Numerous failure mechanisms of total knee arthroplasties (TKA) have been described, including polyethylene (PE) wear, instability, extensor mechanism dysfunction, and progressive radiolucencies.19,32,53 Keating et al30 reported the failure mechanisms of all three components of the Anatomic Graduated Component (AGC) (Biomet Inc, Warsaw, IN) in 4583 knees. They described an osteonecrotic lesion in the tibia seen in 12 patients leading to failure. The tibial component of the AGC total knee replacement is a nonmodular, metal backed, cemented tibial implant with compression-molded PE. We have previously reported a survival rate of this implant to be 98.9% at 10 and 15 years with no knees revised for PE wear or osteolysis.46 The question is if the knees do not fail through the most commonly reported failure mode of PE wear then by what mechanisms does the tibial component fail? The current study includes additional knees operated on until the year 2000 and examines which specific factors are associated with tibial component failure.
MATERIALS AND METHODS
A consecutive series of 8598 Total Knee Arthroplasties (TKA) using the Anatomic Graduated Component System (AGC, Biomet, Inc, Warsaw, Indiana) were done in 5535 patients between 1983 and 2000. Inclusion criteria for this study were a preoperative diagnosis of osteoarthritis and a minimum of 2-year followup, which reduced the cohort to 3152 knees in 2125 patients. Infections were excluded. There were 1281 (60.3%) women and 844 (39.7%) men. The average age was 70.0 years with a range from 33 to 93 years. The mean followup was 5.0 years ranging from 2 to 14.2 years. All of the tibial components examined in the study were cemented and metal backed. All tibial components were manufactured with compression-molded Himont 1900 PE resin onto a Co-Cr endoskeleton resulting in a nonmodular one-piece design. From 1983 to July 28, 1992, the components were sterilized in air, between July 28, 1992 and November 1995 they were gamma irradiated in Argon with a first-generation barrier film, and since November 1995 they have been sterilized with gamma irradiation in Argon with a second-generation barrier film. There were four tibial metal endoskeleton designs used for the study period that varied in thickness and metal configuration but not PE articular surface geometry (Table 1).
Age, Body Mass Index (BMI), overall limb alignment, Knee Society Clinical,27 and Pain Scores were determined preoperatively and during regularly scheduled office visits at 2 months, 6 months, 1, 3, 5, 7, 10, 12, and 14 years postoperatively. Meniscal and chondral degeneration, tibial base plate endoskeleton design, tibial size, femoral size, ligamentous releases, ROM, and laxity were recorded at the time of surgery.
Radiographic evaluation included standing AP, lateral, and sunrise patella views. Radiographs were analyzed for radiolucent lines, tibial component alignment, overall limb alignment and PE wear.14 Anteroposterior radiographs were done standing with standing 14” x 17” cassettes. All radiographic alignment measurements were done using the tibial-femoral anatomic alignment axis using a goniometer. Varus alignments are reflected in negative measurements, and all valgus alignments are represented by positive numbers. Failure was defined as component migration, functional instability, and/or total radiolucencies, and revision because of unremitting and unexplained pain.
Time from implantation to failure related to each mechanism was recorded. Cause of failure was determined from radiographic and clinical findings at the time of revision surgery.
The statistical analysis in this study was primarily driven by the tools and techniques of data mining. Regression tree analysis29,51,54,66 was used to select the most meaningful variables in the study. By examining all possible breaks of the data across every clinical and surgical variable, ie age, BMI, postoperative alignment, postoperative tibial alignment, and others, the procedure selected the most statistically relevant subdivisions of the data set affecting the response, which in this case was prosthesis survival.
Failure time was linearized as in Therneau and Atkinson60 before the regression tree procedure. The results from the regression tree were further verified by Kaplan–Meier survival analysis and rank tests of association of covariates. All reported odds ratios were attained by Cox proportional hazards regression. Algorithms were created in Statistical Analysis Software (SAS) System3,51,54 to further verify the optimal age and the optimal overall alignment given by the regression tree procedure. These optimal ranges, as identified by the SAS algorithms, were chosen based on which range gave the greatest statistical significance in Cox proportional hazards analysis. Other meaningful variables with significance level less than 0.05, such as tibial alignment, were included in each iteration, removing the effect of tibial alignment in determining the effect of age on survival.
