Debonding of the femoral component is the term used to describe the loss of the bond between the metal femoral component and the acrylic cement in a total hip replacement. The first radiographic manifestation of debonding usually is slight subsidence of the femoral component within the cement mantle, which is seen (with most implant designs) as a radiolucent line between the superolateral portion of the prosthesis and the acrylic cement. The importance of this finding has been debated7,10,16-19,22,23,26. Some investigators have thought that debonding initiates the process of aseptic loosening of the femoral component11,12, and they have defined radiographic evidence of debonding as one criterion of femoral loosening2,9,11. Other investigators have suggested that a femoral component with a smooth surface often subsides slightly to a stable position within the cement mantle, and they have not thought that femoral debonding has a negative prognostic importance8,16. This debate has influenced the design of modern prostheses. Some femoral components are given surface treatments to maximize their initial bond with the cement, whereas others are polished and designed to allow debonding and limited subsidence within the cement mantle.
The purpose of the present study was to determine the long-term pattern of debonding, and its impact on survival, of the femoral component of a Charnley total hip arthroplasty.
*No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article. No funds were received in support of this study.
†Department of Orthopedic Surgery, Mayo Clinic, 200 First Street S.W., Rochester, Minnesota 55905.
Three hundred and thirty-three consecutive Charnley total hip arthroplasties were performed at our institution, during 1969 and 1970, in 300 patients; all were contacted for reevaluation at one, two, five, ten, fifteen, and twenty years after the operation or were followed until the time of death or revision. No patient was lost to follow-up. Thus, we were able to analyze a consecutive group of patients (who have been the subject of previous reports1,14,15,25) for whom the long-term clinical fate of the femoral component was known.
A Charnley total hip prosthesis (Thackray, Leeds, England), consisting of an all-polyethylene acetabular component and a stainless-steel smooth flatback femoral stem, was inserted in all patients with use of radiodense methylmethacrylate cement, which was hand-packed into the canal. Pulsatile intramedullary lavage, an intramedullary canal plug, or measures to reduce the porosity of the cement were not used in any patient.
All 333 total hip arthroplasties were analyzed retrospectively. Radiographs made immediately postoperatively, at one to five years after the operation, and at the latest follow-up examination were evaluated for the presence or absence of debonding, which was identified as a radiolucent line between the superolateral portion of the prosthesis and the cement mantle in zone 1 of Gruen et al. on an anteroposterior radiograph of the hip. If a radiolucent line was present, the maximum width was measured and categorized as class I (less than 0.5 millimeter), class II (0.5 to 1.9 millimeters), or class III (2.0 millimeters or more). Poor-quality radiographs or those that did not provide a true anteroposterior view of the proximal part of the femur were excluded. The radiographic analysis was performed by one of us (D. J. B.), who was blinded to the outcome of the operative procedure.
Satisfactory radiographs had been made between one and five years after the operation for 297 of the 333 hips; these 297 hips were included in the study. The remaining thirty-six hips were excluded because either no radiographs or unsatisfactory radiographs had been made or the patient had died less than one to five years after the operation. One hundred and forty-three of the hips were in men and 154 were in women. The mean age at the time of the index total hip arthroplasty was sixty-four years (range, thirty-eight to eighty-five years). Satisfactory radiographs had been made more than five years postoperatively for 220 of the 297 hips. The other seventy-seven hips were not assessed radiographically after five years because the patient had died, radiographs had not been made, or radiographs had been made but were of a quality that precluded accurate analysis.
A radiolucent line that was less than 0.5 millimeter wide was seen between the lateral shoulder of the prosthesis and the cement on the immediate postoperative radiograph of fifty-nine hips. The importance of this radiographic finding is not known, but it is likely that the cement in these hips was never completely apposed to the lateral shoulder of the prosthesis. This group of hips was considered separately to preclude analysis of a mixture of hips that had a radiolucent line immediately after the operation and those in which a radiolucent line subsequently developed. Thus, with the exclusion of these fifty-nine hips, we analyzed 238 hips for which one-to-five-year radiographs were available.
