Background: The role of radiographs in the follow-up of patients with metal-on-metal hip resurfacing (MoMHR) implants is unclear. We investigated whether a combination of patient and radiographic factors predicted MoMHRs with evidence of a pseudotumor.
Methods: We performed a retrospective single-center case-control study of 384 MoMHRs. The pseudotumor group of 130 hips all had evidence of a symptomatic pseudotumor on cross-sectional imaging, with the diagnosis confirmed at revision. The nonpseudotumor group of 254 hips (a subgroup of these hips were previously reported on) all had normal findings on cross-sectional imaging. Radiographs taken immediately prior to revision were assessed in the pseudotumor group and were compared with radiographs taken at the time of normal cross-sectional imaging in the nonpseudotumor group. Two blinded independent observers analyzed the radiographs for signs of failure, with excellent interobserver agreement. Logistic regression modeling identified the patient and radiographic predictors of revision for pseudotumor.
Results: Hips with a pseudotumor more commonly had abnormal findings on radiographs compared to hips without a pseudotumor (80.0% compared with 63.4%; p = 0.001). Patient and radiographic factors predictive of revision for pseudotumor in the multivariable model were female sex (odds ratio [OR], 3.14; 95% confidence interval [CI], 1.85 to 5.35; p < 0.001), high inclination (OR, 1.04 per degree; 95% CI, 1.01 to 1.07 per degree; p = 0.006), acetabular osteolysis (OR, 5.06; 95% CI, 2.14 to 12.0; p < 0.001), femoral osteolysis (OR, 17.8; 95% CI, 5.09 to 62.2; p < 0.001), and acetabular loosening (OR, 3.35; 95% CI, 1.34 to 8.35; p = 0.009). Factors predictive of not having a pseudotumor were anteversion of ≥5° (5° to <10°: OR, 0.31; 95% CI, 0.12 to 0.77; p = 0.012; and ≥10°: OR, 0.32; 95% CI, 0.15 to 0.70; p = 0.004) and heterotopic ossification (OR, 0.19; 95% CI, 0.05 to 0.72; p = 0.015). The final multivariable model was well calibrated (p = 0.589), with good discriminatory ability (area under the curve = 0.801; sensitivity = 74.4%; specificity = 71.7%).
Conclusions: A combination of patient and radiographic factors provided useful information for distinguishing between MoMHRs with and without evidence of a pseudotumor. Surgeons may wish to consider these specific patient and radiographic factors before proceeding with cross-sectional imaging. Radiographs are important when assessing patients with MoMHR implants and should be included in all follow-up protocols.
Level of Evidence: Diagnostic Level III. See Instructions for Authors for a complete description of levels of evidence.
High short-term failure rates are reported for most metal-on-metal hip resurfacing (MoMHR) designs1-3, with many revisions performed following pseudotumor formation4,5. In an attempt to identify pseudotumors early, worldwide medical regulatory authorities recommend regular surveillance for most patients with MoMHR implants6-8.
Radiographs are considered important in the assessment of patients managed with MoMHR, as they provide information on component position, bone quality, and implant fixation9. Radiographs can also identify signs suggestive of implant failure early. Furthermore, in addition to blood metal ion analysis and cross-sectional imaging, radiographs are currently recommended by most7,8 but not all6 medical regulatory authorities. However, given that pseudotumors can be solid or cystic lesions associated with soft-tissue damage and high wear4,5, most clinicians prefer to measure blood metal ion levels and perform cross-sectional imaging rather than use radiographs6-8. Previous studies have reported radiographic risk factors for pseudotumors that include the implantation of acetabular components outside an optimal zone10-12 and substantial reduction in the head-neck ratio following MoMHR13. These studies are limited by their small numbers of revisions for pseudotumor and their assessment of relatively few radiographic parameters10-13. Interpretation of such studies is also complicated by observations that pseudotumors can still occur in optimally positioned MoMHRs11,14. Moreover, it remains unclear whether femoral neck narrowing is a normal physiological process following MoMHR or a clinically important finding, given that neck narrowing is reported both in patients with a well-functioning MoMHR15,16 and in patients who have undergone revision for pseudotumor9,17.
The need for radiographs in the follow-up of MoMHR patients therefore remains unclear. It is important to establish the role of radiographs in MoMHR surveillance given their wide availability and low cost as well as the fact that current follow-up recommendations are not evidence-based but are costly18.
