Abstract
Introduction
This study aimed to describe the epidemiology of the causative organism in a series of primary hip arthroplasties revised for a diagnosis of periprosthetic joint infection (PJI) in England and Wales.
Methods
Patient data from the National Joint Registry (NJR) was linked to microbiology data held by Public Health England (PHE) which identified a series of 248 primary hip arthroplasties revised for PJI between 2003 and 2014. Definitive cultures, isolated at time of revision surgery, were available for all cases. Total hip arthroplasty (n = 239, 96%) and hip resurfacing (n = 5, 2%) were the most commonly performed primary procedures. A two-stage revision was the most common operative management (n = 174, 70%).
Results
202 (81%) cases were infected with a single genus microorganism and the most commonly implicated genus was Staphylococcus species (70% of all single genus infections). Staphylococcus species were also the most commonly identified microorganism in mixed genus infections (74% of patient's cultures). There was a significant difference in microorganism distribution when comparing uncemented vs cement implant fixation, with a higher incidence of Gram-negative infection observed in the uncemented group (p = 0.048, Chi-square).
Conclusions
Both prophylactic and therapeutic antibiotic regimes should be focused on targeting Staphylococci.
Introduction
The incidence of periprosthetic joint infection (PJI) following primary hip arthroplasty is between 0.5 and 2% (1-2-3-4-5). PJI is a devastating complication which can lead to significant patient morbidity and an increased risk of mortality (6-7-8). In contemporary studies, the most frequently implicated organisms confirmed at the time of revision for PJI are staphylococci (coagulase-negative staphylococci and Staphylococcus aureus) (1, 2, 9, 10). The management of PJI often requires multiple surgical interventions, prolonged inpatient admissions and extended antibiotic treatment. The complexity and duration of treatment is associated with significant costs, not only financially, but also in terms of the burden of care for the treating team and the patients. After revision for PJI, patients also have to contend with a decreased functional outcome and reduced implant survival when compared to other indications for revision (8, 11, 12).
In combination with patient and surgical factors, the outcome following revision arthroplasty for PJI is influenced by the ability to isolate organism(s) at time of surgery, the virulence of such an organism and its antimicrobial sensitivity profile (6, 10, 13-14-15-16-17). PJI is a rare event and as such the ability to acquire cases and surgical experience is restricted. Previous work is limited by the number of cases available, with reports typically originating from individual or a combination of a small number of centres (1, 2, 9, 10, 18) from a localised geographical region. As bacterial prevalence varies dependent upon geography the results from these studies may not be representative of the epidemiology for the wider population (10).
We therefore undertook this study to try and ascertain a national perspective and describe the epidemiology for the causative organism in a series of primary total hip arthroplasties revised for a diagnosis of PJI in England, Wales and Northern Ireland. We were specifically interested in determining whether there was significant variation in patient demographics and surgical factors dependent upon the type of infecting organism. This analysis extends previous work done by our group analysing the organisms implicated at the revision in a series of primary knee arthroplasties (19).
Patients and methods
The National Joint Registry (NJR) has prospectively collected information concerning primary and revision hip arthroplasty procedures performed in England and Wales since April 2003 and in Northern Ireland since February 2013. We used the NJRs standard processes for research to request information on all primary hip arthroplasties implanted since the inception of the NJR that subsequently went on to have a revision for infection prior to the 27th February 2014. This request yielded a consecutive series of 1,998 primary arthroplasties revised for a diagnosis of PJI.
