Portal de Periódicos da CAPES

The John Charnley Award

2006; Lippincott Williams & Wilkins; Volume: 453; Linguagem: Inglês

10.1097/01.blo.0000238777.34939.82

1528-1132

Pat Campbell, Paul E. Beaulé, Edward Ebramzadeh, M. LeDuff, Koen De Smet, Zhen Lu, Harlan C. Amstutz,

Orthopedic Infections and Treatments

The reintroduction of surface arthroplasty of the hip with metal-on-metal bearings has the potential to eliminate or substantially reduce long-term wear-induced osteolysis as the major cause of failure. To determine important modes of failure, implant retrieval analysis was done on 98 failed surface arthroplasty components from different manufacturers. Analysis involved sectioning the components, measuring cement mantle thickness and the depth of penetration, histopathology, and measurement of the bearing wear. A finite element model was constructed to examine cement thermal necrosis. Femoral neck fracture and femoral loosening were the main causes of failure. The finite element model showed thermal necrosis could occur when cysts were filled with cement. Histologic observations verified necrosis of interfacial bone, although adaptive remodeling was commonly seen. The amount of cement varied considerably with implant type, and failure mode and was greater in loosened components. Although implant failure is multifactorial, these observations should be a cause for concern in current cementing techniques in which controlling mantle thickness and extent of penetration may be difficult. Optimizing cement technique to avoid leaving the component proud, and to avoid extensive cementation of the femoral head, may be important in reducing some modes of failure. With the application of metal-on-metal bearings, surface arthroplasty is again being performed in a growing number of centers worldwide. Although relatively few procedures have been performed in the United States, thousands of surface arthroplasty components have been implanted in Europe and Australia. We anticipate the problems faced by the first generation of metal-polyethylene surface arthroplasties, primarily related to debris-induced osteolysis caused by polyethylene wear,3,5,6,23 can be overcome by the current generation of low wearing metal-on-metal surface arthroplasties. Short-term clinical followup reports of metal-on-metal surface arthroplasties have been encouraging,4,15-17 although femoral neck fractures4,6,33,34 and femoral loosening2 have been identified as causes of failure. Risk factors in surface arthroplasty highlight the importance of patient selection criteria and good bone quality for implant survival.11 Currently, the role of femoral head vascularity in implant durability is controversial; some surgeons are concerned the posterior surgical approach sacrifices the important extraosseous blood supply to the femoral head,8,10,32 whereas others believe adequate blood supply will be provided intraosseously.21 Although the reduced wear of metal-on-metal bearings is well recognized, there have been concerns that heat-treating the components after casting can lead to higher wear, possibly sufficient to cause osteolysis.15 The unknown long-term consequences of metal wear debris are also a concern.26,28 Despite these concerns, the conservative nature of surface arthroplasty and the restoration of a high degree of function, including the ability to return to sports, make this surgery appealing to young, active patients. In our experience, patients are willing to travel long distances to specialty, high-volume centers, often at their own expense, for this surgical option. With the introduction of any new device there will often be a learning curve as surgeons gain experience and understand the limitations and the factors involved in clinical success and failure. This occurred with the introduction of cementless fixation and it will likely be no different for surface arthroplasty of the hip, which is recognized as a more technically demanding operation than standard total hip replacement. Before the orthopaedic community moves forward with the widespread use of surface arthroplasty, it is paramount we identify mechanisms of failure not fully understood in the previous metal-polyethylene surface arthroplasty experience. Despite many procedures performed over the past several years, failure analysis reports of revised metal-on-metal surface arthroplasty specimens are limited. We examined the failure mechanisms in metal-on-metal surface arthroplasty components submitted to our laboratories over the past 8 years. We compared the findings with failure modes identified in the metal-polyethylene surface arthroplasty era to examine the influence of removing the problem of wear-debris induced osteolysis as a primary mode of failure. The primary goal of these analyses was to understand which failures may be preventable through optimized patient selection and surgical techniques before the widespread reintroduction of surface arthroplasty. Second, we looked for new complications that were unique to this generation of metal-on-metal surface arthroplasty. MATERIALS AND METHODS One hundred forty specimens from four metal-on-metal surface arthroplasty designs were submitted to our implant retrieval