An evaluation of knee arthroplasty fixation in an evolving challenging population

The University of Oxford requires HES admitted patient care dataset (HES APC), mortality and PROMS data for a cohort identified from the National Joint Registry (NJR) for the purpose of an evaluation of knee arthroplasty fixation in an evolving challenging population.

Background:

Over 100,000 primary knee replacements are performed annually in the United Kingdom for end stage knee osteoarthritis of which 95.1% use cemented fixation and 4.9% are cementless (4.9%). The number of knee replacements conducted is expected to increase six-fold by 2030 although this statistic has been questioned by academics. The complication rate following surgery is reported to be up to 11% with younger and obese patients having double the rates of implant failure. Furthermore, up to 20% of patients report clinically significant persistent pain and dissatisfaction with their knee replacements. Worryingly the number of revision procedures has increased by 95% in recent years and have worse outcomes than primary knee surgery despite costing over 10 times more. The commonest causes for revision include aseptic loosening, pain and infection.

Cemented knee replacement surgery is currently the gold standard and provides immediate fixation but has many associated problems including cement embolisms, late implant loosening and complex revision surgery. Furthermore, preliminary studies suggest that thermal necrosis of surrounding structures from cement polymerisation may be linked to acute and chronic pain.

Importantly patient demographics undergoing knee replacement surgery are changing with more obese or younger and active patients requiring surgery. These new demographics have created new challenges given cemented implants have higher failure rates in these groups with the under 65 age group expected to represent the majority of cases by 2030. Additionally, given the rise in life expectancy there is an increasing need for implants to provide a more physiological long-term fixation with current concerns that cemented implants will not last.

Cementless implants currently form under 5% of all primary knee replacements. However, cementless fixation may offer a solution and has many advantages including the elimination of bone cement interface for failure, no thermal necrosis from cement polymerisation and no third body wear from retained or fragmented cement. It is also well recognised that cementless knee replacement surgery has shorter operative times meaning reduced times under anaesthesia, reduced intraoperative blood loss and infection risk all of which reduce the chances of postoperative complications, potentially reducing morbidity and mortality.

There is currently no consensus on which fixation type is best overall and certainly not for population subgroups based on age or body mass index. The majority of studies, which compare outcomes for cemented and cementless knee replacement, have been limited by relatively small numbers of patients.

By analysing linked data from HES and the NJR plus linked fields from the mortality and PROMS datasets, the question of which fixation method is superior overall and for subsets of patients and implant types can be answered. An analysis which propensity matches for patient demographics and surgical caseload is needed to allow for an accurate comparison due to selection bias with younger, healthier fitter patients being more likely to receive cementless implants. Furthermore, this data will allow for comprehensive cost benefit analysis for all implant types to be conducted. These results will then help guide research to improve cementless fixation outcomes.

The study team will use routinely collected data to answer the questions outlined and not only help deliver more patient specific care and improve outcomes but guide surgical provision for healthcare providers and help patients make more informed decisions.

The study team's project is justified by article 6(1)(e) of the GDPR as this data linkage and processing is needed in the public interest to help decide whether cemented or cementless knee replacements are better from a clinical or cost effectiveness perspective, overall and then in different subgroups of the population. The study team's project is justified by article 9(2)(j) of the GDPR as it is in the public interest to know which knee replacements work best for patients to help improve health and social care. This will help develop better care in the NHS and perhaps the world. Only anonymised results will be published. The National Joint Registry and NHS Digital also have systems by which patients can dissent to their data being used. The study team will only be using data for those patients who have not dissented. The study team have full ethical approval and clinical advisory group approval for this project.

The legal basis for the flow of HQIP data from the NJR to NHS Digital is GDPR Articles 6(1)(e) and 9(2)(i) and the duty of confidence is met through support under section 251 of the NHS Act 2006.

Cohort identification:

The data subjects are all patients who have had cemented or cementless knee replacements and have been registered on the NJR between April 2003 to December 2018. All subjects will be over the age of 40 years at the time of their latest hospital episode.

The treatment group will be the patients who have received cementless knee replacements and the control group will be the patients who have received cemented knee replacements.

