The PJI Study: Do Invasive Dental Procedures Cause Prosthetic Joint Infections (PJI)?

The University of Sheffield requires Hospital Episode Statistics (HES) and patient identifiable linking data for use in The PJI Study: Do Invasive Dental Procedures Cause Prosthetic Joint Infections? This is a study of the potential link between Invasive Dental Procedures (IDP) and Prosthetic Joint Infections (PJI). The results of this study will provide much-needed evidence to either support or challenge current UK dental prescribing guidelines, as well as those in use in other parts of the world.

Replacing diseased and worn-out joints with prosthetic joints is one of the great advances of modern medicine. By 2016, 2.9 million (m) joint replacements were performed annually worldwide (including 1.5m hip, 1.1m knee and 100,000 shoulder joint replacements). After loosening of the prosthesis or joint dislocation, infection is the most common complication. Most infections occur in the 3 months following joint replacement and are largely the result of wound contamination at the time of surgery – so called “early infections”. Late peri-Prosthetic Joint Infections (LPJI) – infections occurring 3 months or more after joint placement – are more often attributed to haematogenous (carried by the bloodstream) seeding of bacteria from another site such as the mouth. The cost of treating such infections is high – often 4 to 6 times higher than the cost of the original joint replacement procedure – and the associated morbidity and complications for the patient are also extremely high. This has led orthopaedic surgeons to seek ways to prevent bacterial pathogens entering the circulation (bacteraemia) that could lead to this devastating complication. Most of the focus on bacteraemia prevention in recent decades has been on dental procedures; specifically the concept of giving antibiotic prophylaxis (AP) prior to invasive dental procedures (IDP) in order to prevent bacteraemia as a result of IDP.

There is currently a divergence of guidance between the UK – where AP to patients with prosthetic joints prior to IDP is not recommended, and the USA, Canada and other countries where AP in these circumstances is recommended. Therefore this research is required to on the one hand understand the risks of the current UK guidance and on the other hand to better understand whether AP prior to IDP is warranted, or effective in the prevention of LPJI, due to decades of conflicting advice from dental and orthopaedic surgical professional bodies on the subject because of a lack of evidence in support of the practice. There is no scientific data to support the use of AP to prevent LPJI and there has never been a randomised controlled trial (RCT) of AP. Furthermore, there is little microbiologic data to support a causal link between IDP and LPJI. The American Dental Association (ADA), has therefore recommended that “In general, for patients with prosthetic joint implants, prophylactic antibiotics are not recommended prior to dental procedures to prevent joint infections.” The American Academy of Orthopaedic Surgeons (AAOS), however, has stated that “Given the potential adverse outcomes and cost of treating an infected joint replacement, the AAOS recommends that clinicians consider antibiotic prophylaxis for all total joint replacement patients prior to any invasive procedure that may cause bacteraemia”. This on-going pressure from orthopaedic surgeons for patients with prosthetic joints to receive AP when undergoing IDP means that many patients, on the advice of their orthopaedic surgeon, expect their dentist to provide AP. As a result, many, if not most, American dentists continue to give AP for fear of being considered negligent by their patients and their surgeon.

If there is no association between IDP and LPJI then this would justify the UK position that there is no benefit to the current USA practice of giving AP to the vast majority of prosthetic joint patients for virtually all dental office visits. The continuing use of AP in the USA would therefore represent a large and unnecessary financial burden on individuals and the healthcare system, as well as an unnecessary risk to patients (from adverse drug reactions) and society (from the potential development of antibiotic-resistant bacteria). On the other hand, if a clear link is found between IDP and LPJI, then the UK policy would require re-evaluation because of the significant treatment cost (often 4 to 6 times higher than the cost of the original joint replacement procedure) of LPJI, and this would provide support for the use of AP in the USA to prevent LPJI. If that clear link between IDP and LPJI is found, then it will be necessary to evaluate if there would be cost-savings in the UK from the introduction of AP to prevent LPJI against the cost of treating potentially preventable LPJI cases – policy changes as a result of this study will therefore result in a basis on which to justify cost savings in the UK.