laboratory from 1997 to 2005. Forty-two of the revised components had insufficient clinical information or lacked a femoral component that was suitable for inclusion in the analysis of failure mechanisms, leaving 98 for analysis. There were 50 male and 48 female patients with average ages of 49 years (range 33-66 years) and 51 years (range 24-78 years), respectively. The implants have been described in detail elsewhere.2,9,15,16,29 To assess failure modes of the various designs, the retrieved implants were studied by type and we summarized the design features of these components (Table 1). We examined 58 Conserve® Plus (Wright Medical Technology, Arlington, TN), 23 McMinn (Corin, Cirencester, United Kingdom), 13 Birmingham Hip Replacement (Smith & Nephew, Memphis, TN), and four Cormet 2000 (Corin) components. All femoral components were cemented, and acetabular components were noncemented, except for most of the McMinn components. The implants were inserted using a posterior approach. Most of the 40 patients from our institution had intraoperative photographs of the femoral head taken before and after surgical preparation for implantation.TABLE 1: Surface Arthroplasty FailuresAt revision, the femoral components were resected with a portion of the femoral neck where possible, and were immediately fixed in buffered formalin. In 49 revisions, the acetabular components also were removed. The components were inspected, photographed, and measured for wear depth and clearance using a coordinate measuring machine. Initially, femoral components were sectioned into a variable number of sections to allow inspection of the cement-bone interfaces and access to samples of the bone from various locations for decalcified histologic analysis. Later, a more systematic sectioning protocol was followed to facilitate cement interface analysis on a subgroup of cases (see below). For each of these 98 cases, a mode of failure was determined based on our previously applied criteria.5 Cement fixation and bone histology were quantitatively analyzed in a group of 45 failed metal-on-metal surface arthroplasty implants, including 24 that had failed because of femoral neck fracture or loosening and 21 that failed because of other causes. All 45 implants were sectioned by cutting a 2-mm thick coronal section from the middle of the metaphyseal stem and from the middle of each resulting portion to yield three 2-mm thick slices from the anterior, middle, and posterior portions of the femoral head. Because only specimens with an intact cement-bone interface were suitable for this analysis, implants that had loosened to the point of disassociation of the bone from the cement could not be included. The sections were radiographed and photographed. The following analyses were then performed. The thickness of the cement mantle (defined as the cement layer between the metal and the outer edge of the prepared surface of the bone) and the depth of cement penetration into cancellous bone (ie, the interdigitation of cement into cancellous bone starting from the outer edge of the prepared bone) were measured in a blinded manner in 11 BHR implants, 22 Conserve® Plus implants, 11 McMinns implants, and one Cormet implant. The cut section photographs and microradiographs were scanned into a computer, calibrated to within 0.1 mm, and an image analysis software program (MetaMorph® Version 4.6, Universal Imaging Corp, Downingtown, PA) was used to measure the thickness and area of the cement mantle and the depth and area of penetration of cement into the cancellous bone in 10 sites across the entire section. The data were averaged to give cement mantle thickness, penetration depth, total mantle and penetration areas, and the percentage of the femoral head cross-section within the component occupied by cement (combining the areas of the cement mantle, penetration and cement-filled fixation pegs or cysts) in each of the anterior, middle, and posterior sections. Specimens from all 98 femoral components were examined histologically and the general appearance of the bone and marrow was used to determine if the failure was related to osteonecrosis, fracture, cement interface loosening or infection using standard histopathologic criteria.30 The bone sections were carefully removed from the sectioned metal shell and divided into two or three smaller pieces that included the cement. The pieces were photographed to preserve orientation, then decalcified and embedded in paraffin for routine sectioning and staining with hematoxylin and eosin. Because this involved paraffin embedding, the bone cement was removed, but the bone within the cement and the interfacial soft tissues were preserved. The presence of interfacial membranes and bone necrosis was recorded for each of the 45 cases used for cement analysis. Semiquantitative histologic analysis of features at the proximal cement interface, the middle of the head and the component-femoral neck edges was carried out in a blinded fashion on 25 of these sections, using a modification of the scheme used by Howie et al.24 Bone viability, bone formation, marrow viability, and interface membrane formation were rated as none, low if the feature occupied less than 10% of the 