Local ethical committee approval was obtained prior to the commencement of this study. Written informed consent was obtained from all patients by an independent research assistant. This trial was not registered in a clinical trials registry as there was no requirement to do so at the time the trial commenced. All patients undergoing elective primary total knee arthroplasty with participating surgeons were eligible for recruitment. Patients were excluded if they had a history of infection, refused to provide consent, or had an unstable knee that required a constrained or semiconstrained prosthesis. For patients who refused to provide consent, a metal-backed Kinemax Plus prosthesis (standard treatment at the time) was used.
All patients received the Kinemax Plus cruciate-retaining implant in which the metal components were made of Vitallium (cobalt-chromium) and the polyethylene was ultra-high molecular weight polyethylene that was gamma irradiated in air (Fig. 1). Randomization was performed in the operating room once it had been verified that the patient was a suitable candidate for an unconstrained knee prosthesis. Patients were randomized into one of two groups, either an all-polyethylene or a metal-backed tibial component, on the basis of a set of computer-generated random codes (a block of 30). Operating-room staff opened the sealed envelope and were responsible for providing the surgeon with the specified implant. Given the potential differences in outcomes based on a primary diagnosis of osteoarthritis and rheumatoid arthritis, we stratified the randomization by osteoarthritis and rheumatoid arthritis to ensure we had equal numbers of these patients in both primary study groups. It was impossible to blind the surgeons to the arm of the study as it is part of routine care to show patients their radiographs at the time of follow-up reviews; hence, it was impossible to blind the patients also.
Each surgeon followed the same surgical technique for tourniquet use, bone cuts, and instrumentation regardless of tibial component design. All knees had a minimum tibial component thickness of 10 mm. Patellar resurfacing was not performed, and soft-tissue release was performed as necessary. All implants were cemented with a standardized technique with use of Simplex cement (Stryker Howmedica, Mahwah, New Jersey). All patients received prophylactic antibiotics and thromboprophylaxis in the form of graduated compression stockings. In high-risk patients, pharmacological agents were used for thromboprophylaxis. Postoperatively, patients were mobilized with physical therapy after forty-eight hours and were discharged home when they were able to achieve a straight-leg raise and more than 100° of active knee flexion.
From August 1, 1993, to January 31, 1997, a total of 566 knees (510 patients), with a primary diagnosis of osteoarthritis (80.9%) or rheumatoid arthritis (19.1%), were recruited into this randomized trial. Overall, the ages ranged from fifty to ninety-three years (mean, 69.3 years), and 59% (299) were women. There was no difference in these demographics between the metal-backed and the all-polyethylene tibial component groups (Table I). A total of 293 patients returned for the ten-year review, and Table II shows reasons for less than ten-year follow-up data at the time of these analyses. The two patients who declined to return for a clinical review reported that they had no problems and had not undergone revision surgery. Of the twenty-three patients who repeatedly failed to return for follow-up, the last review was at an average of 6.5 years postoperatively. Two patients were last seen less than two years after surgery, and the remaining patients were last reviewed from five to nine years after surgery.
Patients were examined at three months, twelve months, and then at three, five, eight, and ten years. At each examination, standing anteroposterior and lateral radiographs of the knee were made. They were used to evaluate component position and alignment and to assess for evidence of trauma (for example, a periprosthetic fracture), loosening (for s example, radiolucent lines), and osteolysis, such as bone erosion caused by polyethylene wear particles. At the time of each assessment, serial radiographs were compared with the first postoperative radiograph to determine whether there had been changes in terms of implant position or progression of radiolucent lines. Radiolucent lines of >2 mm were considered to be an indication of potential implant loosening and osteolysis, and patients with progressive osteolysis were subsequently listed for revision. The surgeons noted that osteolysis could be present at the time of revision in the absence of progressive radiolucent lines. For this reason, all serial radiographs of patients who went on to have a revision, and the notes made by the operating surgeon at the time of revision, were reviewed by the senior surgeon (I.M.P.) participating in this trial.
Revisions and reasons for revisions were noted and, for consistency, were verified by the senior surgeon participating in this trial. The revision group also included the patients who were documented as requiring revision but were unfit for surgery. The primary outcome was survivorship of the implant with revision surgery as the end point. For patients who were listed for revision surgery but were medically unfit, the date listed for revision was used.
At all review times, patients who were lost to follow-up were documented as dead (the date of death was noted), unable to return for a review but still willing to participate in the study, moved and unable to participate further, or lost to follow-up (patients who failed to return for a follow-up appointment even after rescheduling new appointments).
Statistical Analysis
The sample size was calculated on the basis of an estimated survivorship of 96% for the overall group at ten years; 250 per group gives a 70% power to detect a difference of +3.3% (96% to 99.3%) or -5.6% (90.4% to 96%) and 80% power for differences of +3.6% (96% to 99.6%) and -6.5% (89.5% to 96%).
