Introduction

Staphylococcus aureus comprises up to two thirds of all pathogens in osteoarticular infections [1], which are difficult to treat because of the organism’s ability to form small colony variants and to grow into biofilms [2]. In a recent report comparing the outcomes of orthopaedic implant infections due to methicillin-susceptible S. aureus (MSSA), methicillin-resistant S. aureus (MRSA) and coagulase-negative staphylococci, overall remission was worst for MRSA (57%), followed by MSSA (72%) and best for coagulase-negative staphylococci (82%) [3]. Kilgus et al. showed that hip arthroplasties infected due to MRSA were treated successfully in only 48% of cases, as compared to 81% with MSSA infection [4]. Salgado et al. attributed a ninefold higher ratio for treatment failure in prosthetic joint infections due to MRSA than due to MSSA [5]. The surgical and medical literature is full of publications regarding staphylococcal osteoarticular infections, especially for MRSA.

In contrast, Pseudomonas aeruginosa is an under-reported Gram-negative pathogen in the orthopaedic domain. P. aeruginosa is at least as clinically virulent as S. aureus and also readily forms biofilms [6, 7] using extensive quorum-sensing abilities [8], develops resistance to administered antibiotics more rapidly than S. aureus and produces a large variety of virulence factors such as type III secretion-dependent cytotoxicity [7, 8]. Unsurprisingly, P. aeruginosa is feared by orthopaedic surgeons and infectious disease specialists with experience in osteoarticular infections. Many of them would attribute a worse prognosis for pseudomonal bone and joint infections compared to S. aureus.

However, there is little literature about P. aeruginosa and osteoarticular infections. P. aeruginosa accounts for an absolute minority of all osteoarticular pathogens [1], and most reports are either case series involving around 20–30 cases [913] or even less [1427]. Moreover, these reports subsume P. aeruginosa under other Gram-negative pathogens [9, 12, 13, 19], including anaerobes [10, 28], and lack adjustment for case mix [9, 10, 19, 21]. Often, outcome comparisons of osteoarticular infections due to P. aeruginosa with Gram-positive pathogens are almost non-existent [10, 28]. Hsieh et al. retrospectively compared arthroplasty infections due to Gram-positive pathogens to the same infections due to Gram-negative pathogens. They found that conservative treatment of Gram-negative infection by debridements was associated with a lower success rate than treating Gram-positive arthroplasty infections (27 vs 47% remission). However, this difference vanished when a two-stage exchange was performed [10].

In this clinical case-control study, we investigate the hypothesis of lower remission risks of pseudomonal osteoarticular infections compared to S. aureus (the most frequent pathogen) and MRSA (the pathogen with the most frequent treatment relapses known so far) by adjusting for case mix of the underlying complex patient population.

Methods

Setting and data collection

The Orthopedic Service of Geneva University Hospitals is a tertiary centre with a septic subunit consisting of 22 acute care beds. A dedicated team of orthopaedic surgeons, infectious disease specialists, diabetologists, nurses and physiotherapists treat patients with osteoarticular infections [29].

Prospectively registered databases from the arthroplasty cohort and septic orthopaedic cohort were searched for osteoarticular infections among adult orthopaedic patients from January 1996 to November 2009. Patients were followed up to 31 July 2011. Forty-two variables for each episode were assessed related to demographic characteristics, immunosuppression, microbiology, surgical and antibiotic treatment modalities and outcomes. The quantitative microbiological procedures were unchanged during the study period and based on the Clinical and Laboratory Standards Institute (CLSI) guidelines [30] and in-house MRSA identification methods [3]. No typing was performed. Since clinical specimens had not been stored, no retrospective analyses of minimal inhibitory concentrations could be performed. The study was supported by the local Ethics Committee (No. 08-017R; 08-029R).

Criteria and definitions

Definition of osteoarticular infection required local signs of infection such as heat, erythema, pus or functional impairment; a medical report; a targeted antibiotic treatment; and the presence of the same pathogen in at least three intra-operative samples. In order to enhance specificity, only cultures which were grown on plates were considered. Co-pathogens were accepted only if S. aureus or P. aeruginosa outnumbered them by at least threefold in intra-operative microbiological specimens, for instance coagulase-negative staphylococci (103 cfu/ml) in one specimen and P. aeruginosa in all four specimens. Only the first episode of bone and joint infection per patient was included; patients presenting with relapsing episodes were excluded. Further exclusion criteria were paediatric patients (≤ 16 years), cure by amputation, involvement of fingers or toes only and a minimal active follow-up of less than 18 months after end of treatment.

