Introduction

Orthopedic device-related infections (ODRI, including joint arthroplasty infections and infections of other orthopedic hardware, such as osteosynthesis) are infrequent, but are potentially severe and costly [1–3]. Staphylococcus aureus, which can persist within the implant site by producing a biofilm or variant microcolonies, is one of the most frequently associated bacteria with ODRI [4–8]. These infections are difficult to cure and relapse can occur many years after the initial episode [1–3, 9–11].

The management of ODRI, whatever the type of ODRI (i.e., joint arthroplasty infections or infections of other orthopedic hardware), globally includes surgery (debridement with or without implant removal) and lengthy antimicrobial therapy [2]. Treatment failure is nine times more frequent in patients with prosthetic joint infections due to hospital-acquired methicillin-resistant S. aureus (MRSA) than in patients suffering from methicillin-susceptible S. aureus (MSSA) infection [12]. MRSA ODRI is considered to be difficult to treat, as: (1) the bacterium is usually resistant to many clinically important non-beta-lactam drugs, such as fluoroquinolones and clindamycin, that have excellent bone penetration and are usually recommended for the treatment of staphylococcal bone and joint infections; and (2) vancomycin, which is largely used to treat MRSA infections, has slow bactericidal activity, and treatment failure is not uncommon, even when strains are fully susceptible (minimum inhibitory concentration [MIC] ≤2 mg/L) [13]. Moreover, it has been recently suggested that pandemic MRSA clones (usually characterized by multilocus sequence typing [MLST]) responsible for such hospital-acquired ODRI might have advantageous virulence properties (such as an enhanced biofilm production, as described for the predominant clone in Brazil) that may facilitate infection and hinder eradication [14–16]. The influence on the outcome of particular clonal characteristics, as well as the pre-therapy vancomycin MIC and the different treatment options, is poorly documented in patients with MRSA ODRI.

The objectives of this study were: (1) to describe the clinical characteristics, surgical and medical therapy, and outcome of patients with MRSA ODRI managed at our institution; (2) to genetically characterize each MRSA isolate and to determine the vancomycin MIC at the onset of therapy in order to identify microbiological and clinical risk factors for treatment failure.

Materials and methods

Patients and setting

Geneva University Hospitals is a 2,200-bed institution admitting about 40,000 patients annually. We conducted a retrospective cohort study of patients who had at least one episode of MRSA ODRI between 2000 and 2008. The databases of the bacteriology laboratory, the orthopedic sepsis cohort study, the arthroplasty cohort study, and the hospital’s administrative coding system were used for patient selection. The study was approved by the local ethics committee, waiving the need for informed consent.

Data collection

Data were collected from medical reports and nursing charts. In order to limit loss to follow-up, patients or their family were contacted by telephone and interviewed about the outcome of their infection. If direct contact was not possible, outcome information was sought through healthcare providers.

Inclusion criteria and definitions

Patients fulfilling all of the following criteria were included in the study: (1) local and/or systemic clinical signs of acute or chronic bone infection (pain and/or tenderness, fever, swelling, heat, erythema, purulent discharge, sinus tract); (2) presence of an implanted device at the site of infection; and (3) MRSA culture from a preoperative specimen (such as aspirated synovial fluid, needle aspirate of a sinus tract, or blood culture associated with clinical evidence that the implant was the primary site of infection) or from intraoperative specimens. Histological confirmation was not required for the diagnosis of bone infection. The infection was considered to be ‘acute’ when symptoms lasted ≤30 days and ‘late’ when occurring more than 30 days prior to admission [2]. Hematogenous infection was diagnosed when the implant site became infected following MRSA bacteremia associated with another initial site of infection. Persistent MRSA infection was recorded if the patient’s clinical status required further surgery five days after initial therapy, with isolation of the same MRSA strain by intraoperative specimen culture. Recurrence was defined as resurgence of the infection with the same MRSA strain after the end of antimicrobial therapy. Treatment failure was recorded in case of persistent infection, recurrence, super infection (infection during treatment of the initial episode), or reinfection (infection after successful treatment of the initial episode) by another pathogen, limb loss, or death from ODRI. Treatment failure involving MRSA was defined as persistent infection, recurrence, limb loss due to MRSA ODRI, or death directly related to MRSA ODRI. The Charlson comorbidity index was calculated as described elsewhere [17]. Combination antimicrobial therapy was defined as a combination of two MRSA-active agents administered for at least one day during the initial ODRI episode. The defined daily dose (DDD) of each administered drug was calculated using current guidelines for the treatment of MRSA bone and joint infections [18].