Life-table survival estimates were used to determine the estimated probability of survivorship at a minimum of ten years after primary total knee arthroplasty. Survivorship analysis was generated with the date of revision used as the end point for failure or, if the patient was subsequently unfit for surgery, then the date when he or she was listed for revision. Patients were censored if they were deceased or lost to follow-up at the time of their last review. The revisions done for reasons other than infection were classified as aseptic failures for these analyses. Survivorship analysis was performed for all revisions, for aseptic failures resulting in revision surgery, and for dichotomous groups on the basis of the type of tibial component design. Univariate analyses of associations between aseptic failure and the binary independent variables, such as diagnosis, sex, and type of tibial component design, were conducted with use of chi-square tests. Univariate analysis of the association between age and aseptic failure was made with use of two-sample t tests. A 5% significance level was maintained throughout these analyses, and all tests were two-sided.
Source of Funding
The research assistant who collected the data for the trial was funded through an education grant from Howmedica (Limerick, Ireland) initially and then by Stryker Howmedica (Newbury, United Kingdom). The study was investigator-led, and the funding sources had no input into the design or conduct of the trial.
A total of twenty-eight knees had a revision for any reason, and nineteen of them were for aseptic failure (Table III). These revisions include three knees in patients who required revision but were medically unfit for surgery and one knee in a patient with a failed replacement (due to rheumatoid arthritis and a medial ligament rupture) who did not return for any appointments. Of the aseptic failures, four were attributed to wear and loosening and three of the four were failures of the tibial component alone. It was clear that osteolysis was not a feature of failure in either arm of the study. A total of 272 knees (92.8%) had radiographs made after 9.5 years. The remaining twenty-one knees (7.2%) had the latest radiographs made between 7.5 and 9.5 years. There were relatively few abnormal radiographic findings at the time of the ten-year follow-up. It is of note that one patient (with the all-polyethylene design) had a lucent zone under the tibial tray at the eight-year follow-up, but no progression was seen at the time of the ten-year follow-up; one knee (with the metal-backed design) had some signs of polyethylene wear at ten years, but it was nonprogressive; one knee (with the metal-backed design) had signs of polyethylene wear at twelve years, but there was no progression and the patient remained asymptomatic at thirteen years; and one knee (in the metal-backed group) had questionable progressive radiolucency at the eleven-year follow-up evaluation, but the patient failed to attend additional reviews. The knees that were revised or listed for revision in this study had no evidence of progressive radiolucent lines.
The ten-year survivorship for the entire group, with revision for any reason as the end point, was 95.3% (95% confidence interval, 93.3% to 96.9%). In the subgroup analysis, ten-year survivorship, with revision for any reason as the end point, was 94.5% (95% confidence interval, 90.4% to 96.8%) for the all-polyethylene components and 96% (95% confidence interval, 92.6% to 97.8%) for the metal-backed tibial components. Ten-year survivorship with aseptic failure as the end point was 97% (95% confidence interval, 93.3% to 98.7%) for the all-polyethylene design and 96.8% (95% confidence interval, 93.6% to 98.4%) for the metal-backed design. No significant difference (p > 0.05) in survivorship was detected between the two designs (Fig. 2). Using a worst-case scenario in which fifty-five knees with less than ten years of follow-up had failed at the time of the last known follow-up, we found that there was no significant difference between the two designs (p = 0.50). In this worst-case scenario, the ten-year survivorship was 82.8% (95% confidence interval, 76.3% to 87.6%) for the all-polyethylene design and 86.8% (95% confidence interval, 81.5% to 90.6%) for the metal-backed design.
Patients who underwent revision for aseptic failure tended to be younger at the time of surgery, and this was of borderline significance (p = 0.051) for the all-polyethylene group; the mean age of the patients who had not had a revision was 68.9 ± 9.2 years, and the mean age of those who had an aseptic revision was 62.5 ± 27.1 years. The difference was not significant in the metal-backed group (p = 0.2); the mean age was 69.9 ± 8.0 years for the patients who had not had a revision and 66.8 ± 6.5 years for those who had had a revision. No significant association was identified between sex or diagnosis and aseptic failure.
This prospective randomized trial of total knee replacements compared the long-term survival of a metal-backed tibial component with that of an all-polyethylene tibial component with use of the Kinemax Plus prosthesis. Our results demonstrate excellent survivorship at ten years, with no significant difference between the two designs. These long-term results reinforce our earlier findings at five to eight years, which also demonstrated no significant difference between the two components13.
The long-term results of this prospective trial are in keeping with other published studies of the Kinemax implant14-16. Wright et al.14 reported excellent ten-year results for a series of 523 metal-backed Kinemax knees with a cumulative survival rate of 96.1%. Back et al.15 reported a cumulative survival rate of 99% at five years for 364 metal-backed Kinemax implants (from an initial sample of 422 implants). This was maintained throughout the study with a cumulative survival of 96.95% at nine years with revision as the end point, and all revisions were either for trauma or for infection. Data from the Norwegian Joint Registry16 included 213 metal-backed Kinemax implants with 98% survivorship at five years.
