The effect of surgeon’s preference for hybrid or cemented fixation on the long-term survivorship of total knee replacement

Background and purpose Recent direct comparative reports suggest that hybrid fixation may have a similar or superior outcome to cemented fixation in total knee replacement (TKR); however, a paucity of long-term data exists. To minimize the confounders of a direct comparison, we performed an instrumental variable analysis examining the revision rate of 2 cohorts of patients based on their surgeon’s preference for cemented or hybrid fixation. Methods Registry data were obtained from 1999 until 2015 for 2 cohorts of patients who received minimally stabilized TKR, defined as those treated by high-volume hybrid fixation preferring surgeons, designated routinely hybrid (RH), and those treated by high-volume cemented fixation preferring surgeons, designated routinely cemented (RC). Results At 13 years, the cumulative percentage revision of the RC cohort was 4.8% (CI 4.1–5.7) compared with 5.5% (CI 3.5–8.7) for the RH cohort. The revision risk for each cohort was the same for all causes (HR =1.0 (CI (0.84–1.20)), non-infective causes, and for infection. This finding was irrespective of patient age or sex, patella resurfacing, and with non-cross-linked polyethylene (NXLPE). The RH cohort who received cross-linked polyethylene (XLPE) had a lower revision risk than the RC cohort with XLPE (HR =0.57 (0.37–0.88), p = 0.01). Interpretation The risk of revision for the patients of surgeons who prefer cemented fixation in minimally stabilized TKR is the same as for the patients of surgeons who prefer hybrid fixation, except when used with XLPE, where hybrid fixation has a lower revision risk.

Background and purpose -Recent direct comparative reports suggest that hybrid fi xation may have a similar or superior outcome to cemented fi xation in total knee replacement (TKR); however, a paucity of long-term data exists. To minimize the confounders of a direct comparison, we performed an instrumental variable analysis examining the revision rate of 2 cohorts of patients based on their surgeon's preference for cemented or hybrid fi xation.
Methods -Registry data were obtained from 1999 until 2015 for 2 cohorts of patients who received minimally stabilized TKR, defi ned as those treated by high-volume hybrid fi xation preferring surgeons, designated routinely hybrid (RH), and those treated by high-volume cemented fi xation preferring surgeons, designated routinely cemented (RC).
Results -At 13 years, the cumulative percentage revision of the RC cohort was 4.8% (CI 4.1-5.7) compared with 5.5% (CI 3.5-8.7) for the RH cohort. The revision risk for each cohort was the same for all causes (HR = 1.0 (CI (0.84-1.20)), non-infective causes, and for infection. This fi nding was irrespective of patient age or sex, patella resurfacing, and with non-cross-linked polyethylene (NXLPE). The RH cohort who received cross-linked polyethylene (XLPE) had a lower revision risk than the RC cohort with XLPE (HR = 0.57 (0.37-0.88), p = 0.01).
Interpretation -The risk of revision for the patients of surgeons who prefer cemented fi xation in minimally stabilized TKR is the same as for the patients of surgeons who prefer hybrid fi xation, except when used with XLPE, where hybrid fi xation has a lower revision risk. ■ The optimum fi xation in total knee replacement (TKR) is controversial, with cemented fi xation remaining the most common method internationally (National Joint Registry, AOA National Joint Registry 2016), compared with hybrid fi xation (cemented tibia and cementless femur) or cementless fi xation of both components. Hybrid fi xation was introduced to overcome the perceived concerns over cementless fi xation of the tibia while attempting to minimize femoral bone loss, decrease operative time, and reduce the polymethylmethacrylate burden of the joint (Wright et al. 1990, Kraay et al. 1991, Faris et al. 2008. While recent reports suggest that hybrid fi xation may have a similar or superior outcome to cemented fi xation (Petursson et al. 2015), a paucity of long-term data exists concerning this method of fi xation in TKR (Nakama et al. 2012). While cemented fi xation of both components has excellent longterm survivorship in national registries (National Joint Registry, AOA National Joint Registry 2016), in the Australian Orthopaedic Association National Joint Replacement Registry (AOANJRR), hybrid fi xation has the lowest revision risk overall when compared with cemented and cementless TKR. However, this revision risk is altered when prosthesis stability is considered. In posterior stabilized (PS) TKR, cemented fi xation has the lowest revision risk. Conversely, in minimally stabilized (MS) TKR there is no difference between hybrid fi xation and cemented fi xation, and both have a lower revision risk compared with cementless fi xation. The reasons why hybrid fi xation has a lower risk for all prosthesis types, but not when PS or MS TKR are considered individually, are uncertain.