Microbiological methods

MRSA was identified according to Clinical and Laboratory Standard Institute (CLSI) recommendations [19]. We retrieved MRSA isolates associated with implant-associated infections that had been stored in skimmed milk/glycerol at −80°C. A dendrogram was constructed for MRSA isolates responsible for the initial infection by using an automated variable number of tandem repeats (VNTR) method (Bioanalyzer Experiments Clustering Software) [20]. Isolates were further genotyped in terms of the accessory gene regulator (agr) allele and SCCmec typing, as appropriate (this analysis was restricted to a minimal number of strains when isolates were considered to be clonal according to the dendrogram) [21–23]. spa typing was performed with the Ridom Staph Type standard protocol (http://www.ridom.de) and the Ridom SpaServer, which assigns spa types (http://spa.ridom.de/index.shtml) and related sequence types (STs). Isolates sharing spa type t041 or relatives, agr type 2, and SCCmec type I were considered to belong to the so-called ‘South German’ clone ST228 [14, 24, 25]. MRSA isolates isolated during persistent or recurrent ODRI were considered to be identical to the isolate responsible for the initial episode when they had a percentage of similitude of 90% or above. The vancomycin MIC was determined for all pre-therapy isolates by using the Mueller–Hinton broth macrodilution method, as recommended by the CLSI [19]. Isolates with vancomycin MICs ≥4 mg/L and <16 mg/L were defined as glycopeptide-intermediate S. aureus (GISA).

Statistical analysis

In the descriptive analysis, the Chi-square test or Fisher’s exact test was used for categorical variables, as appropriate. For the percentage calculation of each variable, the number of missing values were excluded from the denominator. The non-parametric Mann–Whitney test was used for continuous variables. Kaplan–Meier failure curves were compared between groups by using the log-rank test. Independent risk factors for treatment failure involving MRSA (i.e., persistence or recurrence of the MRSA infection) were determined by using a stepwise Cox proportional hazards model. Variables with P-values < 0.15 were included in the multivariate model. Variables were checked for interaction, confounding, and collinearity. To avoid overfitting, a ratio of 10 failures per independent variable was adopted. The model was validated by testing the proportional hazards assumption [26]. Statistical analysis was performed with SPSS software version 15.0 (SPSS, Chicago, IL).

Results and discussion

Patient characteristics

Fifty-two patients met the inclusion criteria. All but two of the patients had previously undergone implant surgery in Geneva. Twenty-three patients (44%) had had joint arthroplasty (18 hip and five knee prostheses) and 29 patients (56%) had other implants (including 23 internal fixation devices, three centromedullar nails, and three external fixation pins) infected by MRSA. The median follow-up was 391 days (range, 4–2,922 days)

Comparison of patient groups

Patients with joint arthroplasty MRSA infection were older (P = 0.009) and had greater comorbidity (P = 0.015) than patients with osteosynthesis material-related MRSA infection (Table 1). Patients with osteosynthesis had longer surgery (P = 0.041) and more frequent emergency surgery (P = 0.038). No significant differences in surgical treatment, antimicrobial therapy, or outcome were noted between these two populations, which were merged for the analysis of risk factors for treatment failure (Table 2).

Table 1 Characteristics and outcome of patients with methicillin-resistant Staphylococcus aureus (MRSA) orthopedic device-related infection (ODRI) and comparison of patients with joint arthroplasty and patients with osteosynthesis
Full size table
Table 2 Comparison of patients with MRSA ODRI with and without treatment failure directly attributed to MRSA
Full size table

Microbiology

MRSA was retrieved from 41 (79%) of the 52 patients with implant-associated infections. The sources were mainly intraoperative specimens (30 isolates), blood (five isolates), aspirated synovial fluid (four isolates), an abscess (one isolate), and sterile aspiration of a sinus tract (one isolate). Thirty-five isolates (85%) were clonally related and shared microbiological characteristics of the South German MRSA clone (spa type t041 or relatives, SCCmec type I, agr type 2; Fig. 1). All of these isolates shared a similar susceptibility pattern, being resistant to gentamicin, erythromycin, lincomycin, and fluoroquinolones. No GISA strain was detected.

