Introduction

Total hip replacement (THR) is one of the most successful operations of our time with nearly 100,000 performed in England, Wales, Northern Ireland, and the Isle of Man in 2017 [1, 2]. They have been shown in general to last about 25 years, but despite this, there is still variation in the survival of implants across the UK [2, 3]. THR components may require changing (via revision surgery) for one of several reasons including infection, wear, loosening, fracture, or instability [2]. The need for future revision surgery can be influenced by preoperative patient factors, implant factors, and surgical factors [4, 5]. It has previously been demonstrated that THR revisions are not as effective in improving pain and function as the primary operation, have a high chance of further revision, and are costly to the health service, as well as resulting in exposure of patients to the additional pain and inconvenience of another operation [2, 6, 7]. Although implant survival is not the only marker of success [8], the cumulative probability of revision of THRs is a readily available outcome measure because of the National Joint Registry for England, Wales, Northern Ireland and the Isle of Man (NJR) a mandatory, national database [9].

The NJR has collected data since 2003 and at the time of writing contains in excess of 1 million records of primary THRs. In 2017, the NJR identified at least 415 different units (hospitals) performing THRs using at least 822 different combinations of femoral stem and acetabular socket [10, 11] and is thought to capture over 95% of primary hip and knee operations and 90% of revisions [2]. Every year, the NJR annual report lists units in which either a higher or lower than expected rate of revision has been observed over the preceding years. The units with a “better than expected” (above the 99.7% confidence limit) revision estimate may offer an opportunity to learn from good practice and potentially reduce variation between units.

The importance of reducing unwarranted variation across the whole National Health Service (NHS) has been highlighted by the recent work of the “Getting it Right First Time” (GIRFT) program [12]. This review has previously highlighted variation in adult elective orthopedic services and makes clear the requirement to learn from good performance [13].

Investigating what may lead one unit to demonstrate better results than others is challenging because of the differing patient populations as well as issues with potential selection bias. Patients may receive different types of implants based on factors such as age and sex, and as a result, outcomes are difficult to interpret. One unit in the NJR, the Royal Devon & Exeter NHS Foundation Trust (RD&E), has been repeatedly identified as having “better than expected” survival outcomes and is widely known for using only one femoral stem (the Exeter V40 femoral stem) in all routine primary THRs regardless of patient factors, thus removing or reducing selection bias [2]. This offers an opportunity to investigate whether “better than expected” survival results observed within this unit were due to a unit effect or because of the implants used.

Methods

We aimed to compare the cumulative revision estimates between the RD&E and the rest of the country to investigate whether “better than expected” outcomes were due to the implants used or because of other unit factors.

The NJR is a mandatory national audit of joint replacement activity. After gaining written consent, operations are reported to the NJR by the healthcare provider at the time of surgery. The dataset consisted of 981,269 linked primary THRs performed in England and Wales between 1 April 2003 and 31 December 2017 with consent for data linkage. Data were censored either by death or administratively on 31 December 2017. After exclusion of THRs with incomplete or inconsistent data or using metal-on-metal bearings, we were left with 664,761 primary THRs, in which osteoarthritis was the only indication for THR. Reasons for exclusion at each stage are shown in Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Reasons for exclusion from analyses.

A sequence is the order of operations recorded in the NJR for any patient. All complete records will start with a primary operation. If a sequence starts with a revision, the primary was performed before the NJR, outside the geographical coverage of the NJR, or data were not submitted to the NJR. THR, total hip replacement; NJR, National Joint Registry for England, Wales, Northern Ireland and the Isle of Man.

https://doi.org/10.1371/journal.pmed.1003291.g001

Results

After exclusions, we were left with 664,761 primary THRs for analysis. The maximum follow-up in the exemplar center group was 13.9 years and was 14.2 years in all other units. The demographics and distribution of the THRs in each group can be seen in Table 1. Of the 6,230 cases performed at the RD&E, there were 83 different recorded “lead” surgeons who performed a range from one THR to 992 THRs included in the study dataset. A total of 68.1% of THRs were performed by 1 of 6 surgeons. The “lead” surgeon may be a consultant (attending), fellow, or higher specialist trainee (resident) operating under the supervision of a consultant.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Demographics and distribution of included total hip replacements.