We investigated whether a combination of patient and radiographic factors predicted MoMHRs with evidence of a pseudotumor. By using the factors identified, a clinical risk scoring tool was developed to predict a patient’s risk of having a pseudotumor.
Materials and Methods
We performed a retrospective single-center multi-surgeon case-control study including 384 MoMHRs implanted in 329 patients (Table I). These hips were divided into pseudotumor (case) and nonpseudotumor (control) groups. All primary MoMHRs were performed between June 1999 and December 2009. During this period, 1,429 MoMHRs in 1,216 patients were implanted at this center, with the outcomes for these patients previously described in detail19.
Pseudotumor Group (130 Hips)
Revision surgery of MoMHRs for pseudotumor has been performed since 2007, when this entity was first recognized4. By August 2015, 231 consecutive MoMHR revisions for any indication were recorded in our prospective clinical database. The pseudotumor group for the present study included all MoMHRs revised for pseudotumor (n = 130; 56% of all revisions). All patients undergoing revision for pseudotumor were symptomatic. Of the 130 hips revised for pseudotumor, 111 (85%) had received the primary MoMHR at our institution and the remainder were referred to our center after undergoing primary resurfacing arthroplasty elsewhere. Prior to revision surgery, patients were evaluated with use of anteroposterior pelvic radiographs, blood metal ion analysis, and ultrasound, with metal artifact reduction sequence magnetic resonance imaging (MARS-MRI) reserved for equivocal or complex cases4,20. The decision to perform revision was made by the patient’s surgeon based on the symptoms and investigative findings.
All pseudotumors were diagnosed on cross-sectional imaging prior to revision surgery and subsequently confirmed intraoperatively. Pseudotumors were defined as cystic, solid, or mixed masses communicating with the hip joint4,21,22. The diagnosis of a pseudotumor was confirmed if there was also histological evidence of lymphocytic infiltrates (including aseptic lymphocytic vasculitis and associated lesions) and a phagocytic macrophage response to metal wear debris, with or without tissue necrosis23-25.
Nonpseudotumor Group (254 Hips)
Following alerts in 2010 and 2012 from the Medicines and Healthcare products Regulatory Agency (MHRA), all symptomatic MoMHR patients underwent clinical examination, anteroposterior pelvic radiographs, blood metal ion analysis, and cross-sectional imaging6,26. In 2007 and 2008, prior to these alerts, we had investigated 201 asymptomatic MoMHRs with anteroposterior pelvic radiographs, blood metal ion analysis, and cross-sectional imaging27.
The nonpseudotumor group for the present study included all patients with nonrevised MoMHRs, regardless of symptoms, with normal findings on cross-sectional imaging (no evidence of a pseudotumor on ultrasound and/or MARS-MRI). The nonpseudotumor group included 254 hips with a MoMHR, of which 128 were symptomatic (median Oxford Hip Score28,29 [OHS] = 32 of 48; interquartile range [IQR], 24 to 38) and 126 were asymptomatic (median OHS = 47 of 48; IQR, 45 to 48). The asymptomatic patient subgroup has previously been reported on27.
Standardized anteroposterior pelvic radiographs for all patients were accessed with use of the hospital’s electronic picture-archiving and communication system (PACS; GE Healthcare). Apart from femoral neck narrowing measurements (which also required use of the radiograph taken immediately following primary MoMHR surgery), all radiographic parameters were assessed with use of a single radiograph for each MoMHR. For the pseudotumor group, the radiograph selected for assessment was the one taken closest to but immediately before the date of revision surgery. This represented a time when the hip was symptomatic and the pseudotumor had already been diagnosed on cross-sectional imaging. In the nonpseudotumor group, the radiograph selected for assessment was taken at the time that cross-sectional imaging excluded a pseudotumor.
Each radiograph was systematically analyzed for the presence or absence of abnormalities previously described in MoMHRs, including component loosening (a radiolucent line of >2 mm in any zone), osteolysis, femoral neck notching, fracture, dislocation, subluxation, impingement, and heterotopic ossification9,30-32. Acetabular component inclination (relative to the pelvic interteardrop line) and anteversion were measured with use of ImageJ software (National Institutes of Health)33. Acetabular components were considered malpositioned if 1 or both parameters were outside the recommended optimal zone for MoMHR (an inclination of 35° to 55° and anteversion of 10° to 30°)11. Femoral neck narrowing was assessed as previously described34. Femoral neck diameter was measured on each radiograph at the junction of the neck and the femoral component, and it was divided by the measured femoral component diameter (allowing correction for magnification). The difference between measurements from the most recent radiograph and the radiograph taken immediately following primary MoMHR allowed calculation of the degree of femoral neck narrowing since the index procedure (expressed as a percentage of the initial neck diameter).
