Skeletal Radiol DOI 10.1007/s00256-016-2346-2
SCIENTIFIC ARTICLE
Long-term radiographic follow-up of bisphosphonate-associated atypical femur fractures Jennifer L. Favinger 1 & Daniel Hippe 2 & Alice S. Ha 3
Received: 10 December 2015 / Revised: 21 January 2016 / Accepted: 1 February 2016 # ISS 2016
Abstract Objective To evaluate the appearance of bisphosphonaterelated femur insufficiency fractures on long-term follow-up radiographic studies and to describe the rate of fracture line obscuration and cortical beak healing over time. Materials and methods In this retrospective study, bisphosphonate-related femur fracture radiographs were reviewed by two radiologists for the presence of a fracture line, callus, and the characteristic cortical beak. Kaplan– Meier curves were used to analyze the time to first indication of healing. Femurs were also subdivided into those who underwent early versus late surgical fixation and those who underwent early versus late discontinuation of bisphosphonate. Clinical data including pain level and medication history were collected. Results Forty-seven femurs with a bisphosphonate-related femur fracture were identified in 28 women. Eighty-five percent took a bisphosphonate for greater than 5 years and 59 % for greater than 10 years. The median time to beak healing was 265 weeks and the median time to fracture line healing was * Alice S. Ha [email protected] Jennifer L. Favinger [email protected] Daniel Hippe [email protected]
1
Department of Radiology, University of Washington, 1959 N.E. Pacific Street, Box 357115, Seattle, WA 98195, USA
2
Department of Radiology, University of Washington, 850 Republication St, Seattle, WA 98109, USA
3
Department of Radiology, University of Washington, 4245 Roosevelt Way NE, Box 354755, Seattle, WA 98105, USA
56 weeks in the 31 femurs with a baseline fracture. No statistically significant difference was identified between surgical fixation and conservative management. Conclusions Bisphosphonate-related fractures demonstrate notably prolonged healing time on long-term follow-up. Keywords Bisphosphonate . Osteoporosis . Fracture healing . Radiography . Femur
Introduction Atypical insufficiency fractures of the femur are a recognized complication of long-term bisphosphonate use [1, 2]. These insufficiency fractures are characterized by a focal thickening of the lateral cortex of the proximal femoral diaphysis, with or without a transverse lucent line [3]. It is thought that bisphosphonates incorporate into bone, impair osteoclast activity, thereby reducing bone remodeling and inhibiting repair of microfractures, resulting in accumulated microdamage, particularly in areas of high mechanical stress [4]. It has also been shown that bisphosphonates, following incorporation, can persist within the bone for greater than 10 years [5]. Atypical femur fractures are rare; the absolute risk of an atypical femur fracture in patient’s taking bisphosphonates is approximately 3.2 to 50 cases per 100,000 person/years [1]. The risk might be higher in patients with long-term bisphosphonate use, potentially up to 100 per 100,000 person/years [1]. Schilcher et al. reported that the risk of atypical femur fractures increases steadily with time of usage, and drops off by approximately 70 % per year since last use once the drug is discontinued [6]. Therefore, the dominant risk factor for an atypical femur fracture is likely continued use, but given the skeletal accumulation of the bisphosphonate, the risk of fracture can remain for years following discontinuation [7]. Given
Skeletal Radiol
this altered bone metabolism, it is not surprising that fracturehealing time would be prolonged. Because it has been shown that bisphosphonates remain in bone for years after drug discontinuation, we hypothesize that these fractures might not heal with the same pattern as normal bone and this atypical appearance will persist on subsequent follow-up imaging. The aim of this retrospective imaging study was to describe the healing patterns as visualized on serial radiographic imaging and estimate healing times of patients presenting with the classic bisphosphonate-associated cortical beak.