Fig. 1
figure 1

Genetic characterization and vancomycin minimum inhibitory concentrations (MICs) of the 41 methicillin-resistant Staphylococcus aureus (MRSA) isolates responsible for orthopedic device-related infection (ODRI) in Geneva between 2000 and 2008. Isolates in bold were considered to belong to the South German clone

Full size image

Surgical treatment and antimicrobial therapy

Surgical treatment consisted of debridement in 47 patients (90%), with implant retention in 28 patients, device explantation in ten patients, and one-stage exchange in nine patients. None of the patients had two-stage exchange, as the four patients scheduled for two-stage exchange experienced treatment failure or had other conditions that prevented reimplantation. Twenty-six patients (50%) received only single-agent antimicrobial therapy (vancomycin alone, cotrimoxazole alone, or vancomycin alone followed by cotrimoxazole or linezolid). The other 26 patients (50%) received combination antimicrobial therapy: 12 patients received rifampin plus fusidic acid (with vancomycin-rifampin as the initial therapy), eight patients received vancomycin plus rifampin, four patients received rifampin plus cotrimoxazole (with vancomycin-rifampin as the initial therapy), and two patients received vancomycin plus cotrimoxazole.

Univariate and multivariate survival analyses

Patients with and without treatment failure are compared in Table 2. The Kaplan–Meier probability estimates of the two-year failure rate were higher when the implant was left in place than when it was removed (log-rank test: P = 0.036; Fig. 2, panel A). The estimates were lower in patients receiving rifampin plus fusidic acid than in patients receiving single-agent therapy and in patients receiving other combinations (log-rank test: P = 0.036 and P = 0.010, respectively; Fig. 2, panel B). In the subpopulation of patients who underwent debridement with implant retention, the two-year probability of treatment failure was 83% with single-agent therapy and 22% with combination therapy (log-rank test: P = 0.020; Fig. 3). The incidence rate of failure involving MRSA in patients who had debridement with implant retention was 3.1 per 100 patient-months in patients treated with single-agent therapy and 1.4 per 100 patient-months in patients treated with an antimicrobial combination, giving an incidence rate ratio of 2.3 (confidence interval [CI], 0.55–13.51; P = 0.11). There was a non-significant trend towards a higher probability of treatment failure in patients infected by the predominant South German clone ST228 in comparison with patients infected by sporadic MRSA strains. There was no difference in the likelihood of failure according to the vancomycin MIC (2 mg/L versus <2 mg/L).

Fig. 2
figure 2

Kaplan–Meier probability estimates for the 2-year failure rate in the 52 patients with MRSA ODRI. A Treated with implant retention (including five patients without surgical debridement) or implant removal. B Treatment with single-agent therapy or with rifampin-fusidic acid or with another combination antimicrobial therapy, regardless of the surgical treatment (*P = 0.010 and **P = 0.036, log-rank test)

Full size image
Fig. 3
figure 3

Kaplan–Meier probability estimates of the 2-year failure rate in the 28 patients with MRSA ODRI after surgical debridement with implant retention, treated with single-agent or combination antimicrobial therapy

Full size image

In multivariate Cox analyses, after exclusion of the five patients who did not receive a surgical debridement, implant retention and single-agent therapy were the only two independent variables associated with treatment failure involving MRSA at two years (hazard ratio [HR], 4.90; 95% CI, 1.32–18.17; P = 0.017 and HR, 4.43; 95% CI, 1.20–16.33; P = 0.025, respectively) (Table 3).

Table 3 Multivariable Cox regression model after exclusion of the five patients who did not receive a surgical debridement: independent factors associated with treatment failure involving MRSA in patients with MRSA ODRI
Full size table

Discussion

In this retrospective cohort study of patients with orthopedic device-related MRSA infection during the period 2000–2008, most isolates belonged to the South German clone (ST228) and were fully susceptible to vancomycin (MIC ≤ 2 mg/L). Only single-agent antibacterial therapy and implant retention were identified as independent risk factors for treatment failure involving MRSA persistence or recurrence at two years, after a median follow-up of 391 days.

Most nosocomial infections worldwide are due to a few hospital-acquired MRSA clones [15]. In our institution, the South German clone is endemic since 1999 [24, 25]. Amaral et al. recently found evidence that the predominant MRSA clone in Brazil exhibited particular virulence properties [16]. In this study, by comparison with sporadic MRSA isolates, isolates belonging to the Brazilian clone were more adhesive and had a higher capacity to produce biofilm in vitro. Here, we detected a non-significant trend toward worse outcome among patients with isolates belonging to the South German clone. Common MRSA isolates are usually fully susceptible to vancomycin (MIC ≤ 2 mg/L), but small increases in the vancomycin MIC, remaining within the range of susceptibility (from 1 to 2 mg/L, for example), were recently shown to influence the outcome of bacteremia [13, 27]. Studies on vancomycin pharmacodynamics revealed that such an MIC increase in the range of susceptibility might have implications for localized orthopedic infections, as vancomycin penetration into bone is not optimal (bone-to-serum ratio 0.3) [28, 29]. However, we found no significant difference in outcome according to the vancomycin MIC.