Data required:

For the cohort described above, data is needed from the HES Admitted Patient Care (APC), Civil Registrations Deaths and Patient Reported Outcome Measures (PROMs) data sets.

For the cohort, NHS Digital would provide a one-off report of linked HES APC, Mortality and PROMs data for the knee replacement cohort provided by NJR. This would provide information on revisions/reoperations, mortality, PROM scores, mortality, length of inpatient stay, costs of admission and patient demographics (comorbidities, BMI, age, gender).

To evaluate the performance of knee replacements the study team need data on the knee status (reoperation, revised, unrevised, death), time to event data, readmissions, length of hospital stay, medical complications, patient reported outcome measures. To allow a fair comparison of cemented and cementless replacements the study team need to match patients on implant factors (component size, knee implant design, year of implantation, bearing type), patient factors (gender, age, BMI), surgical factors (Primary diagnosis, ASA grade, thromboprophylaxis, surgeon’s grade and caseload, bearing type bone graft) and comorbidities (from both HES APC and PROMs datasets). For health economic analyses the study team require PROM scores, EQ-5D scores, HRG codes and OPCS codes for procedures undertaken.

To clarify some of the terminology above:

- The American Society of Anaesthesiologists (ASA) score is a subjective assessment of a patient’s overall health that is based on five classes.

- For information on the EQ-5D score, see: https://digital.nhs.uk/data-and-information/publications/statistical/patient-reported-outcome-measures-proms/patient-reported-outcome-measures-proms-in-england-2011-2012-special-topic-eq-5d-index-analysis.

- Healthcare Resource Groups (HRGs) are standard groupings of clinically similar treatments which use common levels of healthcare resource.

- OPCS codes are operating procedure codes.

The NJR dataset is needed for identifying the cohort and information pertaining to the knee replacement implanted.

Mortality data is needed to perform any survival analysis and is no longer provided by the National Joint Registry. The HRG and OPCS codes of subsequent episodes are essential for any health economic analyses and is only available through HES APC. PROMs data is needed to evaluate the pain and functional outcomes of patients.

The years of HES data requested begins 5 years prior (1997/98) to when the National Joint Registry started collecting data on knee replacements (2003) and finishes with the latest available data at point of production. The HES APC is needed for information on reoperations, mortality and revision. Additionally, HES APC data offers more details on patient demographics such as BMI, gender, Age and comorbidities. The HES-PROMs linked data will provide information on functional outcomes, health economic analyses, and the matching of comparable records to compare outcomes of cemented vs cementless knee replacements.

The NJR will send identifiable data to NHS Digital to allow the data linkage to take place - specifically the NHS number, Date of Birth, Surname, Postcode and Sex. The data (except sex) is not needed for the study team's analysis. CAG and REC have issued support for these items for the purposes of data linkage. The study team only require pseudonymised data to perform analyses.

Data minimisation reasoning:

The study team are only requesting HES data relevant to assessing the outcomes of primary knee replacements for patients over 40 at the time of their latest hospital episode and have limited the data requested to the cohort described. Data is not required for patients who were under 40 years of age when they had knee replacement surgery. However, because the date of surgery can not be determined via the OPCS codes, all patients who had surgery under the age of 40 can not be eliminated from the HES data. The amount of data will be reduced by excluding individuals who were under the age of 40 at the date of their most recent episode in hospital. These data will be removed and excluded entirely.

The study team are requesting data for the years ranging from 1997/98 to the latest available data in 2019/20. The study team will not receive episodes relating specifically to: maternity, alcohol, psychiatry, patient pathway and most geographical fields. The study team have only requested a one off dataset and the study team will not require further data on the data subjects in the future for the purpose of this study.

The University of Oxford had initially tried to limit the data requested to 5 years prior to the date of surgery and by providing a list of OPCS codes to NHS Digital. However, this has created two significant issues.

1) There are known issues with the way knee replacements are coded. Around a quarter of the knee replacements recorded in the NJR data would not be identified if the data extract was limited to only include specific OPCS codes.

Additionally, there is a study published by Middleton et al (2018) which has shown a significant problem with the coding of knee replacement procedures. Therefore, all hospital records (excluding maternity, fetal, alcohol and psychiatry episodes) are needed to prevent this problem.