National Joint Registry dataset
Cases revised for infection were identified within the NJR dataset based on the ‘reason for revision’ information recorded by the surgeon contemporaneously at the time of surgery and subsequently uploaded to the NJR using a specific data collection tool. The recording of a case as infected is therefore based upon preoperative clinical review, relevant biochemical testing and microbiological investigation, as well as any perioperative surgical findings. No additional criteria needed be satisfied (e.g. a microbiological diagnosis) for these revision procedures to be categorised as infected other than that the attending surgeon was of the opinion the cause of revision was infection. Any revision that had the ‘Infection’ box ticked as a reason for revision was included in the analysis irrespective of whether other reasons were chosen. This follows the hierarchical method for ascribing the reason for revision used by the Australian Arthroplasty Register (20) and used by our group previously (21). The NJR data collection forms also collect information on patient demographics (age, gender, American Association of Anaesthesiologists grade (ASA) and Body Mass Index (BMI)), date of primary/revision surgery); in addition to surgical information pertaining to both the primary and revision procedure; as well as implant details for these procedures. This information was identified from the NJR dataset for each primary and revision procedure.
Microorganism dataset and linkage
Having identified the NJR cohort we were then able to link each case to corresponding microbiology data held by Public Health England using patient National Health Service (NHS) number. The Second Generation Surveillance System (SGSS) held at Public Health England is a voluntary laboratory reporting system containing both the Communicable Disease Reporting (CDR) and antimicrobial susceptibility testing (AmSurv) databases. Nearly all microbiology laboratories report their results via CDR and over 75% of laboratories are now reporting to AmSurv. For CDR all organisms specified in Schedule 2 of the Health Protection (Notification) Regulations 2010 should be reported as well as all clinically significant infections for 24 viruses, 39 bacteria and 30 fungi. All clinically significant isolates from sterile sites such as blood, CSF, joint fluids, bone, pleural and pericardial fluids, heart valves, and abscesses in the brain, liver and spleen should be reported. Laboratories are expected to report each week and preliminary reports (with the exception of mycobacteria) should be received by PHE within 14 days of the specimen date in at least 90% of reports, and no later than 21 days. Currently the CDR database contains reports of over 30 million infections. The AmSurv database contains the results of all antimicrobial susceptibility tests performed on an isolate and currently contains over 60 million individual drug-bug results. The Demographics Batch Service (DBS) is used to enable Personal Demographics Service (PDS) to provide the NHS Number and most up to date demographics for the records in SGSS.
Deterministic data linkage of the patient National Health Service (NHS) number from the NJR datasets, stripped of information regarding the operation, to the SGSS microbiology data held by PHE was undertaken at PHE using the data management commands available in STATA version 13. All reported infections and antimicrobial susceptibility results that linked to a particular patients NHS number were retained regardless of the date of infection, the pertinent date being selected during the specific analyses.
All available culture specimen data (at microorganism genus level) were retrieved for each patient from both CDR and AmSurv databases. Data at microorganism species level was not consistently recorded by all contributing laboratories; we therefore focused our analysis at the genus level in order to reduce any potential reporting bias. For each patient a range of microbiological data was available. We therefore applied strict linkage criteria to determine those cultures most likely to represent ‘deep’ samples obtained at the time of revision surgery. This was performed by including only specimens with a recorded collection date which matched the date of revision surgery or the day after in order to allow for a delay in sending of cultures or receipt by the receiving laboratory. Specimens were then restricted to include only those likely to represent deep samples (Fig. 1). Once these criteria had been applied, microbiology data was available for a study population of 248 primary hip arthroplasties revised for infection. There were no patients in this linked dataset who had their revision procedure recorded in Northern Ireland.
Description of the study cohort.
Study cohort demographics
Total hip arthroplasty (n = 239, 96%) and hip resurfacing (n = 5, 2%) were the most commonly performed primary procedures. A two-stage revision was the most common operative management (n = 174, 70%) followed by single-stage revision (n = 62, 25%) and excision arthroplasty (n = 12, 5%). Mean age at time of primary procedure was 67 years (range 18-98 years, SD 11) with the most common indications being osteoarthritis (n = 223, 90%) and avascular necrosis (n = 9, 4%). Mean BMI was 31.6 kg/m2 (19-50 kg/m2, SD 6.1). Primary implant fixation was uncemented in 108 (44%) cases. The majority of patients were ASA grade 1 (n = 46, 19%) or grade 2 (n = 156, 63%). Median time to revision was 1.7 years (<1 month to 10.1 years).