Methods This is a retrospective study with waiver of informed consent approved by our Institutional Review Board. A database search was conducted to select patients between January 2000 and September 2014 who met the following criteria: history of bisphosphonate use, prior radiograph or computed tomography (CT) imaging of the femur for pain, or suspected fracture where a cortical beak was found on imaging, and one or more follow-up radiographs or CT exams. Exclusion criteria included history of high mechanism trauma (e.g., motor vehicle crash), diffuse bone abnormalities such as osteogenesis imperfecta, or known focal bone lesion in the proximal femur. Data was collected from a review of the patients’ electronic medical records, including clinical notes and operative reports. Demographic data collected included patient age, gender, length of bisphosphonate use, type of bisphosphonate, other medications used, and surgery performed. Clinical notes at the imaging intervals were also reviewed for reports of pain related to the fracture. Image analysis Blinded to each patient’s clinical information, the baseline and follow-up radiographs and scout radiographs from CT exams were reviewed by a fellowship-trained musculoskeletal radiologist and a senior radiology resident with a consensus read for presence of a fracture line, callus, and the characteristic cortical beak. Scout images were used of the CT exams only if a radiograph was not available at the same time point. Three separate radiographic indications of healing were defined as the primary outcomes: (1) the cortical beak appearing distinctly less conspicuous, though not necessary absent; (2) the fracture line appearing distinctly less conspicuous, and (3) the presence of callus. The latter two only applied to femurs that had a fracture line at baseline. All outcomes were defined using the prior imaging exam as reference. Partial and complete healing was not differentiated. Figure 1 demonstrates a sample patient with a baseline examination and interval
follow-up; in this patient you can see progression toward healing over the course of 3 years.
Statistical analysis Categorical variables were summarized as count (percentage) and continuous variables were summarized as mean ± standard deviation (SD) or by using the median, inter-quartile range, or range. Radiographic healing outcomes could only be assessed at imaging times, the timing of which varied from patient to patient. These outcomes were treated as intervalcensored observations, where the timing of the outcomes was known only to be within the interval defined by the current and prior imaging examination rather than at an exact time. For example, if healing was observed at the third exam, the date of healing was specified to have occurred between the second and third exams rather than on the date of the third exam. If no healing was observed, the date of healing was specified to have occurred after the last exam (right censored). Kaplan–Meier (KM) curves were estimated from the interval-censored outcome data using the Turnbull expectation-maximization algorithm, which accounts for the additional uncertainty in event timing and the heterogeneity in follow-up intervals [8]. The KM curves were used to estimate healing rates at specific time points and estimate the median healing time. Healing rates were compared between groups of femurs—defined by when bisphosphonates were discontinued or whether surgical fixation was performed— using accelerated failure time regression models based on the Weibull distribution, which can easily accommodate interval-censored data [9]. The robust sandwich estimator was used to calculate standard errors for hypothesis tests to account for dependence between femurs of the same patient. As an exploratory analysis, associations between patientreported pain (coded as a binary variable) at each examination and radiographic healing were analyzed using logistic regression with a random intercept per femur (generalized linear mixed models) [10]. Each examination was classified into one of four categories: (A) baseline examination; (B) examination after baseline but before radiographic healing was observed; (C) examination closest to when radiographic healing was observed, and (D) any examination after radiographic healing was observed (later than category C exam). This categorization was done separately for each type of radiographic healing. Rates of reported pain were compared between periods using Wald tests based on the generalized linear mixed models. No multivariate analysis was performed due to the relatively small sample size. All statistical calculations were conducted with the statistical computing language R (version 3.1.1; R Foundation for Statistical Computing, Vienna, Austria).
