Bisphosphonates (BP) have been available for more than 20 years, possess well-demonstrated antifracture efficacy, and are currently the first-choice osteoporosis treatment. However, some adverse effects have been reported over the past decade, including widespread concern about the incidence of a new type of fracture with characteristic location and radiological appearance in patients on long-term BP (LTB) therapy.1–10
These so-called atypical femoral fractures (AFF) have created the increasing perception of potential harm to bone tissue when chronically exposed to these drugs. They do not meet the classic profile of osteoporotic fragility fractures, have been associated with excessive suppression of bone turnover, and affect sites that are not usually affected by osteoporotic fractures, specifically the subtrochanteric femur.11, 12 Moreover, in these patients all the classical parameters to evaluate the fracture risk seem to be unreliable, especially bone mineral density (BMD) measurements. Actually, AFF patients usually show BMD values in the osteopenia or even normal range and therefore could not be classified as high risk according to densitometric criteria.13, 14
One possibility, given the normal amount of bone mineral in these patients, is the presence of a disorder in the intrinsic material properties of bone tissue. The usual thickening of the cortices in these fractures, despite which a fracture occurs, strongly suggests a deterioration of the material properties of bone at a tissue level. Two recent publications show abnormal nanoindentation in cases of long-term bisphosphonate treatment15 and patients with severely suppressed bone turnover.16
Recently introduced is reference point indentation, a new tool that permits direct in vivo measurement of the bone material properties of patients with these fractures, integrating all their components both at nano and micro scale. Moreover, the technique is suitable for clinical use. We have previously tested patients with hip fracture and controls without fracture using in vivo microindentation, a technique that discriminates between fracture cases and controls, yielding an area under the curve (AUC) of 93%,17 whereas for BMD ranges from 72% to 80%.18–20 Our current aim is to measure the microindentation values in patients with atypical fractures associated with long-term treatment with BP, comparing their results with those of patients with no fractures, patients on LTB therapy without incident fractures, and patients with “typical” osteoporotic hip fractures.
Materials and Methods
Four study groups were defined and recruited between 2008 and 2011 in the Hospital del Mar, Barcelona, and Hospital Reina Sofia, Cordoba, in Spain. Inclusion criteria for patients in each group are detailed below. Exclusion criteria for all groups were previous treatment for osteoporosis (except for AFF and LTB patients) and all-cause secondary osteoporosis (corticosteroids use, a previous diagnosis of advanced renal or liver disease, neoplasia, malabsorption, thyroid or parathyroid disorder, immobilization) or inability to provide consent.
Atypical femoral fracture
AFF were diagnosed after established clinical criteria: fractures in the area below the lesser trochanter and above the distal metaphyseal flare, with a simple transverse or oblique (≤30°) fracture with breaking of the cortex and diffuse cortical thickening of the proximal femoral shaft without previous trauma.12, 21 A full clinical and radiographic assessment with special focus on the history of BP use and other risk factors associated with AFF12 was obtained.
Typical hip osteoporotic fracture
Eligible cases were identified from the orthopedic surgery ward with typical hip osteoporotic fracture, which included pertrochanteric fractures and subcapital fractures. These cases had no history of bisphosphonate or any other antiosteoporosis therapy.
Long-term treatment with bisphosphonates and no incident fractures
Outpatients with 5 or more years of treatment with BPs and no incident fractures were included in this group. Thoracic and lumbar lateral radiography validated the absence of incident subclinical vertebral fracture during the treatment period after the index fracture.
Control group of nonfracture individuals
Postmenopausal women who visited our outpatient clinics or were admitted to the Internal Medicine ward for acute illness and who had no history of fractures, no previous use of antiosteoporosis drugs, and no other cause of bone disease constituted the study controls. No previous history of fracture was required, and a thoracic and lumbar lateral X-ray was obtained to confirm the absence of subclinical vertebral fracture.
Bone microindentation testing (BMT)
The BMT was performed using BioDent (ActiveLife Tech, Inc., Santa Barbara, CA, USA). The complete protocol was previously described.17 In brief, after local anesthesia, the periosteum is scratched and a probe assembly placed on the anterior face of the mid-tibia performs measurements. A 20-cycle indentation at 11N force is performed and the average value of five measurements is recorded. The indentation distances are analyzed by a specific software and three parameters are obtained to use as outcome variables: indentation distance increase (IDI) between the first and last indentation cycle; total distance between the bone surface and the last indentation cycle (Total ID); and creep indentation distance (Creep ID), the progressive indentation distance during the stable force phase of the first indentation cycle at the maximum 11N force.17 The microindentation testing was done after as soon as possible after the fractures. This was a few days for typical fracture cases. Because of the difficulties in locating and transporting the atypical fracture cases, this was typically a few weeks or months for these cases.