Of the 3152 AGC prostheses implanted, there were 41 tibial failures that led to revision (1.3%). All knees with failed tibial components were examined for mechanisms of failure. In the failed knees, 17 (42%) of the knees were in male patients and 24 (58%) of the knees were in female patients. Patients whose procedures failed were younger (mean 66.2 years, range 42-86 years) than those whose knees did not fail (mean 70 years, range 33-93 years) (p = 0.0066, Wilcoxon; p = 0.0067, log rank). Gender (p = 0.7983, log rank) and BMI alone (p = 0.9342, log rank) were not associated with failure. Five surgeons did the procedures and physician differences were not associated with failure.
Four mechanisms of failure were identified: medial tibial bone collapse, ligamentous imbalance, progressive radiolucencies, and pain. The mean time to failure is reported for each of the observed mechanisms (Table 2).
Medial bone collapse was the most common mechanism of failure present in 20 of the 41 failed knees. An example of this mechanism of failure observed for a period of time is shown in Figure 1. All 20 knees were in relative varus alignment preoperatively with a mean of −3.7° varus (range –14° to +6°). The mean overall postoperative anatomic tibiofemoral alignment in this group was +1.6° valgus compared with the mean alignment in the overall group of +3.9o valgus. Regression tree analysis determined that a tibial component with more than 3.0 degrees of varus alignment had increased odds of failure (Hazard Ratio = 17.2, p < 0.0001). Kaplan-Meier Survival Analysis based on tibial component alignment is shown in Figure 2. No preoperative knee in valgus alignment failed through medial bone collapse.
Increased BMI > 33.7 when combined with varus tibial component alignment was associated with a 168-fold increase in failure (p < 0.0001). Kaplan-Meier Survival Analysis based on tibial component alignment and BMI is shown in Figure 3. Although BMI was significant as a sequential variable, it failed to have a significant effect on survival when not combined with varus tibial alignment. Using Kaplan-Meier forward stepwise tests for covariates to examine the order of importance of variables associated with decreased survival, the factors associated with medial bone collapse in order of importance were varus tibial component alignment > 3.0o, BMI > 33.7, and overall varus limb alignment.
The possible protective effect of increased femoral component valgus combined with a varus tibial component was examined. In the tibial components that had more than 3.0o varus alignment (n = 376, 12% of all TKAs) the overall postoperative limb alignment was divided into +1 to +17 valgus and then compared with those with 0° to −5° of varus. The odds of failure in the knees with tibial component 3.0° or more of varus and overall limb alignment from +1 to +17 were reduced from 17.2 to 10.0 times the odds of failure when compared with the group that had a tibial component alignment of less than 3.0° of varus. Therefore, increased femoral component valgus in the face of tibial component varus resulted in a partial reduction in failure but was not fully protective of the strong effect tibial component varus had on decreased survival.
Regression tree analysis determined that an age of 57 years or younger was associated with increased likelihood of failure. The age range of 33 to 56 years had 8.3 times the odds of failure compared with the age range of 57 to 93 years group (p = 0.0019). For knees with tibial component alignment < 3.0° of varus, age is not a significant factor in survival.
Ligamentous imbalance resulted in failures in 13 knees. An example of this mechanism observed for a period of time is shown in Figure 4. The mean preoperative anatomic limb alignment in this group was 8.2° (range –11° varus to +20° valgus). Eight of the knees developed excessive posterior lateral corner rollback with eventual subluxation. Seven of the eight patients with posterior lateral instability had marked valgus preoperative tibiofemoral alignment with a mean of +14°. There were two patients with posterior medial subluxation. Both of these patients had preoperative varus alignments of more than 5°. Two patients had global ligamentous instability and 1 had traumatic posterior cruciate ligament instability.
Using Kaplan-Meier forward stepwise tests for covariates to examine the order of importance of variables associated with decreased survival; the most important factor associated with ligamentous imbalance was preoperative valgus alignment (p < 0.0001).
Progressive radiolucency was the mechanism of failure in six knees. The average postoperative tibial component alignment in this cohort was 3° of varus. Two of the six failures had a varus tilt more than 3.9°. Five of the six (83%) failures in this cohort had preoperative alignments of ≥ 10° varus. None of the knees had osteolysis or PE wear at the time of the revision.