Statistical analysis
The demographic characteristics of 5 defined microorganism groups were compared. These groups were: (1) pure Staphylococcus (single genus); (2) pure Streptococcus (single genus); (3) other Gram-positive infections (single genus); (4) Gram-negative infections (single genus); (5) mixed genus infections. For clarity, a “pure Staphylococcus (single genus)” would include a Staphylococcus aureus pure growth or, for example, a mixed growth but comprising only staphylococci (e.g. Staphylococcus aureus and Staphylococcus epidermidis).
The null hypothesis was that there is no difference in demographic characteristics when comparing microorganism groups. Following assessment of data normality, parametric continuous variables were compared between organism groups using analysis of variance (ANOVA) followed by post hoc testing (Bonferroni) where appropriate. Kruskal-Wallis test was used to compare non-parametric data (e.g. time to revision). Chi-square test was used to compare categorical variable proportions between groups. A p value of <0.05 was deemed to be statistically significant. Data linkage was achieved directly between the NJR and Public Health England. As such, no patient sensitive information (NHS number) was held by Newcastle University as part of this analysis; this study was therefore classed as a service evaluation and ethical approval was not required. After methodological review by Public Health England's confidentiality advisory group, section 251 endorsement was not required.
Results
A total of 302 culture results recorded within 24 hours of revision surgery were identified in 248 patients following linkage of NJR and PHE data. Patient demographics of the study cohort were compared to those for the larger cohort of all infected revisions initially identified from the NJR dataset (n = 1,998) and a cohort of all revision hip arthroplasties performed for all causes recorded by the NJR in 2013 and reported as part of their 2014 annual report (Tab. I) (22). A significantly higher proportion of males were observed in the infected revision group compared to the ‘all cause’ hip revision group (p<0.001, Fisher's exact test). The demographics of the study population were similar to those of the larger cohort of all infected revisions in terms of age (p = 0.07, Student's t-test), BMI (p = 0.34, Student's t-test), gender (p = 0.25, Fisher's exact test) and ASA grade (p = 0.90, Fisher's exact test).
Demographic characteristics of the study cohort were compared to those for the larger cohort of all infected revisions initially identified from our dataset (n = 1,998) and a cohort of all revision hip arthroplasties performed for all causes recorded by the NJR in 2013 and reported as part of their 2014 annual report (22)
During the period of study 159 centres revised at least 1 primary THR for an indication of infection. Of these 67 (42%) submitted data to PHE. These 67 centres undertook 1,216 of the 1998 revisions for infection (61%) during this period and while this suggests there may be reporting bias towards centres with higher volumes it also suggests that, with 60% of operations undertaken within these centres, we have a representative sample of the population of England and Wales.
Microorganism(s) from a single genus were implicated in 202 cases (81%). Mixed genus infections (comprising 2 or more organism genera) occurred in 46 cases (19%). For both single and mixed genus infections, Staphylococci were the most frequent organisms isolated and found in 176 patients overall (71%). Observed frequencies of each of the most common organism genera are shown for single and mixed genus infections (Tabs. II and III). The relative proportions of microorganisms implicated in this series of hip arthroplasties were similar to our previous work reporting a series of 331 knee revisions for PJI using the same linkage methodology (Tab. IV) (19).
Frequency of microorganisms observed in single genus periprosthetic infection (n = 202)
Frequency of microorganisms observed in mixed genus periprosthetic infection (n = 46)
Organism group
Infecting microorganism epidemiology was similar for hip and knee arthroplasties identified from the NJR dataset (p = 0.75, Chi-square) (19).
There were no statistically significant differences in the distributions of patient and surgical characteristics dependent upon the 5 microorganism groups previously described (all p>0.05) (Tab. V). However, when ‘hybrid’ and ‘unrecorded’ cases were excluded, the use of cement fixation (n = 76) (antibiotic impregnated in almost all cases) was associated with a significant difference in observed microorganism distribution compared to uncemented fixation (n = 108) (p = 0.048, Chi-squared) (Tab. VI). This significance was driven by a higher frequency of Gram-negative infection in the uncemented group (n = 15, 14%) compared to cemented (n = 2, 3%) and by a higher frequency of ‘other Gram-positive’ in the cemented group (n = 10, 13%) relative to uncemented group (n = 6, 6%). The frequency of other microorganism groups was otherwise evenly distributed when comparing uncemented and cemented fixation.