The clinical fate of all of the femoral components in this group of patients was known. Kaplan-Meier survivorship analysis13 was performed, with revision of the femoral component because of aseptic loosening or mechanical failure (defined as revision because of aseptic loosening, unequivocal radiographic loosening manifested by a fracture of the cement mantle, or a complete radiolucent line at the bone-cement interface) as the end points. Patients were excluded if the femoral component had been removed because of infection or fracture of the component (the other two reasons for revision of a femoral component in this group of patients). The starting point for the Kaplan-Meier survivorship analysis was the time of the radiographic evaluation performed between one and five years after the operation. This time was chosen to allow a determination of the effects of debonding (once it was radiographically visible) on the subsequent performance of the implant. Differences in survival were compared with use of the log-rank test. A level of p = 0.05 was assumed to be significant. The association between pain and each of three radiographic signs of prosthetic loosening (debonding, a complete radiolucent line at the bone-cement interface, and fracture of the cement mantle) was assessed with use of the chi-square test or the Fisher exact test.
The radiographs made between one and five years after the Charnley total hip arthroplasty showed that a radiolucent line suggesting debonding had developed, since the time of the immediate postoperative radiograph, in 33 per cent (seventy-nine) of the 238 hips. No evidence of debonding was seen in 67 per cent (159) of the hips.
The debonding was class I in 21 per cent (fifty-one) of the hips, class II in 7 per cent (seventeen), and class III in 5 per cent (eleven). There was no significant difference among the four debonding groups (no debonding and classes I, II, and III) with respect to the age (p = 0.44) or the gender (p = 0.09) of the patients. When all hips that had no debonding were compared with all those that had debonding on the one-to-five-year radiograph, we found no significant difference with respect to age (p = 0.61). However, there was a significant difference with respect to gender: the group that had debonding included relatively more men than did the group that had no debonding (p = 0.03).
As mentioned, satisfactory radiographs had been made more than five years postoperatively for 220 hips. From the time of the one-to-five-year radiograph to that of the latest radiograph, the amount of debonding remained unchanged in 62 per cent (137 hips), increased by one class in 30 per cent (sixty-five), increased by two classes in 6 per cent (thirteen), and increased by three classes in 1 per cent (three). In two hips (1 per cent), the amount of debonding apparently had decreased by one class—probably because femoral rotation had changed slightly between the times that the radiographs were made. For the purposes of the analysis, the class of the debonding in these two hips was considered to be unchanged. Evidence of debonding was seen on the latest follow-up radiographs of 39 per cent (forty-seven) of the 120 hips that had had no debonding on the one-to-five-year radiograph.
Kaplan-Meier survivorship analysis demonstrated that, with revision because of aseptic loosening as the end point, the probability of survival of the femoral component (with 95 per cent confidence intervals) at fifteen years after the time of the one-to-five-year radiograph was 94 per cent (89 to 99 per cent) for the hips that had no debonding, 100 per cent (100 per cent) for those that had class-I debonding, and 94 per cent (76 to 100 per cent) for those that had class-II debonding (Fig. 1). The probability of survival of the femoral components that had class-III debonding was markedly worse (43 per cent [14 to 95 per cent]) (Fig. 1). When the hips that had no debonding were combined with those that had a class-I or class-II appearance—that is, when all of the hips in which the radiolucent line was absent or less than 2.0 millimeters wide on the one-to-five-year radiograph were analyzed as a group—the probability of survival at fifteen years was 95 per cent (92 to 99 per cent). This value was significantly higher (p < 0.0001) than the probability of survival of the femoral components associated with a radiolucent line that was at least 2.0 millimeters wide (class-III debonding) (Figs. 2, 3-A, 3-B, 4-A and 4-B).
Kaplan-Meier survivorship analysis also was performed to compare the hips that had no evidence of debonding with those that had evidence of debonding (class I, II, or III) on the one-to-five-year radiograph. The probability of survival at fifteen years (with 95 per cent confidence intervals) was 94 per cent (89 to 99 per cent) for the former group and 90 per cent (81 to 98 per cent) for the latter (p = 0.15) (Fig. 5). With the sample sizes, the power of detecting a difference in proportions of 0.11 was at least 80 per cent (two-sided test at alpha = 0.05).