2) The full set of HES records are needed (excluding maternity, fetal, alcohol and psychiatry episodes) to accurately profile the patients' past medical history. For example, a patient can present for knee trauma/arthroscopy/meniscal injury 10 years before they have a knee replacement and this needs to be factored into the analysis. As these factors affect the underlying indication of knee replacement surgery, limiting this information can mislead the analysis by mixing different surgical indications and lead to erroneous results.

The study team require all pre and post-operative HES episodes (excluding maternity, fetal, alcohol and psychiatry fields) to monitor the knee replacement outcomes over time. Outcomes include revisions, reoperations, medical complications and mortality.

Regarding PROMs data, the study team have only requested the Oxford Knee Score and EQ-5D scores which are needed to assess the outcomes of cemented and cementless knee replacements. This is important to assess the functional outcomes and to perform health economic analyses. The study team have also requested access to PROMs comorbidity data, this is needed for matching of cementless and cemented records in a fair way - i.e. matching records with similar comorbidities. This is assessed by a comorbidity index score and the PROMs comorbidity data will also help assess the quality of reported comorbidities in the HES APC data. PROMs data for the data requested has only been collected from 2009/10 hence the time period in the data requested.

The study team require mortality data to be able to compare mortality between cemented and cementless knee replacements. Dates of deaths are essential for any implant survival analysis.

Deprivation indices are needed for propensity matching comparative groups as this can influence patient outcomes. HRG codes and length of stay are essential for health economic comparisons of cemented and cementless knee replacements.

The study team require data from across England so that the study is adequately powered to compared cemented and cementless knee replacements overall and then to allow appropriate subgroup analyses in different strata of the population (i.e. different age groups etc). Including all regions in England will help by having a mixture of patients with different demographics to allow subgroup analyses. Only data which is needed to assess the objectives outlined above has been requested.

The study team have considered possible methodologies and determined there is no viable alternative method for answering this question. To run a clinical trial is impractical as one would need approximately 10 years follow up and would require an unmanageable sample size to have the same impact as this proposed project. The data available from NHS Digital is not available from the NJR. There are no less intrusive methods of performing the study analysis.

Noting that NHS Digital has previously supplied HES data linked with NJR data for other studies related to knee replacements under separate Data Sharing Agreements, the University of Oxford has considered and ruled out the approach of reusing that data for this study. A new dataset is needed given the applicant needs information on the most recent knee replacements given implant designs are constantly improving in a dynamic implant market. Also, by obtaining information on the newer cementless designs, this will allow comparison to older designs. This project is substantially different to any other NJR-HES linked data projects.

The study team require and have applied for approval for NJR data to release for this specific research purpose. Confirmation of NJR approval will be required prior to the release of this data.

References

------------------

[1] Middleton R, Wilson HA, Alvand A, Abram SG, Bottomley N, Jackson W, Price A. Outcome-based commissioning of knee arthroplasty in the NHS: system error in a national monitoring programme and the unintended consequences on achieving the Best Practice Tariff. Bone Joint J. 2018 Dec;100(12):1572-8.

Data analysis plan:

Data only on primary knee replacements will be sought from 1997/98 to the latest available HES data with data only requested necessary to evaluate knee replacement outcomes. Linked data from Northgate Public services and NHS Digital, will be stored in a secure room in a CCTV monitored, card protected university research building. All analysis will be conducted on a fully encrypted computer. This will ensure that the data will be kept safe and secure.

The study team will propensity match replacement groups based on various factors including gender, age, BMI, primary diagnosis, ASA grade, thromboprophylaxis, surgeon’s grade and caseload, implant component size, bearing type and the use of bone graft.

Using the combined NJR/HES/PROMs/mortality dataset the study team will;

I. Calculate the implant and patient survival of the cemented and cementless implants respectively using Kaplan Meier analysis and compare groups using the log rank test to see which has a superior long-term survival. This comparison requires comorbidity analysis.

II. Analyse mechanisms of implant failure by reporting the incidence of mechanisms of failure.

III. Compare patients early, midterm and long-term PROMs from the cemented and cementless groups. This comparison requires comorbidity analysis.

