Questioning the association between bisphosphonates and atypical femoral fractures

Abstract

Bisphosphonates are the first line treatment for osteoporosis. Structurally, they are stable analogues of pyrophosphate and therefore exhibit a high affinity for bone mineral. They reduce bone loss by attenuating the ability of the osteoclast to resorb bone, decreasing activation frequency and the rate of remodelling. Large prospective randomized placebo-control trials provide unequivocal evidence for a reduction in the incidence of fractures.¹ Impressively, 40 years since their first use in patients, the safety profile of bisphosphonates has been equally re-assuring.² Questions have arisen lately as to whether bisphosphonates could cause atypical fractures, a rare type of atraumatic or minimal trauma femur fracture occurring below the great trochanter. This question has prompted calls for a broader examination of the long-term effects of bisphosphonate use. An attempt by the Food and Drug Administration (FDA) to garner consensus and provide definitive views was not successful.³ This has led to continued anxiety among treating physicians and patients alike, resulting in an overall reduction in prescriptions for bisphosphonates and for osteoporosis therapies in general. Here, we provide an overview of the current data on atypical fractures and bisphosphonate use.

Introduction

Unlike other complications associated with bisphosphonate use, there is an inherent inconsistency between the occurrence of atypical fractures and the use of this class of drugs. Thus, the question, How can a drug that prevents fractures be the cause of fractures? It must be noted nonetheless that bisphosphonates reduce, but do not eliminate fracture risk, and patients being treated with bisphosphonates can and do fracture.

As early as 2005, Odvina et al. reported nine patients who sustained spontaneous non-vertebral fractures after up to 8 years of treatment with alendronate.⁴ Six of these patients also had impaired fracture healing. The authors argued that these fractures are due to overly suppressed bone turnover. In 2008, a phenotypic description of subtrochanteric and femoral diaphyseal fractures was published, with features including (a) cortical thickening in the lateral aspect of the subtrochanteric region, (b) transverse fracture, (c) medial cortical spike, (d) bilateral findings of stress reactions or fractures, and (e) prodromal pain.⁵

These features were observed in subsequent reports and, ultimately, in 2010, became the basis of a formal definition enacted by an American Society for Bone and Mineral Research (ASBMR) Task Force.⁶ More specifically, the ASBMR Task Force stated that, in order for a fracture to be designated as “atypical,” it must include all the following unique radiographic features classified as major criteria, notably that (a) it must be caused by minimal or no associated trauma localized to the subtrochanteric region and femoral shaft; (b) have a transverse or short oblique orientation; (c) have a medial spike when the fracture is complete, and (d) be without comminution. Cortical thickening, a periosteal reaction of the lateral cortex, bilateral prodromal pain, and delayed fracture healing were classified as minor features, together with the presence of co-morbid conditions and concomitant drug exposure. The European Society on Clinical and Economic Aspects of Osteoporosis and Osteoarthritis (ESCEO), and the International Osteoporosis Foundation (IOF) Working Group essentially endorsed the ASBMR's definition.⁷

In 2013, the ASBMR produced a revised case definition for atypical femur fractures considering them as a form of stress or insufficiency fracture. Also, the periosteal stress reaction at the fracture site was upgraded to a major feature. The only mandated criterion relates to the location of the fracture, which must occur “along the femoral diaphysis from just distal to the lesser trochanter to just proximal to the supracondylar flare.” Additionally, four out of five major features must be present for it to qualify as an AFF. Namely, the fracture must be associated with minimal or no trauma; substantially transverse or short oblique in orientation; complete or incomplete (complete fractures may have a medial spike and incomplete fractures must involve the lateral cortex); non-comminuted or minimally comminuted or accompanied by periosteal reaction of the lateral cortex and, finally, presence of localized periosteal or endosteal thickening of the lateral cortex at the fracture site.⁸

