Abstract
Studies suggest an increasing occurrence of atypical femoral fractures with the use of bisphosphonates.
To examine whether the use of bisphosphonates increases the risk for atypical fractures.
Systematic review and meta-analysis.
Literature search of MEDLINE, Embase and Cochrane CENTRAL (1948–June 2013).
Selection criteria: (i) randomized controlled trial or an observational study, (ii) evaluated bisphosphonate therapy versus no treatment and (iii) reported an incidence of subtrochanteric or diaphyseal fracture individually, or a composite of both. Two independent investigators completed study selection, data extraction and validity assessment. The Cochrane Risk of Bias Tool was used to assess the quality of included studies.
Ten (n = 658497) studies were included in the meta-analysis which demonstrated a statistically significant increased risk of subtrochanteric or diaphyseal fracture with bisphosphonate use [adjusted odds ratios (AOR) = 1.99, 95% confidence intervals (CI)= 1.28–3.10] with I2 = 84.3% (95% CI = 73.5%–89.5%) and Egger P = 0.01. Subtrochanteric fractures showed an AOR = 2.71 (95% CI = 1.86–3.95) with I2 = 83.6% (95% CI = 64.3%–90.3%) and Egger’s P = 2.29. Diaphyseal fractures had an AOR = 2.06 (95% CI = 1.70–2.50), I2 = 29.7% (95% CI = 0%–73.7%) and Egger’s P = 1.22.
Results suggest there is an increased risk for atypical fractures associated with bisphosphonates and raises awareness to the potential complications related with bisphosphonates. These findings warrant the comprehensive evaluation of patients before initiating bisphosphonate therapy and highlights the need for additional medical decision analyses in future studies to compare the benefit over potential harms of bisphosphonate therapy.
Introduction
Osteoporosis is a disease characterized by low bone mass resulting in bone deterioration and impaired skeletal strength. It is recognized as the primary cause of fractures and ~10 million individuals over the age of 50 are affected by it in the United States (1). The World Health Organization (WHO) estimates that 1.66 million hip fractures occur worldwide annually and this number is expected to increase by 4-fold by 2050 (2). While the incidence of typical fractures has decreased in postmenopausal women taking bisphosphonates, the occurrence of atypical fractures have increased over the last decade (3). Therapy with bisphosphonates is considered first line for preventing and treating osteoporosis and has been shown to reduce the incidence of typical hip fractures in large multinational randomized controlled trials (RCTs (4–7)). Typical hip fractures often occur in the femoral neck and intertrochanteric regions whereas atypical fractures occur in the subtrochanteric and diaphyseal regions. Patients with atypical fractures present with radiologic features of thickening at the femoral cortex, presence of a transverse fracture and a cortical break (8).
There are numerous case reports and studies of atypical fractures associated with the long-term use of bisphosphonates, however, most are observational in nature and have conflicting results (9–12). In a retrospective case–control study of postmenopausal women, 15 of 41 subtrochanteric/femoral shaft fracture cases were found to be taking a bisphosphonate, a significantly greater proportion compared with 9 out of 82 intertrochanteric/femoral neck fractures [odds ratio (OR) 4.44, 95% confidence intervals (CI) 1.77–11.35, P = 0.002 (9)]. However, some registry-based analyses conclude that there is only a small absolute risk of atypical fracture occurrence (10–12). Moreover, a secondary analysis on three large RCTs: the Fracture Intervention Trial (FIT), the FIT Long-Term Extension (FLEX) trial and the Health Outcomes and Reduced Incidence in Alendronic Acid Once Yearly (HORIZON) Pivotal Fracture Trial (PFT) concluded that the occurrence of atypical fractures was rare, even among women who had been treated with bisphosphonates for as long as 10 years, though the authors acknowledged that the study was underpowered (13).
Observational studies and large trials with small event rates yielded conflicting results, so a systematic review of both RCTs and observational studies was conducted to examine whether the use of bisphosphonates increases the risk for atypical fractures.
