoms is a known factor for TF regardless of the microorganism involved, but is particularly pertinent in the case of SA-PJI [2, 6, 12,13,14]. Establishing a threshold is complex. The 21-day threshold for performing DAIR recommended by the guidelines [9] was based on one small-scale study experience [19], and a delay of > 2 days in patients with SA-PJI treated with beta-lactams was associated with increased TF in another study [2] but not when a rifampicin-fluoroquinolone combination was used [6,7,8,9,10,11,12,13,14, 16]. In our cohort, we found an increased risk for each day of delay, supporting the recommendation that debridement should be performed as early as possible.

Rifampicin in combination with other antibiotics (mainly fluoroquinolones) was reported to increase cure rates in a small randomized trial [19] and in observational studies [5, 6, 12, 20,21,22]. However, two recent meta-analyses found controversial results on the role of rifampicin: one found no benefit in staphylococcal infections [23] and the other only a limited impact [24]. The studies included in the meta-analysis had a considerable risk of selection and immortal time bias. In our study, on the other hand, we found that rifampicin combinations were associated with a protective effect, even after performing sensitivity and landmark analyses.

Although previous studies have found higher TF rates in PJIs caused by MRSA compared with susceptible strains [3, 15], our data did not demonstrate this association, which is in line with more recent observations [5, 6]. Whether this is due to the use of anti-MRSA drugs with good bioavailability and anti-biofilm activity, such as linezolid, would require further studies.

Finally, we also analysed the predictors of TF after all procedures had been performed. Apart from other factors, hip fracture increased the risk of TF, probably reflecting the frailty of the patients affected. In these patients, the first surgical approach was crucial, since initial TF was followed by functional failure in all cases (data not shown). In a previous multicentre cohort study of patients with hip PJIs, fracture was also associated with clinical failure and worse functional prognosis [25]. The risk of TF also increased when additional joint surgery not due to persistent infection was performed; this may have been due to tissue damage, delayed healing or facilitation of bacterial superinfections.

This study has some limitations that should be considered when interpreting the results. The sample size was too small to investigate predictors of other initial surgical procedures and may have been insufficient to detect additional predictors of TF; its retrospective design limited the available variables; residual confounding is also possible; we did not collect data about ethnicity of the patients; finally, some changes in management may have occurred during the study period. Some strengths include that it is a multinational study, the definitions of TF including clinical and functional aspects and the good predictive ability of the models developed.

Conclusions

In conclusion, we observed that a considerable proportion of SA-PJIs failed after initial surgical treatments, although a substantial part of them were recovered with further procedures. Significant functional loss should be considered alongside clinical failure, and the importance of certain risk factors for TF was confirmed, including the protective role of rifampicin-based treatment in DAIR.