Larger Femoral Periprosthetic Bone Mineral Density Decrease Following Total Hip Arthroplasty for Femoral Neck Fracture Than for Osteoarthritis: A Prospective, Observational Cohort Study Tobias Mann,1 Thomas Eisler,2 Henrik Bode´n,2 Olle Muren,2 Andre´ Stark,2 Mats Salemyr,2 Olof Sko¨ldenberg2 1 Department of Orthopaedics and Rehabilitation, University of Rochester, Rochester, New York, 2Division of Orthopaedics, Department of Clinical Sciences at Danderyd Hospital, Karolinska Institute, Stockholm S-18288, Sweden
Received 6 September 2014; accepted 31 October 2014 Published online 8 February 2015 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jor.22783
ABSTRACT: Studies on patients with degenerative joint disease of the hip show that femoral periprosthetic bone mineral decreases following total hip arthroplasty. Scarcely any osteodensitometric data exist on femoral neck fracture (FNF) patients and periprosthetic bone remodelling. In two parallel cohorts we enrolled 87 patients (mean age, 72 12 years; male:female ratio, 30:57) undergoing total hip arthroplasty for either primary osteoarthritis (OA) of the hip (n ¼ 37) or for an acute FNF (n ¼ 50) and followed them for a mean of 5.4 years. Outcomes were bone mineral density (BMD) changes in the periprosthetic Gruen zones 1–7, the incidence of periprosthetic fractures and clinical outcome. The bone mineral loss in the fracture group was more than twice that of the osteoarthritis group, 16.9% versus 6.8% (p ¼ 0.004). The incidence of periprosthetic fractures was 12% (6/50) in the fracture cohort compared with none (0%) in the OA cohort (p ¼ 0.03). Periprosthetic bone mineral loss following total hip arthroplasty is significantly greater in patients who are treated for acute FNF than in OA patients. This decrease of BMD follows a different pattern with the FNF patients losing larger proportions of bone in Gruen zones 1, 2, 6, and 7 while the OA patients tend to have larger losses only in zones 1 and 7. ß 2015 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 33:504–512, 2015. Keywords: total hip arthroplasty; femoral neck fracture; osteoarthritis; bone mineral density; periprosthetic fracture
Periprosthetic bone mineral decrease adjacent to femoral stems is considered to be a consequence of stress shielding following total hip arthroplasty (THA).1–3 The most commonly expressed hypothesis is that the higher modulus of elasticity of the metallic implant (in relation to bone) bears the majority of the load, thus shielding the surrounding bone from stress and leading to atrophy, in accordance with Wolff’s law.4,5 This local decrease in bone mass in the proximal diaphysis of the femur seems to progress more rapidly than natural, age-related bone loss1 and could potentially explain the incidence of late occurring periprosthetic femoral fractures.6,7 However, all studies on bone mineral density (BMD) change are from cohorts with degenerative joint disease and few studies have shown that periprosthetic BMD decrease has any systematically important clinical consequence. Late-occurring periprosthetic fractures following THA (i.e., fractures occurring after the immediate postoperative period) are more commonly observed in patients with acute femoral neck fractures (FNF) compared to elective procedures for hip osteoarthritis (OA).8 The aim of the current study was to compare the periprosthetic bone loss in patients with OA to that of patients with a FNF and to determine if this has clinical consequences in the form of periprosthetic fractures.
Conflicts of interest: None. Grant sponsor: Ulla and Gustaf Ugglas Stiftelse; Grant sponsor: Ake Wiberg stiftelse; Grant sponsor: Loo and Hans Ostermans Stiftelse; Grant sponsor: Sven Nore´n. Correspondence to: Olof G. Sko¨ldenberg (T: þ46-8-12357245; F: þ46-8-7551476; E-mail:
[email protected])