Methods
Search strategy
A comprehensive systematic literature search of Pubmed, Embase and the Cochrane Library was conducted from the inception of each database to June 2013. A search strategy with keywords and medical subject headings related to bisphosphonate therapy and atypical fractures were used to identify relevant articles. No restrictions to languages were applied. In addition, a manual search of included references was performed to identify any additional relevant studies. Finally, authors were attempted to be contacted when appropriate. No additional data were received. For further details of the search strategy, see Supplemental Appendix A (14).
Study selection and data abstraction
Studies were included in the analysis if: (i) it was either an RCT or an observational study which adjusted for confounders, (ii) evaluated bisphosphonate therapy versus no treatment and (iii) reported an incidence of either subtrochanteric (identified through radiographic reports of femur fractures below the lesser trochanter or ICD-9 codes) or diaphyseal fracture (identified through radiographic reports of femur fractures above the distal metaphysis or ICD-9 codes) individually, or a composite of both subtrochanteric fracture and diaphyseal fractures. We restricted our eligibility of observational studies which attempted to adjust for confounders because observational studies are susceptible to biases. Consequently, by only extracting and analyzing results from observational studies which attempted to adjust for confounders we hoped to maximize the integrity of our results.
Two independent investigators completed study selection, data abstraction and validity assessment for all studies using a standardized data abstraction tool. Disagreements were resolved by discussion. Pertinent data from each study was recorded independently by two investigators with an emphasis on: study design characteristics, sample size, inclusion and exclusion criteria, baseline population characteristics, type and duration of bisphosphonate therapy, study control for potential confounders (i.e. randomization, multiple regression, matching) and patient demographics. The outcome of interest was the occurrence of subtrochanteric or diaphyseal fractures reported either individually or as a composite outcome.
Of note, the original publication for Park-Wyllie (12) reported their estimates as three separate cohorts (short-term use of bisphosphonates, moderate term use, and long term use compared to transient use). As such, we extracted the data as given by the original publication and reported it as three separate entries.
Quality rating
Quality rating of studies was performed by two investigators using methodology recommended by the Evidence-based Practice Center Method Guide (15,16). Each study was assessed for the following individual criteria: comparable study groups at baseline, detailed descriptions of study outcomes, blinding of outcome assessors, intent to treat analysis, description of participation withdrawals (percent follow-up) and potential conflict of interest. Additionally, RCTs were assessed for randomization technique and allocation concealment. Observational studies were assessed for sample size, participant selection method, exposure measurement method, potential design biases and appropriateness of methods of control for confounding. Studies were then given an overall ranking of ‘good’, ‘fair’ or ‘poor’.
Data synthesis and analysis
The adjusted estimated effects [OR, hazards ratio (HR) or relative hazard (RH)] and associated variances were pooled using a DerSimonian and Laird random-effects model (a variation on the inverse variance method, which incorporates an assumption that the different studies are estimating different, yet related, treatment effects) was used (17,18). Results are reported as adjusted odds ratios (AOR) and accompanying 95% CI. Heterogeneity was evaluated with I2 statistic (19). Value ranges between 0% and 100% with values of 25%, 50% and 75% representing low, moderate and high levels of statistical heterogeneity, respectively. A low level of heterogeneity is desired. Publication bias was assessed using the Egger’s P value (14). The statistical analysis was performed by using StatsDirect software version 2.7.9. A P value of <0.05 was considered statistically significant for all analyses.
Due to the limited number of studies available, the decision to pool RCT and adjusted observational data together was made. The observational study results are strengthened by their reporting of adjusted effect sizes, thereby decreasing the chance for confounding to occur. Of note, the analyses evaluating subtrochanteric fractures alone or diaphyseal fractures alone only included data from observational studies.
Results
Our search strategy yielded 3232 citations and after screening for inclusion and exclusion of citations, 10 studies (10–13,20–25) (n = 658497 patients) met all inclusion criteria and qualified for full-text screning (Fig. 1). Articles that did not meet the inclusion criteria were excluded for the following reasons: not an RCT or observational study, did not evaluate bisphosphonate therapy versus no treatment or did not report the incidence of subtrochanteric or diaphyseal fracture individually, or as a composite of subtrochanteric fracture and diaphyseal fracture as an endpoint.