4× field of view, moderate if it occupied 10% to 50%, and high if more than 50% of the field of view showed the feature. Bone viability was judged by the presence of osteocyte nuclei in most lacunae of the bone trabeculae. Image analysis software (Image One, West Chester, PA) was used to position a 100-point grid over the field of view. The proportion of each feature was calculated by recording which feature lay at the intersecting points of the grid. The thickness of interfacial membranes was measured in 10 places using a calibrated caliper function. The clinical and radiographic histories of each case were reviewed together with the results of the above analyses to determine the factors associated with their failures. Statistical analysis was done using SPSS version 11.5 analytic software (SPSS, Inc, Chicago, IL). Means and standard deviations of all continuous variables were calculated for each subcategory of failure mode, implant type, etc., and then were compared using student's t tests. When there were three or more categories, analysis of variance was used. For categorical variables, such as gender or implant type, chi square analysis was used to compare the ratio of failures in each subcategory. Logistic regression analysis then was used to assess the relative effects of each independent variable (when all variables were considered in the same analysis) on the risk of failure (loosening or fracture) compared with survival until revision. One hundred five components out of the 140 in our archived collection were measured for wear prior to the destructive sectioning procedure, which was not always performed in the same fashion. Consequently, only 57 components used in the failure analyses above were measured. Twenty-seven specimens were revisions of one side only, and 39 pairs were available for the analysis of clearance and wear. Revisions were performed from less than 1 month to 110 months. Wear depth was measured with a coordinate measuring machine (BMT 504, Mitotoyo, Aurora,IL) at 300 to 400 points over the surface of the implant. The original clearance between the ball and cup was calculated from this data. Linear regression analyses were performed to examine correlations between clinical and implant factors with wear depth. The cement thickness analysis led to the observation that, despite being designed to have a 1-mm or less space for a cement mantle within the femoral head, the thickness of this layer often was much greater. Cement penetration was variable, often extending well into the cancellous bone, and femoral head cysts were almost always filled with cement. A finite element model was constructed using MSC.PATRAN modeling software (MSC Software Corp, Santa Ana, CA) and analyzed using ABAQUS software (ABAQUS, Inc, Pawtucket, RI) to evaluate the temperature within the bone adjacent to the cement. The three-dimensional finite element model had a metal shell with a diameter of 46 mm and a maximum thickness of 7 mm at the apex. A bone cement layer 1.5-mm thick was chosen. To simplify the calculations, the cross-sections of the cancellous and cortical bone beyond the extent of the metal shell were modeled as circular. This basic model was meshed with eight-node heat transfer elements. The femoral shell was represented by 1734 elements, the cement layer by 1734 elements, the reamed cancellous bone by 5340 elements, and the cortical shell by 600 elements. The stem-cement and bone-cement interfaces were modeled as completely bonded. The thermal parameter analyses were performed on five configurations of the cement layer: (1) no cement penetration into the cancellous bone (ie, the basic model); (2) 1.5 mm penetration; (3) 6 mm penetration; (4) 6 mm penetration plus a cement-filled head cyst of 1 cm3; and (5) 6 mm penetration plus a cement-filled head cyst of 2 cm3. The depth of cement penetration and the presence of a cement-filled cyst were modeled by changing the properties of the corresponding elements of the basic model from those of cancellous bone to those of cement. Consequently, the overall mesh density and distribution were the same for the five configurations of the cement layers. The thermal properties of metal, bone cement, cancellous bone, and cortical bone were based on previous publications.25,27 The heat dissipation from the surfaces of the metal shell and cortical bone was modeled by applying convection with an ambient temperature of 23°C. The analysis combined transient heat transfer analysis of the finite element model and the calculations outside the finite element model for the heat generation during cement polymerization. The amount of heat generated in the cement layer was calculated using the formulas developed for Surgical Simplex cement by Baliga et al:7 where α is the degree of polymerization, S is the rate of heat generation per unit volume and Qtot is the total amount of heat per unit volume, which is 1.55 × 108 J/m3. The thermal analysis was performed in a series of time steps. The initial rate of the polymerization was taken as 10−5, as suggested by Baliga et al,7 which led to a heat flux of 35.4 W/m3 for all of the cement elements, and a duration of 43.8 seconds for the first step, using a user-complied