There were two failures in one patient because of the presence of inexplicable pain. There was increased uptake on bone scan under the tibial components. Because the pain was not caused by any noticeable reason, the measurements were removed from the analysis of the failure mechanisms. The pain did not improve with revision of either knee.
The mechanisms of TKA failure leading to revision have been examined and include PE wear, component loosening, instability, infection, arthrofibrosis, malalignment, extensor mechanism dysfunction, and patellar problems.15,53,64 Reductions in PE wear through enhanced locking mechanism and manufacturing methods have been proposed and examined but may not eliminate the most common failure mechanisms of PE wear17,24,34 and osteolysis39 in the long term. The AGC tibial component has an extremely low rate of revision because of PE wear.13,16,49,65 This is thought to be because of the combined effects of the nonmoldular design and the direct compression molded PE resin.
A considerable amount of work has been done examining the material properties and clinical biomechanics of the implant side of the tibial component of the TKA. This body of work has examined PE peak stresses,56 use of metal backing,1 tibial stem length, and articular geometry. The majority of this work has been with finite element analysis.5–8,11,20,22,42–45,52,62 In general, these results by finite element analysis have shown that increasing articular congruity, addition of metal backing to the tibial component, and increasing the thickness of the PE all decrease PE stress.
In regards to the bony side of the construct, it has been proposed also through finite element analysis that stress in the cancellous bone is decreased with addition of a metal tray and a peg5 and that alignment has a considerable effect on proximal tibia loading.21,33 The contact stresses in the loaded total knee replacement have been examined with digital pressure sensor film38 and finite element analysis.35 Contract stresses were found to increase with flat geometry and varus tibial alignment. This investigation was conducted focusing PE wear as a failure mechanism and not necessarily focusing on bony collapse. Furthermore, in a mathematical rolling model, PE stresses decrease as the thickness of the PE increases.62 Taylor et al58 proposed the concept of medial cancellous overload, where cancellous bone stresses were examined with further finite element analysis.57 The found that the best bone was in the central portion of the condyles and not at the edge where most of the abnormal loading conditions occur.
All these previous studies have examined the factors that negatively effect PE yield strength and result in failure of tibial implants through a wear-based mechanism. Our clinical studies however have not shown PE wear to be a failure mechanism with this implant design.46 If PE wear is not a clinical failure mechanism for a nonmodular implant, then why do the knees fail?
Postoperative limb alignment plays a crucial role in the long-term survival of a TKA.4,36,59 Varus postoperative limb alignment has been shown to be associated with failure of TKAs when compared with limbs in normal alignment. Ritter et al47 reported on a series of posterior cruciate condylar TKAs with 27 of 38 (71%) tibial failures found to be in varus alignment. Aglietti et al2 found that any tibial component with a varus tilt of more than 2° had a considerably greater occurrence of radiolucent lines. They also concluded that optimal fixation of the tibial component occurred with a tibial cut of 90°. The current study found increased failure associated with tibial component alignment more than 3.0° of varus and with overall limb alignment in relatively less valgus.
A recent biomechanical study from Green et al21 found that varus tibial component alignment was associated with increased posteromedial and anteromedial tibial surface strain. The increased strain from tibial component varus alignment, observed in this biomechanical study, compared with a neutral tibial cut may indeed lead to medial cancellous bone overload and eventual failure through medial tibial collapse as occurred in 20 knees in this series. The clinical manifestation of this process was described by Keating et al30 and begins with an osteolytic lesion in the medial tibia that may heal, but may progress to failure. Varus alignment is thought to considerably increase edge loading and may overload the medial compartment and result in failure through either cancellous fatigue or a possible osteonecrosis pathway. Pathologic specimens from revision surgery have not yet clarified the pathologic mechanism of failure. Although pathologic specimens were not obtained in every case, several of the cases, medial tibial collapse did indeed reveal osteonecrotic bone in the medial tibial lesion.
In this study, which involved large retrospective databases, data mining techniques, such as regression tree and iterative algorithm analysis, paired with traditional analysis such as Cox regression led to a better understanding of the questions under study than has been previously reported in the medical literature.
