Introduction

Native joint septic arthritis (NJSA) is considered as a medical emergency and a potentially life-threatening condition. Absence or delay of appropriate treatment can lead to irreversible joint destruction, disability or death. In a previous French retrospective study, the risk of an adverse joint outcome after a NJSA occurred in about 30%.1 2 Recently, a sixfold risk of knee arthroplasty within the 15 years following NJSA has been reported in patients who had undergone arthroscopic knee washout for NJSA compared with the age-matched general population controls.3 Older age, comorbidities and underlying rheumatic diseases have been identified as predictor of poor outcome. Also, even in recent cohorts, mortality remained high, around 5%–9%, occurring mainly within the first 90 days,1 3–5 and increased with age.6

NJSA is a rare condition, with an incidence ranging from 4 to 19/100 000 patient-year.4 5 7 8 Also, diagnosis could be challenging, particularly when no microorganism is identified. The optimal treatment (antibiotic and surgical procedures) is debated and heterogeneous depending more on practitioner habits than on evidence-based medicine. In 2017, this study was launched by French Rheumatology Society Bone Joint infection working group for analysing current practice and serve as bases for the French guidelines established in parallel.9 The previous guidelines dated back from 1991 in France and 2006 in Great Britain.10 11 Since then, the management and antibiotic use have considerably changed, but no formal consensus existed on the way to manage these patients. Also, multiple audits suggested that too often patients with NJSA did not receive the most accurate care.12

We, therefore, conducted this retrospective national survey in rheumatology departments, to describe ‘in real life’, the current management of NJSA and identify possible factors associated with mortality and adverse joint outcomes.

ResultsPatients

Fifty-two centres, 28 (53.8%) general hospitals and 24 (46.2%) tertiary care centres (university hospitals) reported 380 patients of which 362 were included in this study (figure 1 and online supplemental data S2-S3). Patients came to rheumatology department after direct admission in 280 cases (78.0%) or from other departments or medical institutions in 79 cases (22.0%).

Figure 1

Flowchart and strategy for microbiological identification. NJSA, native joint septic arthritis.

Patient characteristics are summarised in table 1. A pre-existing arthropathy of the affected joint was observed in 97 (28.3%) patients. Only nine (2.6%) patients had an inflammatory rheumatic disease involving the joint affected by NJSA. Of note, 87 (28.8%) patients had received antibiotics in the 3 months prior to hospitalisation, including 39 (12.9%) after NJSA onset and before bacteriologic investigations.

Table 1

Clinical and microbiological characteristics of the 362 patients with native joint septic arthritis

Clinical, biological and radiological presentation

Joint effusion was present in the vast majority of cases: 287 (83.9%), while fever was observed in only 149 (43.6%). Median joint pain visual analogue scale (0 to 10) at diagnosis was 7 (0–10) (table 1).

The most frequent presentation of NJSA was monoarticular arthritis (n=319, 89.6%) and mainly affected the knee (n=160, 38.9%) (table 1). Raised C reactive protein levels were observed in 351 (99.4%) patients, with a mean of 208.0 (±124) mg/L. Of the 312 patients with an initial radiography, only 50 (16.0%) had radiological features of NJSA.

Diagnosis and microbiological characteristics

Among the 362 patients (figure 1), a microorganism was identified in 340 patients (93.9%), by either synovial fluid analysis (n=272, 80.0%) or another bacteriological sample (n=68, 20.0%). PCR of ribosomal DNA 16s (PCR rDNA,16s) performed on 47 synovial fluids was positive in only nine cases (19.1%), including eight patients for whom other microbiological samples were positive. Thus, microbiological identification relies solely on a positive PCR in only one patient. Synovial fluid puncture was not performed in 43 patients because microbiologic diagnosis was obtained either by another bacteriologic sampling (n=24; 55.8%), by a surgical procedure (n=2; 4.7%) or because the involved joint could not be aspirate (n=17; 39.5%).