IV. Determine the incidence of serious medical complications following surgery including thromboembolism (pulmonary embolism or deep vein thrombosis), blood loss and transfusions, myocardial infraction, stroke and mortality. The study team will compare these rates between groups using multivariate analyses.

V. Conduct subgroup analysis on different patient subgroups and implant types.

VI. Conduct cost effective analysis for the propensity matched cemented and cementless groups using a lifetime Markov model.

Data Controller statement:

The University of Oxford is the sole data controller and processor, given the University of Oxford are solely determining, through the study team, the way the data is being processed. Northgate Public Services are providing the NJR dataset and identifiers for the cohort only, but not taking further part in the research. The University of Oxford is the data controller responsible for determining the purpose for all data flows described and will be the data controller for the data received.

Linkage process:

Northgate Public Services will send the study team de-identified data on all patients who have had a knee replacement from April 2003-December 2018 with an NJR ID number (which is not patient identifiable) on an encrypted device. They will simultaneously send their identifiable patient data (NHS Number, Date of Birth, Surname, Sex, Postcode) to NHS Digital and ask them to link the HES dataset (inpatient data, patient reported outcome measures and mortality data) to the NJR dataset and then remove any patient identifiers prior to sending anything to the study team. NHS Digital will then only send the study team de-identified data which would include the NJR ID and linked HES inpatient data. The study team will then be able to merge the datasets obtained from NHS Digital to the previously received dataset from Northgate Public Services using the NJR IDs (which are de-identified). The linked dataset will then be analysed by the study team in the University of Oxford.

Funding:

The Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences at the University of Oxford is the sole data controller and will receive, process and analyse the data and subsequently publish the findings. The NJR is part of the national audit programme of the Healthcare Quality Improvement Partnership (HQIP) and is managed by Northgate Public Services Ltd – which will provide the cohort of identifiers to NHS Digital for the data request.

The project has been instigated and is being undertaken by a substantive employee of the Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences at the University of Oxford. This individual has an employment contract with the University of Oxford and is a Royal College of Surgeons (RCS)/NJR Fellow which allows the individual to undertake work in their home institution, but their salary funding is provided by the RCS. Therefore, RCS is providing funding for the person who will undertake the work but is not providing funding specifically for the purpose of this work and has not determined that it should be undertaken. The purpose of the investigation has been determined solely and entirely by employees of the University of Oxford. Objectives of the work are focused on informing knowledge in the field of knee replacement care delivery within the NHS to meet recognised needs. The purpose of the work has received external review by the NJR Research Sub-Committee as part of the approvals process to undertake the work, as it includes the use of NJR data. Such review ensures that the proposed work falls within the NJRs thematic areas of interest.

Assessment of how the data will be processed to meet the proposed purpose has been undertaken by the senior team members (University of Oxford employees), who have extensive experience in the field.

The results from this research will be disseminated both to the Academic world and to patients. Each will be discussed in turn.

Academic world dissemination

---------------------------------------------

The results from this study will be submitted to prominent peer reviewed journals and conferences.

Journals which the study team plan to submit the findings to will include; The Lancet, The BMJ, Journal of Bone and Joint Surgery. Conferences the study team will submit this work to will include the American Academy of Orthopaedic Surgeons, the EFORT congress meeting and the British Orthopaedic Association meeting. The results will also be put on the Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences website.

To maximise the impact of the study research the study team will monitor for conferences with themes on using big data to help change clinical practice. The study team will always try to select high impact journals first with wide authorship. Additionally, the study team have a twitter account to help share the publication to enhance the altmetric score for published articles.

The results will not identify individuals and contain only data in aggregate form (tabulations and figures showing analysis results at the minimum level of detail required) in line with the HES Analysis Guide relating to the data being shared under this Agreement. In the year following receipt of the data the following reports, analyses and presentations are planned:

Specific data Analyses and papers include:

1. A comparison of the clinical outcomes of cemented and cementless knee replacements. This will also involve comparing total and unicompartmental knee replacements (Target date for completion June 2020).

2. An analysis of the mechanisms of failure of cemented and cementless knee replacements (Target date for completion October 2020).