# 2015 Orthopaedic Research Society. Published by Wiley Periodicals, Inc.
504
JOURNAL OF ORTHOPAEDIC RESEARCH APRIL 2015
MATERIALS AND METHODS Study Design and Setting This observational, prospective, cohort study was performed at the Orthopedic Department of Danderyd Hospital in collaboration with the Karolinska Institute and the Department of Clinical Sciences in Stockholm, Sweden between 2005 and 2013 (inclusion period 2005–2008). The STROBE (STrengthening the Reporting of OBservational studies in Epidemiology) guidelines were followed.9 The study was conducted in accordance with the ethical principles of the Helsinki declaration10 and was approved by Ethics Committee of the Karolinska Institute. Participants Starting 2005, two parallel patient cohorts were followed at our institution as part of two separate studies on bone remodeling studies.11,12 We included OA and FNF patients scheduled for THA who gave their oral and written consent to participate. Additional inclusion criteria were applied to the OA patients: A diagnosis of primary hip OA, with failed conservative management. For the FNF patients, additional inclusion criteria were a diagnosis of a Garden III or IV13 FNF. We excluded patients with hypocalcaemia, as well as patients who were taking medications that are known to affect bone metabolism (e.g., systemic steroids or bisphosphonates). No patients were allowed bisphosphonates or other bone-active drugs during the study and ever-use of bisphosphonates was an exclusion criteria. Patients who were deemed to be unsuitable for the operation by the anesthesiologist because of multiple co-morbidities and patients who were unsuitable for study participation for any other reason were also excluded. Surgery All operations were performed by one of four experienced arthroplasty surgeons (HGB, MOS, OM, or OGS). A standard posterior approach was used, and the posterior capsule and short external rotators were meticulously repaired. An
LARGER FEMORAL PERIPROSTHETIC BONE MINERAL DENSITY
uncemented titanium alloy (Ti-6Al-4V), straight, tapered, femoral component with a hydroxyapatite (HA)-coated surface was used (Bi-Metric HA, Biomet, Warsaw, IN). A 28 mm (diameter) modular chromium-cobalt head was used for the OA patients, and a 32 mm head was used for the patients with FNFs. This head was combined with either a cemented acetabular cup (ZCA Longevity, Zimmer, Warzaw, IN) or an uncemented press-fit cup (Trilogy Longevity, Zimmer, Warzaw, IN). The patient was mobilized using standard physiotherapy program, and immediate full weight-bearing was encouraged with the use of crutches. Outcome Variables All patients were followed with repeated follow-ups until death or until September 1, 2013 (mean, 5.4 years; range, 0.2–7.1 years). The outcome variables were change in BMD in the periprosthetic regions, clinical outcome, the occurence of periprosthetic fractures, and frequency of reoperations. Exposure was defined by group (OA vs. FNF) and confounders included BMD measured pre- and post-operatively in the
505
periprosthetic area and in the lumbar vertebras, gender, and stem size of the femoral implant. Data Sources
Bone Mineral Density Changes BMD was measured postoperatively at 2 days, 3, and 6 months, and 1, 2, and 4 years (Fig. 1). Periprosthetic BMD was measured in the seven Gruen zones using dual energy X-ray absorptiometry (DPX-L; Lunar, Madison, WI) in the frontal plane.1,14 The change in BMD was calculated by dividing the most recent measured BMD value by the value measured at postoperative day 2. This value was then expressed as a percentage change. To assess the bone quality of the operated hip, preoperative BMD15 of the operated hip was measured in the OA patients. Because this measurement was not possible in the patients with FNF, the healthy contralateral hip was measured in these patients at postoperative day 2 and used as a proxy for preoperative BMD of the operated hip. To assess the general bone quality of each treated patient and to study
Figure 1. Diagram of the patient flow through the study. DXA, dual-energy X-ray absorptiometry. †Withdrawn from the study because of illness or dementia. ‡Analyzed for complications and reoperations. *Twelve patients had 4-year clinical follow-up data, but we were unable to obtain a complete DXA scan follow-up because three had sustained periprosthetic fractures before the last follow-up and because nine were too frail to attend follow-ups at the hospital and were interviewed in a nursing home. JOURNAL OF ORTHOPAEDIC RESEARCH APRIL 2015
506
MANN ET AL.
change over the study period, the BMD of the L1 through L4 vertebrae were measured postoperatively at 2 days and at 2 and 4 years. Subjects with missing BMD data at any of the follow-up visits were analyzed by carrying the last observation forward. This process was performed for three patients who sustained periprosthetic fractures before the 4-year follow-up because fracture healing would affect the measurements.
Other Outcomes At each visit, hip function was evaluated with use of the Harris hip score,16 and health-related quality of life was assessed using the EuroQoL (EQ-5D) questionnaire.17 We used Swedish unique personal id-numbers to identify all hip related complications during the study period. Our own database was searched for the occurrence of reoperations and adverse events prior to September 1, 2013. The Swedish death registry was used to verify patient mortality. We used the Swedish Hip Arthroplasty Registry to search for reoperations and adverse events treated elsewhere in Sweden, but no such cases were found.