The probability of survival of the femoral component also was calculated with mechanical failure (revision because of aseptic loosening, a visible fracture of the cement mantle, or a complete radiolucent line at the femoral bone-cement interface) as the end point. The probability of survival at fifteen years was 91 per cent for the hips that had no debonding, 90 per cent for those that had class-I debonding, and 93 per cent for those that had class-II debonding. Because so few femoral components with a class-III appearance had survived to this end point, it was not possible to calculate the probability of survival of those components at fifteen years. However, the probability of survival at eight years was only 50 per cent, which was significantly poorer than the probability in the other groups (p < 0.0001). With use of the same end point, the probability of survival at fifteen years for the hips that had no evidence of debonding on the one-to-five-year radiograph was 91 per cent (85 to 97 per cent) compared with 87 per cent (77 to 98 per cent) for those that had evidence of debonding (class I, II, or III) (p = 0.08). Thus, as was found with revision because of aseptic loosening as the end point, the probability of survival with mechanical failure as the end point was slightly poorer for the hips that had evidence of debonding, but the difference was not found to be significant with the numbers available. With these sample sizes, the power of detecting a difference in proportions of 0.13 was at least 80 per cent (two-sided test at alpha = 0.05).
When the fifty-nine hips that had evidence of class-I debonding on the immediate postoperative radiograph were considered separately (to preclude analysis of a mixture of hips that had a radiolucent line immediately after the operation and those in which a radiolucent line developed subsequently), the survivorship curves were indistinguishable from those of the other 238 hips. Also, the Kaplan-Meier survivorship analyses were performed with and without this group of fifty-nine hips, and the results were indistinguishable. (The Kaplan-Meier survivorship curves in Figures 1, 2, and 5 do not include the fifty-nine hips.)
Statistical analysis was performed to determine whether the radiographic finding of debonding was associated with pain. Hips that had radiographic signs other than debonding, such as a continuous radiolucent line at the bone-cement interface or evidence of a fracture of the cement mantle, were excluded from this analysis. With the number of patients in the present study, no association was identified between pain and debonding at the one-to-five-year evaluation (p = 0.310) or at the latest clinical and radiographic evaluation (p = 0.579). In contrast, a positive association was identified between moderate or severe pain and the presence of a complete radiolucent line at the bone-cement interface of the femoral component (p < 0.001 at the one-to-five-year radiographic evaluation and p < 0.001 at the latest follow-up evaluation) and between moderate or severe pain and a fracture of the femoral cement mantle (p < 0.001 at the one-to-five-year radiographic evaluation and p = 0.05 at the latest follow-up evaluation).
The importance of a radiolucent line at the superolateral portion of a femoral component that was inserted with cement is controversial yet consequential. Substantially different rates of loosening have been reported in different studies of the same type of femoral implant depending on whether or not this finding was considered to be a criterion of loosening15,24. Thus, an analysis of the relationship between this radiographic finding and the clinical performance of the femoral component is of value. Furthermore, most investigators have believed that this radiographic finding is evidence that the femoral component has debonded from the surrounding cement mantle, and the impact of such debonding on subsequent prosthetic performance has important implications for prosthetic design. Currently, some femoral components are designed with surface treatments and a shape to maximize the prosthesis-cement bond and to minimize the likelihood of debonding, whereas others are polished and designed to allow debonding and limited subsidence of the prosthesis. The goal of the present study was to examine the effect of the early appearance of a radiolucent line at the interface of the superolateral portion of the prosthesis and the cement on the subsequent survival of a smooth Charnley femoral component.
In our study, debonding alone did not have a significant effect on the long-term survival of the femoral component when the radiolucent line at the superolateral border of the prosthesis was less than two millimeters wide during the first one to five years after the operation. The probability of survival of the components with early debonding was slightly poorer than that of those without early debonding, although the difference was not found to be significant with the numbers available. The difference was primarily the result of the poor probability of survival of the small group of components that had class-III debonding (a radiolucent line with a maximum thickness of 2.0 millimeters or more) on early radiographs. The data from the present study and from others4,6,24,27 suggest that radiographic evidence of debonding develops around many prostheses of the smooth Charnley design during the first twenty years that the implant is in situ. Most hips that had radiographic evidence of debonding within the first five years after the operation had little or no additional prosthetic subsidence during many subsequent years. However, it is important to note that, in a small group of these hips, the radiographic appearance of debonding of the femoral component was the first manifestation of loosening and of ultimate clinical failure.