3. A comparison of early and late cementless knee replacements (Target date for completion January 2021).

4. A cost effectiveness comparison of cemented and cementless knee replacements (Target date for completion October 2021).

Dissemination of results to patients

----------------------------------------------------

The study team have already established a patient focus group in Oxford who helped the research team devise their research protocol. The focus group felt that studies comparing fixation were essential and important to patients and they helped the study team determine the outcome measures relevant to patients. This patient focus group will meet on a six monthly basis to discuss the results of the work and offer ideas to help direct the research. Additionally, they will offer suggestions on how best to disseminate the results of the study to other patients.

The study team also intend to run a research open day in the Nuffield Orthopaedic Centre in which members of the public can come to ask questions about the study teams current research and our future directions.

Finally, the study team have a twitter account @OOEC which will be used to tweet summaries of the results. This will be accessible to all twitter members including patients and the general public. They will also be able to message the twitter account for any questions they have regarding the research. The study team have a designated individual who regularly checks the account and currently has 10.2K followers.

The University of Oxford requires HES admitted patient care dataset (HES APC), mortality and PROMS data for a cohort identified from the National Joint Registry (NJR) for the purpose of an evaluation of knee arthroplasty fixation in an evolving challenging population.

Background:

Over 100,000 primary knee replacements are performed annually in the United Kingdom for end stage knee osteoarthritis of which 95.1% use cemented fixation and 4.9% are cementless (4.9%). The number of knee replacements conducted is expected to increase six-fold by 2030 although this statistic has been questioned by academics. The complication rate following surgery is reported to be up to 11% with younger and obese patients having double the rates of implant failure. Furthermore, up to 20% of patients report clinically significant persistent pain and dissatisfaction with their knee replacements. Worryingly the number of revision procedures has increased by 95% in recent years and have worse outcomes than primary knee surgery despite costing over 10 times more. The commonest causes for revision include aseptic loosening, pain and infection.

Cemented knee replacement surgery is currently the gold standard and provides immediate fixation but has many associated problems including cement embolisms, late implant loosening and complex revision surgery. Furthermore, preliminary studies suggest that thermal necrosis of surrounding structures from cement polymerisation may be linked to acute and chronic pain.

Importantly patient demographics undergoing knee replacement surgery are changing with more obese or younger and active patients requiring surgery. These new demographics have created new challenges given cemented implants have higher failure rates in these groups with the under 65 age group expected to represent the majority of cases by 2030. Additionally, given the rise in life expectancy there is an increasing need for implants to provide a more physiological long-term fixation with current concerns that cemented implants will not last.

Cementless implants currently form under 5% of all primary knee replacements. However, cementless fixation may offer a solution and has many advantages including the elimination of bone cement interface for failure, no thermal necrosis from cement polymerisation and no third body wear from retained or fragmented cement. It is also well recognised that cementless knee replacement surgery has shorter operative times meaning reduced times under anaesthesia, reduced intraoperative blood loss and infection risk all of which reduce the chances of postoperative complications, potentially reducing morbidity and mortality.

There is currently no consensus on which fixation type is best overall and certainly not for population subgroups based on age or body mass index. The majority of studies, which compare outcomes for cemented and cementless knee replacement, have been limited by relatively small numbers of patients.

By analysing linked data from HES and the NJR plus linked fields from the mortality and PROMS datasets, the question of which fixation method is superior overall and for subsets of patients and implant types can be answered. An analysis which propensity matches for patient demographics and surgical caseload is needed to allow for an accurate comparison due to selection bias with younger, healthier fitter patients being more likely to receive cementless implants. Furthermore, this data will allow for comprehensive cost benefit analysis for all implant types to be conducted. These results will then help guide research to improve cementless fixation outcomes.

The study team will use routinely collected data to answer the questions outlined and not only help deliver more patient specific care and improve outcomes but guide surgical provision for healthcare providers and help patients make more informed decisions.