All radiographs were assessed by 2 independent observers (G.S.M. and K.D.) in a random sequence, with both blinded to all clinical information, including the study group. For the presence or absence of different radiographic abnormalities, interobserver agreement was excellent (Cohen kappa statistic, 0.88 to 1.00)35. Any discrepancy regarding the presence or absence of abnormalities was settled by the senior author (H.G.P.), with this final assessment used for analyses. For continuous radiographic data, intraclass correlation coefficients between observers were excellent: 0.979 (95% confidence interval [CI], 0.955 to 0.990) for inclination, 0.968 (95% CI, 0.947 to 0.988) for anteversion, and 0.941 (95% CI, 0.861 to 0.987) for femoral neck narrowing. The mean of the 2 observer measurements was used for continuous radiographic variables.
The study outcome of interest was a binary variable: MoMHR with or without a pseudotumor. The influence of patient factors (sex, age, and implant design) and radiographic factors (noted earlier) were assessed between groups. Numerical data were compared between groups using either unpaired t tests (parametric data) or the Wilcoxon rank-sum test (nonparametric data), with categorical data compared using either the chi-squared test with Yates’ correction or Fisher’s exact test.
Logistic regression modeling was used to identify predictors of outcome. Univariable models explored the association between each predictor and outcome. For continuous predictors, linearity was assessed using likelihood ratio tests, with data categorized if the relationship between a predictor and the outcome was nonlinear. A multivariable logistic regression model was formulated using backward selection, with patient and radiographic predictors retained in the final model if p < 0.10. Regression diagnostics were assessed to ensure that all assumptions underlying the model were met36,37.
Internal validation of the final multivariable model was performed, including calibration, discrimination, and bootstrapping (see Appendix)37-39. Patient and radiographic factors from the final multivariable model were formulated into a clinical risk-scoring tool, with each factor assigned a weighting on the basis of its respective regression coefficient38,40. The calculated overall score represents a patient’s risk of having evidence of a pseudotumor, with higher scores associated with increased risk (see Appendix). P values of <0.05 were considered significant.
Patient Factors (Table I)
Compared with the nonpseudotumor group, the pseudotumor group was younger (p = 0.0286), was more commonly female (63.9% versus 38.2%; p < 0.001), and had a longer follow-up time (mean, 5.8 versus 4.8 years; p < 0.001). There were significant differences in MoMHR implant design between groups (p = 0.022).
Radiographic Factors (Table II)
The pseudotumor group was significantly more likely to have abnormal radiographic findings compared with the nonpseudotumor group (80.0% versus 63.4%; p = 0.001). The abnormalities more frequently observed in the pseudotumor group compared with the nonpseudotumor group were acetabular osteolysis (21.5% versus 4.3%; p < 0.001), femoral osteolysis (20.0% versus 1.6%; p < 0.001), acetabular component loosening (13.1% versus 4.7%; p = 0.003), higher acetabular inclination (mean, 49.5° versus 46.0°; p = 0.0013), femoral fracture (3.1% versus 0%; p = 0.013), dislocation (2.3% versus 0%; p = 0.038), and subluxation (2.3% versus 0%; p = 0.038). Acetabular component anteversion (mean, 15.1° versus 14.7°; p = 0.697) and femoral neck narrowing (mean, 5.0% versus 4.5%; p = 0.556) were not different between the pseudotumor and nonpseudotumor groups. The nonpseudotumor group was significantly more likely to have heterotopic ossification compared with the pseudotumor group (15.7% versus 2.3%; p < 0.001).
Logistic Regression (Table III)
Five factors (female sex, high acetabular component inclination, acetabular osteolysis, femoral osteolysis, and acetabular component loosening) significantly predicted being in the pseudotumor group in both the univariable and multivariable logistic regression analyses, and 2 factors (acetabular component anteversion of ≥5° and heterotopic ossification) significantly predicted being in the nonpseudotumor group. Young patient age at the time of radiography (p = 0.044) was a significant predictor of being in the pseudotumor group in the univariable analysis, although it was not significant in the multivariable analysis.
Internal Validation of the Final Multivariable Model
The final multivariable model was well calibrated (p = 0.589; Fig. 1) and demonstrated good discriminatory ability, with an area under the curve (AUC) of 0.801 (95% CI, 0.752 to 0.849; sensitivity = 74.4% and specificity = 71.7%; Fig. 2). Bootstrap validation of the final model provided a bias-corrected AUC of 0.784.