Radiology Digital anteroposterior and lateral radiographs (Bucky Diagnostics; Philips, Eindhoven, The Netherlands) of the hip were obtained postoperatively at day 2 and at 1, 2, and 4 years. To ensure that all components were well fixed, the migration of the femoral stem was measured from the digital radiographs using Ein-Bild-Roentgen-Analysis Femoral Component Analysis (EBRA-FCA) software (University of Innsbruck, Innsbruck, Austria).18 The presence of radiolucent lines between the bone and implant in the zones, as described by Charnley-DeLee and Charnley,19 around the acetabular component were also recorded. Heterotopic ossification was evaluated according to the Brooker classification.20 The periprosthetic fractures were classified at the time of fracture on radiographs, according to the Vancouver classification.21
Study Size Prior to the study, we conducted a power analysis on the basis of data from the original studies.11,12 To our knowledge, there are no published clinical studies analyzing periprosthetic BMD decrease and FNFs and the risk of future periprosthetic fractures. However, a cadaveric study showed an increased incidence of periprosthetic fractures in older specimens with low global periprosthetic BMD.22 In addition, women who sustain fragility fractures have been shown to have BMD in the hip that is 9% lower than women who do not sustain fragility fractures.23 Based on these studies, we assumed that an approximately 10% decrease in overall (Gruen zones 1–7) periprosthetic BMD would be clinically relevant with respect to future periprosthetic fractures. The analysis showed that a total of 70 patients (35 in each group) were required at the 2 year follow-up to provide 90% power to detect a difference of 10% in BMD decrease between groups of patients with low and high BMD, assuming a standard deviation of 13%11,12 and considering a two-sided pvalue of 0.05 to be significant. We, therefore, planned to include approximately 90 patients in the study to allow for losses. The study was not powered to evaluate the incidence of periprosthetic fractures. Statistical Methods We used the unpaired Students t-test and Levene´s test for comparison of BMD data between the OA and FNF cohort. We used the paired t-test to compare postoperative values with follow-ups within the whole cohort. Nonparametric tests were used for clinical outcome scores. We used a multiple linear regression analysis to analyze which factors influenced bone loss in the entire periprosthetic region at 4 years. Factors separately investigated in the analysis was exposure (OA vs. fracture) and covariates who, from other studies, are known to influence bone remodeling around implants (preoperative bone mineral density in the hip,24 sex,25 age, body mass index, and stem size).26 Pearson´s Chi-squared test was used to compare the categorical variables between the
Table 1. Baseline Characteristics of the Patients
Agea (yr) Sexb Male Female Weighta (kg) Heighta (cm) Body mass indexa (kg/m2) ASA classificationb 1 or 2 3 or 4 Charnley classb A B C Bone mineral density, total hip (g/cm2)a Bone mineral density, L1–L4 vertebrae (g/cm2)a [37] Stem size in mm 9–11/12–14/ 15–17 (no.)
Osteoarthritis (n ¼ 37)
Fracture (n ¼ 50)
60 7
81 5
16 (43) 21 (57) 86 17 173 9 28 5
14 (28) 36 (72) 66 14 168 8 24 4
34 (92) 3 (8)
29 (58) 21 (42)
16 (43) 16 (43) 5 (14) 1.049 0.188 1.294 0.270 11/25/1
31 (62) 5 (10) 14 (28) 0.775 0.136 1.050 0.254 13/22/15
ASA, American Society of Anesthesiologists. aValues are given as the mean and the standard deviation. bValues are given as the number of patients with the percentage in parentheses.
JOURNAL OF ORTHOPAEDIC RESEARCH APRIL 2015
LARGER FEMORAL PERIPROSTHETIC BONE MINERAL DENSITY
groups. The Wilcoxon signed rank test was used to compare clinical outcome scores within the groups because theses were not normally distributed. The level of significance was set at p 0.05. We used SPSS 21.0 for Mac (IBM, New York, NY) for all analyses.