The study team's project is justified by article 6(1)(e) of the GDPR as this data linkage and processing is needed in the public interest to help decide whether cemented or cementless knee replacements are better from a clinical or cost effectiveness perspective, overall and then in different subgroups of the population. The study team's project is justified by article 9(2)(j) of the GDPR as it is in the public interest to know which knee replacements work best for patients to help improve health and social care. This will help develop better care in the NHS and perhaps the world. Only anonymised results will be published. The National Joint Registry and NHS Digital also have systems by which patients can dissent to their data being used. The study team will only be using data for those patients who have not dissented. The study team have full ethical approval and clinical advisory group approval for this project.

The legal basis for the flow of HQIP data from the NJR to NHS Digital is GDPR Articles 6(1)(e) and 9(2)(i) and the duty of confidence is met through support under section 251 of the NHS Act 2006.

Cohort identification:

The data subjects are all patients who have had cemented or cementless knee replacements and have been registered on the NJR between April 2003 to December 2018. All subjects will be over the age of 18 years.

The treatment group will be the patients who have received cementless knee replacements and the control group will be the patients who have received cemented knee replacements.

Data required:

For the cohort described above, data is needed from the HES Admitted Patient Care (APC), Civil Registrations Deaths and Patient Reported Outcome Measures (PROMs) data sets.

For the cohort, NHS Digital would provide a one-off report of linked HES APC , Mortality and PROMs data. This would provide information on revisions/reoperations, mortality, PROM scores, mortality, length of inpatient stay, costs of admission and patient demographics (comorbidities, BMI, age, gender). NHS Digital will use OPCS codes supplied by the study team to search for knee surgeries in the HES APC data and extract 5 years of HES episodes prior to surgery for each patient record and all subsequent episodes after the surgery until 2019.

To evaluate the performance of knee replacements the study team need data on the knee status (reoperation, revised, unrevised, death), time to event data, readmissions, length of hospital stay, medical complications, patient reported outcome measures. To allow a fair comparison of cemented and cementless replacements the study team need to match patients on implant factors (component size, knee implant design, year of implantation, bearing type), patient factors (gender, age, BMI), surgical factors (Primary diagnosis, ASA grade, thromboprophylaxis, surgeon’s grade and caseload, bearing type bone graft) and comorbidities (from both HES APC and PROMs datasets). For health economic analyses the study team require PROM scores, EQ-5D scores, HRG codes and OPCS codes for procedures undertaken.

To clarify some of the terminology above:

- The American Society of Anaesthesiologists (ASA) score is a subjective assessment of a patient’s overall health that is based on five classes.

- For information on the EQ-5D score, see: https://digital.nhs.uk/data-and-information/publications/statistical/patient-reported-outcome-measures-proms/patient-reported-outcome-measures-proms-in-england-2011-2012-special-topic-eq-5d-index-analysis.

- Healthcare Resource Groups (HRGs) are standard groupings of clinically similar treatments which use common levels of healthcare resource.

- OPCS codes are operating procedure codes.

The NJR dataset is needed for identifying the cohort and information pertaining to the knee replacement implanted.

Mortality data is needed to perform any survival analysis and is no longer provided by the National Joint Registry. The HRG and OPCS codes of subsequent episodes are essential for any health economic analyses and is only available through HES APC. PROMs data is needed to evaluate the pain and functional outcomes of patients.

The years of HES data requested begins 5 years prior (1997/98) to when the National Joint Registry started collecting data on knee replacements (2003) and finishes with the latest available data at point of production. The HES APC is needed for information on reoperations, mortality and revision. Additionally, HES APC data offers more details on patient demographics such as BMI, gender, Age and comorbidities. The HES-PROMs linked data will provide information on functional outcomes, health economic analyses, and the matching of comparable records to compare outcomes of cemented vs cementless knee replacements.

The NJR will send identifiable data to NHS Digital to allow the data linkage to take place - specifically the NHS number, Date of Birth, Surname and Postcode. This data is not needed for the study team's analysis and the study team will never receive this data. The study team only require pseudonymised data to perform analyses.

Data minimisation reasoning:

The study team are only requesting HES data relevant to assessing the outcomes of primary knee replacements and have limited the data requested to the cohort described. The study team are year range from 1997/98 to the latest available data in 2019/20. The study team will not receive fields relating specifically to: maternity, alcohol, psychiatry, patient pathway and most geographical fields. The study team have only requested a one off dataset and the study team will not require further data on the data subjects in the future for the purpose of this study.