Hybrid fi xation may not be appropriate for all patients, particularly in patients with osteoporosis, osteonecrosis, complex deformity, rheumatoid arthritis, or inaccurate bone resection (Scott 2012). These factors may bias registry data against cemented fi xation when directly compared. Conversely, hybrid fi xation may be used more commonly in younger active patients, which may bias registry data against hybrid fi xation.
Previous registry studies have performed direct comparison of hybrid versus cemented TKR survivorship rates. In contradistinction, we performed an instrumental variable analysis based on surgeon preference for different prosthesis fi xation options rather than the actual prosthesis received. This technique compares the revision rate of all primary minimally stabilized TKR undertaken by high-volume surgeons who preferred hybrid fi xation TKR to those undertaken by highvolume surgeons who preferred cemented TKR. The rationale for this instrumental variable approach is that it has the capacity to remove the confounding by indication or disease severity between hybrid and cemented fi xation that is not possible by directly comparing hybrid and cemented TKR implant registry revision rates (Vertullo et al. 2017).
Our primary hypothesis was that there would be no difference in the revision rate when the 2 patient cohorts were compared. Our secondary hypothesis was that there would be no difference in the revision rate with sub-analysis based on age, sex, type of polyethylene, and patella resurfacing.

Study design
2 groups of surgeons, who performed more than 50 TKR per year, differing by their femoral fi xation preference were selected to perform an instrumental variable survivorship analysis (Newhouse and McClellan 1998;Stukel et al. 2007 references missing) with surgeon preference serving as the instrument, using data from the Australian Orthopaedic Association National Joint Replacement Registry (AOANJRR). A revealed preference for a femoral fi xation option was defi ned as choosing to utilize it greater than 90% of the time, based on prior studies relating to knee implant choice by surgeons (Vertullo et al. 2017). Hence, a hybrid preferring (HP) surgeon used hybrid fi xation at least 90% of the time, and the patient cohort treated by those surgeons has been termed routinely hybrid (RH). A cemented preferring (CP) surgeon used cemented fi xation at least 90% of the time and the patient cohort treated by those surgeons has been termed routinely cemented (RC).
This study included all MS primary TKR undertaken for osteoarthritis (OA) with fi xed cemented tibial components and cemented or cementless femoral components, undertaken by the 2 groups of surgeons and reported to the registry, irrespective of patella resurfacing. PS TKR, mobile bearing TKR, cementless (cementless femur and tibia), reverse hybrid (cemented femur and cementless tibia) TKR, and TKR with a higher than anticipated risk of revision in the AOAN-JRR were excluded, as were non-osteoarthritic patients such as those with rheumatoid arthritis, or osteonecrosis (AOA National Joint Registry 2016). Data for the 2 patient cohorts and their treating surgeons were obtained from the AOAN-JRR from September 1, 1999, until December 31, 2015. The AOANJRR commenced data collection in 1999 and includes data on more than 98% of arthroplasty procedures performed nationally since 2002 (AOA National Joint Registry 2016). The AOANJRR collects information on prosthesis type by catalogue and lot number, as well as cement used for each component by catalogue and lot number. Intended component fi xation method is confi rmed by linking component data to an internally developed comprehensive international prostheses library, validated with both manufacturers and other registries. If the actual component fi xation method is not recorded at time of surgery (approximately less than 1% of TKR), the absent information is then obtained from the hospital. This linking of actual and intended fi xation ensures almost complete accuracy in determining and verifying the fi xation used in every procedure.
The AOANJRR defi nes MS prostheses as those that have a fl at or dished tibial articulation regardless of congruency, hence this group includes cruciate retaining and ultracongruent polyethylene options. PS prostheses provide additional posterior stability, most commonly using a peg and box design. Cross-linked-polyethylene (XLPE) was defi ned as ultra-highmolecular-weight polyethylene that has been irradiated with high-dose (≥ 50 kGy) radiation, regardless of re-melting or annealing (de Steiger et al. 2015).