Any association between IDP and LPJI can only be studied in the UK, where AP prior to IDP has never been advocated for patients with prosthetic joints, and where a national dataset exists; such a study in the USA is compromised by the widespread use of AP – which would tend to hide any association. Indeed, since 2000 the British National Formulary has stated that dentists or physicians should NOT give AP to patients with prosthetic joints. Consequently, there is no history of AP use prior to IDP for patients with prosthetic joints in the UK. This means that any association between invasive dental procedures and LPJI will be fully exposed in the UK data, enabling this study to provide the data to settle the confusion once and for all, and provide the evidence for, or against, any association between invasive dental procedures and LPJI. Professional organisations, guideline committees, dentists, orthopaedic surgeons and their patients can then use this evidence to make decisions about the value of prescribing AP prior to IDP for patients with prosthetic joints

Processing of personal data for this study is justified under GDPR Article 6 (1) e: processing is necessary for the performance of a task carried out in the public interest, due to the current lack of evidence on which dentists can base their prescribing decisions, the huge, potentially unnecessary, financial burden of widespread routine use of AP across the USA, and the significant, again potentially unnecessary, risk to patients (from adverse drug reactions) and global society (from the potential development of antibiotic-resistant bacteria) that must arise from widespread, routine antibiotic use. Processing of personal data concerning health is also justified under GDPR Article 9 (2) j: processing is necessary for scientific research purposes and shall be proportionate to the aim pursued; this will be by far the largest study of the purported association between IDP and LPJI ever undertaken, and will therefore have far greater statistical power to detect an association than any previous study, as well as being the first study performed in a population where AP prescribing does not confound the outcome of the study by hiding or reducing the likelihood of detecting an association between IDP and LPJI. As per GDPR Article 9 (2) j, the processing of personal health data for this study will also respect the essence of the right to data protection and provide for suitable and specific measures to safeguard the fundamental rights and the interests of the data subject, as detailed within the privacy notice webpages for the University of Sheffield and for this specific project.

This study will use HES data to identify all patients who had a hospital admission for a joint infection between 1st April 2010 and 31st March 2017, and also retrieve additional HES data for admissions relating to these patients which include a procedure code indicating a prosthetic joint replacement, covering the period prior to the joint infection as far back as possible (1st April 2000 to 31st March 2017), in order to retrieve details of any prior prosthetic joint replacement procedures for as many of the patients as possible. This will enable identification of the type of joint replaced for each patient, facilitating sub-analysis of the data by type of joint replacement (hip vs knee vs other type of joint). This historical HES data detailing the timing of each patient’s joint replacement procedure will also enable the exclusion of patients with an early prosthetic joint infection (within 3 months of joint replacement) from the study, as these early infections are largely the result of wound contamination at the time of surgery and this study therefore focuses on LPJI only. Using personally identifying details, the group of patients identified via the HES data will be linked to routinely-collected dental data held in the NHS Business Services Authority Dental Information Services database, to identify which patients had an IDP in the period preceding a PJI. Patient identifiable data will be processed and linked by NHS Digital and NHS Business Services Authority Dental Information Services (the Data Processor) only, with pseudonymised HES and dental data then being passed to the study team at the University of Sheffield (the Data Controller) for analysis. The University of Sheffield will not receive patient identifiable data.

The primary objective of the study is to perform a case-crossover design study to quantify the incidence of IDP in the 3 months immediately preceding an LPJI diagnosis (case periods) and compare this with the incidence of IDP in earlier 3-month periods for the same patient, i.e. 3-6, 6-9, 9-12 months before the LPJI diagnosis (matched control periods; each patient acts as their own control) - to see if there is any association between IDP and LPJI. If there is a link between IDP and LPJI, the expectation is that the frequency of IDP would be significantly higher in the 3 months immediately preceding a LPJI (case periods) than in earlier 3-month (matched control) periods. Alternatively, if there is no link between IDP and LPJI, the expectation is that there would be no significant difference in the incidence of IDP in the 3 months immediately preceding a LPJI (case periods) and earlier 3-month (matched control) periods.

In addition, as a secondary objective the study will include a case-control design study comparing the frequency of courses of dental treatment that DO (cases) and DO NOT (controls) involve an IDP in the 3 months immediately preceding a LPJI. This will enable the study to further test if there is an association between IDP and LPJI using a different analytical/statistical approach to that used in the primary objective. If there is a link between IDP and LPJI, the expectation is that there would be a significantly higher incidence of cases than controls in the three months immediately preceding a LPJI. Alternatively, if there is no link between IDP and LPJI, the expectation is that there would be no significant difference between cases and controls.