Excessive medial tibial loading does appear to be the final pathway of failure seen in this study. Factors that increase medial edge loading, such as varus tibial component alignment, relative overall varus limb alignment, and the combination of increased BMI > 33.7 and varus tibial component alignment were found to be associated with medial bone collapse. Interestingly, neutral alignment and increased BMI were not associated with failure. This also correlates well with the biomechanical work of Green et al21 who found that neutral alignment of the tibial component resulted in more balanced or centralized loading patterns that may protect the tibia and possibly reduce failure through medial collapse. Perillo-Marcone et al43 reported that, between 2.5° and 5° of varus tibial alignment, there was a 41% increase in cancellous bone stresses. The increase in bone stress was not caused by an increase in load, but is because of the decreased area in which the load is applied.
Hsu et al26 attempted to establish an optimal tibial angle for the Kinematic prosthesis, which is also a metal-backed tibial component. Their data led them to conclude that a tibial angle of 2° varus was optimal for this particular component. In the current series, regression tree analysis determined that components with more than 3.0° of tibial component varus had a higher incidence of failure (p < 0.0001). This value falls into the range of increased stress on the cancellous bone presented by Perillo-Marcone et al,43 and the bone collapse could be indicative of the increased stress placed on the bone presented in both biomechanical studies.
The effects of BMI on survival was analyzed sequentially with the alignment of the tibial component and found to be a considerable combined variable that affected survival. However, when this index was examined without the condition of tibial alignment, no decrease in the survival of the prosthesis was found. Several studies22,41,55 have also found that the survival of a TKA in obese patients is not reduced compared with nonobese patients for 4 to 10 year followup. These studies included both PCL- retaining and PCL-substituting prostheses with cemented and cementless fixation. The affect of alignment and increased BMI was not analyzed in these studies. In the current study, varus alignment and BMI > 33.7 was highly associated with failure and further supports a medial cancellous overload pathway that may be protected by neutral alignment.
The possible protective nature of compensating for the negative effect of varus tibial component alignment > 3.0o through increasing femoral component valgus to reduce overall varus limb alignment is an important question. In this study, increased femoral valgus had a partial protective effect but did not completely eliminate failure associated with tibial component varus. This is seen in the hazard ratio being reduced from 17.2 to 10.0, but not to zero, when comparing the alignment group that had a tibial component alignment of < 3.0o of varus to the knees with a tibial component > 3.0o of varus, but a more valgus overall alignment.
Young age was associated with failure through the medial bone collapse mechanism but not failure in all groups. Regression tree analysis identified 57 years as an age below which survival was negatively effected. Vazquez-Vella Johnson et al61 analyzed demographic factors associated with failure with the AGC total knee replacement and found that age under 60 years was associated with decreased survival. They also found obesity more associated with failure. One limitation of this study was a small sample size in the group with combined obesity and age < 60 of only 6 patients. Hofmann25 found that PE exchange was not infrequent in younger patients with uncemented modular TKAs, but fixation was reliable.
Ligamentous imbalance or instability was the second most common cause of tibial component failure in this study. Posterolateral subluxation was highly associated with preoperative valgus alignment averaging +14 of valgus in knees that failed through this mechanism. The etiology of this failure mechanism may be a result of rotational imbalance in the knee as a result of extraarticular deformity associated with a valgus knee deformity such as a planovalgus foot. Relatively flat coronal plane geometry with less articular restraint may be associated with instability as reported with the Miller-Gallante II TKR.40 Joint line position has also been shown to affect stability of a TKR.37 The effect of laxity on long-term ligamentous imbalance leading to revision was not specifically analyzed in this study. Feng et al17 observed that ligamentous imbalance increases translational and rotational motion in prostheses with flat-on-flat articulating surfaces. This increased motion causes increased stresses on the articular surfaces, further subluxation, and increased stress on the periphery of the PE. In an 11-year followup study of the AGC, Emerson et al13 reported no failures caused by ligamentous imbalance and no observed tibial-femoral subluxations. They concluded this to be because of medial eminence and anterior flare, which help to provide some constraint of the articulating surfaces, preventing edge loading.