Bone mineral density measurement
Dual-energy X-ray absorptiometry (DXA) with a Hologic QDR 4500 SR Bone Densitometer (Hologic, Inc., Waltham, MA, USA) was used to measure BMD at the nonfractured hip in all atypical fracture cases, in 8 controls, in 9 typical fractures, and in all LTB patients.
The Committee on Human Subjects Research at the Municipal Institute of Medical Research (IMIM; Parc de Salut Mar, Barcelona, Spain) approved the study, and written informed consent was obtained from each participant after a full explanation of the purposes and characteristics of the study.
Subject characteristics and microindentation variables are expressed as mean (SD) or percentage. Normality of continuous variables was assessed by Q-Q plots. Analysis of covariance was used to obtain and compare age-adjusted means.
A p value of less than 0.05 was considered to indicate significance. All reported p values are two-sided. The analysis was performed using SPSS for Windows, version 15.0, 2006 (SPSS Inc., Chicago, IL, USA).
The study included 70 women (6 cases of AFF, 38 with typical osteoporotic hip fracture [26 pertrochanteric fractures and 12 subcapital fractures], 6 with LTB use and no incident fracture, and 20 controls). Table 1 shows the main clinical characteristics and BMD results of the enrolled patients. The control individuals without fractures were significantly younger than the women in the fracture groups. The average BP exposure was 5.5 years (range 5 to 12 years) for the AFF group and 5.4 years (range 5 to 8 years) for those with LTB use and no fracture. Among the patients with atypical fracture, all major clinical characteristics defined by the ASBMR task force were present.12 Four of those six cases also presented minor clinical features (localized periosteal reaction and bilateral symptoms). One patient had sustained a previous contralateral atypical femoral fracture. None of the AFF cases had a history of exposure to glucocorticoids. There were no differences in BMD at the lumbar spine or total hip between the LTB group and controls or AFF, whereas AFF showed significantly higher BMD at the hip than typical osteoporotic fractures.
Table 1. Main Clinical Properties and Bone Mineral Density (BMD) Results of the Enrolled Patients (Only Significant Differences Shown)
Long-term BP (LTB)
Atypical fractures (AFF)
Total ID = total indentation distance; IDI = indentation distance increase); long-term BP = long-term bisphosphonate treatment; N/A = not available.
After adjusting by age, IDI values were significantly different between AFF and controls (19 ± 3 versus 13 ± 2, p < 0.05, mean ± SD) and between typical fractures and controls (18 ± 5 versus 13 ± 2, p < 0.05).
Likewise, Total ID values differed significantly between AFF and controls (46 ± 4 versus 36 ± 6, p < 0.001) and between typical fractures and controls (47 ± 12 versus 36 ± 6, p < 0.001) (Fig. 1 and Table 1).
Among the subgroups of typical osteoporotic fractures, both pertrochanteric fractures and subcapital hip fractures presented no differences in Total ID or in IDI.
Finally, there were no significant differences in the Creep ID value between groups.
In the AFF patients, where DXA was measured, BMD values were in the range of osteopenia at lumbar spine in 5 of the 6 patients and only one was within the limit value for osteoporosis (Table 2). With respect to the four total hip measurements (one patient had had a bilateral hip replacement), three were normal values and one mild osteopenia.
Table 2. BMD Values in Patients With Atypical Femoral Fracture
BMD lumbar spine (g/cm2)
T-score lumbar spine
BMD total hip (g/cm2)
T-score total hip
BMD = bone mineral density; f = bilateral hip prosthesis replacement; N/A = not available.
The current study used in vivo microindentation to assess the material properties of bone tissue in patients with atypical fractures after long-term treatment with BP. The indentation distance, an inverse estimate of crack growth toughness, was comparable with that observed in patients with severe osteoporosis and typical hip fracture, even though AFF cases had BMD levels in the osteopenia range. In patients without AFF after LTB exposure, the indentation values were between those of fracture groups and controls. To the best of our knowledge, this is the first in vivo study of bone material properties at a tissue level in these groups of patients.