If the study finds a link between IDP and LPJI, three sub-analyses will also be performed:

1. A sub-analysis to see if the link involves all types of prosthetic joint or particularly affects specific types of prosthetic joint (hip vs knee vs other type of joint). Procedure codes from the historical HES data allow sub-analysis of the data by the type of joint replaced.

2. A sub-analysis to examine the risk of developing LPJI with different types of IDP (extractions vs dental scaling vs endodontic treatment). Because each type of IDP is recorded separately in the NHS Business Services Authority dental database, the study can look at the link between LPJI and each of these IDP individually.

3. The oral bacteria most commonly implicated in LPJI are oral viridans group Streptococci, (OVGS), and these are the organisms that are principally targeted by amoxicillin AP. If there is a link between IDP and LPJI, it should be strongest, therefore, for those cases of LPJI caused by OVGS, and weak or non-existent for cases of LPJI caused by bacteria not typically associated with the mouth, e.g. Staphyloccoci. The study will therefore use the diagnosis codes from the HES data for the PJI admission for each patient to perform a sub-analysis of the link to IDP of cases of LPJI caused by OVGS, and for comparison Staphylococci and other types of Streptococci.

The PJI study will link national data on courses of dental treatment (NHS Business Services Authority Dental Information Services database) and hospital admissions for Prosthetic Joint Infections (PJI) (Hospital Episode Statistics (HES) database) to investigate if there is a link between invasive dental procedures and the development of these infections in individuals who have prosthetic joints.

NHS Digital will identify within the HES inpatient admissions data all admissions between 1st April 2010 and 31st March 2017, with any of a specified list of diagnosis codes (present in any diagnosis field) indicating a joint infection (ICD-10 codes M00.0, M00.1, M00.2, M00.8, M00.9, or T84.5). NHS Digital will then undertake an internal linkage exercise with data from the MRIS Personal Demographics Service, to produce a full set of patient identifying information for these admitted patients (NHS number, surname, forenames, date of birth, sex, full address, and postcode). From the full HES and patient identifiable data, NHS Digital will create two datasets, linked by a unique study ID (encrypted HES ID) for each patient. These are described below.

Dataset 1: patient-level, identifiable data. This dataset will contain a full set of patient identifiers for all study patients, and their encrypted HES IDs, but no HES clinical data. This dataset will be sent to NHS Business Services Authority (NHSBSA) Dental Information Services in Eastbourne, England. NHSBSA will use the supplied patient identifiers to identify study patients within their own database, and retrieve all their dental treatment records from 1st April 2009 - 31st March 2017. Included in these records will be the date of any course of dental treatment and whether the treatment included any extraction, endodontic treatment or a scale and polish. NHSBSA will create Dataset 3 (pseudonymised dental treatment records data) by removing the patient identifiers from the retrieved dental treatment records (but retaining the encrypted HES ID provided by NHS Digital). NHSBSA will send Dataset 3 to the University of Sheffield study team. After the University of Sheffield study team have confirmed to NHSBSA that they have checked Dataset 3 and it is suitable for analysis, NHSBSA will securely destroy their copy of Dataset 1, (patient identifiers received from NHS Digital), notifying NHS Digital of data destruction.

Dataset 2: pseudonymised record-level clinical and operational HES data. This dataset will contain clinical and operational HES data for all inpatient admissions between 1st January 2000 and 31st March 2017, for all of the patients within the group already identified as described above (by discharge diagnosis ICD10 code on an admission between 1st April 2010 and 31st March 2017); this historical data is for the purpose of identifying the dates of any previous joint replacement procedures for each patient, and the specific type of joint replaced in each case. All records will include the patient's encrypted HES ID, but no patient identifying information. This dataset will be sent by NHS Digital to the University of Sheffield study team in Sheffield, England. The University of Sheffield study team will link Dataset 2 (pseudonymised HES inpatient data) to Dataset 3, (pseudonymised dental treatment data provided by NHSBSA for the patients identified in Dataset 1), using the encrypted HES ID (common to both datasets) for each patient; this linked data will be used for the study analysis.