Perhaps the ligamentous imbalance seen in the current investigation resulted from rotational imbalance of the knee or relative tightness of the posterolateral or posteriormedial structures or residual PCL tightness. Posterior instability was infrequent with the AGC, but has been described by Waslewski et al63 as a cause of revision in some PCL-retaining TKAs with improper flexion gap balance. Perhaps additional articular constraint may be warranted in knees with excessive preoperative valgus deformity or with considerable extraarticular deformity in the foot or hip. Rorabeck50 and Fehring and Valadie16 have described the possible need for additional constraint in revision of a TKR with clinical instability.
Sharkey et al53 found aseptic loosening to be the second most common mechanism of failure. Cemented fixation of AGC nonmodular components appear to be quite durable for a long period of time with only six knees failing through progressive radiolucencies and loss of fixation. We did not observe PE wear and osteolysis in any of these knees. The mobile bearing knee replacement is typically thought to decrease the amount of stresses on the PE and at the bone–implant interface. The decreased amount of stress at the bone–implant interface would decrease the occurrence of component loosening; however, this appears to be theoretical, because Kim et al31 showed there was no difference in aseptic loosening between fixed bearing knees and mobile bearing knees in the same patient. It remains unclear whether or not a mobile bearing articular surface would decrease medial bone collapse and ligamentous imbalance. Revision TKR for aseptic loosening has been reported by Friedman et al18 Cement fixation proved reliable in this investigation with only six knees failing through progressive radiolucencies and aseptic loosening. Reduction in radiolucent lines through bony preparation48 may enhance long-term survival. The natural history of radiolucencies has been reported as well.12
The AGC prosthesis is a PCL-sparing TKA with a relatively flat-on-flat coronal plane geometry and an anterior lip to limit posterior translation. Feng et al17 reported that posterior cruciate sparing prostheses with flat-on-flat femoral tibial articulation have a higher wear rates because of increased stresses on the PE, leading to revision. The AGC component, however, has not seen the increased failure rates noticed in other flat-on-flat articulating prosthesis. Osteolysis and PE wear are seen primarily in modular total knee designs and may lead to early revision.15,39 The source of the wear particles may be from the articular surface and the undersurface or backside of the modular PE. Polyethylene wear has been reported in PCL-retaining knees,31 PCL-substituting knees,39 and mobile bearing knees.31 Mobile bearing designs are thought to decrease loosening, PE wear, and PE stresses, however, Buechel10 reported an osteolysis rate requiring revision of 1.8%. Hartford et al23 reported increased loosening in an uncemented mobile bearing design. In the current study, no failures were seen to be caused by either PE wear or osteolysis.
The AGC prosthesis has been changed in both endoskeleton design and sterilization technique since 1983. The prosthesis has been sterilized in three different environments throughout the evolution of the product. In this series, there is no difference in the number of failures for each of the three sterilization environments. There have been four different designs of the nonmodular component’s tibial metallic endoskeleton, mainly differing in the thickness of the metal tray, and design was not associated with increased failure in this study.
The primary forces throughout the limb that lead to preoperative varus or valgus deformity may play a considerable role in the failure mechanism for both medial bone collapse and ligamentous imbalance. The current investigation revealed that preoperative deformity considerably affects the postoperative failure mechanism. Varus knees were more likely to fail through medial bone collapse as shown in Figure 4, whereas valgus knees were more likely to fail through ligamentous imbalance as shown in Figure 3.
Limitations of the current study are a small selection bias, because one implant design was used in the majority of, but not all, patients treated at our institution. Furthermore, the patients lost to followup may have had an effect on outcome; however, Joshi et al28 concluded that outcome in nonattenders does not necessarily have poor results.
Four distinct mechanisms of failure of a nonmodular compression molded tibial component have been identified. These include: (1) medial tibial bony collapse, (2) ligamentous imbalance, (3) progressive radiolucencies, and (4) pain. Preoperative deformity plays an important role in the mechanism of failure of the tibial component. In general, preoperative varus knees failed by medial compartment overload, whereas preoperative valgus knees were more likely to fail because of posterolateral compartment subluxation and ligamentous imbalance. This emphasizes for the surgeon the importance of deformity correction at the time of TKA and assessment of other variables, which may have precipitated the preoperative deformity of the knee. The combined effect of varus tibial component alignment > 3.0o and increased BMI > 33.7 supports the concept of a medial tibial cancellous bone overload mechanism resulting in collapse, which was the most common failure mechanism observed with this TKA design. Residual varus limb alignment and age < 57 years are also associated with this failure mechanism. Increased femoral component valgus in the face of a tibial component positioned in > 3.0o of varus enhances survival but does not fully eliminate the negative effect of tibial component varus > 3.0o on the long term survival of a TKA. Intraoperative alignment and ligament balance remain crucial factors in TKA component longevity.
