Bisphosphonates and Fractures of the Subtrochanteric or Diaphyseal Femur
NEJM March 24 2010|
"We conclude that the risk of fracture of the subtrochanteric or diaphyseal femur associated with bisphosphonate use is very low, even in women with osteoporosis who received bisphosphonates for up to 10 years. Given the limitation of a small number of events and wide confidence intervals, these studies did not show a significant increase in the risk of subtrochanteric or diaphyseal fracture with 3 to 4 years of use of alendronate or zoledronic acid or up to 10 years of alendronate use."
Dennis M. Black, Ph.D., Michael P. Kelly, M.D., Harry K. Genant, M.D., Lisa Palermo, M.A., Richard Eastell, M.D., Christina Bucci-Rechtweg, M.D., Jane Cauley, Ph.D., Ping Chung Leung, M.D., Steven Boonen, M.D., Ph.D., Arthur Santora, M.D., Anne de Papp, M.D., Douglas C. Bauer, M.D., for the Fracture Intervention Trial and HORIZON Pivotal Fracture Trial Steering Committees
From the University of California at San Francisco, San Francisco (D.M.B., M.P.K., H.K.G., L.P., D.C.B.); University of Sheffield, Sheffield, United Kingdom (R.E.); Novartis Pharmaceuticals, East Hanover, NJ (C.B.-R.); University of Pittsburgh, Pittsburgh (J.C.); Chinese University of Hong Kong, Hong Kong (P.C.L.); Katholieke Universiteit Leuven, Leuven, Belgium (S.B.); and Merck Research Laboratories, Rahway, NJ (A.S., A.P.).
Background - A number of recent case reports and series have identified a subgroup of atypical fractures of the femoral shaft associated with bisphosphonate use. A population-based study did not support this association. Such a relationship has not been examined in randomized trials.
Methods - We performed secondary analyses using the results of three large, randomized bisphosphonate trials: the Fracture Intervention Trial (FIT), the FIT Long-Term Extension (FLEX) trial, and the Health Outcomes and Reduced Incidence with Zoledronic Acid Once Yearly (HORIZON) Pivotal Fracture Trial (PFT). We reviewed fracture records and radiographs (when available) from all hip and femur fractures to identify those below the lesser trochanter and above the distal metaphyseal flare (subtrochanteric and diaphyseal femur fractures) and to assess atypical features. We calculated the relative hazards for subtrochanteric and diaphyseal fractures for each study.
Results - We reviewed 284 records for hip or femur fractures among 14,195 women in these trials. A total of 12 fractures in 10 patients were classified as occurring in the subtrochanteric or diaphyseal femur, a combined rate of 2.3 per 10,000 patient-years. As compared with placebo, the relative hazard was 1.03 (95% confidence interval [CI], 0.06 to 16.46) for alendronate use in the FIT trial, 1.50 (95% CI, 0.25 to 9.00) for zoledronic acid use in the HORIZON-PFT trial, and 1.33 (95% CI, 0.12 to 14.67) for continued alendronate use in the FLEX trial. Although increases in risk were not significant, confidence intervals were wide.
Conclusions - The occurrence of fracture of the subtrochanteric or diaphyseal femur was very rare, even among women who had been treated with bisphosphonates for as long as 10 years. There was no significant increase in risk associated with bisphosphonate use, but the study was underpowered for definitive conclusions.
Several case series have described cases of "atypical" subtrochanteric and diaphyseal fractures of the femoral shaft and have suggested that the risk may be increased in long-term users of bisphosphonates.1,2,3,4,5,6,7,8,9,10,11,12,13 Descriptions of associated atypical characteristics vary but have been described as a simple transverse or oblique (<30 degrees) fracture with cortical beaking and diffuse cortical thickening.2 Clinical associations that have been mentioned in some reports include low-level trauma, prodromal pain or evidence of a previous stress fracture, contralateral changes, and the use of specific concomitant drugs (including corticosteroids),9,10 antiresorptive drugs (including hormone-replacement therapy),10 and proton-pump inhibitors.13 However, these case reports and series have not had either a controlled or a prospective design and have not systematically examined risk factors.