Secondary evaluation of data from pivotal clincial trials

Randomized clinical trials powered to estimate the risk of atypical femoral fractures in bisphosphonate users have neither been conducted nor are likely to be ever conducted because of the rarity of these fractures, associated high costs, and the almost unreal scope of patient numbers required to show significant differences. Secondary analysis of the results of three large, randomized bisphosphonate trials, namely the Fracture Intervention Trial (FIT), the FIT Long-Term Extension (FLEX) trial, and the Health Outcomes and Reduced Incidence with Zoledronic Acid Once Yearly-Pivotal Fracture Trial (HORIZON-PFT), with an overall number of 14,195 women, is the closest we have come to in utilizing prospective data.⁹

This analysis attempted to identify and characterize in terms of atypical features those fractures that were below the lesser trochanter and above the distal metaphyseal flare, namely subtrochanteric and diaphyseal femur fractures. From a total of 284 fractures sustained, 12 of them in 10 patients could be classified as subtrochanteric or diaphyseal femur fractures. Their combined rate was 2.3 per 10,000 patient-years. Compared with placebo, there was no significant increase in the risk of atypical or any fracture associated with bisphosphonate use. One important limitation of this study was a lack of access to patients’ radiographs, which would have allowed a more complete characterization of the atypical features. Another limitation was that two of the three placebo-controlled trials, namely FIT and HORIZON-PFT, lasted only 3 to 4.5 years, restricting the ability to draw definitive conclusions on whether the risk of fracture might increase with longer exposure. Overall, even 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.

Epidemiological studies using radiographs

Several retrospective epidemiological studies have investigated the possible association between bisphosphonate exposure and atypical subtrochanteric or diaphyseal femur fractures. Importantly, only six of these studies had access to radiographs. A case-control study from a single Swiss trauma centre covering a 12 year period (1999 to 2010) identified 477 individuals aged 50 years or older who were admitted with a fracture of the subtrochanteric or femoral shaft area.¹⁰ Using the ASBMR criteria, they assigned 39 patients to the atypical fracture group and the remaining 438 to the classical fracture group. A control sample of 200 healthy individuals without femoral fracture was selected randomly. Consistent with studies from pivotal clinical trials, bisphosphonate treatment led to a 47% reduction in risk of classic fracture (odds ratio [OR]: 0.5; 95% CI: 0.3–0.9). The incidence rate of atypical fractures increased with the duration of use but was low, notably 32 cases per million person-years. Furthermore, a contralateral fracture occurred in 28.2% of atypical cases and in 0.9% of classic cases (OR: 42.6; 95% CI: 12.8–142.4). 32 of the 39 (82.1%) patients with an atypical fracture had been treated with a bisphosphonate compared with 28 (6.4%) in the classic fractures group (OR: 66.9; 95% CI: 27.1–165.1) and 11.5% in the group without fracture (OR: 35.2; 95% CI: 13.9–88.8). The OR for atypical versus classic fracture did not change after adjustments for age, sex, and the use of vitamin D, corticosteroids, or proton-pump inhibitors. With that said, the confidence intervals for ORs were very widely dispersed, and even though a statistically significant difference was computed, a type 2 error was not excluded. A major inherent weakness of this study (and many of the others discussed below) relates to there being no information on confounding variables, including BMD, prior femur fracture, other risk factors for fracture and metabolic bone disease, bone turnover markers.