The authors thank the Joint Replacement Surgeons of Indiana Research Foundation.
1. Adalberth G, Nilsson KG, Bystrom S, Kolstad K, Milbrink J: Low-conforming all-polyethylene tibial component not inferior to metal backed component in cemented total knee arthroplasty: Prospective, randomized radiostereometric analysis study of the AGC total knee prosthesis. J Arthroplasty 15:783–792, 2000.
2. Aglietti P, Buzzi R: Posteriorly stabilized total-condylar knee replacement: Three to eight years’ follow-up of 85 knees. J Bone Joint Surg 70B:211–216, 1988.
3. Allison PD: Survival Analysis for SAS System. Cary, North Carolina, SAS Institute Inc., 1995.
4. Bargren JH, Blaha MD, Freeman MAR: Alignment in total knee arthroplasty: Correlated biomechanical and clinical observations. Clin Orthop 173:178–183, 1983.
5. Bartel DL, Burstein AH, Santavicca EA, Insall JN: Performance of the tibial component in total knee replacement. J Bone Joint Surg 64A:1026–1033, 1982.
6. Bartel DL, Bicknell VL, Wright TM: The effect of conformity, thickness, and material on stresses in ultra-high molecular weight components for total joint replacement. J Bone Joint Surg 68A:1041–1051, 1986.
7. Bartel DL, Burstein AH, Toda MD, Edwards DL: The effect of conformity and plastic thickness on contact stresses in metal-backed plastic implants. J Biomech Eng 107:193–199, 1985.
8. Bartel DL, Rawlinson JJ, Burstein AH, Ranawat CS, Flynn Jr WF: Stresses in polyethylene components of contemporary total knee replacements. Clin Orthop 317:76–82, 1995.
9. Breiman L, Friedman J, Olshen R, Stone C: Classification and Regression Trees. Belmont, California, Wadsworth International Group, 1984.
10. Buechel Sr FF: Long-term follow-up after mobile-bearing total knee replacement. Clin Orthop 404:40–50, 2002.
11. D’Lima DD, Chen PC, Colwell Jr CW: Polyethylene Contact Stresses, Articular Congruity, and Knee Alignment. Clin Orthop 392:232–238, 2001.
12. Ecker ML, Lotke PA, Windsor RE, Cella JP: Long-term results after total condylar knee arthroplasty: Significance of radiolucent lines. Clin Orthop 216:151–158, 1987.
13. Emerson Jr RH, Higgins LL, Head WC: The AGC total knee prosthesis at average 11 years. J Arthroplasty 15:418–426, 2000.
14. Ewald FC: The Knee Society Total Knee Arthroplasty roentgenographic evaluation and scoring system. Clin Orthop 24:9–12, 1989.
15. Fehring TK, Odum S, Griffin WL, Mason JB, Nadaud M: Early failures in total knee arthroplasty. Clin Orthop 392:315–318, 2001.
16. Fehring TK, Valadie AL: Knee instability after total knee arthroplasty. Clin Orthop 299:157–162, 1994.
17. Feng EL, Stulberg SD, Wixson RL: Progressive subluxation and polyethylene wear in total knee replacements with flat articular surfaces. Clin Orthop 299:60–71, 1994.
18. Friedman RJ, Hirst P, Poss R, Kelley K, Sledge CB: Results of revision total knee arthroplasty performed for aseptic loosening. Clin Orthop 255:235–241, 1990.
19. Furnes O, Espehaug B, Lie SA, et al: Early failures among 7,174 primary total knee replacements: A follow-up study from the Norwegian arthroplasty register 1994-2000. Acta Orthop Scand 73:117–129, 2002.
20. Garg A, Walker PS: The effect of the interface on the bone stresses beneath tibial components. J Biomech 19:957–967, 1986.