Skeletal Radiol
Fig. 1 Healing of an atypical femur fracture. Multiple antero-posterior radiographs of the right femur in a 74-year-old female with thigh pain who had been taking alendronate for 9 years demonstrate an atypical femoral fracture on the baseline examination (a) with a transverse lucent
line compatible with an incomplete fracture. On the 1-year follow-up examination (b), the cortical beak remains, but the fracture line has healed. On the 3-year follow-up examination (c), the cortical beak has become less conspicuous, suggesting healing
Results
The rates of radiographic healing are summarized in Table 2 and Fig. 2. The median time to beak healing was 265 days for the 47 femurs. Only 2.4 % showed beak healing by 6 weeks and 22 % by 6 months. In the group of 31 femurs with a fracture at baseline, the median time to fracture line healing was 56 weeks. No fracture line healing was seen by 6 weeks and 45 % demonstrated fracture line healing by 6 months. Only 27 % had callus formation detected by 6 months. Results were little changed when the three patients on steroids at baseline were excluded. Femurs were grouped by whether bisphosphonate was discontinued early (by 1 week after baseline or earlier) or late (later than 1 week after baseline). There were no statistically significant differences in radiographic healing times between groups, but there were some trends. Compared with the femurs where bisphosphonate was discontinued late (n = 32), femurs where bisphosphonate was stopped early (n = 15) had faster beak healing (6-month healing: 24 vs. 13 %, p = 0.17). In the 31 fractured femurs, those where bisphosphonate was discontinued early (n = 23) had faster fracture line healing (6month healing: 53 vs. 15 %, p = 0.073; Fig. 3) and faster callus formation (6-month healing: 31 vs. 0 %, p = 0.59) than those who stopped later (n = 8). Similarly, femurs were grouped by whether their first surgical fixation procedure was performed early (by 1 week after
Twenty-eight patients met the study inclusion criteria. Both femurs met the study inclusion criteria in 19 patients, providing 47 femurs for analysis. Of these 47 femurs, 31 (66 %) from 24 patients had a fracture line or displaced fracture visible at baseline (seven patients with bilateral fractures). Patient characteristics are summarized in Table 1. All patients were female and ages ranged from 46 to 85 years (mean, 65 years). Eightyfive percent of the women took bisphosphonate for 5 years or more and 59 % for more than 10 years. Eleven patients (39 %) had stopped taking bisphosphonate by the time of the baseline imaging exam while all but one discontinued bisphosphonate during the study period, though that patient was lost to followup (censored) after only 42 weeks. Three patients (representing five femurs) were on steroids at the time of their baseline exam. Seventeen (61 %) patients initiated teriparatide therapy during the study period, all after discontinuation of bisphosphonates. The 47 femurs were followed by a total of 284 radiologic exams; 267 (94 %) were radiographs and 17 (6 %) were CT. The total number of exams per patient ranged from two to 14 (median, 6; inter-quartile range, 5–7). The total follow-up time ranged from 149 days to 9.3 years (median, 1.7 years) and the median time interval between exams was 78 days.
Skeletal Radiol Table 1 Patient characteristics and history (n = 28)
No. (%) or mean ± SD Age, years
65.1 ± 8.7
Bisphosphonate*
Stopped bisphosphonates
Total duration of bisphosphonate use
Surgical fixation (earliest)
Alendronate
26 (92.9)
Ibandronate Zoledronate
3 (10.7) 2 (7.1)
On or before first day of study
11 (39.3)
During study period
16 (57.1)
No/after study period†
SCIENTIFIC ARTICLE
Long-term radiographic follow-up of bisphosphonate-associated atypical femur fractures Jennifer L. Favinger 1 & Daniel Hippe 2 & Alice S. Ha 3
Received: 10 December 2015 / Revised: 21 January 2016 / Accepted: 1 February 2016 # ISS 2016
Abstract Objective To evaluate the appearance of bisphosphonaterelated femur insufficiency fractures on long-term follow-up radiographic studies and to describe the rate of fracture line obscuration and cortical beak healing over time. Materials and methods In this retrospective study, bisphosphonate-related femur fracture radiographs were reviewed by two radiologists for the presence of a fracture line, callus, and the characteristic cortical beak. Kaplan– Meier curves were used to analyze the time to first indication of healing. Femurs were also subdivided into those who underwent early versus late surgical fixation and those who underwent early versus late discontinuation of bisphosphonate. Clinical data including pain level and medication history were collected. Results Forty-seven femurs with a bisphosphonate-related femur fracture were identified in 28 women. Eighty-five percent took a bisphosphonate for greater than 5 years and 59 % for greater than 10 years. The median time to beak healing was 265 weeks and the median time to fracture line healing was * Alice S. Ha [email protected] Jennifer L. Favinger [email protected] Daniel Hippe [email protected]
1
Department of Radiology, University of Washington, 1959 N.E. Pacific Street, Box 357115, Seattle, WA 98195, USA
2
Department of Radiology, University of Washington, 850 Republication St, Seattle, WA 98109, USA
3
Department of Radiology, University of Washington, 4245 Roosevelt Way NE, Box 354755, Seattle, WA 98105, USA
56 weeks in the 31 femurs with a baseline fracture. No statistically significant difference was identified between surgical fixation and conservative management. Conclusions Bisphosphonate-related fractures demonstrate notably prolonged healing time on long-term follow-up. Keywords Bisphosphonate . Osteoporosis . Fracture healing . Radiography . Femur