The study team at the University of Sheffield will analyse the data according to the study protocol and report the results through peer-reviewed journals and conference proceedings. Plans for further dissemination to the relevant interested public in the UK will be made in the light of the results of the study, and with advice from the study team's patient and public representative colleagues.

Identifiable patient data will only be used for linking purposes and it will only be transferred, in the form of Dataset 1, between the two NHS organisations (NHS Digital and NHSBSA). The study team at the University of Sheffield will not receive any patient identifiable data; they will only receive Datasets 2 and 3, from both of which all patient identifiable information will have been removed (and replaced with the encrypted HES ID in both datasets to enable data linking without identification of patients). There will be no attempt made by the study team to re-identify individuals.

The pseudonymised HES inpatient data will be stored and processed solely at the University of Sheffield, on a secure virtual machine maintained by the University, which is accessible only by specified individuals from specified IP locations using University-owned and maintained computers. The data will only be accessed by individuals within the University of Sheffield study team - all of whom are contracted employees of the University of Sheffield and have received University of Sheffield training in data protection, information governance and confidentiality. The team's research collaborators at The Carolinas Healthcare System (Charlotte, North Carolina USA), the Mayo Clinic College of Medicine (Rochester, Minnesota USA) and the OrthoCarolina Research Institute (Charlotte, North Carolina USA), fulfil an advisory role only, providing valuable input during the selection of ICD-10 codes to use for condition identification, advice on the analysis of causal organisms for sub-analysis of PJI, (including identification of likely oral organisms and their associated ICD-10 codes), and additional clinical perspective on the study. For clarity, research collaborators will see the end results of University of Sheffield analysis and may be involved in subsequent health policy decision making in the USA, but they

- do not have an active decision-making role

- have no control over the analysis or processing of the data

- will not have any access to the data.

University of Sheffield are therefore the sole data controller; no other organisations, including the funding organisation or any research collaborators, will have any decision-making powers over the study or will handle the data. University of Sheffield will be the sole organisation with responsibility for determining the purpose for which and the means by which data is processed, and therefore no organisation other than University of Sheffield will have data controllership over the data provided under this data sharing agreement.

All organisations party to this agreement must comply with the Data Sharing Framework Contract requirements, including those regarding the use (and purposes of that use) by “Personnel” (as defined within the Data Sharing Framework Contract i.e.: employees, agents and contractors of the Data Recipient who may have access to that data).

The precise timing and publication strategy for the results of the study will depend on when the research team receives the necessary data from NHS Digital and NHSBSA and the speed with which the team can analyse it. It will also depend on what the results show and when the results become available bearing in mind the submission deadlines for different scientific meetings.

It is likely that the results will be of interest to key dental opinion leaders in the UK and in other countries, whatever the outcome. The largest and most important dental research meeting each year is the International Association for Dental Research (IADR). This meeting is by far the most important research meeting each year for key opinion leaders within UK dentistry and their international colleagues. More key opinion leaders within UK dentistry attend this meeting each year than any other. The first IADR meeting after the team are likely to have the results is scheduled for 23rd-26th June 2021. However, the deadline for submission of abstracts for that meeting is usually in November of the year before. It is possible therefore, that the results of the study will not be available in time for that deadline. Nonetheless, the focus will be on meeting this deadline if at all possible.. If it is not possible to meet this deadline, the research team may, as an alternative, target the 2021 meetings of the British and Irish Society for Oral Medicine or the British Society for Oral and Dental Research. However, these meetings are much smaller than the IADR meeting and attended by fewer UK dental key opinion leaders. Also, the dates, location and abstract submission deadlines of these meetings are not yet available.