Atypical femoral fractures have characteristics very distinct from “typical” fragility fractures.22 They occur in the subtrochanteric femur, an anatomical region that contains the strongest parts of this bone and is unlikely to fracture after a low-energy trauma, even in advanced osteoporosis.23 Their association with LTB has raised the concern of a paradoxical negative effect of these drugs on bone tissue properties.24–26 The rationale for this deleterious action is that neither the normal or near-normal BMD values nor the thickening of the cortices reported in most of these cases would justify such a severe fracture. Moreover, the healing process is much longer and more problematic than for the typical osteoporotic fracture. Therefore, a negative effect of bisphosphonates on intrinsic properties of the bone tissue (a major component of bone quality) has been strongly suggested.
The obvious question is why all patients treated with BP do not present this impairment in bone properties. The number of patients treated without apparent complication is much larger than the number of observed AFF.2, 3, 8, 11, 27–29 Data from a case-control study showed a prevalence of atypical femoral fractures as low as 1.1% of all femoral fractures and, although more frequent in BP users, they also occurred in patients never treated with BP.30 Therefore, this problem seems to be restricted to a small minority of treated patients by an undetermined mechanism, although several hypothetical explanations have been suggested.27
Nonetheless, the main clinical problem is to clarify if this is a generalized undesirable effect of these drugs on bone tissue material properties or a special idiosyncratic condition in a small subset of patients.15 Bala and colleagues reported an increased degree of mineralization associated with lower crystallinity in trabecular bone in LTB-treated postmenopausal women. Moreover, microhardness and elastic modulus were decreased by nanoindentation measurement. They conclude that BPs alter the quality of the bone matrix and compromise micromechanical properties. In another study using nanoindentation, Tjhia and colleagues16 concluded that patients with AFF and severely suppressed bone turnover had greater resistance to plastic deformation at the cortical level.
Our analysis differs from these two studies in several key aspects. We measured the material properties in vivo and in a weight-bearing cortical bone, a more comparable region to where the AFF would occur. More important, microindentation differs from nanoindentation in the measured target. Nanoindentation is performed at the bone structural unit (BSU) level and therefore is restricted to an individual bone package unit with particular conditions of remodeling, age, collagen maturity, and crystallinity. Microindentation integrates the overall components of bone tissue, both at the nano and micro level, which contribute to overall mechanical competence. It captures the levels of bone porosity and heterogeneous osteons, the relationship between bone tissue components and increased bone stress, and the interfaces between osteons, microdamage, mineral, collagen, noncollagen proteins, and other components. Therefore, in a single measurement, microindentation assesses the capability of all these elements to dissipate energy in response to a mechanical challenge.
Accordingly, we entertain the hypothesis that microindentation induces the separation of mineralized collagen fibrils, the intimate mechanism of initiation, and the propagation of cracks evolving to fracture.17 In fact, when analyzed by electronic microscopy, our technique opens microcracks that can be superimposed on those observed in experimental bone fractures;17, 31, 32 the more microcracks the indenter opens, the more fragile is the bone and the higher are the indentation values. This effect is much closer to the actual conditions of fracture.
As a result, our data suggest that LTB therapy has a distinct effect in AFF patients. A potential interfering effect of glucocorticoids, usually associated with AFF, plays no role in our patients because they have not been exposed to these components. Despite the relatively high BMD levels in our AFF cases, bone material properties at the tissue level are similar to that observed in typical osteoporotic fractures. In contrast, patients on LTB without AFF are similar to that for nonfracture controls, suggesting that the effect of the drugs on the bone tissue is not negative in the average patient. Although their values were not statistically different from AFF and typical hip fractures, we believe that this is because of a limited statistical power.