Introduction Atypical insufficiency fractures of the femur are a recognized complication of long-term bisphosphonate use [1, 2]. These insufficiency fractures are characterized by a focal thickening of the lateral cortex of the proximal femoral diaphysis, with or without a transverse lucent line [3]. It is thought that bisphosphonates incorporate into bone, impair osteoclast activity, thereby reducing bone remodeling and inhibiting repair of microfractures, resulting in accumulated microdamage, particularly in areas of high mechanical stress [4]. It has also been shown that bisphosphonates, following incorporation, can persist within the bone for greater than 10 years [5]. Atypical femur fractures are rare; the absolute risk of an atypical femur fracture in patient’s taking bisphosphonates is approximately 3.2 to 50 cases per 100,000 person/years [1]. The risk might be higher in patients with long-term bisphosphonate use, potentially up to 100 per 100,000 person/years [1]. Schilcher et al. reported that the risk of atypical femur fractures increases steadily with time of usage, and drops off by approximately 70 % per year since last use once the drug is discontinued [6]. Therefore, the dominant risk factor for an atypical femur fracture is likely continued use, but given the skeletal accumulation of the bisphosphonate, the risk of fracture can remain for years following discontinuation [7]. Given
Skeletal Radiol
this altered bone metabolism, it is not surprising that fracturehealing time would be prolonged. Because it has been shown that bisphosphonates remain in bone for years after drug discontinuation, we hypothesize that these fractures might not heal with the same pattern as normal bone and this atypical appearance will persist on subsequent follow-up imaging. The aim of this retrospective imaging study was to describe the healing patterns as visualized on serial radiographic imaging and estimate healing times of patients presenting with the classic bisphosphonate-associated cortical beak.
Methods This is a retrospective study with waiver of informed consent approved by our Institutional Review Board. A database search was conducted to select patients between January 2000 and September 2014 who met the following criteria: history of bisphosphonate use, prior radiograph or computed tomography (CT) imaging of the femur for pain, or suspected fracture where a cortical beak was found on imaging, and one or more follow-up radiographs or CT exams. Exclusion criteria included history of high mechanism trauma (e.g., motor vehicle crash), diffuse bone abnormalities such as osteogenesis imperfecta, or known focal bone lesion in the proximal femur. Data was collected from a review of the patients’ electronic medical records, including clinical notes and operative reports. Demographic data collected included patient age, gender, length of bisphosphonate use, type of bisphosphonate, other medications used, and surgery performed. Clinical notes at the imaging intervals were also reviewed for reports of pain related to the fracture. Image analysis Blinded to each patient’s clinical information, the baseline and follow-up radiographs and scout radiographs from CT exams were reviewed by a fellowship-trained musculoskeletal radiologist and a senior radiology resident with a consensus read for presence of a fracture line, callus, and the characteristic cortical beak. Scout images were used of the CT exams only if a radiograph was not available at the same time point. Three separate radiographic indications of healing were defined as the primary outcomes: (1) the cortical beak appearing distinctly less conspicuous, though not necessary absent; (2) the fracture line appearing distinctly less conspicuous, and (3) the presence of callus. The latter two only applied to femurs that had a fracture line at baseline. All outcomes were defined using the prior imaging exam as reference. Partial and complete healing was not differentiated. Figure 1 demonstrates a sample patient with a baseline examination and interval
follow-up; in this patient you can see progression toward healing over the course of 3 years.
Statistical analysis Categorical variables were summarized as count (percentage) and continuous variables were summarized as mean ± standard deviation (SD) or by using the median, inter-quartile range, or range. Radiographic healing outcomes could only be assessed at imaging times, the timing of which varied from patient to patient. These outcomes were treated as intervalcensored observations, where the timing of the outcomes was known only to be within the interval defined by the current and prior imaging examination rather than at an exact time. For example, if healing was observed at the third exam, the date of healing was specified to have occurred between the second and third exams rather than on the date of the third exam. If no healing was observed, the date of healing was specified to have occurred after the last exam (right censored). Kaplan–Meier (KM) curves were estimated from the interval-censored outcome data using the Turnbull expectation-maximization algorithm, which accounts for the additional uncertainty in event timing and the heterogeneity in follow-up intervals [8]. The KM curves were used to estimate healing rates at specific time points and estimate the median healing time. Healing rates were compared between groups of femurs—defined by when bisphosphonates were discontinued or whether surgical fixation was performed— using accelerated failure time regression models based on the Weibull distribution, which can easily accommodate interval-censored data [9]. The robust sandwich estimator was used to calculate standard errors for hypothesis tests to account for dependence between femurs of the same patient. As an exploratory analysis, associations between patientreported pain (coded as a binary variable) at each examination and radiographic healing were analyzed using logistic regression with a random intercept per femur (generalized linear mixed models) [10]. Each examination was classified into one of four categories: (A) baseline examination; (B) examination after baseline but before radiographic healing was observed; (C) examination closest to when radiographic healing was observed, and (D) any examination after radiographic healing was observed (later than category C exam). This categorization was done separately for each type of radiographic healing. Rates of reported pain were compared between periods using Wald tests based on the generalized linear mixed models. No multivariate analysis was performed due to the relatively small sample size. All statistical calculations were conducted with the statistical computing language R (version 3.1.1; R Foundation for Statistical Computing, Vienna, Austria).