RESULTS Study Participants We enrolled 87 patients (37 OA/50 FNF, 30 males, 57 females; mean age SD, 72 12 years) into the study (Fig. 1, Table 1). The patients in the FNF cohort were, compared to the OA group, older, had a larger proportion of females, had a lower body mass index, a lower BMD of the hip and lumbar spine, and a higher American Society of Anesthesiologists-score. Twelve patients, all of whom had been diagnosed with FNF as an indication for primary surgery, died during the study. Bone Mineral Density The periprosthetic BMD decreased by 12.3 10.2% (mean standard deviation) up to 4 years. However, there was only a minor decrease of the systemic BMD in the lumbar spine (1.2 5.8%) versus the postoperative value, and this change was not statistically significant (p ¼ 0.124). The patients in the femoral neck fracture group experienced, on average, more than double the amount of bone mineral loss in all measured periprosthetic regions up to 4 years than did patients in the hip osteoarthritis cohort, with an absolute difference between the groups of 11.2% (95% confidence limits[95%Cl] 4.6 to 17.8), p ¼ 0.002 (Fig. 2, Table 2). In both groups, most bone mineral was lost in the two proximal most regions, zones 1 and 7. The largest difference in BMD decrease between the two
507
groups was found in zones 2 and 6, where, in zone 2, the FNF patients lost on average almost nine times more than did the patient in the OA cohort, 22.1% vs. 2.5%. In zone 6 the loss was almost four times greater; 23.4% versus 6.2%. The rate of periprosthetic bone loss was greatest during the first 2 years following surgery and continued during the observation period (Tables 2, and 3). In the linear regression analysis, the strongest predictor for the loss of periprosthetic bone was preoperative low BMD in the hip (p ¼ 0.002) (Fig. 3) and a diagnosis of FNF (p ¼ 0.034). Sex, age, body mass index, and stem size did not significantly influence the periprosthetic bone loss (Table 4). Clinical Results Six patients (12%) in the femoral neck fracture group and none in the osteoarthritis group sustained an early periprosthetic fracture (p ¼ 0.029) (Fig. 2) occurring at a mean of 3.4 years (range, 2.1–5.7) postoperatively. Prior to their fractures, the patients who sustained periprosthetic fractures had lost 14.2% (95% confidence interval [CI] 6.6–21.9%) more bone compared to the patients who did not sustain fractures (Fig. 4). Three of the fractures (one Vancouver B2, two Vancouver C) required major surgery, and three undisplaced Vancouver B1 fractures were treated conservatively with protected weight-bearing. All fractures healed. All stems were stable, according to the EBRA-FCA analysis and there were no stem revisions, except for the hip with the Vancouver B2-fracture. Nine hips dislocated, six of which had recurrent dislocations, and one hip’s acetabular component was revised to a
Figure 2. Line graph illustrating the mean (95% CI) change in BMD around the femoral stem (zones 1–7) for patients with osteoarthritis (OA) (solid line) and femoral neck fracture (FNF) (dashed line). The arrows indicate the time for the six periprosthetic fractures in the FNF group. The change in lumbar spine BMD is also indicated by the gray lines. JOURNAL OF ORTHOPAEDIC RESEARCH APRIL 2015
508
MANN ET AL.
Table 2. Change in Periprosthetic Bone Mineral Density Outcome Change in BMD 3 months 6 months 1 year 2 years 4 years Change in BMD 3 months 6 months 1 year 2 years 4 years Change in BMD 3 months 6 months 1 year 2 years 4 years Change in BMD 3 months 6 months 1 year 2 years 4 years Change in BMD 3 months 6 months 1 year 2 years 4 years Change in BMD 3 months 6 months 1 year 2 years 4 years Change in BMD 3 months 6 months 1 year 2 years 4 years Change in BMD 3 months 6 months 1 year 2 years 4 years
Osteoarthritis Group zone 1–7 (%)a 3.4 4.1 4.0 5.0 4.5 4.9 5.1 4.9 6.8 6.5 zone 1 (%)a 5.9 9.0 9.7 9.9 14.5 11.2 17.7 13.1 20.5 12.6 zone 2 (%)b (%)a 0.1 17.3 0. 1 17.3 1.1 18.6 2.7 17.4 2.5 21.6 zone 3 (%)a 4.3 6.6 3.7 7.3 4.4 7.0 2.9 5.5 3.4 7.8 zone 4 (%)a 0.9 8.4 0.8 9.2 0.7 8.9 1.3 7.4 3.3 9.3 zone 5 (%)a 0.9 5.7 0.7 6.8 2.8 7.4 2.2 6.4 2.8 7.0 zone 6 (%)a 3.9 6.1 4.5 6.7 3.8 7.8 4.7 8.7 6.2 11.9 zone 7 (%)a 10.3 10.9 13.1 11.7 16.1 12.0 18.1 14.9 22.3 15.4
Fracture Group
Difference (95% CI)
p-Value
6.5 6.8 8.6 9.2 12.3 10.3 16.8 12.1 18.0 12.9
3.1 4.6 7.9 11.7 11.2
(0.6–5.5) (1.4–7.7) (4.4–11.3) (7.6–15.7) (4.6–17.8)
0.011 0.005