At the time of the ten-year follow-up, we reviewed a total of 293 patients and had a low number of patients (twenty-three) who had been lost to follow-up. There were only twenty-eight revisions, which included nine revisions for infection and only one of which occurred within the first year. The majority (twelve) of the nineteen revisions for aseptic failure occurred in the first five years and included the four patients with subsequent trauma. For the trauma patients, failure occurred secondary to the injury and included one tibial fracture sustained in a road traffic accident and two supracondylar femoral fractures and one periprosthetic femoral fracture sustained in falls. In addition, two aseptic failures were revised in the next five to ten years, and four knees were revised after ten years, predominantly for polyethylene wear and loosening.
As patients may be asymptomatic, serial radiographs are essential to assess the progression of osteolytic changes and to assess any changes in the position of the prosthesis that can be indicative of loosening. In the knees that were revised or listed for revision in this study, there was no evidence of progressive radiolucent lines or osteolysis. The importance of radiolucent lines and whether they indicate poor surgical technique have been debated for many years. Results from several clinical studies have indicated that nonprogressive radiolucent lines of <2 mm are not indicative of failure and are not associated with clinical symptoms17-22. In contrast, radiolucent lines of >2 mm which are progressive are commonly associated with failure20. In this study, any patient with evidence of progressive asymptomatic radiolucent lines of >2 mm was listed for revision.
It has previously been suggested that total knee arthroplasty should be used with caution in younger patients23 and that the all-polyethylene design was considered to be unsuitable for younger, more active patients and should be reserved for older, less active patients24. In our study, we found a trend for the patients who had undergone revision for aseptic failure to be younger at the time of surgery, and this reached borderline significance in the subgroup analysis of the all-polyethylene implants (62.5 compared with 68.9 years, respectively; p = 0.051). However, the overall revision rate for the all-polyethylene group was extremely low. Ranawat et al.24, in a study of thirty-eight patients with fifty-four knee replacements, reported a failure rate of 1.8% at an average of five years after arthroplasty with an all-polyethylene implant in younger, more active patients with osteoarthritis. The average age of all patients in that study was fifty-seven years.
Gioe et al.25,26 compared an all-polyethylene tibial component with a metal-backed tibial component in a prospective randomized study of 290 consecutive patients (316 knees) who were sixty years old or more. They also found no significant difference between the two designs at ten years, with the remaining 147 patients (120 had died, twenty-two had a revision, and one was lost to follow-up) demonstrating survivorship, with revision for any reason as the end point, of 91.6% in the all-polyethylene component group and 88.9% in the metal-backed component group. They highlighted the potential for considerable cost savings with the all-polyethylene tibial component as has also been identified by other authors27.
Metal-backed components have been associated with problems of so-called backside wear and osteolysis28. The renewed interest in an all-polyethylene component feeds the continuing debate surrounding these components, although many published studies, both small and large, have found no difference between the metal-backed component and the all-polyethylene component in terms of survivorship26. Apel et al.29 recommended the use of all-polyethylene over metal-backed components, recognizing the cost differential as early as 1991. Despite the lack of clinical findings to show that the metal-backed design is superior, and given that it is more expensive, it is surprising that currently only 3.9% of the 60,000 primary total knee replacements in England and Wales had an all-polyethylene tibial component1.
A limitation of this study is that the surgeon and patient were not blinded to the type of tibial component. The lack of blinding may have been a factor in the decision to proceed with revision in the patients undergoing revision when there was not a clear indication, such as infection, trauma, or failure of the implant (for example, loosening and wear). However, there were only eight knees (four with stiffness and four with instability) in which this occurred, and these were evenly distributed between the two tibial component designs.
Some surgeons may consider that the decision to use a cruciate-retaining implant and not to resurface the patella in all patients regardless of diagnosis is a limitation of this study. This is in contrast to current trends in the United States for routinely resurfacing the patella and the use of posterior stabilized implants in patients, especially those with inflammatory arthritis. The decision not to resurface the patella was based on published results from our institution that have supported this approach30. We found that, if accurate patellofemoral balance was obtained as a result of correct tibiofemoral alignment at the time of knee arthroplasty, there was not a notable prevalence of patellofemoral problems. In this study, there were no revisions or reoperations for patellar complications. Additionally, the knees in only five patients with inflammatory arthritis were revised (three for infection, one for stiffness, and one for trauma). No knee had late instability recorded because of a deficient posterior cruciate ligament.
A further limitation of this study is the use of survivorship analysis, which does not identify what proportion of patients may have clinical failure (that is, severe pain and functional limitation). However, our study has many strengths and this is evident when our study is compared with other studies that have compared an all-polyethylene tibial component and a metal-backed tibial component25,26,29,31-34 (Table IV). It is clear that our work adds to the literature as it describes a broad age range and a greater proportion of patients with inflammatory arthritis. Additionally, we believe it has the largest number of patients reviewed at ten years, providing the strongest results in terms of statistical power to detect the superiority of one tibial component design compared with the other. To determine equivalence between the two designs would require a much larger sample size than in our study.
The long-term results of this prospective, randomized controlled trial demonstrate excellent survivorship for both the metal-backed and the all-polyethylene designs, with no significant difference noted between them. These findings are strengthened by the fact that we had a large cohort of patients with a very high rate of follow-up at ten years. Finally, this study demonstrates the suitability of the all-polyethylene component for both older, less active patients as well as younger, more demanding patients. 