Comparison of microorganism groups
= Statistical analysis compared indication of osteoarthritis vs all other indications (Chi-square).
Microorganism groups by fixation type
Frequency and row percentage, n = 184, p = 0.048, Chi-square.
Discussion
The current study demonstrates that, in a national cohort of primary hip arthroplasties revised for a diagnosis of infection, 81% of cases were infected with a single genus microorganism and the most commonly implicated genus was Staphylococcus species (70% of all single genus infections). Staphylococcus species were also the most commonly identified microorganism in mixed genus infections where they were identified in 74% of patient's cultures. There was a significant difference in microorganism group distribution when comparing uncemented fixation vs use of bone cement (antibiotic loaded in almost all cases), with a higher incidence of Gram-negative infection observed when using an uncemented technique. There were no other observed differences between microorganism groups for the distribution of a number of patient and surgical demographics including age, BMI, gender, ASA grade, indication for primary procedure and primary implant characteristics.
The number of cases for which microbiology data was available (248) represented a small proportion (12%) to the total cohort (1,998) which may limit the generalisability of our results. However, despite the low rate of data linkage, our dataset represents one of the largest reported cohorts of hip arthroplasties which were deemed infected by the judgment of the attending surgeon and further substantiated with microbiological data. There are several factors we believe may contribute to the low rate of data linkage. Most importantly, reporting of microbiological results to the national databases is performed on a voluntary basis. In addition we used strict inclusion criteria to identify samples to help ensure they were taken at time of surgery.
Secondly, there may be miscoding of cases as infected within the NJR database, as surgeons do not need a confirmed microbiological diagnosis to select this reason on the NJR data collection tool. Surgeons may therefore record a case as infected based on a clinical suspicion that later turns out to be incorrect. For that reason, the number of cases recorded as infected within the NJR may be an overestimation of the true number and highlights the value of linking to additional data sources that hold information that can enhance and clarify the information recorded within the NJR. As such, we would be wary of any analysis of registry data that implies a case is revised as infected without additional microbiological confirmation. Other reasons for the lower than expected number of cases include culture-negative specimens not being recorded in the CDR and AmSurv databases, despite selecting ‘infection’ as a cause for revision not all surgeons send cultures at time of revision surgery.
We attempted to identify specific patient and surgical demographic traits that were associated with infection with a specific organism genus. Uncemented fixation appeared to be associated with an increased relative frequency of Gram-negative infections as compared to cemented fixation. This finding is an important consideration for the revision surgeon, who should consider a Gram-negative aetiology in patients presenting with infection following uncemented primary fixation. Studies examining Gram-negative PJI are few in number, however, there is some evidence to suggest that they may occur in older patients, are more challenging to treat and associated with a lower rate of success of revision surgery in comparison to Gram-positive infections (23, 24). The use of antibiotic bone cement as a factor in reducing PJI has been described previously and, in the present study, antibiotic bone cement was used in more than 98% of cases of cemented fixation, with the most common form being gentamicin impregnated (‘Palacos’). Gentamicin is an aminoglycoside antibiotic that is highly active against Gram-negative organisms which, we hypothesise, may offer reduced susceptibility to such infections (25, 26).
Unfortunately the relative number of infections for organisms other than Staphlococcus species was small (e.g. Streptococcus n = 20, pure other Gram-positive n = 20, Gram-negative n = 20). Our comparative analyses were therefore likely to be underpowered to detect anything other than very strong associations which may explain the observed results. For example the infections with Gram-negative organisms occurred in ASA 3 or 4 patients in 40% of cases. This compares to a figure of 20% or less in all other organism groups. In addition, the time to revision for Gram-negative organism was shorter than for staphylococcal and streptococcal infections (0.9 years vs 2.1 and 1.9 years respectively). While not significant in this analysis, these findings are of interest, as they suggests increasing co-morbidity and frailty may be linked to a greater risk of Gram-negative infection which may in turn increase the risk of earlier revision.