Microbiological species identified are listed in table 1. Enterobacterales NJSA were older compared with other NJSA (mean age in years 75 (±15) vs 63 (±19), p<0.001). Of note, 37 patients (14.7%) had concomitantly a crystal and a microorganism identified in the synovial fluid. Blood culture was performed in 346 patients (95.6%) and was positive in 156 (45.2%) patients. In the 90 patients for whom microorganism was not isolated by joint puncture (puncture not performed or with negative culture), microorganisms were most frequently detected by blood cultures in 43 (47.8%) (figure 1). The most common mode of contamination was haematogenous (n=118, 61.5%), while direct iatrogenic inoculation was recorded in 26 cases (13.5%) (table 1). Echocardiography was performed to detect infective endocarditis (IE), in 257 patients (80.0%)), with finally confirmed infective endocarditis (IE) in only 12 (3.3%). Ten patients had concomitant bacteraemia. Staphylococcus aureus was the most common pathogen involved in IE (n=7, 58.3%). No risk factor for IE based on age, comorbidities and microbiological features as been identified (data not shown).

Therapeutic management

In 356 patients (63.2%), antibiotic therapy was not started until the microorganism was identified either by Gram stain or a positive culture on synovial fluid or in the blood cultures. An infectious specialist advice was requested in 275 cases (75.9%), either initially (n=224, 61.9%) or subsequently because of worsening (n=51, 14%).

Antibiotic therapy was very heterogeneous in terms of antibiotic type, duration and route of administration (table 2). For example, among the 147 methicillin-susceptible Staphylococcus aureus NJSA, 19 different antibiotic combinations were used as the first-line treatment after microbiological findings, and 165 (45.6%) patients had to change at least two times antibiotic treatment. Use of antibiotic as monotherapy at initiation was more frequent in case of Streptococcus sp compared with Staphylococcus sp (52.4% vs 34.8%, p=0.006). Only 25 (6.9%) patients were initiated with orally administered antibiotic therapy. Among them, 12 subsequently switched for parenteral antibiotics because of microbiological findings (n=7), clinical or biological worsening (n=4), poor adherence (n=1) or infectious specialist advice (n=1).

Table 2

Therapeutic management of NJSA in all patients and according to the identified micro-organism

Overall, the mean duration of antibiotics was 46.8 (±22.0) days including a mean of 17.2 (±15.4) days of intravenous route (table 2).

Among all, 171 patients (48.4%) underwent a surgical procedure of 227 joints (table 2), mainly for knee NJSA (n=104). The most frequently used procedure was surgical lavage (n=162, 94.7%) associated or not with a synovectomy. Primary surgical management was decided at the diagnosis start of NJSA management in n=111 (65.7%) or after failure of initial conservative approach or worsening in n=58 (34.3%). Joint immobilisation was performed in n=129 (43.8%).

Outcomes

Mean length of hospital stay was 22.6 days (±16.3) (table 2). During this period, 91 patients (28.3%) had complications including 19 transfers in intensive care unit (online supplemental data S5).

One-year mortality was 9.2% (28/305 patients) rising to 21.3% (17/80 patients) in patients over 80 years old (online supplemental data S6). No association with mortality was observed based on microbiological characteristics or site of NJSA. Antibiotic duration was shorter in patients who died (p=0.029). However, this parameter was not included in the multivariate analysis since shorter duration was linked to the early deaths (ie, before the end of antibiotic treatment). In multivariate analysis, older age (adjusted OR (aOR): 1.08 [95% CI: 1.04 to 1.13], for each additional year), higher Charlson’s index (aOR: 1.30 (95% CI 1.06 to 1.58), for each additional point), presence of bacteremia (aOR: 4.02 (95% CI 1.35 to 11.99)), use of antibiotics in the 3 months prior to hospitalisation (aOR: 3.32 (95% CI 1.11 to 9.87)) were associated with 1-year mortality (table 3). Among patients having had antibiotics in the 3 months prior to hospitalisation, only those for whom antibiotics were started after NJSA onset and before microbiological investigations had a higher risk for mortality (alternative model, p<0.002). Staphylococcus aureus were associated with an increased risk of mortality, compared with Streptococcus sp (aOR: 7.24 (95% CI 1.26 to 41.68)). Adverse joint outcome at 1 year, evaluated in 278 patients, was observed in 125 patients (45.0%) (Description of local joint outcome in online supplemental data S5). No factor was associated to adverse joint outcome in multivariate analysis (table 4).

Table 3

Univariate and multivariate analysis of factors associated with mortality in the following year

Table 4

Univariate and multivariate analysis of factors associated to an adverse joint outcome

Culture-negative septic arthritis

Overall, 22 (6.1%) were considered culture-negative NJSA despite careful assessment (table 5). Compared with bacteriological proven NJSA, only use of antibiotic in the previous 3 months (aOR: 2.67 (95% CI 1.02 to 6.95)) was significantly associated with culture-negative NJSA in multivariate model adjusted on age (table 5).