program. These initial values were put into the finite element model to run the first step of the analysis. The results of the first step then were used to calculate the duration of the second step, magnitude of the heat flux, and accumulated rate of the polymerization for individual elements, again using the user-complied program. Lastly, the values were put into the finite element model to perform the second step of analysis. This cycle was repeated until the cement polymerization was complete. RESULTS The main reasons for failure were femoral neck fracture (28) and femoral loosening (23). Acetabular loosening accounted for 18 cemented socket failures of the first generation McMinn components and for failure in 10 porous-coated sockets (Table 2). Acetabular component malalignment causing mechanical problems such as impingement, edge loading, or subluxation resulted in the failure of four Birmingham Hip Replacement components, one Cormet 2000 component, and one Conserve® Plus component. Sepsis caused five failures. Eight failed arthroplasties originally attributed to "other causes" included one recurrent late dislocation, one case of effusion and joint swelling, and six revisions performed for unexplained pain. Retrieval analysis found femoral loosening and collapse (one case), osteonecrosis (two cases) and an extensive lymphocytic infiltration of the tissues suggestive of a metal sensitivity immune response (three cases), with some of these cases having more than one of these findings.TABLE 2: Clinical Details of the 98 RetrievalsFractures were the main cause of short-term failure (Fig 1). Trauma was reported in only a few cases, but most fractures occurred suddenly. Twenty-three occurred less than 6 months after surgery (median, 2 months). Longerterm femoral neck fractures occurred in five cases (median time to failure, 12.4 months) and were caused by extensive osteonecrosis of the femoral head, shown by the complete lack of viability of the bone and marrow. There was no repair to the original cut surfaces, suggesting the ischemia occurred at the time of surgery.Fig 1A: E. Illustrations show a 56-year-old woman with end-stage osteoarthritis of the right hip. (A) A preoperative anteroposterior radiograph shows excellent bone quality. (B) A postoperative anteroposterior radiograph shows the implants in good position. (C) Two months after receiving a Conserve® Plus implant, the patient had a fracture of the femoral neck. (D) This illustration shows a cut section through the retrieved femoral head. The component stem bent during the fracture event. (E) A light micrograph of an area of new bone at the component edge consistent with repair of a previous stress fracture or similar damage, through which the fracture occurred. This feature was seen in most of the short-term fractures. (Stain, hematoxylin and eosin; original magnification ×40).Femoral loosening occurred from 17 to 100 months after surgery. The degree of loosening and loss of underlying bone varied. Radiographically, femoral loosening was associated with a sclerotic and/or radiolucent line around the short stem that often appeared 2 or more years before failure (Fig 2A). Three loosening failures were associated with an area of ongoing bone repair in the middle of the femoral head (Fig 2B). Histologic features were similar to a nonunion. Five of the 21 failures caused by femoral loosening were associated with complete loss of fixation and femoral head shape because the proximal bone had been replaced by thick fibrous tissue that was often partly necrotic. In the nondissociated cases, loosening was associated with fibrous membrane formation between the cement and bone, ranging from 50 μm to 5 mm. The membranes were vascularized and contained variable numbers of macrophages, giant cells, and particulate bone cement. Bone adjacent to the membrane often was undergoing active osteoclastic erosion (Fig 3).Fig 2A: C. A 43-year-old woman had developmental dysplasia of the right hip. The femoral component loosened 39 months after the resurfacing procedure using a Conserve® Plus implant. (A) An anteroposterior radiograph taken shortly before revision surgery shows a wide radiolucency around the short metaphyseal stem of the femoral component. (B) The cut section of the retrieved femoral component is shown. (C) A corresponding microradiograph shows the component was left proud, and also shows the presence of a nonunion within the viable femoral head. There was ongoing bone repair at this site.Fig 3: A light micrograph shows a piece of bone undergoing osteoclastic resorption by the membrane that had formed between the cement and the bone in a Conserve® Plus component after 70 months. Scalloping of the bone is obvious on the top surface. There has been new woven bone formation around a central core of older, dead bone (healing osteonecrosis). (Stain, hematoxylin and eosin; original magnification ×100).Eighteen cases from a first generation metal-on-metal surface arthroplasty (McMinn design) were revised in association with socket-cement failure, including socket-cement dissociation from a bearing cemented in a large-diameter acetabular component of an older design. These generally were later term failures (5 