To minimise the amount of data received the study team have supplied the OPCS codes for primary knee replacements to allow NHS Digital to find the actual primary knee operations and the dates of surgery within the cohort. The study team then only obtain 5 years of HES data, preoperatively, from the primary surgery date (excluding maternity, fetal and psychiatry fields) of each record. This will minimise the data on each patient and will reduce disproportionate pre-surgical episodes for different individuals. The preoperative admissions are crucial to allow accurate profiling of patients’ past medical histories for comorbidities. This information is needed as cemented and cementless knee replacements will need to be matched on comorbidities.

The study team require all post-operative HES episodes (excluding maternity, fetal and psychiatry fields) to monitor the knee replacement outcomes over time. Outcomes include revisions, reoperations, medical complications and mortality.

Regarding PROMs data, the study team have only requested the Oxford Knee Score and EQ-5D scores which are needed to assess the outcomes of cemented and cementless knee replacements. This is important to assess the functional outcomes and to perform health economic analyses. The study team have also requested access to PROMs comorbidity data, this is needed for matching of cementless and cemented records in a fair way - i.e. matching records with similar comorbidities. This is assessed by a comorbidity index score and the PROMs comorbidity data will also help assess the quality of reported comorbidities in the HES APC data. PROMs data for the data requested has only been collected from 2009/10 hence the time period in the data requested.

The study team require mortality data to be able to compare mortality between cemented and cementless knee replacements. Additionally, the cause of death is important to investigate whether fixation predisposes to a mechanism of death. Dates of deaths are essential for any implant survival analysis.

Deprivation indices are needed for propensity matching comparative groups as this can influence patient outcomes. HRG codes and length of stay are essential for health economic comparisons of cemented and cementless knee replacements.

The study team require data from across England so that the study is adequately powered to compared cemented and cementless knee replacements overall and then to allow appropriate subgroup analyses in different strata of the population (i.e. different age groups etc). Including all regions in England will help by having a mixture of patients with different demographics to allow subgroup analyses. Only data which is needed to assess the objectives outlined above has been requested.

The study team have considered possible methodologies and determined there is no viable alternative method for answering this question. To run a clinical trial is impractical as one would need approximately 10 years follow up and would require an unmanageable sample size to have the same impact as this proposed project. The data available from NHS Digital is not available from the NJR. There are no less intrusive methods of performing the study analysis.

Noting that NHS Digital has previously supplied HES data linked with NJR data for other studies related to knee replacements under separate Data Sharing Agreements, the University of Oxford has considered and ruled out the approach of reusing that data for this study. A new dataset is needed given the applicant needs information on the most recent knee replacements given implant designs are constantly improving in a dynamic implant market. Also, by obtaining information on the newer cementless designs, this will allow comparison to older designs. This project is substantially different to any other NJR-HES linked data projects.

The study team require and have applied for approval for NJR data to release for this specific research purpose. Confirmation of NJR approval will be required prior to the release of this data.

Data analysis plan:

Data only on primary knee replacements will be sought from 1997/98 to the latest available HES data with data only requested necessary to evaluate knee replacement outcomes. Linked data from Northgate Public services and NHS Digital, will be stored in a secure room in a CCTV monitored, card protected university research building. All analysis will be conducted on a fully encrypted computer. This will ensure that the data will be kept safe and secure.

The study team will propensity match replacement groups based on various factors including gender, age, BMI, primary diagnosis, ASA grade, thromboprophylaxis, surgeon’s grade and caseload, implant component size, bearing type and the use of bone graft.

Using the combined NJR/HES/PROMs/mortality dataset the study team will;

I. Calculate the implant and patient survival of the cemented and cementless implants respectively using Kaplan Meier analysis and compare groups using the log rank test to see which has a superior long-term survival. This comparison requires comorbidity analysis.

II. Analyse mechanisms of implant failure by reporting the incidence of mechanisms of failure.

III. Compare patients early, midterm and long-term PROMs from the cemented and cementless groups. This comparison requires comorbidity analysis.