There is potential for reporting bias from laboratories contributing to CDR and AmSurv databases. Currently, reporting to these databases is voluntary and there is inconsistency between laboratories in the way that specimen types are reported. Patients identified within the microbiology databases may therefore be a ‘selected’ group. It is therefore reassuring that the population of patients for whom samples were available were demographically similar to the entire cohort of infected revisions identified from the NJR (Tab. I).
As with most registries, the NJR provides limited data on the outcomes of revision surgery. The focus for registries is the monitoring of primary implants and revision is typically viewed as the endpoint. It is fundamental to informed reporting to have further information about the outcome of revision procedures and, for revisions for infection, this should include complete data pertaining to the microbiological interpretation of biological samples retrieved in the perioperative period.
In comparison with findings from this study, previous studies examining geographic variations in organism profile have also found high rates of staphylococcal infection (1, 9, 10). Aggarwal et al (10) examined tissue cultures taken at the time of revision for infection from a cohort of 1,670 primary hip and knee arthroplasties performed at two centres in Germany and the United States (US). The overall incidence of pure staphylococcal infection was 51% at the US centre and 52% at the European centre. They observed significant variation in the distribution of infecting organism between these centres with higher rates of coagulase-negative staphylococci infection in the European centre (39% vs US 20%) and a higher rate of methicillin-resistant Staphylococcus aureus (MRSA) in the US centre (48% of all Staphylococcus aureus vs European 13%). The significant difference in infecting organism profile observed between centres demonstrates the geographical variability further highlighting the need for population level data analysis employed by the present study.
In their examination of 41 primary knee and 34 primary hip arthroplasties revised in the United Kingdom for PJI, Philips et al found staphylococci to be implicated in 65% of cases (coagulase-negative Staphylococcus = 36%, Staphylococcus aureus = 25%, MRSA = 4%) (1). They reported a mean time to diagnosis of 1.2 years which was lower than the median time to revision observed in the present study (1.6 years). This is probably due to the additional time taken to get patients to surgery following their initial diagnosis. A superficial wound infection treated with antibiotics before deep infection developed occurred in 40% of patients, and in 39% of patients potential aetiological factors for PJI were identified (perioperative sepsis, urinary catheterisation etc.). In comparison to the present study more than one organism was identified in 31% of cases. However, direct comparison with the incidence reported in our study was not possible as data concerning microorganism species was not consistently reported to PHE meaning we were only able to report mixed genus infections which is likely to be an underestimation of the number of true ‘polymicrobial’ infections.
The present study is the largest multicenter analysis of infecting organism in PJI following primary hip replacement. National sampling demonstrated that staphylococci are the most frequently isolated organism species responsible for PJI following primary hip arthroplasty. Despite linking two large national databases there was a substantial volume of missing data (approximately 88% in this study). The nomination of ‘infection’ as a primary indication for revision in the absence of definitive tissue confirmation is therefore questionable. The increasing global threat of antibiotic resistance necessitates close examination and interpretation of epidemiological data at population level in order to guide appropriate prophylactic and therapeutic management strategies.
Footnotes
Acknowledgement
We thank the patients and staff of all the hospitals in England, Wales and Northern Ireland who have contributed data to the National Joint Registry. We are grateful to the Healthcare Quality Improvement Partnership (HQIP), the NJR Research Sub-committee, staff at the NJR Centre and Public Health England (PHE) for facilitating this work. The authors have conformed to the NJR's standard protocol for data access and publication. The views expressed represent those of the authors and do not necessarily reflect those of the National Joint Registry Steering Committee or the Health Quality Improvement Partnership (HQIP) who do not vouch for how the information is presented.
Financial support: None.
Conflict of interest: None.