years average, range 2 months to 131 months) and in most cases, the femoral components were radiographically stable. We present an example of a nonfailed femoral head revised after 10 years (Fig 4).Fig 4A: C. This 53-year-old woman had all-cemented McMinn metal-on-metal resurfacing for osteoarthritis of the right hip. The prosthesis failed on the acetabular side but the femoral component was well fixed at the time of conversion to total hip replacement. (A) A cut section of the retrieved femoral component and the corresponding microradiograph (B) shows the presence of large, cement debris-filled cysts that have eroded the bone. The dark staining of the tissue is a cutting artifact. (C) A light micrograph of the cement-bone interface illustrating the close proximity of the cement to the bone and marrow. The scalloped surfaces indicate previous osteoclastic remodeling. Parts of the bone lack nuclei but the bone is mostly viable in this long-term retrieval. (Stain, hematoxylin and eosin; original magnification ×40).Six revisions were performed for poor acetabular position that led to impingement, edge loading, and pain. The acetabular angles ranged from 56° to 85°. The total maximum wear depth was measured in four revisions and ranged from 14 μm after 10 months to 261 μm at 17 months. A focal stripe wear pattern found on the femoral component in three revisions was consistent with the subluxation and edge loading of the components (Fig 5).Fig 5A: C. A 54-year-old woman received a 46-mm Birmingham Hip Replacement component for osteoarthritis of the left hip. Postoperatively, the patient had an abnormal gait and general discomfort. Anterior-posterior (A) and lateral radiograph (B) show the acetabular component had been implanted at 85°. At revision surgery 13 months after implantation, the joint fluid was gray, and there was metal staining of the surrounding tissues. (C) This wear plot shows wear depth of the femoral component, which measured 100 μm (the maximum wear depth of the acetabular component was 37 μm; the diametral clearance was 202 μm).Cement penetration did not differ among failure modes. Cement mantle thickness and area and the depth and area of cement penetration were different in those components that failed because of femoral-related problems compared with nonfemoral failures (Fig 6). Specifically, the average cement mantle thickness in cases that failed from femoral loosening averaged 2.90 ± 1.8 mm, whereas failures attributable to neck fracture averaged 2.1 mm ± 0.9 (p = 0.03) and those with nonfemoral failures averaged 2.3 mm ± 1.1 (p < 0.04) (Table 3).Fig 6A: C. A box plot demonstrates the thickness of cement plotted by the mode of failure with p values. (A) A box plot shows cement mantle thickness in millimeters, (B) cement penetration depth in millimeters, and (C) amount of cement in the section (percentage), ie, mantle, penetration, and cement-filled fixation pegs and cysts.TABLE 3: Results of Cement AnalysesAverage cement mantle area and thickness were different among the implant types, with Conserve® Plus implants having a larger average cement mantle (2.7 mm ±1.1) and Birmingham Hip Replacements having the highest average depth of penetration (6.7 mm ± 3.9; p < 0. 001) (Fig 7). There was a tendency for cement penetration to be lower (R = 0.48, p < 0.003) when the cement mantle thickness was greater.Fig 7A: B. A box plot shows (A) cement mantle thickness in millimeters and (B) cement penetration depth in millimeters.The total percentage of the femoral head sections occupied by cement (mantle, cement-filled fixation pegs or cysts, and penetration combined) ranged from 11% to 89% and was more (p = 0.001) in loose failures compared with all other modes of failure. Cement-filled cysts were more prevalent (p < 0.025) in cases that failed by femoral loosening compared with nonfemoral failures. Bone was necrotic in areas penetrated by cement. There was a higher incidence of necrotic bone below the cement zone, especially in the short-term cases in which necrosis of bone and marrow extended up to several millimeters from the cement; however, the edges of the component, where cement tended to be minimal, had the least (p < 0.01) necrotic bone. Healing osteonecrosis, where new bone had formed around a necrotic trabecular core, was common near the cement and was associated with abundant blood vessels, but was sparse in most of the other regions examined. Fibrous membranes were uncommon around deeply penetrated cement; however, this interface was more likely to be characterized by fibrotic marrow and bone trabeculae having healing osteonecrosis (Fig 8). Bone resorption usually was associated with areas of woven bone formation or with nearby membranes containing particulate bone cement. Active bone resorption was commonly seen in femoral loosening, but also was common in the unfailed femoral heads from cemented socket failures because the cement debris-filled granulomas invaded the bone through the cement-implant interface. Osteolysis and large defects containing cement debris-filled macrophages were seen within the femoral heads of four of these acetabular failures.Fig 8: A light micrograph shows histologic features of the cement-bone interface