More recently, a population-based registry evaluated hip and subtrochanteric and diaphyseal femur fractures14 and compared their incidence in alendronate users with that in nonusers. All three fracture types were more common in alendronate users. Given the known efficacy of alendronate for a reduction in the rate of hip fracture, the higher rates of all three types of fracture among alendronate users were attributed to the increased use of alendronate among high-risk patients rather than an increase in risk associated with alendronate. This confounding by indication is a bias often seen in nonrandomized studies.15 The authors concluded that there was no evidence of an increased risk of fracture of the subtrochanteric or diaphyseal femur with either short-term or long-term use of a bisphosphonate. However, population-registry studies have limitations, including their observational, nonrandomized nature, a limited ability for covariate adjustment, and incomplete details with respect to fractures (e.g., precise morphologic features and the presence or absence of trauma).
Randomized, controlled trials may provide unbiased information about femoral-shaft fractures. We evaluated the effect of bisphosphonates on such fractures by reevaluating femur fractures from three large, controlled, blinded, randomized trials of bisphosphonates with more than 14,000 patients and more than 51,000 patient-years of follow-up for up to 10 years.
Data from three large, placebo-controlled, randomized trials indicate that the risk of fracture of the subtrochanteric or diaphyseal femur associated with the use of bisphosphonates was extremely low. Although the point estimates of the relative risk with bisphosphonate use suggested no association, confidence intervals were wide because of the small number of events. In addition, in the FLEX trial (the largest and longest randomized extension study of long-term bisphosphonate use), despite an average of 10 years of alendronate use, the risk of fracture of the subtrochanteric or diaphyseal femur was low, with no significant increase among women who continued to receive alendronate for 10 years, as compared with those who discontinued the drug, although again confidence intervals were wide. In all three studies, atypical features could not be fully assessed, since radiographs were not generally available.
The low risk of femoral-shaft fracture is consistent with recent results from a Danish population-based registry.14 That study showed low rates overall, and women receiving alendronate had no higher risk than that caused by their underlying osteoporotic fragility. A recent study that was based on national discharge and medical-claims data in the United States reported similarly low rates of femoral-shaft fracture,22 which made up less than 10% of all hip fractures. In the three studies that we reviewed, even among women with up to 10 years of bisphosphonate exposure, the risk of fracture of the subtrochanteric or diaphyseal femur ranged from one to six cases per 10,000 patient-years. If we had been able to assess atypical characteristics (e.g., cortical thickness and fracture morphology) and limit our consideration to atypical fractures, we probably would have excluded some additional events; this would have further reduced the risk. This assumption is supported by one study showing that only one quarter of subtrochanteric and diaphyseal fractures had atypical features23 and another study showing that approximately one third of low-energy femoral-shaft fractures were transverse or oblique with a small angle.24 Although we can confidently conclude that absolute rates of such fractures are low, wide confidence intervals preclude definitive conclusions regarding the relative risk of treatment.
In the studies that we analyzed here and in other studies, bisphosphonates have been shown to decrease the rates of both vertebral and nonvertebral (including hip) fractures in women with osteoporosis.17,18,20,25 Among women with osteoporosis in the FIT25 and HORIZON-PFT20 trials,25 the estimated numbers of women who would need to be treated were 90 to prevent one hip fracture, 35 to prevent one nonvertebral fracture, and 14 to prevent one morphometric vertebral fracture (Table 3). Thus, we estimated that treating 1000 women with osteoporosis for 3 years would prevent about 100 fractures: 71 vertebral fractures and 29 nonvertebral fractures, including 11 hip fractures.