PRISMA diagram. Flowchart selection processes of excluding articles based on inclusion criteria.
PRISMA diagram. Flowchart selection processes of excluding articles based on inclusion criteria.
Study and patient characteristics
The 10 articles included in this study were all published between 2009 and 2013. Nine of which were observational trials (three case control, five cohort, one nested case control) and one RCT. Mean age identified in each study was between 70 years old and 83 years old. The mean percent of females who participated in each study was between 83% and 100%. In total, the meta-analysis evaluated 658 497 unique patients. Full details of patient characteristics are presented in Table 1 and Supplementary Table S1 (14). Studies were given an overall ranking of ‘good’, ‘fair’, or ‘poor’ based on the grading criteria mentioned above. Of the nine observational trials included in this study, seven were given an overall ranking of ‘fair’, and two were given an overall ranking of ‘good’. The RCT included in this study was assigned an overall ranking of ‘good’.
Baseline characteristics of studies and their patients included in meta-analysis
| Study | Design | Intervention | Participants: age (mean), gender, BMI | Previous fall and fracture (%) |
|---|---|---|---|---|
| Erviti (25), N = 264 | CCS | Bisphosphonate | 82.2, 100% female, 29.4 | 20.5 |
| Control | 82.2, 100% female, 29.1 | 8.2 | ||
| Napoli (22), N = 9704 | PCS | Bisphosphonate | 71.7, 100% female, 26.4 | 36.9 |
| Atypical fx | 74.2, 100% female, 26.8 | 44.4 | ||
| Meier (23), N = 477 | CCS | Bisphosphonate | —, 92.3% female, — | — |
| Control | —, 72.8% female, — | — | ||
| Hsiao (24), N = 11278 | RCS | Alendronate | 77.3, 100% female, — | 9.9 |
| Untreated control | 77.5, 100% female, — | 9.9 | ||
| Park-Wyllie (12), N = 4296 | NCCS | Case | 83a, 100% female, — | 67 |
| Control | 83a, 100% female, — | 5.5 | ||
| Schilcher (11), N = 322 | CCS | Bisphosphonate | —, 100% female, — | 16.9 |
| Control | —, 100% female, — | 36.5 | ||
| Vestergaard (20), N = 414,245 | CS | Bisphosphonate | —, —, — | — |
| Control | —, —, — | — | ||
| Abrahamsen (10), N = 197835 | CS | Alendronate | 69.8, 82.8% female, — | 29.7* |
| Non-alendronate | 69.8, 82.8% female, — | 16.1 | ||
| Black (13a), N = 6459 | RCT | Alendronate 5 or 10mg PO QDay | —, 100% female, 24.9–25.5 | 42.6 |
| Placebo | —, 100% female, 24.9–25.0 | 42.2 | ||
| Black (2010b), N = 1099 | RCT | Alendronate 5 or 10mg PO QDay | —, 100% female, 25.8–25.9 | 22.8§ |
| Placebo | 73.7, 100% female, 25.8 | 24.2 | ||
| Black (2010c), N = 7765 | RCT | Zoledronic acid 5mg IV yearly | 73.1, 100% female, 25.4 | — |
| Placebo | 73.0, 100% female, 25.1 | — | ||
| Abrahamsen (21), N = 15561 | CS | Alendronate | 73.1, 90.2% female, — | 100 |
| Control | 73.1, 90.2% female, — | 100 | ||
| Abrahamsen (2009b), N = 10284 | CS | Etidronate | —, —, — | 100 |
| Control | —, —, — | 100 |