Table 5

Univariate and multivariate analysis of factors associated to culture-negative native joint septic arthritis

Discussion

We here report the results of nationwide multicentric survey including a large number of patients with NJSA. Even in this recent study, NJSA remains a serious condition with high rate of 1-year mortality (9.2%), or adverse joint outcome (45.0%). Management of this condition was very heterogeneous, reflecting the lack of evidence or consensus and the absence of recent guidelines at the time we conducted this study. We identified age, comorbidities, bacteraemia, antibiotic use before bacteriological sampling and Staphylococcus aureus (compared with Streptococcus sp) as being associated with mortality. Antibiotic use before bacteriological sampling was also associated with a higher risk of culture-negative NJSA. But, none of these factors was associated to adverse joint outcome.

In the literature, case-definition of septic arthritis usually relies on modified Newman criteria.13 However, we intentionally chose broader inclusion criteria, in line with our main objective, being more representative of the real-life NJSA management. To our knowledge, multicentric surveys are unusual in the field of NJSA, studies being mostly monocentric and retrospective. Thus, our study was more representative of the nationwide variety of management of NJSA.

Clinical and microbiological characteristics of our patients were closed to those found in other studies.1 4 Of note, 28.3% of NJSA had pre-existing arthropathy on the affected joint, mainly osteoarthritis (18.9%), and crystal arthropathy (8.2%). Rheumatoid arthritis (RA), commonly described as a risk factor for NJSA,14 15 was found in only 4% of patients, and was even lower if we considered only cases where RA affected the same joint as the NJSA. This result is line with a 2%–4% rated observed in more recent cohorts1 4 5 16 could be explained by a better management of RA with less articular damage.

The most frequent causative organisms were Staphylococcus sp and Streptococcus sp as previously described in literature.1 17

Diagnotic strategy to obtain microbiological documentation was very homogenous and congruent with guidelines.9 11 18 Our study supports the systematic implementation of joint aspiration and blood cultures. Interestingly, blood cultures were very effective in obtaining bacteriological diagnosis, being positive in almost half of all the patients, including those with no organism identified in the joint or for whom joint aspiration could not be performed. In previous studies, frequency of bacteraemia was rarely reported but accounted for 24% to 46% of NJSA.1 15 19–21 Likewise, there was limited data on frequency of infective endocarditis until a recent study reporting a rate of 3.7%,22 consistent with our findings.

The frequency of culture-negative NJSA observed here (5.5%) was in the lowest values of literature (4% to 17%),4 23–25 may be explained by the extensive microbiological searches performed in our patients. The only risk factor identified being antibiotic use in the previous 3 months, this highlights the importance of not giving antibiotics prior to microbiological sampling when NJSA is suspected.

By contrast, therapeutic management was very heterogeneous and ensues from the insufficiency of evidence in literature and the absence of recent guidelines at the time of the study. The widely variable length of admission to hospital after disease onset highlights that NJSA may be misrecognised or underappreciated by primary physicians. Antibiotic treatment was mostly started after knowledge of microbiological findings and had an average overall duration of 6 weeks, in line with recent studies in large joint NJSA reporting an average duration of 5 to 7 weeks.1 5 Even if the trend is to shorten antibiotic duration, optimal duration for antibiotic treatment in NJSA is uncertain. For uncomplicated hand small joint NJSA with surgical management, Gjika et al 16 reported that 2 weeks were sufficient compared with 4-week antibiotic therapy. But, we caution against generalisation of these results outside small joint NJSA.26 Our study population was very different with 91% of large joint involvement, and occurring mostly after a haematogenous seeding. Nevertheless, long-term exposure to antibiotics increases bacterial resistance. Currently, a French nationwide trial initiated by members of these study group in collaboration with infectious disease specialists aims to evaluate whether a shorter antibiotic treatment (3-week treatment) is safe and not inferior to the conventional 6-week treatment in NJSA (SHASAR, NCT03716921). Of note, in this trial, as in the OVIVA trial, the duration of antibiotic in the control arm is of 6 weeks, demonstrating that it is still the reference standard. Nevertheless, we now know from the OVIVA trial, that the duration intravenous course should be shorter than the 15 days, we here observed since even in patients with complex bone or joint infections to 7 days was sufficient.27