Remission was defined as complete clinical and laboratory resolution of the former infection. Immunosuppressed patients were considered those with active neoplasm, insulin-dependent diabetes mellitus, Child class C cirrhosis, dialysis or those requiring chronic steroid therapy (prednisolone equivalence of 15 mg/day). Multi-resistance among P. aeruginosa strains meant resistance to at least four classes of antibiotic agents.

Statistical analyses

Group comparisons were performed using the Pearson χ2 test, Fisher’s exact test or the Wilcoxon rank sum test, as appropriate. Logistic regression analyses determined associations with remission. Independent variables with a p value ≤0.2 in univariate analysis were added stepwise in the multivariate analysis to adjust for case mix. We included five to ten predictor variables per outcome event [31]. The following variables were introduced into the final model independently of their association in univariate analysis: P. aeruginosa vs S. aureus, age, duration of antibiotic treatment and number of surgical interventions. Since comparison of P. aeruginosa with S. aureus (MSSA & MRSA) might be different from the comparison of P. aeruginosa with MRSA alone, all analyses were repeated for S. aureus and MRSA separately. Key variables were checked for collinearity and interaction, the latter by Mantel-Haenszel estimates. A p value ≤ 0.05 (two-tailed) was considered significant. Stata software (9.0, StataCorp., College Station, TX, USA) was used.

Results

Infections and patients

A total of 131 episodes of osteoarticular infections in 131 adult orthopaedic patients were retrieved. The median age was 63 years and 53 were women. Thirty-seven patients (28%) were immunocompromised: 17 due to diabetes mellitus, nine by advanced cirrhosis due to alcoholism, one by dialysis, one by cancer and nine by chronic steroid medication. The infected implants in the arthroplasty group included total hip (n = 22) and total knee (n = 16) prostheses. In the fracture fixation devices group, the infected implants included plates (n = 28), intramedullary nails (n = 14), external fixation devices (n = 7), hip screws (n = 3), other screws (n = 2), patellar cerclage wire (n = 1) and spondylodesis material (n = 1). Additionally, we retrieved seven episodes of septic native joint arthritis and 30 episodes of osteomyelitis without implant. Crude group comparisons of demographic and treatment characteristics according to the pathogen are displayed in Table 1.

Table 1 Group comparisons between osteoarticular infections (P. aeruginosa vs S. aureus and P. aeruginosa vs MRSA)

Pathogens

According to study definitions, we included osteoarticular infections due to S. aureus and P. aeruginosa only. With 111 infectious episodes, S. aureus was overwhelmingly predominant compared to P. aeruginosa (20 episodes). In 37 cases, S. aureus was resistant to methicillin. In 21 S. aureus infections, there was a co-pathogen with at minimum a threefold lesser quantitative growth in cultures: coagulase-negative staphylococci (four), streptococci (seven), Proteus spp. (five), Klebsiella spp. (two), Escherichia coli (one), Enterobacter cloacae (one) and Enterococcus faecalis (one). Co-pathogens were seen in six pseudomonal infections: coagulase-negative staphylococci (four), Enterobacter spp. (one) and Klebsiella spp. (one). With the exception of two multi-resistant P. aeruginosa episodes, all pseudomonal infections were susceptible to most antipseudomonal antibiotic agents, 11 of which were susceptible to every antipseudomonal antibiotic tested.

Among the S. aureus cases, all but two were surgical site infections, while the origin of P. aeruginosa infections was heterogeneous: open fractures (seven), surgical site infections (five), haematogenous infections (two; one of which was in a patient with sickle cell anaemia), IV drug abuse (one), unknown (two) and pressure ulcer (three). Two patients had prior empiric antibiotic therapy before diagnosis. We failed to establish epidemiological links between the different cases of pseudomonal or MRSA infections.

Surgical treatment

All patients underwent surgical treatment with a median of two surgical interventions in the operating theatre (range one to six surgeries). Pseudomonal and staphylococcal surgical treatments did not differ in the number of interventions (median two vs two interventions, p = 0.43) (Table 1).

Antibiotic treatment

All patients received systemic antibiotic therapy. No antibiotics or detergents [32] were added to the irrigating solutions. Gentamicin beads were used in six cases (three for S. aureus and three for P. aeruginosa) for a median duration of three weeks. The median duration of total antibiotic therapy was eight weeks (range six–14 weeks) for the entire study population. The median duration of antibiotic treatment tended to be longer for P. aeruginosa (87 days), compared to 46 days for S. aureus infections, and 58 days for MRSA, although the differences were not significant (all p > 0.05) (Table 1).