A Swedish group in a similarly designed, but nationwide study, reviewed 1234 radiographs (out of 1271) of women aged 55 years or older who sustained a subtrochanteric or femoral shaft fracture in Sweden in 2008.¹¹ For the estimation of the relative and absolute risk of atypical fractures, 1.5 million women of a same age and residents of Sweden in 2008 were used as a reference population of which 83,811 women had received bisphosphonates. Of the 1234 patient radiographs reviewed, 59 cases assigned as having atypical femoral fracture. These patients were compared with 263 controls, who had suffered subtrochanteric or shaft fractures, but without atypical features. It was noted that a significantly higher percentage of the cases had been exposed to bisphosphonates (78% or 46 patients) compared to controls (10%). The estimated age-adjusted relative risk (RR) of atypical fracture was 47.3 (95% CI: 25.6–87.3). The absolute risk (AR), however, was only 5 cases per 10,000 patient-years (95% CI: 4–7). On average yearly, this would translate into one fracture out of 2000 treated patients. The risk was higher with longer exposure [OR: 1.3 (95% CI: 1.1– 1.6) per 100 prescribed daily doses]. One major criticism of this study is that there was indication bias that the authors demonstrated but made no attempt to account or adjust for in the nonuser group in the cohort study, making the 47.3 RR value likely spurious. Another issue relates to the short, three-years (2005–2008) span of available data during which patients were on medications or suffered from co-existing conditions. It is almost certain that a significant number of patients included in the study had been taking bisphosphonates for much longer periods prior to July 2005. That could explain the early onset of risk despite the apparently short exposure to bisphosphonates reported, and could also result in an overestimation of incidence.

The same data were re-analyzed by the same group where they changed their case capture for atypical femoral fractures to utilize the ASBMR definition.¹² This resulted in changes to a number of findings in the original paper. Applying the ASBMR criteria,⁶ they would have uncovered 80 atypical femoral fractures, not 59, and that 49 would have been on bisphosphonates (changing the percentage of atypical femoral fractures in patients on bisphosphonates from 78% to 61%), meaning that 31 (39%), not 13 (22%), would have been atypical femoral fractures in nonusers. These changes in the numbers would drastically change the RR and OR calculations originally made by the authors.¹² This recent publication also illustrates the problem in evaluating information across studies where different criteria were employed to identify cases.

A Dutch single-center cohort study over an 11-year period (Jan 1997–Dec 2007) assigned 10 patients as having atypical features of the 63 cases of subtrochanteric and femoral shaft low trauma fractures that were examined for fracture site and its features.¹³ Five out of these 10 cases had no history of exposure to a bisphosphonate, while the other five were current or past bisphosphonate users. In contrast, 51 of the 53 patients with an ordinary subtrochanteric/femoral shaft fracture without atypical features were not treated with bisphosphonates. When all 63 cases were matched individually for age and gender with 126 patients with low-energy hip fractures, bisphosphonates were used by 9.5% of cases and by 8.7% of controls (OR: 1.10; 95% CI: 0.39–3.06). Furthermore, the duration of bisphosphonate use was comparable between the two groups (cases: 54 ± 35 months vs. controls: 54 ± 52 months, P = 0.53). Radiographically, there was no significant difference between cases and controls when cortical thickness was measured 5 cm below the lesser trochanter (P = 0.71). However, there was a significant negative correlation between cortical thickness and age (R = -0.281, P < 0.001), as has been noted before.

Electronic medical records and radiographs from women ≥50 years of age and men ≥65 years of age, covering 13.5 years (January 1996 to June 2009) were reviewed at Kaiser Permanente Northwest (Oregon, U.S.) for the occurrence of atypical fractures. Seventy-five atypical femoral fractures meeting at least major ASBMR's criteria and 122 non-atypical femoral shaft fractures were identified.¹⁴ Notably, 70% of the atypical femur fractures had never taken bisphosphonates. As with the Swiss, Danish, and Dutch studies, a low, largely stable rate of 5.9 atypical femoral shaft fractures per 100,000 person-years (95% CI, 4.6–7.4) was reported. Hip fractures appeared to decline, whereas the incidence of non-atypical femoral shaft and distal femur fractures remained flat. In contrast, analysis of five years registry data (January 1, 2007 to December 31, 2011) from Kaiser Permanente Southern California estimated the age-adjusted incidence rates for an atypical fracture at 1.78/100,000/year (95% confidence interval [CI], 1.5–2.0) after bisphosphonate exposure up to 1.9 years, increasing to 113.1/100,000/year (95% CI, 69.3–156.8) with exposure from 8 to 9.9 years.¹⁵ In this cohort study of 1,835,116 individuals, 142 patients had atypical fracture adjudicated by examination of radiographs; 128 cases had bisphosphonate exposure, some of them for less than a year.