| Study | Design | Intervention | Participants: age (mean), gender, BMI | Previous fall and fracture (%) |
|---|---|---|---|---|
| Erviti (25), N = 264 | CCS | Bisphosphonate | 82.2, 100% female, 29.4 | 20.5 |
| Control | 82.2, 100% female, 29.1 | 8.2 | ||
| Napoli (22), N = 9704 | PCS | Bisphosphonate | 71.7, 100% female, 26.4 | 36.9 |
| Atypical fx | 74.2, 100% female, 26.8 | 44.4 | ||
| Meier (23), N = 477 | CCS | Bisphosphonate | —, 92.3% female, — | — |
| Control | —, 72.8% female, — | — | ||
| Hsiao (24), N = 11278 | RCS | Alendronate | 77.3, 100% female, — | 9.9 |
| Untreated control | 77.5, 100% female, — | 9.9 | ||
| Park-Wyllie (12), N = 4296 | NCCS | Case | 83a, 100% female, — | 67 |
| Control | 83a, 100% female, — | 5.5 | ||
| Schilcher (11), N = 322 | CCS | Bisphosphonate | —, 100% female, — | 16.9 |
| Control | —, 100% female, — | 36.5 | ||
| Vestergaard (20), N = 414,245 | CS | Bisphosphonate | —, —, — | — |
| Control | —, —, — | — | ||
| Abrahamsen (10), N = 197835 | CS | Alendronate | 69.8, 82.8% female, — | 29.7* |
| Non-alendronate | 69.8, 82.8% female, — | 16.1 | ||
| Black (13a), N = 6459 | RCT | Alendronate 5 or 10mg PO QDay | —, 100% female, 24.9–25.5 | 42.6 |
| Placebo | —, 100% female, 24.9–25.0 | 42.2 | ||
| Black (2010b), N = 1099 | RCT | Alendronate 5 or 10mg PO QDay | —, 100% female, 25.8–25.9 | 22.8§ |
| Placebo | 73.7, 100% female, 25.8 | 24.2 | ||
| Black (2010c), N = 7765 | RCT | Zoledronic acid 5mg IV yearly | 73.1, 100% female, 25.4 | — |
| Placebo | 73.0, 100% female, 25.1 | — | ||
| Abrahamsen (21), N = 15561 | CS | Alendronate | 73.1, 90.2% female, — | 100 |
| Control | 73.1, 90.2% female, — | 100 | ||
| Abrahamsen (2009b), N = 10284 | CS | Etidronate | —, —, — | 100 |
| Control | —, —, — | 100 |
BMI, body mass index; CS, cohort study; CA, cancer; CCS, case–control study; NCC, nest case–control study; PCS, prospective cohort study; PPI, proton pump inhibitor; RCS, retrospective cohort study; RCT, randomized control trial; RD, renal disease; —, not reported.
aMedian.
*P < 0.001.
Quantitative data synthesis
Upon meta-analysis, bisphosphonate use was associated with a statistically significant increased risk of developing the composite outcome of subtrochanteric or diaphyseal fracture (AOR = 1.99, 95% CI= 1.28–3.10) (Fig. 2) when RCTs and observational trials were combined. A high level of heterogeneity [I2=84.3% (95% CI = 73.5%– 89.5%)] as well as publication bias (Egger’s P = 0.01) was detected for the composite outcome when RCTs and observational trials were combined. Likewise, a subanalysis of the composite outcome of subtrochanteric or diaphyseal fractures using only observational studies yielded similar results [AOR = 2.08, 95% CI: 1.29–3.35; I2=88.5 (95% CI = 80.4%–90.3%)].
Composite forest plot of the composite outcome and subtrochanteric or diaphyseal fractures.
Composite forest plot of the composite outcome and subtrochanteric or diaphyseal fractures.