There were no clear preferences for the choice of antibiotic agents. For staphylococcal parenteral therapy, flucloxacillin, amoxicillin/clavulanic acid and vancomycin were the most frequently used agents and ciprofloxacin/rifampicin and clindamycin for oral treatment. In 44 staphylococcal infections (40%), rifampin was used in combination therapy at a median dose of 1,200 mg/day. The choices were culture based.

For pseudomonal infection, parenteral therapy consisted of ceftazidime (n = 5), cefepime (n = 4), imipenem (n = 2), piperacillin (n = 1) and polymyxin B (n = 1). Seven episodes were treated by oral ciprofloxacin 500–750 mg b.i.d. from the start. Thirteen episodes revealed a combined initial parenteral antimicrobial treatment for P. aeruginosa. Including rifampin treatment for S. aureus, 50 episodes (38%) were treated with a combination therapy during the majority of the antibiotic administration time. Three recurrent P. aeruginosa strains (3/20; 15%) became resistant to one of the antimicrobial agents used during initial treatment (twice to ciprofloxacin and once to imipenem). Except for one case there was no ciprofloxacin resistance among pseudomonal isolates at the beginning of therapy.

Outcomes

Overall, 100 episodes (76%) remained in remission after a median active follow-up time of 4.0 years (range 1.5–13 years). Stratified by the pathogen, a tendency for a lower remission rate was found for pseudomonal infections as compared to S. aureus (12/20 vs 88/111; p = 0.06). This was similar when P. aeruginosa was compared with MRSA alone (12/20 vs 30/37; p = 0.08) (Table 1).

We further stratified remission rate according to the type of infection. In all three groups of infection, P. aeruginosa tended towards a lower remission rate than S. aureus, but none of these results were statistically significant: 2/5 vs 20/25 [odds ratio (OR) 0.2, p = 0.1] for osteomyelitis without implant, 2/3 vs 3/4 (OR 0.7, p = 1.0) for septic native arthritis and 8/12 vs 61/78 (OR 0.5, p = 0.5) for implant-related infections as a group. Of note, among the seven episodes of pseudomonal infections treated with oral ciprofloxacin from the start, five were cured and two failed to respond. This proportion was not different from P. aeruginosa episodes treated with initial parenteral therapy (5/7 vs 6/13; p = 0.58).

Multivariate analysis

Because of the heterogeneity of the study population and differences in crude group comparison between MRSA and P. aeruginosa infections (Table 1), we performed two multivariate logistic regression analyses to adjust for this case mix. The first analysis compared P. aeruginosa with S. aureus as a group and the second P. aeruginosa with MRSA.

No demographic or therapeutic variable showed a significant association with remission in either of the multivariate analyses, although pseudomonal infection tended towards a lower remission rate with an OR of 0.4 (p = 0.08)(Table 2). While the number of surgical interventions tended to be less associated with remission, the duration of antibiotic treatment did not influence outcome. Less than six weeks of total antibiotic treatment had the same outcome as six–12 weeks [OR 0.8, 95% confidence interval (CI) 0.3–2.6] or more than 12 weeks (OR 0.4, 95% CI 0.1–1.7). Our statistical model yielded a non-significant goodness-of-fit test (p = 0.4; p = 0.2 for MRSA comparison) and a receiver-operating characteristic (ROC) curve value of 0.72 (ROC curve value 0.83 for MRSA comparison), highlighting an acceptable appropriateness of our final models.

Table 2 Variables associated with remission in logistic regression analysis

Discussion

Formally, we report a trend towards a lower remission rate for orthopaedic infections due to P. aeruginosa compared to S. aureus or to MRSA with an adjusted OR of 0.4 for remission (2.5 for failure) in multivariate analysis. Although this result was formally statistically insignificant, we detected this tendency in all analyses performed.

Our failure risk of 24% is congruent with the 20% [12], 21% [13], 25% [27], 27% [14] or 30% [9] failure rates reported for pseudomonal osteoarticular infections, although lower (5% [11]; 7% [21]) or higher failure rates ranging from 33% [23], 35% [24], 38% [22] to 39% [15] also exist in small case series from the 1980s. Development of resistance among failure cases to the antimicrobial agents used during therapy occurred in 15%, which is congruent with the 13–24% risk after a prolonged antibiotic monotherapy during three–four months [9, 24]. Hence, we exclude a major setting bias. The only particularity of our study population is the relatively small proportion of haematogenous infections due to S. aureus (2 of 111), which has been reported more frequently in another tertiary centre of Switzerland [33].