in the middle of the femoral head in a Birmingham Hip Replacement implant with deeply penetrated cement after 1 month. The marrow has been replaced by granulation tissue and several blood-filled vessels nearby indicate this area is undergoing repair. The paler stained bone is dead, but is surrounded by a seam of newly formed bone (healing osteonecrosis) (Stain, hematoxylin and eosin; original magnification ×40).We summarized wear depth and clearance measurements (Table 4). The wear depth varied from undetectable (< 2 μm) to 164 μm. Diametral clearances varied from 123 to 400 μm. Components that failed because of femoral fracture or loosening had similar wear to those that failed because of acetabular loosening, malposition, infection or other reasons, but the study may have been underpowered to address this point.TABLE 4: Results of Wear MeasurementsThe Birmingham Hip Replacements, which were revised mostly for acetabular malposition, had higher average wear and clearance than the other types (p < 0.001). The effect of time in vivo on wear was small for the total group, which ranged from less than one month to more than 10 years, but when examined for cases retrieved at less than 2 years in vivo, wear in the Conserve® Plus, McMinn, and Cormet types was higher (p < 0.05) than those components in vivo longer than 2 years. When the modeled bone cement mantle was 1.5 mm thick and there was no cement penetration into the femoral head (or only 1.5 mm of penetration), the temperatures throughout the bone and cement remained below body temperature (Fig 9). In contrast, with 6 mm penetration or a cement-filled cyst of 1 cm3 or 2 cm3, the peak temperatures within the bone were 55°C, 61°C, and 66°C, respectively. The temperature in the bone remained above 50°C for 37, 64, and 117 seconds respectively with the above cement configurations. The temperature in the bone was above 50°C to a depth 0.5 mm below the cement-bone interface with 6 mm of penetration, 1.25 mm with 6 mm penetration plus a cement-filled cyst of 1 cm3, and 2 mm with 6 mm penetration plus a cement-filled cyst of 2 cm3.Fig 9: Results of the finite element model of temperatures reached during cement polymerization within a cemented femoral surface replacement. Three conditions were modeled, from left to right: 1.5 mm penetration, 6 mm penetration; and 6 mm penetration and a cement-filled cyst of 2 cm3. The red portions indicate a temperature high enough to damage surrounding bone.DISCUSSION The introduction of metal-on-metal bearings in surface arthroplasty has nearly eliminated aseptic loosening caused by particulate bearing wear debris. Consequently, achieving lasting fixation and avoiding mechanical failures are the primary requirements for a successful and durable surface arthroplasty. Although cement is used for femoral fixation in most of the resurfacing designs currently in use, the method and timing of cement application and the amount of the cement mantle and bone penetration are variable. This study showed wide variation in cement in failed and nonfailed femoral heads using thickness and area measurements. The practice of filling cysts with cement, the problem of unseated components, and the pressurization of low viscosity cement into bone of variable quality can result in large amounts of cement in the resurfaced femoral head. In some cases in this retrieval group, almost complete filling of the bone had occurred. We found the amount of cement was greater in loosened femoral components. This should be a cause for concern in light of the current cementing techniques where the control of the cement mantle thickness and extent of penetration may be difficult to achieve, as well as the common practice of cementing cystic lesions. Trabecular bone structure is optimized to withstand the dynamic stresses applied to the natural femoral head, but the depth of cement penetration that could affect this subtle balance, such as by preventing bone from remodeling according to new stresses when an implant is present, is unknown. Additional studies are required to determine how cement-filled bone behaves under changing stresses or conditions. One of the concerns of excessive cement use is thermal necrosis of the surrounding bone.7,18,19,31 The results of the present FEA indicated when deep cement penetration was combined with a relatively small cement-filled cyst, the peak temperatures within the femoral head interfacial bone were high enough, and the durations were long enough to cause thermal bone damage. The model underestimated the amount of cement seen in many of the retrievals. Because it did not account for practices such as extensive cold saline irrigation of the prepared bone surfaces and the seated component, only real-time temperature probe measurements of the cement bone interface will confirm if the modeled temperature changes occur in vivo. Bone and marrow necrosis were observed up to several millimeters deep to cement-filled cancellous bone or cysts in short-term failures, but did not seem to cause failure. New bone formation within close proximity to the cement also was noted in some cases. The bone-cement interfaces in several of the longer-term specimens with deep cement showed