Before this study, the senior author (T.P.G.) performed 830 HRAs since 1999 [11]. Therefore, by most published criteria, he had already surpassed the learning curve of hip resurfacing procedures prior to this study. Beginning in July 2006, we routinely recorded bone mineral density with the use of a dual energy x-ray absorptiometry (DEXA) scan prior to metal-on-metal HRAs. After September 2008, we started treating low bone density patients with alendronate. In this retrospective study, 373 consecutive metal-on-metal HRAs were implanted in 346 patients by the senior author between July 2006 and September 2008. Bone mineral density data (T-score) were recorded for all of these cases, and none of the cases were treated with alendronate. Two patients (two hips) died from unrelated causes. Because their two-year follow-up information was available, they were still included in the study. 233 (67%) patients were men, the average age was 52 ± 8 years (range: 23 to 76 years), the average body mass index was 27 ± 4 kg/m2 (range: 18 to 43 kg/m2), and the average DEXA scan T-score was 0.09 ± 1.4 (range from -2.4 to 6.7). The primary diagnosis was osteoarthritis in 290 hips (78%) followed by dysplasia in 52 hips (14%), osteonecrosis in 14 cases (4%), post-traumatic arthritis in 8 cases (2%), Legg-Calve-Perthes in three cases (0.8%), slipped capital femoral epiphysis in three cases (0.8%), post-infection in one case (0.3%), rheumatoid arthritis (RA) in one case (0.3%), and ankylosing spondylitis in one case (0.3%). Pre-operative demographic information, Harris hip scores, and intra-operative technical data were routinely collected in this study. Follow-up visits were requested at six weeks, one year, two years, and every other year thereafter post-operatively. The average length of follow-up in the present study was 30 ± 6 months (range: 24 to 47 months). Post-operative information including post-operative Harris hip scores, visual analog scale (VAS) pain scores on regular days and on worst days, UCLA activity scores, complications, and failures were prospectively recorded for all patients. Anteroposterior and lateral radiographs were also routinely analyzed at each follow-up visit. All of the above data were maintained in a computerized database, OrthoTrack (Midlands Orthopaedics, p.a., Columbia, SC). Institutional review board approval (IRB) was obtained for this study.
The senior surgeon used a previously described [12] posterior, minimally invasive surgical approach on all cases. In 77% of these cases, a Biomet ReCap™ cemented femoral component (Biomet, Warsaw, IN, USA) was used while in the remaining 23%, a ReCap™ fully porous coated femoral component was used. The average femoral component size was 50 ± 4 mm (range: 40 to 60 mm). Fully porous coated Magnum™ acetabular components were used in all cases, and their outer diameter sizes were 6 mm larger than the corresponding femoral component. The average acetabular inclination angle was 42° ± 7° (range: 19° to 61°).
A paired t-test was utilized to compare the statistical difference between the pre- and post- operative HHS score. Kaplan-Meier survivorship curves [13] were calculated using femoral failure, acetabular failure, or both for any reason as the end points, respectively, in order to analyze the success rates of up to four-year follow-up in this study. However, the primary endpoint studied was any femoral failure that occurred before two years post-operatively. This included all femoral neck fractures and all less acute femoral failures that were evident clinically or radiographically before two years. If a patient was revised or had radiographic signs of femoral failure at up to three years post-operatively, they were included as an early failure if their symptoms or radiographic abnormalities were present prior to two years post-operatively. All of the following statistical analyses used only early femoral failure for any reason as the end point. Multivariable logistic regression models were generated to identify significant risk factors for early femoral failure after metal-on-metal HRA. In this logistic regression model, early femoral failure was a categorical variable and defined as the outcome. Age, gender, diagnosis, body mass index, T-score, femoral implant fixation type, and the size of the femoral component were each defined as explanatory variables. These explanatory variables of age, gender, body mass index, T-score, and the size of the femoral component were initially included as categorical variables grouped with different thresholds according to our experience or suggested by previous references [8, 9, 14, 15], as well as numerical variables. Different multivariable logistic regression models were tested by changing the types and thresholds of these variables in order to find the best regression model to predict the early femoral failures. The final regression model determined whether these five variables should be treated as category variables and, if so, what the thresholds should be. First, a full factorial regression model including all seven factors was generated to help us predict the possibility of early femoral failure. Covariates, which did not contribute significantly to the model fit with the significance level α = 0.05, were excluded from the present model. Then, a reduced regression model was built to evaluate which independent factor had the strongest effect on the failures. Possibilities for femoral failures within the ranges of these independent risk factors were predicted based on this reduced model and plotted to determine their effects. Finally, the significant risk factors were regrouped with different thresholds. Mosaic plots were depicted and Chi-square analyses were performed to evaluate the thresholds of each risk factor and thecombined factors in order to provide more meaningful information for surgeons for clinical use.