Of note, infections due to P. aeruginosa benefited from the same median number of surgical interventions and an enhanced median duration of 87 days of antibiotic administration compared to 46 days for S. aureus and still tended towards more treatment failures. Conversely, duration of post-debridement antibiotic administration was unrelated to remission. Less than six weeks of antibiotic treatment yielded the same remission as six–12 or more than 12 weeks. Our data suggest that the tendency towards a lower remission rate in pseudomonal osteoarticular infections cannot be compensated for by simple prolongation of post-debridement antibiotic therapy. This finding is congruent with recent retrospective data suggesting that a total antibiotic duration of maximally six weeks post-debridement [3, 16, 20, 34], or regimens with an early switch to oral antibiotics with good oral bioavailability and bone penetration [11], are as effective as prolonged parenteral regimens [14, 35]. Longer treatment durations, e.g. for more than three months [9, 14, 20, 33] or longer [13], did not enhance remission when compared in the literature. Dan et al. treated 22 cases of pseudomonal osteomyelitis with oral ciprofloxacin with various durations ranging from four to 12 weeks. Remission rates remained the same or were even enhanced with short regimens (95%) [11].

We cannot comment on the effect of combination antibiotic therapy for pseudomonal bone and joint infections, since few of our episodes had been treated with monotherapy. Although a more favourable outcome of combined regimens is still under debate for severe pseudomonal infections such as bacteraemia [7] or pneumonia [7], so far no study could really provide convincing evidence for the benefits of a targeted combined regimen for these infections once the results of the susceptibility testing became available, let alone for osteoarticular infections. Indeed, orthopaedic and infectious diseases studies reporting a dual therapy for P. aeruginosa [13, 14] seem to have similar remission incidences than reports with single-agent treatment [15, 26]. In one randomised comparative trial, seven cases of biopsy-proven P. aeruginosa osteomyelitis were treated with oral ciprofloxacin, and six patients received a standard parenteral therapy including an aminoglycoside plus an antipseudomonal beta-lactam during six–eight months. Although it appeared that patients on standard therapy did better (two vs four failures), the difference was statistically insignificant and numbers were too small to permit a valid evaluation [13, 25]. In another randomised study comparing oral ciprofloxacin to parenteral ceftazidime and amikacin, Gentry and Rodriguez reported equal efficacy of the regimens (77% remission vs 79%) [27].

Our study has limitations: (1) It was a retrospective, single-centre study with a small and heterogeneous population. The small sample size may hide differences that would have been seen if the analyses had been performed with a larger patient population. However, such a sample size has never been achieved for osteoarticular infections due to P. aeruginosa to the best of our knowledge. According to our estimation, a study population of at least 120 P. aeruginosa patients would be required. (2) Recurrences of osteomyelitis after several years, if not decades, have been reported [36] and there is no internationally accepted minimal follow-up duration. In the literature, minimal follow-up times range from three [16] to six months [21, 25, 26] or one [12, 22, 23] to two years [9, 10]. Hence, we consider our individual minimal and median follow-up times of 1.5 and four years as a strong point of our study. (3) Patients with failures treated in another hospital may have been undetected. However, Geneva University Hospitals are by far the largest and only public hospitals in the area; we therefore consider this possibility as low. (4) The completeness of initial surgical debridement as well as the consideration of tissue vascularity and bone necrosis are of paramount importance [23]. Surgeons know how extensively they had performed this debridement. There are no retrospective possibilities to estimate the completeness of such a debridement. (5) Staphylococcal small colony variants are associated with persistence of bone infections [2]. We could not assess their presence, since our specimens had not been stored. (6) Despite adjustment for case mix, we cannot completely exclude therapeutic decision bias. Patients, who were doing less well, might have deserved a longer antibiotic treatment. Likewise, patients infected due to P. aeruginosa may have special circumstances that explain this unusual pathogen. Indeed, a large part of our Pseudomonas cases were due to open fractures. (7) The period of data collection is from 1996 to 2009. During this period, surgical technique and knowledge about osteoarticular infections might have evolved.

Conclusions

Despite a similar number of surgical interventions and a trend towards longer antibiotic treatment, osteoarticular infections due to P. aeruginosa may tend towards a lower remission rate than infections due to S. aureus or MRSA. This assumption has existed for 20 years but has not yet been proven [37]. Our single-centre study warrants confirmation in large multicentre cohorts. If such larger databases confirm a worse outcome for P. aeruginosa osteoarticular infections, specific research on treatment strategies and prevention is needed. In today’s recommendations, only the choice of the antimicrobial agent [38] is pathogen dependent, but not its duration or the surgical approach, with a few exceptions for tuberculosis, fungal infection or nocardiosis [39].