Skeletal Radiol
Fig. 1 Healing of an atypical femur fracture. Multiple antero-posterior radiographs of the right femur in a 74-year-old female with thigh pain who had been taking alendronate for 9 years demonstrate an atypical femoral fracture on the baseline examination (a) with a transverse lucent
line compatible with an incomplete fracture. On the 1-year follow-up examination (b), the cortical beak remains, but the fracture line has healed. On the 3-year follow-up examination (c), the cortical beak has become less conspicuous, suggesting healing
Results
The rates of radiographic healing are summarized in Table 2 and Fig. 2. The median time to beak healing was 265 days for the 47 femurs. Only 2.4 % showed beak healing by 6 weeks and 22 % by 6 months. In the group of 31 femurs with a fracture at baseline, the median time to fracture line healing was 56 weeks. No fracture line healing was seen by 6 weeks and 45 % demonstrated fracture line healing by 6 months. Only 27 % had callus formation detected by 6 months. Results were little changed when the three patients on steroids at baseline were excluded. Femurs were grouped by whether bisphosphonate was discontinued early (by 1 week after baseline or earlier) or late (later than 1 week after baseline). There were no statistically significant differences in radiographic healing times between groups, but there were some trends. Compared with the femurs where bisphosphonate was discontinued late (n = 32), femurs where bisphosphonate was stopped early (n = 15) had faster beak healing (6-month healing: 24 vs. 13 %, p = 0.17). In the 31 fractured femurs, those where bisphosphonate was discontinued early (n = 23) had faster fracture line healing (6month healing: 53 vs. 15 %, p = 0.073; Fig. 3) and faster callus formation (6-month healing: 31 vs. 0 %, p = 0.59) than those who stopped later (n = 8). Similarly, femurs were grouped by whether their first surgical fixation procedure was performed early (by 1 week after
Twenty-eight patients met the study inclusion criteria. Both femurs met the study inclusion criteria in 19 patients, providing 47 femurs for analysis. Of these 47 femurs, 31 (66 %) from 24 patients had a fracture line or displaced fracture visible at baseline (seven patients with bilateral fractures). Patient characteristics are summarized in Table 1. All patients were female and ages ranged from 46 to 85 years (mean, 65 years). Eightyfive percent of the women took bisphosphonate for 5 years or more and 59 % for more than 10 years. Eleven patients (39 %) had stopped taking bisphosphonate by the time of the baseline imaging exam while all but one discontinued bisphosphonate during the study period, though that patient was lost to followup (censored) after only 42 weeks. Three patients (representing five femurs) were on steroids at the time of their baseline exam. Seventeen (61 %) patients initiated teriparatide therapy during the study period, all after discontinuation of bisphosphonates. The 47 femurs were followed by a total of 284 radiologic exams; 267 (94 %) were radiographs and 17 (6 %) were CT. The total number of exams per patient ranged from two to 14 (median, 6; inter-quartile range, 5–7). The total follow-up time ranged from 149 days to 9.3 years (median, 1.7 years) and the median time interval between exams was 78 days.
Skeletal Radiol Table 1 Patient characteristics and history (n = 28)
No. (%) or mean ± SD Age, years
65.1 ± 8.7
Bisphosphonate*
Stopped bisphosphonates
Total duration of bisphosphonate use
Surgical fixation (earliest)
Alendronate
26 (92.9)
Ibandronate Zoledronate
3 (10.7) 2 (7.1)
On or before first day of study
11 (39.3)
During study period
16 (57.1)
No/after study period†