Time to fi rst revision was the principal outcome measure, with revision being defi ned as any procedure that involves the insertion, removal, and/or replacement of a prosthesis. Reasons for revision and the type of revision were reported for procedures undertaken by both groups of surgeons. Further analyses based on patient's age, sex, patella resurfacing, and the type of polyethylene were also undertaken. Analysis of surgeon practice public/private mix, years of contribution to registry, number of TKR in registry, and hospital arthroplasty volume was also undertaken.

Statistics
Kaplan-Meier estimates of survivorship were used to estimate the time to the fi rst revision, with right censoring for death or closure of the database at the time of analysis. The unadjusted cumulative percentage revision (CPR) of the primary arthroplasty, along with 95% confi dence intervals (CI), was calculated using unadjusted point-wise Greenwood estimates. Hazard ratios (HR), calculated using Cox proportional hazard models and adjusted for age and sex, were used to make statistical comparisons of the rate of revision between the 2 cohorts. All tests were 2-tailed at the 5% level of signifi cance. The analysis was performed using SAS version 9.3 (SAS Institute Inc., Cary, NC, USA).

Ethics, funding, and potential confl icts of interest
The AOANJRR is approved by the Australian Federal Government as a Declaration of Quality Assurance Activity under section 124X of the Australian Federal Health Insurance Act, 1973. All investigations were conducted in accordance with ethical principles of research (the Helsinki Declaration II). No funding was received specifi c to this study and there are no competing interests to declare.

Results
There were 39,623 primary TKR that met the inclusion criteria, undertaken by 108 surgeons, with 30,544 cemented TKRs and 9,079 hybrid TKRs (Figure 1). Most surgeons were cemented preferring (87%) and they undertook 77% of the included procedures. The hybrid preferring surgeons each performed on average more of the included TKR compared with the CP surgeons, with a mean of 649 TKR/surgeon compared with a mean of 325 TKR/surgeon, respectively (Table 1). The CP surgeons undertook hybrid fi xation in 0.7% of their TKR and the HP surgeons undertook cemented fi xation in 2.8% of their TKR.
The demographics of each surgeon group had some differences, with the proportion of surgeons who worked in both private and public settings being higher in the HP surgeons (93%) compared with the CP surgeons (70%). The CP surgeons had contributed to the registry for more years (mean 9.6 years) compared with the HP surgeons (mean 6.2 years). Otherwise, the mean number of TKR in the registry, volume of all TKR/year and respective hospital arthroplasty volume was comparable for each group (Table 2).
The 5 most common diagnoses at revision were similar in each cohort (Figure 3) (Table 4, see Supplementary data). The types of revision were similar between the cohorts.
When the effects of age were examined, the revision risk for the RC cohort who were less than 65 years was similar to the RH cohort who were less than 65 years (HR = 1.1 (1 = 0.83-1.42)). Similarly, the revision risk for patients older than 65 years (HR = 0.96 (0.76-1.21)) was similar between the 2 cohorts.
Stratifi cation of patients into males and females aged less than 65 years (Figure 4, see Supplementary data) and greater than 65 years revealed the same revision risk both for males in each cohort, and for females in each cohort.
XLPE usage was more common in the RC cohort (47%) than in the RH cohort (29%). When the effects of XLPE were examined, the RH cohort with XLPE had a lower revision

Number at risk (Figure 2)
Year RC RH    risk than the RC cohort with XLPE (HR = 0.57 (0.37-0.88)) ( Figure 5). Revision risk with non-cross-linked-polyethylene (NXLPE) was the same between the cohorts. Patella resurfacing was more common in the RC cohort (61%) compared with the RH cohort (51%). When the effects of patella resurfacing were examined, both cohorts had the same revision risk with and without patella resurfacing ( Figure  6, see Supplementary data).
When the revisions for infection in each cohort were stratifi ed by sex there was no difference between males and females in each cohort (Figure 7).