Surgery is the cornerstone of the treatment of implant-associated orthopedic infections. Retention of the implant is considered, nowadays, as a possible surgical option. This surgical procedure: (1) has to be performed if the pathogen is fully susceptible to antimicrobial agents; (2) has to be reserved for patients with a duration of symptoms <3 weeks and with a stable implant without soft-tissue damage nor sinus tract involvement; and (3) requires a rigorous debridement [2, 3]. For staphylococcal ODRI, only a few studies are available and most of them included methicillin-susceptible isolates. For instance, in the study performed by Brandt et al. that included 33 patients with S. aureus prosthetic joint infections treated by debridement with prosthesis retention, the two-year probability of treatment failure was 69%, but only one isolate was methicillin-resistant [30]. More recently, Marculescu et al. found a two-year treatment failure rate of 40%, but, again, only one of the 32 S. aureus infections was due to MRSA [31]. Even though data on MRSA ODRI are lacking, complete implant removal is strongly recommended for MRSA ODRI [1–3]. Our study, which exclusively involved MRSA, definitively demonstrated that implant removal is required for the treatment of MRSA ODRI, as implant retention clearly emerged as an independent risk factor for treatment failure.

Antimicrobial therapy should always be combined with surgery for the treatment of ODRI [1–3]. Only one randomized double-blind placebo-controlled trial has demonstrated the superiority of combination therapy with rifampin (plus ciprofloxacin) over single-agent therapy (ciprofloxacin) for the treatment of staphylococcal ODRI [32]. It is noteworthy that, in this study, only two isolates (two coagulase-negative staphylococci) were resistant to methicillin. Since this landmark study, and since it has been demonstrated that rifampin was also effective on bacteria embedded in biofilm, rifampin-based combinations have been considered as standard therapy for MRSA ODRI [2, 3, 33]. Few studies have compared different rifampin-based regimens in staphylococcal ODRI. Drancourt et al. demonstrated that a combination of rifampin and fusidic acid or ofloxacin was similarly effective and well tolerated during staphylococcal ODRI, but all of the isolates were methicillin-susceptible [34]. To our knowledge, different rifampin-based regimens have not been compared in MRSA ODRI. In our study, it is noteworthy that none of the patients who received combination therapy with rifampin plus fusidic acid experienced treatment failure. Controlled trials are needed to confirm the superiority of the rifampin-fusidic acid combination for the treatment of MRSA ODRI.

This study has some limitations. First of all, the combination of joint arthroplasty infections with other orthopedic hardware infections is criticable. Indeed, the type, the surface of the hardware, the long-term surgical implications, and the outcome might be different in these two subgroups of patients. However, our cohort of patients is microbiologically homogeneous, as all of them were infected with MRSA, and guidelines for the initial treatment of joint arthroplasty or other orthopedic hardware infections are globally similar (i.e., surgery including debridement with or without implant retention with antimicrobial therapy) [2]. Secondly, our study has the inherent limitations of all retrospective observational cohort studies. This was a single-center study, and the surgical and medical management of ODRI likely evolved during the eight-year study period. Thirdly, many patients were considered to be lost to follow-up at two years, as a quarter of them, enrolled after 2006, did not have a complete follow-up at the end of the study in 2008. Finally, patients with treatment failure occurring in another hospital may have been undetected. However, since the Geneva University Hospitals is, by far, the largest hospital in the area, we consider this latter possible selection bias as minimal.

In contrast, to our knowledge, this is the first study that examined microbiological and clinical risk factors of treatment failure specifically for MRSA ODRI. Indeed, the few previous reports of risk factors for treatment failure in staphylococcal ODRI included mainly MSSA isolates.

In conclusion, we observed a treatment failure rate of 35% in a cohort of patients with orthopedic device-related MRSA infection. Implant retention and single-agent antimicrobial therapy were the only independent risk factors for treatment failure. Microbiological factors, such as infection by the South German clone and a vancomycin MIC of 2 mg/L, were not associated with treatment failure. Therapy using a combination of antimicrobials should be considered for patients with MRSA ODRI, especially when implant removal is not feasible.