On the basis of our studies, we estimated that bisphosphonate treatment would result in an annual rate of 2.3 subtrochanteric or diaphyseal fractures per 10,000 patient-years in untreated women with osteoporosis. This rate is consistent with those of the Danish registry study14 and a recent population-based study.22 According to the hypothetical relative risks, the increase in the number of such fractures per 1000 women who are treated for 3 years would be 0.3 events if the relative risk were 1.5, 0.7 events if the relative risk were 2.0, and 1.4 events if the relative risk were 3.0 (Table 3). When balanced against the prevention of 100 fractures, the hypothetical risk is quite small and would be even smaller if the risk were limited to fractures with atypical features.
The Danish observational study showed no relationship between the use of bisphosphonates and the risk of femoral-shaft fracture,14 consistent with our findings. However, these results contrast with those of some case series. For example, one case series2 identified 20 women who had fractures with atypical features (e.g., transverse morphology, cortical thickening, or a low-level trauma), 19 of which were diagnosed in women receiving bisphosphonates. The lack of a control group (together with an inconsistent definition of atypia across case studies) increases the challenge of interpreting discrepancies among the studies. But results from case series suggest that patients taking bisphosphonates who have persistent thigh pain, evidence of a stress fracture, or a contralateral femur fracture should be carefully evaluated for further testing and possible interventions.
It has been suggested that an increased risk of atypical femoral-shaft fracture associated with the use of bisphosphonates may be confined to a small subgroup of patients.3,9 This hypothesis suggests that future studies should attempt to identify risk factors for subtrochanteric or diaphyseal fracture in patients receiving bisphosphonates. Case reports have suggested several risk factors, particularly the concomitant use of corticosteroids.9,10 Other proposed risk factors include concurrent use of proton-pump inhibitors13 and markedly suppressed bone turnover, perhaps owing to the use of multiple antiresorptive medications.8,10 In the HORIZON-PFT trial, one woman in the placebo group who had a femoral-shaft fracture was also receiving injected calcitonin; in the zoledronic-acid group, one woman was also receiving raloxifene, and another had a relevant history of taking multiple antiresorptive medications. In the FIT trial, one woman in the alendronate group had low levels of biochemical markers of bone turnover, but in the FLEX trial, such levels were all well within the premenopausal reference range. Most of the initial case series did not include data on the use of concurrent medications.5,23 Other characteristics that have been suggested as risk factors include a younger age at the time of bisphosphonate initiation, a low pretreatment level of bone turnover, and normal bone mineral density.10 However, none of the characteristics that have been postulated to increase the risk of femoral-shaft fracture among women receiving bisphosphonates have been systematically or prospectively studied.
Our results from the FLEX trial did not show a significant increase in the risk of fracture of the subtrochanteric or diaphyseal femur after 10 years of alendronate use, despite suggestions that a longer duration of use may increase risk. Data regarding duration from the case studies have been mixed. The mean duration in the cases reported by Neviaser et al.2 was 7 years, but the duration of therapy has varied from 2 to 4 years in other reports,8,10,12 suggesting that long-term use of bisphosphonates may not be necessary to increase risk. The Danish registry study showed no increase in risk with long-term use (>10 years). It would be important to identify the effect of the duration of bisphosphonate use so that it could be weighed against benefits in clinical decision making regarding continuing therapy.
Although most of the reported cases of atypical femoral-shaft fracture among bisphosphonate users have occurred in women receiving alendronate (the most commonly used bisphosphonate), cases of such fractures also have been reported among users of ibandronate and risedronate.8,10,12 Furthermore, we found cases of femoral-shaft fracture in patients receiving zoledronic acid, as well as in those receiving placebo, so there is no definitive evidence or theoretical reason to believe that any possible increase in risk would be limited to alendronate.
Rare conditions such as atypical femoral-shaft fractures are difficult to study. The ideal study would be a long, placebo-controlled, randomized trial involving tens of thousands of patients, a study design that is neither feasible nor ethical, given the known efficacy of bisphosphonates to reduce the overall risk of fracture. More rigorous observational studies with the use of a population database with both archived fracture radiographs and medication information might be more feasible.