https://doi.org/10.1371/journal.pmed.1003291.t001

Crude analyses

The crude 10-year cumulative revision estimate of all THRs implanted at the RD&E was 1.7% (95% CI 1.3–2.3). In all other units, the 10-year cumulative revision estimate for all THRs was 2.9% (95% CI 2.8–3.0; log-rank test P < 0.001); for just THRs using cemented stems, it was 2.6% (95% CI 2.5–2.7; log-rank test P = 0.007), and for just THRs using the Exeter V40 femoral stem, it was 2.3% (95% CI 2.2–2.4) (log-rank test P = 0.05). Net revision estimates calculated using 1-Kaplan–Meier curves can be seen in Fig 2; the number of hips at risk at all time points for all analyses can be seen in S3 Table.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Unadjusted 1-Kaplan–Meier revision estimates of total hip replacements in each subgroup.

Comparison of the all-cause construct revision estimates of total hip replacements performed in the exemplar center compared with those performed in all other hospitals in the National Joint Registry.

https://doi.org/10.1371/journal.pmed.1003291.g002

Adjusted analyses

A FPSA model was fitted for all THRs using the Exeter V40 femoral stem and showed excellent “goodness of fit” (S1 Fig). Comparison of time-dependent and proportional hazards models suggested the time-dependent model showed better fit (S2 Fig). After adjustment for age, sex, and ASA and allowing for time varying effects, the relative revision estimates of each subgroup of THRs (THRs using a cemented stem or THRs using the Exeter V40 femoral stem) modeled for a 68-year-old, female patient, are shown in Fig 3A, Fig 3B and Fig 3C.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. FPSA adjusted for age, sex, and American Society of Anesthesiology score.

Results presented for a 68-year-old female patient with an American Society of Anesthesiology score of 2. FPSA, flexible parametric survival analysis.

https://doi.org/10.1371/journal.pmed.1003291.g003

The femoral stem was paired with 9 different acetabular components in the RD&E, and 99% of these THRs used 1 of only 3 cups. In other units, the Exeter V40 femoral stem was paired with 111 different acetabular components. Fig 3D shows a comparison of the 2 groups if analyses are restricted to stem/cup combinations used at the RD&E. This restricted analysis compares 6,227 performed in the RD&E with 148,295 THRs performed elsewhere. After 13.9 years, there is a discrepancy in restricted mean survival time (RMST) of 0.02 years (95% CI −0.02 to 0.07; P = 0.33). A P value of 0.33 suggests there is little or no evidence of any difference in survival of THRs after 13.9 years between those implanted in the RD&E compared with elsewhere when the same implants were used. Fig 4 shows how the difference in RMST between the 2 subgroups changes over time. RMST is reported in years and estimates the difference in life expectancy of the THR between the 2 exposure groups, i.e., the extra time a THR lasts because it was implanted in the RD&E. The difference in RMST changes slightly over time, most notably at 2 time points, 3 years and at 10 years. This may reflect particular modes of failure such as loosening of cups at 10 years, potentially due to cementation technique. It should be noted that for the majority of time reported, the confidence intervals cross the null value of 0. Graphically, it appears that for a short period there may be a transient center effect; however, analysis of the entire period shows no such center effect.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Difference in RMST between total hip replacements performed at the RD&E and in all other hospitals combined when using the same implants as those used at the RD&E.

A demonstration of how the difference in RMST varies between the 2 exposure groups using the flexible parametric model adjusted for age, sex, and American Society of Anesthesiologists score and allowing for time varying effects. RMST, restricted mean survival time; RD&E, Royal Devon & Exeter.

https://doi.org/10.1371/journal.pmed.1003291.g004

Discussion

After 13.9 years, both crude and adjusted cumulative revision estimates showed better implant survival when THRs were performed at the RD&E compared with elsewhere in the country. In analyses adjusted for age, sex, and ASA score, these differences attenuated after restricting to only cemented implants and disappeared when only THRs using the same implant combinations as the RD&E were analyzed. This suggests that implant choice is responsible for the “better than expected” results at the RD&E and not unit (or surgeon) factors.