When a radiolucent line that was 2.0 millimeters or more in width was evident between the superolateral shoulder of the prosthesis and the cement within five years after the arthroplasty, the likelihood that the femoral component would ultimately fail and need to be revised was increased substantially. Marked early subsidence of the smooth Charnley femoral component, therefore, can be considered to have a strong negative impact on the likelihood of a satisfactory long-term result. We believe that these data support the idea that debonding of a smooth Charnley femoral component followed by minimum subsidence to a relatively stable position within the cement mantle occurs frequently and is compatible with good clinical function. Debonding followed by pronounced subsidence of the prosthesis in the cement mantle is evidence that the cement is unlikely to provide satisfactory support for the implant, and the risk of revision because of symptomatic loosening substantially increases.
Debonding was not associated with pain in the hip unless there were other radiographic signs of prosthetic loosening, such as a complete radiolucent line at the bone-cement interface or a fracture of the cement. This finding is consistent with that reported in other studies4,6,16,19,22-24 and suggests that, although radiographic evidence of debonding of the femoral component occurs in association with certain prosthetic designs, the prostheses function well and cause no pain. Our inability to detect an association between debonding and pain or to find a strong association between debonding and subsequent failure necessitating revision suggests that debonding alone with subsidence of less than 2.0 millimeters should not be considered to be analogous to loosening of a smooth Charnley femoral component.
We recognize that a radiolucent line at the superolateral portion of the femoral component, which we termed debonding, may in actuality have several different etiologies. In most instances, we believe that the radiolucent line represents a space between the prosthesis and the cement caused by a slight subsidence of the prosthesis within the cement mantle. Without stereophotogrammetric methods5, it is not possible to measure subsidence of the prosthesis accurately enough to prove that the amount of subsidence is associated directly with the width of the radiolucent line. It is also possible that, in some instances, the formation of the radiolucent line was due to plastic deformation of the femoral component as those components were made with a ductile stainless-steel alloy. However, no hip in the present study had visible radiographic evidence of plastic deformation of the prosthesis. In any case, if such a deformation had occurred, it also would have been associated with debonding when a radiolucent line was present between the prosthesis and the cement. Finally, a slight radiolucent line (class-I debonding) may, in some hips, represent a radiographic artifact attributable to the so-called Mach effect. This thin artifactual line can occur when a radiograph is made of a radiodense object physically juxtaposed to an object that is less radiodense. If the Mach effect had occurred, interpretation of this radiographic finding as debonding would have led to an overestimation of the number of hips that had class-I debonding and an underestimation of the number that had no debonding. However, it would not have dramatically altered the conclusions of the study, as the survivorship curves for the hips that had no debonding and those that had class-I debonding were similar. Differences in the rotation of the lower limb and the angle of the x-ray beam with the prosthesis-cement interface theoretically can affect the measured size of the radiolucent line and thus the classification of the debonding. For this reason, only good-quality anteroposterior radiographs were accepted for the present study. If small rotational changes had led to notable inaccuracies in measurement, the measured size of the radiolucent line on later radiographs would have decreased with time in many hips because of random inaccuracies. In fact, this phenomenon was observed in only two hips (with a measured change of only one debonding class in both), suggesting that, with use of standardized positioning of the patient and radiographic technique, the radiolucent line can be consistently measured with the accuracy required to classify the debonding of the femoral component.
No fracture of the cement mantle was evident radiographically in most hips that had debonding. This finding possibly can be explained by the creep of the acrylic cement under physiological conditions. Alternatively, in some instances, fractures of the cement may have been present but were not visible on radiographs because of their small size or their obliquity with respect to the x-ray beam.
The present study involved Charnley total hip prostheses with a smooth femoral component, a tapered trapezoidal stem, and a small collar. Bundy and Penn demonstrated that a prosthesis-cement bond formed—possibly because of molecular interactions—even when an implant was smooth, as was the case in our study. Such a prosthesis may provide conditions that frequently allow debonding and limited subsidence to a stable position without major adverse consequences. The results of the present study should not be extrapolated to prostheses of very different designs as it appears that different designs behave differently when debonding occurs20. In a study involving femoral components with a large collar, a roughened surface finish, and a rounded cross section, Mohler et al. reported rapid clinical failure due to pain in the hip and osteolysis that developed after debonding occurred. In contrast, the present study and others4,6,16,19,22-24 have demonstrated that some stems with a smooth surface finish, tapered shape, angular cross section, and small or absent collar often tolerate debonding for years without causing clinical problems. This finding suggests that the clinical effect of debonding of the femoral component is specific to the prosthetic design and is probably affected by the geometry and the surface finish of the component as well as by the characteristics of the surrounding cement mantle.