Although our study was based on randomized trials, it has important limitations. First, we did not have access to radiographs and therefore could not evaluate the atypical features of these fractures. Second, although the studies that we evaluated were large, there were insufficient numbers of events to make definitive conclusions. Third, by design, the use of many potentially confounding medications, including corticosteroids, was excluded from all three studies. The use of other antiresorptive medications was allowed in the HORIZON-PFT trial but excluded in the FIT and FLEX trials. Finally, the two placebo-controlled trials lasted only 3 to 4.5 years and thus could not assess whether the risk of fracture increased with treatment duration. Although the FLEX trial included women who were treated up to 10 years, all women had received bisphosphonates for at least 3 years; therefore, the trial could address only the question of bisphosphonate continuation versus discontinuation.
We conclude that the risk of fracture of the subtrochanteric or diaphyseal femur associated with bisphosphonate use is very low, even in women with osteoporosis who received bisphosphonates for up to 10 years. Given the limitation of a small number of events and wide confidence intervals, these studies did not show a significant increase in the risk of subtrochanteric or diaphyseal fracture with 3 to 4 years of use of alendronate or zoledronic acid or up to 10 years of alendronate use.
Subtrochanteric and Diaphyseal Fractures
In all three trials, there were 283 hip or femur fractures (Figure 1). Excluded from the review were 135 femoral-neck or subcapital fractures and 13 periprosthetic, pathologic, or high-energy fractures and 1 hip fracture with insufficient information to classify the location. Of the remaining 134 fractures, 12 were fractures of the subtrochanteric or diaphyseal femur that occurred in 10 women (for whom only one radiograph was available). With 12 fractures during 51,287 patient-years in the three studies, the combined rate was 2.3 per 10,000 patient-years.
In the FIT trial, we reviewed 31 fractures, of which a majority were either intertrochanteric or extended from the intertrochanteric to the subtrochanteric region (Figure 1). Only two fractures met the criteria for fracture of the subtrochanteric or diaphyseal femur (one in each study group), with an overall rate of 0.8 per 10,000 person-years. The relative hazard among subjects receiving alendronate, as compared with those receiving placebo, was 1.03 (95% confidence interval [CI], 0.06 to 16.46) (Table 1). Details regarding individual patients are shown in Table 2. Neither of the two women reported having received previous antiresorptive therapy. Levels of biochemical markers for serum C-terminal telopeptide of type 1 collagen (CTX) and serum procollagen type 1 N-propeptide (P1NP) for the women who received alendronate were below the lower limit of the premenopausal range, although the level of bone-specific alkaline phosphatase (BSAP) was within the premenopausal range. One woman was 69 years of age, and the other was 75. Neither reported a history or current use of corticosteroids or proton-pump inhibitors or any diseases affecting bone metabolism.
For the HORIZON-PFT trial, we reviewed 84 hip or femur fractures in 82 women. Five women had six fractures that met the location criteria for fracture of the subtrochanteric or diaphyseal femur; of these women, three were receiving zoledronic acid (2.8 per 10,000 person-years) and two were receiving placebo (1.9 per 10,000 person-years) (relative hazard for zoledronic acid, 1.50; 95% CI, 0.25 to 9.00) (Table 1). The mean age of the five women was 71 years and 2 years below the study mean of 73 years. In three women receiving zoledronic acid, one had simultaneous fractures of both femoral shafts, one had a transverse fracture but no beaking and no focal or generalized cortical thickening on radiography, and one had a spiral fracture on radiography, although the morphologic features were not described as "atypical" in case reports.
For these five fracture events, three events in women receiving zoledronic acid occurred after minimal trauma (a fall from standing height or less), and two events in women receiving placebo occurred after a fall from a stair, step, or curb. Two women (one receiving zoledronic acid and one receiving placebo) reported taking concurrent antiresorptive therapy (raloxifene and injectable calcitonin, respectively), and one woman (who had two simultaneous fractures) reported having previously taken a bisphosphonate (>4 years before baseline) and hormone-replacement therapy. No women reported previous use of corticosteroids. Two women reported having hip pain at some point before the fracture, and the woman with simultaneous fractures had reported "bone pain" at two previous annual visits. Biochemical markers were not assessed in these patients.