We are unaware of any studies to date investigating the reasons why 1 unit achieves better THR survival than others. This study suggests that when attempting to improve implant survivorship, units performing THR, particularly those with “lower than expected” implant survival, should focus attention on choice of implant rather than other factors. The use of implants without evidence of good long-term survival should be limited to well-controlled and monitored studies or experiments. Although this study has focused on 1 single femoral stem (the Exeter V40 femoral stem), we believe that the observed high survival would be reproducible with other well performing implants. Previous work by Deere and colleagues has compared the survival of implant combinations after 10 years and provides a reference to demonstrate other implant combinations with low revision rates [5]. The NJR annual report provides a list of units with “better-than-expected” survival results as well as survival estimates for individual stem/cup combinations and can act as a reference document to units wishing to review their implant selection. These findings are of relevance to surgeons, commissioners, and most importantly, patients when deciding whether to, where, and when to have a THR. Patients should be encouraged to ask surgeons about the long-term survival evidence for the implant they plan to use.

The strength of this study stems from the high number of patients included and the use of a linked, national database with high capture of revision procedures. The lack of selection bias in choice of femoral stem at the RD&E (our reference unit) is another strength. The data in this study are, however, observational, and conclusions regarding causality must be interpreted with caution. Our outcome is revision surgery, and although important, it is not the only marker of success of a THR. Patient reported outcomes such as pain and function have not been assessed and patients may have been unsuitable or unwilling to undergo a revision operation and as such a failure may have been misclassified as a success. We made no attempt to restrict by bearing surfaces in this study, which may be a contributing factor in the longevity of a THR; this would, however, have created several subgroups, which we wished to avoid so we could maintain sample size. We would expect the complexity of cases to be generally representative of the UK population given the NHS referral system and have additionally made attempts to adjust for potential confounders; however, there may still be some residual confounding for variables with incomplete data or not captured by the dataset. There are likely to be sequential hip replacements performed on different sides within the same patient included in this study. We treated each hip as an individual case. There is a risk of failure of one THR leading to subsequent failure of the other side in cases of infection; however, given that this is also the case from other joint replacements (e.g., knee) or from other conditions leading to a higher propensity to infection, we felt this risk was negligible and therefore did not exclude these from analyses. The use of complete case analysis over multiple imputation for handling missing covariates was also a potential weakness and may result in a loss of power by restricting the sample size. Given the distribution of missing data (S1 Table) and the large numbers offered by the registry, we felt that a complete case analysis was suitable for this study, and any reduction in power would be negligible. We cannot exclude the possibility that better surgeons may choose prostheses with lower revision rates.

The fact that the results seen at the RD&E were achieved with 83 different lead surgeons supports the theory that the implant is the driver of improved survival results rather than the skill of the individual surgeon. If this observed association is indeed true, the use of implants with evidence of good survival should be encouraged throughout the health service. Further work may focus on the effect of a change in implant use of a single hospital/unit on survival results in time-series analyses. Although implant selection appears to be associated with improved survival in THR, it is yet to be seen whether the same phenomenon is true in other branches of medicine heavily reliant on implantable devices. Further research in other areas is warranted to investigate this effect.

Conclusion

In this study, we found evidence suggesting that implant selection is associated with the long-term survival of THRs rather than factors specific to a high-performing unit. Surgeons, commissioners, and patients should use this information when considering THR.

Supporting information

Acknowledgments

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.

The Healthcare Quality Improvement Partnership (HQIP) and/or the NJR take no responsibility for the accuracy, currency, reliability, and correctness of any data used or referred to in this report, nor for the accuracy, currency, reliability, and correctness of links or references to other information sources and disclaims all warranties in relation to such data, links, and references to the maximum extent permitted by legislation.

HQIP and NJR shall have no liability (including but not limited to liability by reason of negligence) for any loss, damage, cost or expense incurred or arising by reason of any person using or relying on the data within this report and whether caused by reason of any error, omission, or misrepresentation in the report or otherwise. This report is not to be taken as advice. Third parties using or relying on the data in this report do so at their own risk and will be responsible for making their own assessment and should verify all relevant representations, statements, and information with their own professional advisers.

The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.