Bisphosphonate therapy was also associated with a statistically significant risk of developing subtrochanteric fractures (AOR = 2.71, 95% CI: 1.86–3.95) (Fig. 2). A high level of heterogeneity [I2 = 83.6% (95% CI = 64.3%–90.3%)] was identified, however, publication bias was not detected (Egger’s P = 0.14) for the outcome of subtrochanteric fracture. An analysis evaluating the risk of diaphyseal fractures associated with bisphosphonate use also proved to be statistically significant (AOR = 2.06, 95% CI: 1.70–2.50) (Fig. 2) with a medium level of heterogeneity [I2 = 29.7% (95% CI = 0%–73.7%)]. Publication bias was not detected (Egger’s P = 0.29) for the outcome of diaphyseal fracture. The authors of this study rank the overall strength of evidence presented here as being moderate based on the Evidence-based Practice Centers (EPCs) method guide (14). Consequently, the authors’ are moderately confident that the estimate of effect presented here lies close to the true effect for the evaluated outcomes (14).
Discussion
We examined whether bisphosphonate therapy increased the risk for developing atypical fracture in 658497 patients with osteoporosis from 10 clinical studies. Systematic evaluation of these conflicting studies indicated that the composite risk of developing either a subtrochanteric or diaphyseal fracture was statistically increased (AOR = 1.99, 95% CI = 1.28–3.10). However, the results were associated with a high level of heterogeneity [I2 = 84.3% (95% CI = 73.5%–89.5%)] and publication bias (Egger’s P = 0.01). Similar results were obtained when bisphosphonate therapy and the risk for either subtrochanteric of diaphyseal fractures were evaluated individually (subtrochanteric AOR = 2.71, 95% CI = 1.86–3.95; diaphyseal AOR = 2.06, 95% CI = 1.70–2.50). A high level of heterogeneity was associated with subtrochanteric fractures [I2 = 83.6% (95% CI = 64.3%–90.3%)], whereas a medium level of heterogeneity was associated with diaphyseal fractures [I2=29.7% (95% CI = 0%–73.7%)]. Outcomes for either subtrochanteric or diaphyseal fractures were not associated with publication bias.
According to the American Society for Bone and Mineral Research (ASBMR) the exact pathogenesis of atypical fractures related to bisphosphonate use remains unclear (26). The increased risk of atypical fracture may be explained by the over suppression of bone turnover rate or an accumulation of microdamage (27). The normal physical stresses in activities of daily living contribute to microdamage in the bone, which typically trigger the initiation of bone remodeling (24). With the inhibition of bone resorption by bisphosphonate use, the microdamage may not be adequately repaired and allowed to accumulate (24). This accumulation then leads to brittle bone and can increase risk of fracture with low-impact activity (24).
Previous studies have concluded mixed results about the risks associated with atypical fractures and bisphosphonate therapy. Although the results from this meta-analysis indicate a statistically significant increase in atypical fractures with bisphosphonate therapy, it does have its limitations. A high level of statistical heterogeneity and the presence of publication bias are limitations of this study. Various formulations of bisphosphonate therapy combined with different comorbidities within each study group could have contributed to the clinical heterogeneity present in this study. Furthermore various durations of follow-up and different study designs within each article could have contributed to the methodological heterogeneity of this study. We could not ascertain whether patients with shorter exposure to bisphosphonates were less susceptible to the risk of atypical fracture, in the absence of patient-level data. However, a number of studies suggest that a longer duration of use of bisphosphonates is associated with higher risk of atypical fracture (12,24). It should also be noted that whether the results obtained in this study is due to a class effect or any specific bisphosphonate drug could not be answered.
The results of this meta-analysis demonstrated a statistically significant increased risk of subtrochanteric or diaphyseal fracture with bisphosphonate therapy and is clinically significant because clinicians should be aware that such an association exists and monitor patients for possible signs and symptoms of atypical fracture while the patient is on bisphosphonate therapy. Despite the widespread use of bisphosphonate therapy, there are still many questions surrounding its use that are yet to be answered. Future research should be designed to determine if there are specific diseases or patient characteristics which contribute to the increased risk of atypical fracture with bisphosphonate use and to clarify its mechanism. In the meantime, this study should serve as a reminder to clinicians about the importance of discussing with their patients about the risks associated with bisphosphonate use and atypical fractures.
Declaration
Funding: none.
Ethical approval: not required at our institution for systematic reviews.
Conflict of Interest: none.
Supplementary material
Supplementary material is available at Family Practice online.