In the FLEX trial, we reviewed 19 hip or femur fractures, and 4 fractures (occurring in 3 women) met the criteria for fracture of the subtrochanteric or diaphyseal femur: 2 of 662 women receiving alendronate (6.3 per 10,000 patient-years) and 1 of 437 women receiving placebo (4.7 per 10,000 patient-years; relative hazard in the alendronate group, 1.33; 95% CI, 0.12 to 14.67). All women had previously taken alendronate in the FIT trial (mean duration at FLEX baseline, 5 years).19 The two women receiving alendronate had fractures more than 1000 days after randomization but had discontinued therapy 3 years earlier. None of the three women reported previous use of antiresorptive drugs or corticosteroids. One woman receiving alendronate (Subject 8 with two fractures) reported the use of a proton-pump inhibitor. In all three women with fractures in the location of interest, biochemical markers of bone turnover before fracture were substantially above the lower reference value for premenopausal women. The average age of the women (80 years) was about 7 years above the mean in the study. Three fractures occurred after a fall from standing height, while the fourth was due to trauma not associated with a fall.
We examined the incidence of fracture of the subtrochanteric or diaphyseal femur in three large, randomized trials: two that studied the use of oral alendronate (a placebo-controlled phase 3 trial and a randomized extension to 10 years of follow-up) and a placebo-controlled trial of zoledronic acid (infused annually). The trials used similar protocol-specified procedures for original collection and classification of fractures. We reviewed the fractures, using identical methods for the three studies, to identify and describe femoral-shaft fractures and assess atypical features if possible.
Evaluation of Three Trials
Fracture Intervention Trial (FIT)
In this trial, 6459 women over the age of 65 years with a bone mineral density T score of less than -1.6 were recruited16 and randomly assigned to receive daily alendronate (at a dose of 5 mg for 2 years and 10 mg thereafter) or placebo and were followed for 3 to 4.5 years.16,17,18 Relevant data included demographic characteristics, bone mineral density scores, the use of concomitant medications, previous osteoporosis therapies, adverse events, and levels of biochemical markers of bone turnover.
FIT Long-Term Extension (FLEX) Trial
After the FIT trial ended, women were offered open-label alendronate for 1 year, after which a randomized extension trial began.19 A total of 1099 women who were originally assigned to receive alendronate underwent a second randomization to receive alendronate at a daily dose of 5 mg (in 30% of subjects) or 10 mg (in 30% of subjects) or placebo (in 40% of subjects) and were followed for 5 more years. At baseline in the FLEX trial, the mean duration of alendronate therapy was 5 years; by the end of the trial, the mean duration was 10 years for women who were assigned to the continued alendronate group.
HORIZON Pivotal Fracture Trial
In the Health Outcomes and Reduced Incidence with Zoledronic Acid Once Yearly (HORIZON) Pivotal Fracture Trial (PFT), 7736 women were randomly assigned to receive an annual infusion of 5 mg of zoledronic acid or placebo and were followed for 3 years.20 Inclusion criteria included an age between 65 and 85 years and a bone mineral density T score of less than -2.5 or less than -1.5 in women with a vertebral fracture. Other information included previous osteoporosis medications and concurrent use of all prescription medications, bone mineral density score, and biochemical markers of bone turnover in approximately 20% of the subjects.
Fracture Reporting and Adjudication
The procedures for central adjudication of clinical fractures at the University of California, San Francisco (UCSF), were almost identical in the three studies. Subjects were specifically queried about fractures at study visits. With positive reports, investigators at the clinical site obtained documentation of all fractures, including those associated with trauma. The primary adjudication document was usually either a preoperative radiology report or the surgical medical record. Original radiographs were not generally obtained unless documentation was insufficient or inconclusive.
In the FIT and FLEX trials, hip fractures were classified as involving the femoral neck, the intertrochanteric region, or "miscellaneous hip." In the HORIZON-PFT trial, a fifth category of subtrochanteric fracture was added. Classification was based on the determination of community radiologists, as interpreted by the central adjudicator at UCSF and was not standardized across clinical centers. These procedures were developed to classify types of hip fractures, not to classify femoral-shaft fractures.