Discussion
The advantages and disadvantages of cement fi xation have been debated for decades; however, the choice of TKR fi xation typically remains the preference of the surgeon, characteristically founded in efforts to maximize the long-term outcomes of their patients (Kobs and Lachiewicz 1993), with loosening and lysis remaining the dominant reasons for revision.
In this analysis, patients of surgeons who preferred hybrid fi xation had the same long-term risk of revision compared with the patients of those surgeons who preferred cemented fi xation. This fi nding was irrespective of patient's age or sex, and whether the patella was resurfaced or not. When reasons for revision were stratifi ed into non-infective and infective, there was no difference between the 2 cohorts overall. There was also no difference between the 2 cohorts for those who received NXLPE, but the patients of surgeons who preferred hybrid fi xation and received XLPE had a lower revision risk than the patients of surgeons who preferred cemented fi xation and received XLPE.
Cement disease (Jones and Hungerford 1987) was described in 1987, suggesting particles of polymethylmethacrylate were the primary cause of osteoclast-induced failure at the prosthesis-bone interface. More recently, the theory of cement disease has been discarded in favor of wear-particle induced lysis and loosening, which primarily focuses on bearing surface generated particles rather than those from the fi xation interface (Harris 1994). Furthermore, cement has been suggested as a possible protective barrier, or seal, to wear particle-laden synovial fl uid ingress into the prosthesis¬-bone interface (Harris et al. 1996). Our results are not in keeping with either the cement disease theory or cement as a seal theory.
Despite a paucity of supporting clinical data, cementless fi xation of the femur and tibia has been recommended as the optimum biologic TKR fi xation solution for younger at-risk patients when compared with cemented fi xation (Dorr 2002). However, in registry studies, the long-term revision risk of cementless fi xation is higher than cemented and hybrid fi xation and in smaller clinical series outcomes of cementless fi xation of the femur and tibia remain similar, or inferior to, cemented fi xation of the femur and tibia (Pulido et al. 2015, Dalury 2016. It is for this reason we did not examine cementless fi xation of both components in this analysis. Cementless fi xation of the tibia does not reduce the revision risk or migration when compared with cemented fi xation in radiostereometric trials (Carlsson et al. 2005), in registry studies (Graves et al. 2016), or clinical series (Behery et al. 2016). Given that the tibia remains the component most at risk for failure (Voigt and Mosier 2011), hybrid fi xation was introduced as a pragmatic alternative to employ the advantages of 2 differing fi xation philosophies (Kraay et al. 1991, Petursson et al. 2015. Hybrid fi xation has similar outcomes to cemented fi xation in direct comparisons (Pelt et al. 2013), clinical series (Choi et al. 2012, McLaughlin andLee 2014), and in some registry reports, a lower revision risk (Petursson et al. 2015, AOA National Joint Registry 2016. Petursson et al. (2015) performed a registry review of 3 different fi xed and mobile bearing TKR designs, reporting that 1 of 3 three designs examined had a lower revision risk in the hybrid version. When this prosthesis, mainly performed at one high-volume hospital, was excluded, there was no difference between hybrid and cemented fi xation in their direct comparative analysis, in keeping with our results.
We specifi cally examined the effects of age and sex on the revision risk in each cohort. Cemented fi xation for older females may have had some advantage due to lower femoral bone density; however, we found no difference with hybrid fi xation in females over 65 years, consistent with other authors (Nakama et al. 2012, Dalury 2016. Similarly, hybrid fi xation may be of advantage to younger active males, but there was no difference between fi xation types for males under 65 years. We also assessed whether hybrid fi xation lowered the infection risk given recent registry data suggesting a lower rate of revision for infection with certain TKR designs and bearing materials (Vertullo et al. 2017). Theoretically, a reduced burden of cement, cement particulate, and third-body wear could favorably alter the local immunomodulation of the joint environment (Spaan et al. 2013); however, our results suggest no advantage exists when only 1 major component is cemented.
In registry studies, the use of cross-linked polyethylene in TKR lowers the risk of loosening and lysis when compared with non-cross-linked polyethylene (de Steiger et al. 2015), presumably via a reduction in particle-related osteolysis. In our series, when the effect of XLPE was examined, it resulted in a 43% lower revision risk in the HP cohort compared with the CP cohort. It remains uncertain if this is due to an additive effect of XLPE when used with a lesser volume of cement, or some other unrecognized confounders such as patient selection or femoral component design.