Overall, 45% of patient had adverse joint outcome within 1 year, in line with literature ranging from 24% to 49%.1 15 16 Surgical drainage accounted for the half of our population, in line with the literature (37%–83%).1 5 28 However, evidence regarding the need for a systematic surgical drainage or its best timing is scarce. A non-operative approach of NJSA management has not been yet prospectively studied. By contrast with Flores-Robles et al who did not observe worst outcomes with medical or surgical management,29 in our cohort, patients undergoing surgery had higher rate of adverse joint outcome. However, we could not exclude an indication bias, since patients undergoing surgery probably had a more serious condition. Effectively, when focusing on systematic surgical joint drainage decided as a part of initial management, surgery was no more associated with worst outcome. Regarding surgical modalities, arthroscopy or arthrotomy, having the same effectiveness,30–33 some authors support that arthrotomy could be associated to a worst joint outcome and a longer recovery.30 31 We here found the same trend, even though non-significant.

Mortality raised to 9.2% in our study, in the same range of studies mainly including large joint NJSA,3 5 19 22 but higher than in studies with different recruitments reports (around 6%).1 4 Small NJSA have different epidemiology and lower mortality.5 As expected,3 15 19 age and Charlson’s comorbidity index were independently associated with 1-year mortality. Also, patients with bacteremia had a fourfold higher risk of mortality, as previously reported.19 Mortality raised to 18% in patients with IE. We are the first to report that Staphylococcus aureus NJSA had a worst prognosis with a nearly sevenfold higher risk of death as compared with Streptococcus sp. In contrast to previous report, polyarticular involvement was not associated with increased mortality.2 15 19 For the first time, we report that use of antibiotics in the previous 3 months before admission was associated with mortality. This observation might reflect the frailty of patients who are more likely to contract multiples infectious diseases, and/or that antibiotics prescribed before microbiologic samples might complicate the diagnostic and therapeutic management of NJSA. Besides, antibiotics for NJSA started before microbiological samples could have been prescribed in frailty patients with more comorbidities.

Our study had important limitations, such as its retrospective nature. Outcomes, obtained from medical records by each participating centre, might lack of standardisation and precision. Nevertheless, it also has some strength such as its large sample size and representativeness of management of NJSA in rheumatology departments of both tertiary care centres and general hospitals. Since all patients were recruited in rheumatology departments, we cannot exclude a recruitment bias compared with surgical recruitment, notably older patients with more comorbidities. However, in France, management of NJSA relies on multidisciplinary approach formalised in many centres by the CRIOAc (‘Centre de Référence des Infections Ostéo-articulaires Complexes’ - reference center for complex joint and bone infections) healthcare network,34 as observed here, with frequent involvement of infectious specialist (>75%) and orthopaedic surgeons (surgical management in ~50% of the patients).

In this study, NJSA has serious consequences with 9.2% mortality and 45.0% of adverse joint outcome within the year, increasing in older patients with comorbidities. This study emphasises that no antibiotic should be given before microbiological diagnosis. Management of NJSA was very heterogeneous across the different centres. Since then, new recent guidelines of the French society of rheumatology on management of NJSA have been published9 and might have improved harmonisation of NJSA management. Finally, whether antibiotic duration might be shortened or not, should be analysed in prospective randomised controlled trials.

Acknowledgments

The authors thank all of the physicians who made inclusion of patients possible: Baptiste Glace, Laetitia Dunogeant, Constance Beaudoin, Bernadette Saint Marcout, Fadela Daoued, Moana Sulpice, Guillaume Servant, Elsa Cattelain-Lopez, Fabienne Le Guilchard, Lucia Andras, Gérald Rajzbaum, Armelle Lamour, Sandrine Guis, Francis Berenbaum, Pascal Claudepierre, Maxime Breban, Jean Louis Legrand, Martine Gayraud, Ahmed Benmansour, Christian Lormeau, Ourida Ighilhariz, Philippe Gaudin, Christian Cadene, Gilles Bolla, Stéphanie Paupiere, Elisabetta Lanciano, Yannick Allanore, Bruno Fautrel, Marie Hélène Guyot, Loïc Le Dantec, Thierry Schaeverdeke, Eve David Vaudey, Véronique Belin. We are grateful to Jean-Marie Gouin (Public Health and Epidemiology Informatics) for assistance in data collection. We are in debt to Véronique Gordin for logistic assistance in relation with this project.