Clinical Risk Scoring Tool (Table IV)
A clinical points-based risk tool for identifying patients with evidence of a pseudotumor was developed using the final multivariable model. High overall scores represented an increased risk of pseudotumor. Validation of the overall risk score model demonstrated that it had good discriminatory ability (AUC = 0.796; 95% CI, 0.747 to 0.845). The optimal risk score threshold for identifying MoMHRs with evidence of a pseudotumor was 18 points or more (95% CI, 11.9 to 24.1 points), which had 80.8% sensitivity and 65.2% specificity.
Our study demonstrated that a combination of patient and radiographic factors provided useful information for distinguishing between MoMHRs with and without evidence of a pseudotumor. Patient and radiographic factors predictive of MoMHRs with evidence of a pseudotumor included female sex, acetabular component malposition, acetabular osteolysis, femoral osteolysis, acetabular loosening, and the absence of heterotopic ossification. Surgeons may wish to consider these factors before proceeding with cross-sectional imaging.
Our findings suggest that radiographs form an important part of the assessment of MoMHR patients. The high AUC of the final model (0.801; bias-corrected, 0.784) confirms that a combination of patient and radiographic factors was useful for distinguishing between MoMHRs with and without evidence of a pseudotumor. Previous studies are limited by their assessment of only a few radiographic factors, such as cup position or neck narrowing10-13,34. By contrast, the current study assessed all major radiographic parameters, and it is further strengthened by having a large control group with both symptomatic and asymptomatic patients who had no evidence of a pseudotumor on cross-sectional imaging. These patients with nonrevised MoMHRs are typical of the many patients currently under regular surveillance worldwide18.
The final multivariable model identified female sex as the only patient factor significantly predicting MoMHRs with evidence of a pseudotumor. This finding is consistent with the literature1,2,12,41 and further highlights the importance of stratifying MoMHR patients by sex for surveillance18,42. Although we recognize that certain MoMHR designs have higher failure rates1,2,12, the present study did not identify implant design as a significant predictor of pseudotumor revision. This may reflect the limited number of different implant designs and the small numbers of certain implant designs that were included in the present study. However, we agree with previous recommendations that it is important to consider implant design when making clinical decisions about MoMHR patients18,42.
Similar to the authors of previous studies10-12, we identified high inclination and inadequate anteversion (<5°) as predictors of pseudotumor revision. Hard-on-hard bearings have low tolerance for acetabular component positioning outside an optimal zone, with such malposition associated with edge-loading, high bearing wear, and early failure11. However, the relationship between acetabular component position and pseudotumors is complex. Our study and 2 others10,11 observed that MoMHRs revised for pseudotumor more commonly have acetabular components positioned in inadequate anteversion. Nonetheless, we recognize that other authors report excessive anteversion to be more important12. In addition, pseudotumors in patients with adequately positioned acetabular components have been reported in both our study (46% of the pseudotumor group; Table II) and previous studies11,12,43. Furthermore, a number of patients with well-functioning MoMHRs have malpositioned acetabular components15,16,44,45. Although 63% of our nonpseudotumor group had abnormal findings on radiographs, which appears high, acetabular component malposition was the primary reason for these abnormalities. When all radiographic abnormalities apart from malposition are considered, only 19% of our nonpseudotumor group had abnormal findings on radiographs. Other studies have similarly reported abnormal radiographic findings (excluding malposition) in up to 25% of nonrevised MoMHRs15,16,44,45. These observations suggest that pseudotumor development is multifactorial and not solely dependent on acetabular component malposition. It is therefore important to assess radiographs for other signs suggestive of failure.
Osteolysis (femoral and acetabular) and acetabular component loosening were highly predictive of MoMHRs with pseudotumors in this study. Previous reports observed intraoperative osteolysis in up to 33% of MoMHRs revised for pseudotumor and acetabular loosening in up to 28%17,46. Furthermore, extensive osteolysis can require complex reconstruction, which may contribute to poor short-term outcomes following revision for pseudotumor21,47,48. Surgeons must therefore carefully inspect MoMHR radiographs for subtle osteolysis or acetabular loosening and arrange further investigations as necessary, as early identification of pseudotumors may improve patient outcomes following revision.