Other considerations besides the epidemiological data support the hypothesis that the vast majority of the LTB-treated population is not at increased risk for AFF even after long periods of treatment. The observed cortical thickening has been suggested as a compensatory mechanism opposing bone tissue properties deterioration,33, 34 a phenomenon not typically seen after long treatments with alendronate.35 Moreover, AFF has been observed in patients not treated at all with BP36 and in monogenic diseases such as pycnodysostosis.37 Furthermore, there are no differences between typical and AFF in spite of the fact that AFF cases have BMD values in the range of osteopenia. Considering all these findings, which are consistent with our data, it may be suggested that the group of patients with atypical fractures has some underlying condition of the bone that impairs its material properties, its response to the drug, or both. It has been suggested that the atypical fractures are a phenotype associated with an underlying genetic condition that suffers a clear alteration in material properties under the effect of BP treatment, which leads to these fractures.38, 39
Our results also have a potential future implication for clinical practice because BMD monitoring does not detect patients on bisphosphonates at risk of AFF. Alerting clinical changes, mainly local pain and, if explored, radiographic or scintigraphic alterations, occur when the bone damage in the subtrochanteric region is already established. Therefore, after 3 to 5 years on treatment, when the question arises of continuing or stopping the therapy, no clinical data are really available to inform medical judgment. If our results are further replicated and prospectively demonstrated, bone material properties testing by microindentation might be a method to decide whether the patient is not showing biomechanical improvement despite BMD increase or if tissue properties are progressively improving, which would reflect a continuous positive effect of the drug that can be further increased.
Our work has both strengths and weaknesses. Microindentation makes a direct measurement of crack growth toughness at the tissue level, is feasible in vivo, and appears to be suitable for clinical use. The technique takes no more than 5 minutes and is reproducible and completely painless. Given that actual microscopic fractures are produced, the technique is supposed to directly assess the fracture propensity of bone. There is no other technique currently available to directly measure the intrinsic bone tissue “quality” without invasive sampling. However, some limitations should be acknowledged. First, the experience with the technique is still very limited and the results, therefore, must be considered as preliminary. Further replication of our results and validation of the value of the technique in other series of AFF, long-term bisphosphonate exposure as well as other clinical situations are needed. Moreover, the low incidence of atypical fractures makes it very difficult to collect a large number of cases. This limited sample size gives a low statistical power and some differences could have been missed, which again raises the need for wider series and replication. Furthermore, we cannot exclude some preexisting idiosyncratic problem in bone material properties in the AFF cases because there are no pre-BP baseline measurements in the treated groups. Similarly, we can only indirectly imply the positive effect of the treatment in our LTB patients because of the lack of baseline assessment.
The very recent development of bone material property testing with microindentation makes some of these limitations unavoidable. Likewise, the practice of microindentation is still limited to a few centers, and wider experience is needed to further validate its performance in clinical assessment. Furthermore, microindentation measurements are obtained in cortical bone, a compartment that only recently has been considered a key factor in fragility fractures;35 notably, AFF cases suffer a problem precisely located in cortical femur. If cortical bone is affected by LTB treatment, it seems plausible that the tibial measurement would reflect the bone material properties in the subtrochanteric-diaphysis region where the fracture occurs.
In summary, bone material properties at a tissue level, as measured by microindentation, is deteriorated in patients with AFF, well beyond what BMD indicates. This deterioration is similar to that for classical fragility fractures of the hip; no significant differences in material properties measurements were seen between the patients with typical and atypical fracture, but both were significantly different from controls without fracture of the hip, whereas LTB values were generally in between (not statistically different though). Our results suggest that a general, intrinsic effect of BP causing this decrease in tissue properties seems unlikely because this decrease was not observed in patients without AFF after long-term treatment with these drugs. There were trends but not significant differences between patients on long-term bisphosphonates and the other groups. Further studies are needed to understand the paradoxical effect of increased BMD and decreased bone material properties at the tissue level in patients with AFF.
PH is a member of Active Life Scientific, which sells the Biodent product line of RPI instruments for research use only at present. If the Biodent or future RPI instruments from Active Life Scientific have a future clinical application, this author could benefit financially.
We thank the Fondo de Investigaciones Sanitarias (PI07/90912) and the RETICEF (RD06/0013/1009) of the Instituto Carlos III (fondos FEDER). Carlos II Health Institute, Science and Innovation Ministry and NIH grant RO1 GM065354 are also acknowledged. The authors thank Elaine M Lilly, PhD, for helpful advice and critical reading of the manuscript.
Authors' roles: Study design: ADP and RGF. Study conduct: RGF, ADP, and XN. Patient data collection: LM, ET, JMQ, LP, and RGF. Data analysis: XN, GY, and RG. RGF and ADP performed the technique, and RGF, ADP, XN, NGG, GY, and PKH interpreted the data and discussed the results. Statistical analyses: ADP and RGF. Writing and drafting manuscript: ADP and RGF. Revising the manuscript critically for important intellectual content: XN, LM, ET, LP, and PKH. All authors approved the final version of the manuscript. RGF and ADP take responsibility for the integrity of the data analysis.