If the data shows that invasive dental procedures are not associated with prosthetic joint infections, this would validate the UK guidance, and prove that LPJI costs can not be reduced by giving AP. It would also be of impact to American orthopaedic surgeons, since they currently recommend that patients with prosthetic joints should have antibiotic prophylaxis before undergoing invasive dental procedures – and this result would suggest that this practice should stop. . Alternatively, if analysis showed that invasive dental procedures were strongly linked to prosthetic joint infections, then the results would be of considerable importance to a UK orthopaedic audience and may result in a change to the current UK guidance. The results obtained from the study will therefore play an important part in the decision about which orthopaedic meeting at which to present the data, particularly as there is no large international orthopaedic meeting. . However, the dates, location and abstract submission deadlines of these meetings are not yet available. Most meetings will not permit data to be presented that has already been presented at another meeting of the same type. Therefore, although it may be possible to present at one dental and one orthopaedic surgery meeting, it would not be feasible to present at more than one dental or orthopaedic surgery meeting.

With regard to publication of the results, the team can only do this in one journal. It will therefore be important to target the journal likely to give the study the biggest exposure and impact. Generally, dental journals have greater impact and wider readership than orthopaedic surgery journals. Therefore, the first choice would likely be the Journal of Dental Research. If that should fail, the next choice would likely be determined by the outcome of the research. If it showed a strong link between invasive dental procedures and prosthetic joint infections, the British Dental Journal would be the second choice, since the data would suggest that the current UK practice of not giving antibiotic prophylaxis to those with prosthetic joints undergoing invasive dental procedures could be wrong. If on the other hand, the data showed that there was no link between invasive dental procedures and prosthetic joint infection, the team would probably prioritise the Journal of the American Dental Association, as a second choice, since this would suggest that the current US practice of giving antibiotic prophylaxis to these patients is likely to be wrong.

Final decisions about where to present and publish the data will need to be made, therefore, once the results of the study are known. Successful publication would help bring it to the attention of guideline committees responsible for advising clinicians about the use of antibiotic prophylaxis prior to invasive dental procedures, to prevent late peri-prosthetic joint infections.

Plans for further dissemination to the relevant interested public in the UK will be made in the light of the results of the study, and with advice from the study team's patient and public representative colleagues.

All publications and presentations resulting from the study will contain only aggregate data with small numbers suppressed in line with the HES Analysis Guide, and will contain no record-level data.

The results of this study will be of considerable interest to dentists, orthopaedic surgeons, family practitioners and patients. It will also be of interest to guideline committees around the world.

If the research confirms an association between Invasive Dental Procedures (IDP) and Late peri-Prosthetic Joint Infections (LPJI), this would be seen as further validation of the view that Antibiotic Prophylaxis (AP) might have a role in preventing the disease - although further studies would be needed to confirm this. Such a result would

provide support for the use of AP before IDP to prevent LPJI, would provide evidence to support a re-evaluation of the UK advice that AP should not be given to prosthetic joint patients undergoing IDP, and would provide support for the current US guidance. The results would also provide further justification for the need for a randomised controlled trial to assess the efficacy of AP for IDP, which would determine to what extent the UK costs for treating LPJI could be offset by an investment in the use of AP for IDP in patients with prosthetic joints.

Alternatively, if the research demonstrates no association between IDP and LPJI, it would suggest that IDP are not a clinically significant cause of LPJI and it would provide support for the UK guidance not to give AP before IDP, and that there would be little or no cost savings in LPJI in the UK by giving AP before IDP. In addition, it would provide support for the view expressed by the American Dental Association Council on Scientific Affairs that AP is unlikely to be effective in preventing LPJI. Indeed, it would provide strong evidence to support guidance that the use in the USA of AP prior to IDP should cease as a means of preventing LPJI. Implementation of this guidance would remove a huge financial burden on patients and healthcare systems (the cost of providing AP for IDP is estimated at $59,640,000 annually in the USA), as well as removing unnecessary risk to patients (from adverse drug reactions) and global society (from the potential development of antibiotic resistant bacteria), resulting from the current widespread and routine use of AP for IDP in countries such as the USA and Canada where AP prior to IDP is a common practice for patients with prosthetic joints.

In addition, there is a wider benefit to the UK health system in a body of credible research which uses HES as a powerful dataset on which to build an evidence base which can be applied globally to benefit patients, lower costs and reduce antibiotic use. However, there is currently no data available on potential cost savings in the UK were antibiotic use to be reduced as a result of this study

The study collaborators, researchers outside the University of Sheffield, and the funding organisation will not have influence on the outcomes nor suppress any of the findings of this research.

Yielded Benefits is not a requirement for new applications.