The clinical importance of femoral neck narrowing remains unclear. In well-functioning MoMHRs, narrowing has been observed in up to 77% of cases, with up to 28% having >10% narrowing15,16,30,34. Longitudinal studies have reported that femoral neck narrowing stabilizes in well-functioning MoMHR patients within 5 years16,30,34. Therefore, it has been suggested that neck narrowing is a normal process reflecting early bone remodeling13,15. However, neck narrowing has also been reported in up to 26% of MoMHRs revised for pseudotumor9,17. We observed similar degrees of femoral neck narrowing in MoMHRs with and without pseudotumors. This suggests that femoral neck narrowing does not necessarily represent an underlying pseudotumor. However, if narrowing is observed, it must be interpreted in the context of other abnormalities.
Radiographic heterotopic ossification was more common in MoMHRs without evidence of a pseudotumor. The heterotopic ossification rate in the nonpseudotumor group (15.7%) was much lower than previous observations (up to 59%)49. As heterotopic ossification is more common in males49, we suspect that the higher rates observed in our nonpseudotumor group are related to significantly more males having well-functioning MoMHRs15,16,41. We recognize that other factors may also contribute to differences in heterotopic ossification rates, including surgical approach and nonsteroidal anti-inflammatory drug use during the postoperative period. However, these factors were not assessed because of a lack of medication data and the risk of overfitting our multivariable model. It is possible that patients with heterotopic ossification have a lower pseudotumor risk because the associated stiffness and reduced hip motion may decrease the risk of edge-loading and subsequent pseudotumor formation; however, this requires further investigation.
Current worldwide follow-up recommendations for MoMHR patients have recently been reported to be costly and not evidence-based18. The use of radiographs during follow-up has been somewhat overlooked because of the use of blood metal ion analysis and cross-sectional imaging18, with some authorities not specifying a role for radiographs6. The final study model and clinical risk scoring tool contained relevant patient and radiographic factors that are useful for distinguishing between MoMHRs with and without evidence of a pseudotumor. Our study therefore demonstrates that radiographs comprise an important part of the assessment of MoMHR patients, and we urge all regulatory authorities to include radiographs in their follow-up recommendations. Our findings may be particularly useful in centers where follow-up resources may need to be rationed given the costly nature of MoMHR surveillance18 as well as in centers where access to blood metal ion analysis and cross-sectional imaging is limited. However, radiographs should not be considered a substitute for performing blood metal ion analysis and obtaining cross-sectional imaging, given that 20% of our revised MoMHRs had normal findings on radiographs despite having histologically confirmed pseudotumors. Furthermore, our findings require validation prior to any clinical implementation50.
This study has limitations, such as being retrospective and potentially not applicable to other MoMHR designs. Furthermore, the radiographs assessed in the pseudotumor group were taken significantly later after the primary MoMHR compared with nonpseudotumor patients, which should be considered when interpreting our findings. Given limitations with radiographic data (49% of MoMHRs had adequate-quality immediate postoperative radiographs), this study cannot make conclusive statements about femoral neck narrowing. It is important to also acknowledge that this study only predicts the presence or absence of a pseudotumor at the time of radiographic assessment, and not the subsequent development of a pseudotumor, which would require a longitudinal study. Our final model requires validation in an external cohort; however, robust internal validation techniques were employed and the final model was not overfitted.
In conclusion, a combination of patient and radiographic factors provided useful information for distinguishing between MoMHRs with and without evidence of a pseudotumor. Surgeons may wish to consider these patient and radiographic factors predictive of pseudotumor (including female sex, acetabular component malposition, osteolysis, acetabular loosening, and the absence of heterotopic ossification) before proceeding with cross-sectional imaging. Radiographs are important when assessing MoMHR patients and should be included in the follow-up recommendations issued by all regulatory authorities.
Detailed descriptions of the statistical methods for internal validation of the final multivariable model and for the development of a clinical risk scoring tool are available with the online version of this article as a data supplement at jbjs.org.
Note: The Royal College of Surgeons of England and Arthritis Research UK provided 1 of the authors with funding to undertake this research.
Investigation performed at the Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, Nuffield Orthopaedic Centre, University of Oxford, Oxford, United Kingdom
Disclosure: One of the authors received funding from Arthritis Research UK and the Royal College of Surgeons of England in support of this research. On the Disclosure of Potential Conflicts of Interest forms, which are provided with the online version of the article, one or more of the authors checked “yes” to indicate that the author had a relevant financial relationship in the biomedical arena outside the submitted work.
- Copyright © 2017 by The Journal of Bone and Joint Surgery, Incorporated