evidence this damage healed without membrane formation. This may reflect the ability of the remaining bone to withstand some degree of insult, but this ability may be overwhelmed when several accumulative insults occur. Healing can only occur with the preservation or restoration of an adequate femoral head blood supply. In this series, there were seven osteonecrosis-induced failures. Ongoing bone resorption, interfacial membrane formation and cement-interface instability were noted in few cases with deep cement penetration. In contrast, membranes between bone and cement were seen where there was little or no cement penetration, and there often was active bone resorption, presumably from the bone-cement-related histiocytic tissue. We observed thick interfacial membranes, bone resorption, and transformation of bone into fibrous tissue in several loose femoral metal-on-metal surface arthroplasty components. Similar features were reported in loosening failures of metal-polyethylene surface arthroplasty. Howie et al23 proposed aseptic loosening of cemented metal-polyethylene surface arthroplasties started with the accumulation of polyethylene particles at the cement-bone interface, leading to bone resorption, membrane formation, and eventual fibrous transformation of the femoral head bone as loosening progressed. In the absence of polyethylene debris, the most likely factors initiating interfacial membrane formation are mechanical instability and thermal necrosis. Mjoberg31 postulated migration of resurfacing components measured with roentgen stereophotogrammetry resulted from instability because of the fibrous tissue layer that replaced thermally damaged bone under the polymerizing bone cement. Femoral neck fracture after hip resurfacing is a risk for the patient, and the mechanisms precipitating this failure mode include patient-related, surgery-related, and biomechanical factors.6,11,19,33 The technical difficulty associated with the surface arthroplasty also may be a factor in neck fracture. Several fractures in this retrieval series occurred with surgeons new to the procedure. Cut sections and cement mantle thickness measurements showed the implants were sometimes left not fully seated. This puts the femoral neck at risk in several ways. The surgeon may apply extra pressure or hammer blows to seat the proud component, leading to stress fractures if there is weak bone or component misalignment. Reamed bone may remain uncovered and could act as a stress riser, and the thick mantle may reduce bone cement penetration for fixation, as suggested by our analysis. Although healing of stress fractures occurring under metal-on-metal surface arthroplasty components has been reported,14 bone undergoing repair after stress fracture is a site of weakness, and our histologic analysis showed that fractures occurred in these areas. One of the major improvements in this era of metal-on-metal bearings is the reduction in wear provided by metal-on-metal. The wear of these failed components generally was low, with the exception of poorly functioning implants, particularly when acetabular malpositioning resulted in steep cup angles (> 55° of abduction).12 This may be a unique complication of metal-on-metal surface arthroplasty, but it should be preventable with education and careful attention to surgical technique and implant placement. Ten of the components submitted to our laboratories were revised for unexplained pain and implant retrieval analysis was helpful in finding explanations in most of those cases. Metal sensitivity was the suspected cause of pain in three revisions. This has been a concern since cobalt chromium implants were introduced.20 Recent reports have shown some individuals have an immunological reaction to wear products made of this alloy.1,35 Although this is a rare complication, efforts should be continued to devise a means of preimplantation testing.22 In 1982, in a paper presented at a symposium on surface arthroplasty, Clarke noted "the main flaws to be overcome in realizing the potential success of the double-cup arthroplasty procedure are failures due to femoral cup loosening, acetabular cup loosening, and femoral-neck fractures. The clinical uncertainties include the selection of a suitable patient with adequate bone stock and the technical difficulties associated with (1) reaming the acetabulum adequately, (2) reaming down onto the neck without violating it, and (3) anchoring the components securely by interdigitation of acrylic cement."13 Twenty-four years later, with resurfacing arthroplasty being used more widely than ever, we face some of the same questions and issues, but progress has been made. The combination of careful and thorough clinical followup, laboratory studies including wear simulation, implant design modeling, and retrieval studies will provide useful information in the understanding of the success and failure of the new generation of surface arthroplasties. Acknowledgments The authors thank Dr. Joseph Mirra, Mr. Bill McGarry, Dr. Fred Dorey, Mr. Billy Lundergan, Dr. Brook Wager, Dr. Roel De Haan, Dr. Harry McKellop, Ms. Cathy Fischl and the technicians at Impath, and all of the surgeons who contributed specimens for analysis.