21. Green GV, Berend KR, Berend ME, Glisson RR, Vail TP: The effects of varus tibial alignment on proximal tibial surface strain in total knee arthroplasty: The posterior medial hotspot. J Arthroplasty 17:1033–1039, 2002.
22. Griffin FM, Scuderi GR, Insall JN, Colizza W: Total knee arthroplasty in patients who were obese with 10 years follow-up. Clin Orthop 356:28–33, 1998.
23. Hartford JM, Hunt T, Kaufer H: Low contact stress mobile bearing total knee arthroplasty: Results at 5 to 13 years. J Arthroplasty 16:977–983, 2001.
24. Heck DA, Clingman JK, Kettelkamp DG: Gross polyethylene failures in total knee arthroplasty. Orthopedics 15:23–28, 1992.
25. Hofmann AA, Heithoff SM, Camargo M: Cementless total knee arthroplasty in patients 50 years or younger. Clin Orthop 404:102–107, 2002.
26. Hsu H, Garg A, Walker PS, Spector M, Ewald FC: Effect of knee component alignment on tibial load distribution with clinical correlation. Clin Orthop 248:135–144, 1989.
27. Insall JN, Dorr LD, Scott RD, Scott WN: Rationale of The Knee Society clinical rating system. Clin Orthop 248:13–14, 1989.
28. Joshi AB, Gill GS, Smith PL: Outcome in patients lost to follow-up. J Arthroplasty 18:149–153, 2003.
29. Kalbfleisch JD, Orentice RL: The Statistical Analysis of Failure Time Data. New York, John Wiley and Sons Inc, 1980.
30. Keating EM, Meding JB, Faris PM, Ritter MA: Long-term follow-up of nonmodular total knee replacements. Clin Orthop 404:34–39, 2002.
31. Kim YH, Kook HK, Kim JS: Comparison of fixed-bearing and mobile-bearing total knee arthroplasties. Clin Orthop 392:101–115, 2001.
32. Knutson K, Lewold S, Robertsson O, Lidgren L: The Swedish knee arthroplasty register: A nation-wide study of 30,003 knees 1976-1992. Acta Orthop Scand 65:375–386, 1994.
33. Lewis JL, Askew MJ, Jaycox MS: A comparative evaluation of tibial component designs of total knee prostheses. J Bone Joint Surg 64A:129–135, 1982.
34. Lewis P, Rorabeck CH, Bourne RB, Devane P: Posteromedial tibial polyethylene failure in total knee replacements. Clin Orthop 299:11–17, 1994.
35. Liau J, Cheng C, Huang C, Lo W: The effect of malalignment on stresses in polyethylene component of total knee prostheses: Finite element analysis. Clinical Biomechanics 17:140–146, 2002.
36. Lotke PA, Ecker ML: Influence of positioning of prosthesis in total knee replacement. J Bone Joint Surg 59A:77–79, 1977.
37. Martin JW, Whiteside LA: The influence of joint line position on knee stability after condylar knee arthroplasty. Clin Orthop 259:146–156, 1990.
38. Matsuda S, Whiteside LA, White SE: The effect of varus tilt on contact stresses in total knee arthroplasty: A biomechanical study. Orthopaedics 22:303–307, 1999.
39. Mikulak SA, Mahoney OM, de la Rosa MA, Schmalzried TP: Loosening and oseteolysis with the press-fit condylar posterior-cruciate-substituting total knee replacement. J Bone Joint Surg 83A:398–403, 2001.
40. Mitts K, Muldoon MP, Gladden Jr M, Padgett DE: Instability after total knee arthroplasty Miller-Gallante II total knee: Five- to 7-year follow-up. J Arthroplasty 16:422–427, 2001.
41. Mont MA, Mathur SK, Krackow KA, Loewy JW, Hungerford DS: Cementless Total Knee Arthroplasty in Obese Patients. J Arthroplasty 11:153–156, 1996.
42. Paganelli JV, Skinner HB, Mote Jr CD: Prediction of Fatigue Failure of a Total Knee Replacement Tibial Plateau Using Finite Element Analysis. Orthopaedics 11:1161–1168, 1988.
43. Perillo-Marcone A, Barrett DS, Taylor M: The Importance of Tibial Alignment: Finite Element Analysis of Tibial Malalignment. J Arthroplasty 15:1020–1027, 2000.