Review of Fractures
In 2008, after the publication of reports of atypical femoral-shaft fractures, we began a systematic review of fracture documentation. We used available radiographic and surgical reports to review fractures that originally had been classified as occurring in the hip (with the exception of those involving the femoral neck) and all other femur fractures. The review was performed by a radiologist who was not aware of study-group assignments and who specifically evaluated the location and any atypical features of fractures.
Location of Fractures
In our study, the region of interest was a fracture confined to the area below the lesser trochanter and above the distal metaphyseal flare (hereafter termed subtrochanteric and diaphyseal femur fractures). We excluded all periprosthetic fractures, pathologic fractures (i.e., those occurring in bone weakened by another disease process), and fractures associated with high-energy trauma. Other fracture categories included intertrochanteric only, intertrochanteric to subtrochanteric, and distal metaphysis.
Assessment of Atypia
For fractures meeting the criteria for location, we used radiographs (which were rarely available) for morphologic evaluation. Morphologic measurements included transverse or oblique versus spiral, abnormal cortical thickening (focal or generalized), cortical beaking, medial spiking, and any contralateral similarities. When radiographs were not available, atypical features in the radiology report were recorded.
The original studies were designed by the steering committees in collaboration with the sponsors (Merck for the FIT and FLEX trials and Novartis for the HORIZON-PFT trial). UCSF investigators were responsible for the collection and adjudication of fracture data for all three studies. All other aspects of data collection were the responsibility of UCSF investigators for the FIT trial and of the sponsors for the FLEX and HORIZON trials. In each study, investigators developed publication guidelines that specified full participation of academic investigators together with sponsors in the publication of results. The two lead academic authors designed the study we report here and wrote the initial draft of the manuscript with substantial input from the other authors. Data analyses were performed solely by members of the UCSF group, who had full access to the study data. Representatives of the sponsors reviewed the final drafts of the manuscript. All authors vouch for the completeness and accuracy of the data.
For each trial, we calculated the risk per 10,000 person-years according to treatment and the relative hazard of fracture of the subtrochanteric or diaphyseal femur for subjects receiving bisphosphonates, as compared with control subjects. We also tabulated selected characteristics for individual subjects, including age, trauma level, and the duration of and adherence to treatment. When available, we compared biochemical-marker values for subjects who were receiving treatment but before the occurrence of a fracture with reference ranges for premenopausal women.21
Supported by Merck and Novartis.
Dr. Black reports receiving grants (to UCSF) from Merck, Novartis, Amgen, and Roche and travel reimbursements from Merck and Novartis; Dr. Genant, receiving consulting and lecture fees and grant support from Merck and Novartis; Ms. Palermo, receiving support from Nycomed for data analysis; Dr. Eastell, serving on advisory boards for Procter & Gamble and the Medical Research Council and receiving consulting fees from Amgen, AstraZeneca, GlaxoSmithKline, Medtronics, Nastech, Nestle, Fonterra Brands, Novartis, Ono Pharma, Osteologix, Pfizer, Lilly, Sanofi-Aventis, Tethys, Unilever, Unipath, Inverness Medical, Ortho Clinical Diagnostics, OSI Prosidion, and Roche, lecture fees from Takeda, Lilly, Amgen, National Osteoporosis Foundation, Procter & Gamble, and GlaxoSmithKline Nutrition, and travel expenses from Novartis; Dr. Bucci-Rechwig, being a full-time employee of and having an equity interest in Novartis; Dr. Cauley, receiving consulting fees from Novartis; Drs. Santora and de Papp, being full-time employees of and having equity interests in Merck; and Dr. Bauer, receiving consulting fees from Amgen and lecture fees from Merck. No other potential conflict of interest relevant to this article was reported. Disclosure forms provided by the authors are available with the full text of this article at NEJM.org.