IV. Determine the incidence of serious medical complications following surgery including thromboembolism (pulmonary embolism or deep vein thrombosis), blood loss and transfusions, myocardial infraction, stroke and mortality. The study team will compare these rates between groups using multivariate analyses.

V. Conduct subgroup analysis on different patient subgroups and implant types.

VI. Conduct cost effective analysis for the propensity matched cemented and cementless groups using a lifetime Markov model.

Data Controller statement:

The University of Oxford is the sole data controller and processor, given the University of Oxford are solely determining, through the study team, the way the data is being processed. Northgate Public Services are providing the NJR dataset and identifiers for the cohort only, but not taking further part in the research. The University of Oxford is the data controller responsible for determining the purpose for all data flows described and will be the data controller for the data received.

Linkage process:

Northgate Public Services will send the study team de-identified data on all patients who have had a knee replacement from April 2003-December 2018 with an NJR ID number (which is not patient identifiable) on an encrypted device. They will simultaneously send their identifiable patient data (NHS Number, Date of Birth, Surname and Postcode) to NHS Digital and ask them to link the HES dataset (inpatient data, patient reported outcome measures and mortality data) to the NJR dataset and then remove any patient identifiers prior to sending anything to the study team. NHS Digital will then only send the study team de-identified data which would include the NJR ID and linked HES inpatient data. The study team will then be able to merge the datasets obtained from NHS Digital to the previously received dataset from Northgate Public Services using the NJR IDs (which are de-identified). The linked dataset will then be analysed by the study team in the University of Oxford.

Funding:

The Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences at the University of Oxford is the sole data controller and will receive, process and analyse the data and subsequently publish the findings. The NJR is part of the national audit programme of the Healthcare Quality Improvement Partnership (HQIP) and is managed by Northgate Public Services Ltd – which will provide the cohort of identifiers to NHS Digital for the data request.

The project has been instigated and is being undertaken by a substantive employee of the Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences at the University of Oxford. This individual has an employment contract with the University of Oxford and is a Royal College of Surgeons (RCS)/NJR Fellow which allows the individual to undertake work in their home institution, but their salary funding is provided by the RCS. Therefore, RCS is providing funding for the person who will undertake the work but is not providing funding specifically for the purpose of this work and has not determined that it should be undertaken. The purpose of the investigation has been determined solely and entirely by employees of the University of Oxford. Objectives of the work are focused on informing knowledge in the field of knee replacement care delivery within the NHS to meet recognised needs. The purpose of the work has received external review by the NJR Research Sub-Committee as part of the approvals process to undertake the work, as it includes the use of NJR data. Such review ensures that the proposed work falls within the NJRs thematic areas of interest.

Assessment of how the data will be processed to meet the proposed purpose has been undertaken by the senior team members (University of Oxford employees), who have extensive experience in the field.

This results from this study will be disseminated nationally and internationally in peer reviewed journals and conferences. Examples of journals the study team will submit to will include; The Lancet, The BMJ, Journal of Bone and Joint Surgery. Examples of conferences the study team will submit this work to will include the American Academy of Orthopaedic Surgeons, the EFORT congress meeting and the British Orthopaedic Association meeting. The results will also be put on the Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences website.

To maximise the impact of the study research the study team will monitor for conferences with themes on using big data to help change clinical practice. The study team will always try to select high impact journals first with wide authorship. Additionally the study team have a twitter account to help share the publication to enhance the altmetric score for published articles.

The results will not identify individuals and contain only data in aggregate form (tabulations and figures showing analysis results at the minimum level of detail required) in line with the HES Analysis Guide relating to the data being shared under this Agreement. In the year following receipt of the data the following reports, analyses and presentations are planned:

Data Analyses and papers:

1. A comparison of the clinical outcomes of cemented and cementless knee replacements. This will also involve comparing total and unicompartmental knee replacements (Target date for completion June 2020).

2. An analysis of the mechanisms of failure of cemented and cementless knee replacements (Target date for completion October 2020).

3. A comparison of early and late cementless knee replacements (Target date for completion January 2021).

4. A cost effectiveness comparison of cemented and cementless knee replacements (Target date for completion October 2021).

Yielded Benefits is not a requirement for new applications.