When possible confounders were reduced using instrumented variable methodology, our analysis did not demonstrate superior survivorship with hybrid fi xation overall. Consequently, it remains uncertain whether the extra cost of cementless femoral components is justifi ed by the reduced operative time (Petursson et al. 2015) and possibly decreased bone loss given the excellent long-term results of cemented fi xation of the femur.
By comparing revision risk based on surgeon fi xation preference, we believe we have addressed concerns related to the potential for selection bias that may arise in a direct comparison of all cemented and hybrid TKR. To our knowledge this is the fi rst time registry data have been used to investigate the outcome of surgeon preference in TKR fi xation rather than directly comparing the long-term revision risk of hybrid and cemented TKR.
This study was specifi cally designed to address major confounders that may introduce bias between the hybrid and cemented fi xation. The impact of potential differences due to age, gender, and primary diagnosis have been considered. In addition, prosthesis-specifi c factors such as the use of posterior cruciate stabilization, mobile bearing, patellar resurfacing, and cross-linked polyethylene were also considered. Only patients with osteoarthritis were included as other diagnoses such as rheumatoid arthritis are more likely to have a higher incidence of osteoporosis and consequently a higher use of cemented TKR. Mobile bearing TKR were excluded as they have a known higher revision risk compared with fi xed bearing, and potentially could have a detrimental interaction with cementless femoral fi xation (AOA National Joint Registry 2016). PS TKR were excluded as they have a higher revision risk than MS TKR and have a higher revision risk with hybrid fi xation than cemented fi xation (Vertullo et al. 2017).
As this is a registry analysis, some specifi c clarifi cations are important. First, a registry analysis differs from a clinical trial, in that while it can identify and monitor comparative national outcomes it cannot assign causality. Nonetheless, to optimize TKR survivorship, it is not vital to know why there is a difference between options, just that one exists, allowing all stakeholders to make shared informed decisions (Graves 2010). Second, another issue is unrecognized confounders. Unrecognized selection bias or confounding may have occurred, but, by focusing on the surgeon's stability preference rather than the actual prosthesis used, this risk is minimized. Randomized controlled trials can reduce this selection bias; however, the current RCT literature showing no survivorship difference between hybrid and cemented fi xation is underpowered to show a difference and has inadequate follow-up (Nakama et al. 2012). Surgeons with less surgical experience may prefer cemented fi xation, but we restricted our analysis to surgeons who perform over 50 TKR per year to remove performance bias, as this has been previously cited as a large enough volume to exclude surgeon inexperience (Abdel et al. 2011). While the hybrid fi xation preferring surgeons had on average more TKR per surgeon in the analysis than the cemented fi xation preferring surgeons, the cemented fi xation preferring surgeons performed more TKR per year on average and overall had performed a greater number of TKR. The AOANJRR collects level 1 and 2 data, hence comorbidities, patient-recorded outcome measures, and prosthesis alignment data are not collected. The AOANJRR only recently commenced recording ASA and BMI, and hence these factors could not be included in this analysis.
A possible limitation with any registry-based analysis is the data's accuracy and validation. In the AOANJRR, after an initial capture rate of 96.8% (AOA National Joint Registry 2016), a sequential multi-level matching process against health department unit record data is undertaken, resulting in an almost complete dataset of primary and revision knee replacement in Australia.
In summary, there was no overall difference in the revision risk for the patients of surgeons who prefer hybrid fi xation in minimally stabilized TKR, compared with the patients of surgeons who prefer cemented fi xation. Only when the effects of alternative bearing surfaces were examined had the patients of surgeons who preferred hybrid fi xation and utilized XLPE a 43% reduction in revision risk. Table 4, Figure 4 and Figure 6 are available as supplementary data in the online version of this article, http://dx.doi.org/ 10. 1080/17453674.2018.1449466 CJV designed the study, developed the methodology, performed the analysis, and wrote the manuscript. YP developed the methodology and performed the analysis. SEG and PLL collected the data, developed the methodology, performed the analysis, and wrote the manuscript.