44. Rakotomanana RL, Leyvraz PF, Heegaard JH, Rubin PJ: A Finite Element Model For Evaluation of Tibial Prosthesis-Bone Interface In Total Knee Replacement. J Biomech 25:1413–1424, 1992.
45. Rawlinson JJ, Bartel DL: Flat medial-lateral conformity in total knee replacements does not minimize contact stresses. J Biomech 35:27–34, 2002.
46. Ritter MA, Berend ME, Meding JB, et al: Long-term follow-up of Anatomic Graduated Components posterior cruciate-retaining total knee replacement. Clin Orthop 388:51–57, 2001.
47. Ritter MA, Faris PM, Keating EM, Meding JB: Postoperative Alignment of Total Knee Replacement: Its Effect on Survival. Clin Orthop 299:153–156, 1994.
48. Ritter MA, Herbst SA, Keating EM, Faris PM: Radiolucency at the bone-cement interface in total knee replacement : the effects of bone-surface preparation and cement technique. J Bone Joint Surg 76A:60–65, 1994.
49. Ritter MA, Worland R, Saliski J, et al: Flat-on-flat, nonconstrained, compression molded polyethylene total knee replacement. Clin Orthop 321:79–85, 1995.
50. Rorabeck CH: Managing instability following total knee replacement. Orthopedics 24:903–904, 2001.
51. SAS/STAT User Guide: The Lifetest Procedure, Cary, NC, SAS Institute Inc., 1990.
52. Sathasivam S, Walker PS: Computer Model to Predict Subsurface Damage in Tibial Inserts of Total Knees. J Orthop Res 16:564–571, 1998.
53. Sharkey PF, Hozack WJ, Rothman RH, Shastri S, Jacoby SM: Why are total knee arthroplasties failing today? Clin Orthop 404:7–13, 2002.
54. SPLUS 6.1, Guide to Statistics: Vol. 2, Classification and Regression Trees, Seattle, WA, Insightful Corporation, 2001.
55. Stern SH, Insall JN: Total Knee Arthroplasty in Obese Patients. J Bone Joint Surg 72A:1400–1403, 1990.
56. Szivek JA, Anderson PL, Benjamin JB: Average and Peak Contact Stress Distribution Evaluation of Total Knee Arthroplasties. J Arthroplasty 11:952–963, 1996.
57. Taylor M, Tanner KE, Freeman MAR: Finite element analysis of the implanted proximal tibia: a relationship between the initial cancellous bone stresses and implant migration. J Biomech 31:303–310, 1998.
58. Taylor M, Tanner KE: Fatigue failure of cancellous bone: a possible cause of implant migration and loosening. J Bone Joint Surg 79B:181–182, 1997.
59. Tew M, Waugh W: Tibiofemoral alignment and the results of knee replacement. J Bone Joint Surg 67B:551–556, 1985.
60. Therneau TM, Atkinson EJ: An Introduction to recursive partitioning using the rpart routine. Technical Report 61, Mayo Clinic, Section of Statistics, 1997.
61. Vazquez-Vela Johnson G, Worland RL, Keenan J: Norambuene: Patient Demographics as a Predictor of the Ten-year Survival Rate in Primary Total Knee Replacement. J Bone Joint Surg 85B:52–56, 2003.
62. Waldman SD, Bryant JT: Dynamic Contact Stress and Rolling Resistance Model for Total Knee Arthroplasties. J Biomech Eng 119:254–260, 1997.
63. Waslewski GL, Marson BM, Benjamin JB: Early, incapacitating instability of posterior cruciate ligament-retaining total knee arthroplasty. J Arthroplasty 13:763–767, 1998.
64. Windsor RE, Scuderi GR, Moran MC, Insall JN: Mechanisms of failure of the femoral and tibial components in total knee arthroplasty. Clin Orthop 248:15–19, 1989.
65. Worland RL, Johnson G, Alemparte J, et al: Ten to fourteen year survival and functional analysis of the AGC total knee replacement system. Knee 9:133–137, 2002, Erratum In: Knee 10:303, 2003.
66. Zhang H, Singer B: Recursive Partitioning in the Health Sciences. New York, Springer-Verlag, 1999.