Epidemiological studies without radiographs

Epidemiological studies from Denmark, the United States, and Canada, despite their larger size, had no access to radiographs. This prevented the identification of the atypical features identified by the ASBMR. There is also no ICD code allocated to atypical fractures, and therefore some studies even the low-force category could not be separated from the fractures due to high-energy trauma. The outcomes from these studies to no surprise are contradictory.

Two different groups used an adequate number of cases and 4-fold larger number of controls than cases from the Danish National Hospital Discharge Register and the National Prescription Database. Vestergaard et al. assessed the risk in 103,562 patients, both men and women, treated with bisphosphonates and other osteoporosis medications matched for age and gender with 310,683 controls.¹⁶ Abrahamsen and co-workers used a total of 44,704 cases and 168,642 controls matched for gender, age, date, and location of the fracture, but restricted their observations to alendronate users only.^{17, 18} Both studies concluded that subtrochanteric and diaphyseal fractures are likely to be due to osteoporosis, rather than to a bisphosphonate or other anti-resorptive treatment. Vestergaard in particular found that there was an increased risk of fracture in alendronate users that existed prior to their start of alendronate, and found no relationship between duration of use and fracture, thus implicating underlying disease.¹⁶

A population-based, nested case-control study further assessed the risk of fractures in 205,466 Canadian (Ontario) women aged ≥68 years who initiated therapy with an oral bisphosphonate, namely alendronate, risedronate, or etidronate, between April 1, 2002 and March 31, 2008. The patients were followed up until March 31, 2009.¹⁹ During this period, 716 women sustained a subtrochanteric or femoral shaft fracture and 9723 women sustained a typical osteoporotic fracture of the intertrochanteric region or femoral neck. The 716 women were matched for age and cohort entry period with 3580 controls from the cohort who were not hospitalized with a subtrochanteric or femoral shaft fracture. The authors found no increase in risk for subtrochanteric or femoral shaft fractures in both short (100 days to 3 years) and intermediate (3 to 5 years) term bisphosphonate users. However, in the long-term users (≥5 years), the adjusted OR increased to 2.74 (95% CI: 1.25 to 6.02); a subtrochanteric or femoral shaft fracture occurred in 71 (0.13%) during the subsequent year and 117 (0.22%) within 2 years. Overall, the absolute risk of these fractures was low. The study can be criticized for its unbalanced risk factors applied between groups. Furthermore, the risk may have been inflated due to lower mortality among bisphosphonate users.²⁰

We examined the U.S. healthcare data (MarketScan) for the incidence of fragility fractures at all sites of the femur, including the relatively rare, true atypical fractures, in relation to the duration and amount of alendronate or risedronate exposure for up to 8 years.²¹ This cohort study comprising 287,099 patients (93% women) compared to those with higher versus lower degrees of compliance; 764 subtrochanteric and femur shaft fractures and 2,769 hip fractures out of 3,655 incident cases of femoral fracture were identified during 917,741 person-years of follow-up. The HRs for subtrochanteric/shaft fractures were 1.05 (CI: 0.87–1.26), 0.89 (CI: 0.60–1.33) for risedronate, and 0.99 (CI: 0.92–1.05) and 1.05 (CI: 0.92–1.20) for alendronate, respectively. Also, the HRs (95% CI) for overall femoral fractures associated with each additional year of alendronate or risedronate therapy within first 5 years and beyond were not significantly different. In general, risk reduction appeared smaller for subtrochanteric and femur shaft versus classical femoral neck and intertrochanteric fractures. This difference did not reach significance when it was assessed for a time-dependence or an increased risk of these fractures overall with prolonged therapy.

Using the same large medical claims database MarketScan and the National Hospital Discharge Survey, Nieves et al. studied the epidemiology of femoral fractures from 1996, the year alendronate was introduced in the United States, to 2006 among individuals aged 50 years or older.²² The annual hip fracture incidence declined from about 600/100,000 to 400/100,000 person-years. Subtrochanteric, femoral shaft, and lower femur rates remained stable, each approximately 20 per 100,000 person-years. Furthermore, no significant trends were found in any of these fractures during the more recent years of 2002–2006. Interestingly, ~80% of femoral fractures, including subtrochanteric fractures, cases were not bisphosphonate users. Another study analysed national trends in hip fractures and medication use in the elderly US population 65 years and older. Covering almost the same period (1996–2007), but with different U.S. data sources, namely the Nationwide Inpatient Sample and the Medical Expenditure Panel Survey, Wang and Bhattacharyya compared subtrochanteric fragility fractures with typical hip fractures in femoral neck and intertrochanteric regions. The age-adjusted rates for typical hip fractures decreased by 31.6% among women and by 20.5% among men. However, the rates for subtrochanteric fragility fractures remained unchanged among men (P = 0.34), but increased 20.4% among women. The annual percentage increase in these rates was 2.1% (95% CI: 1.3–2.8, p<0.001). During the study period, bisphosphonate use increased from 3.5 to 16.6% in women, but reached only 2.3% in men. Furthermore, the prevalence of bisphosphonate use preceded the increase in the proportion of subtrochanteric fractures. Using age-adjusted rates, the authors estimated that for approximately every 100 reductions in typical femoral neck or intertrochanteric fractures, there was an increase of one subtrochanteric fragility fracture.²³

Finally, in a large population-based and propensity score-matched cohort study, new users of the oral bisphosphonates alendronate, risedronate, and etidronate were compared with new users of either raloxifene or calcitonin nasal spray. The authors estimated the incidence rates and risk of subtrochanteric and diaphyseal femoral fractures as well as the risk associated with treatment duration.²⁴ These data were collected from low-income elderly Medicare beneficiaries participating in Pharmaceutical Assistance programs with comprehensive pharmacy coverage in Pennsylvania and New Jersey (U.S.) from January 1996 through December 2006. In the 33,815 patients included in the study there were 104 subtrochanteric or diaphyseal femur fractures with no significant difference in the risk between bisphosphonate and raloxifene or calcitonin use (HR =1.03; 95% CI 0.70–1.52). Thus, these results could suggest that osteoporosis itself and not its treatment regimens increase the risk for atypical fractures.

In summary, therefore, secondary analysis of three pivotal prospective clinical studies with over 14,000 women did not reveal an association between bisphosphonate use and atypical femur fractures. Case-control studies in which radiographs were available provided an incidence of atypical fractures as low as 3 cases per 100,000 patient-years of bisphosphonate use. However, most studies confirmed that bisphosphonate use reduced the incidence of typical osteoporosis-related fractures. Epidemiologic studies in which radiographs were not available were contradictory, but had the advantage of larger cohorts of patients. These again showed mostly no association between bisphosphonate use and the incidence of subtrochanteric and femur diaphyseal fractures (not classified as atypical by the ASBMR radiographic criteria). In cases where there was an association, the incidence was particularly low.

Bone biopsy data

The majority of bone samples were obtained from the iliac bone and not from the femur. Although in generalized bone diseases such as osteoporosis, trans-iliac bone biopsy is considered representative of distal parts of the skeleton, this may not be true for atypical fractures. This difference was demonstrated in a rheumatoid arthritis patient on alendronate for 8 years who developed spontaneous bilateral subtrochanteric/diaphyseal fractures.²⁵ The osteoclast number was six times higher in the femoral cortex (~1 cm above the fracture) than in iliac cancellous bone, with the cells having all the morphological characteristics of active osteoclasts. Furthermore, suppression of bone turnover has not been a universal finding in bisphosphonate-treated patients with atypical fractures who underwent a bone biopsy, despite the fact that therapeutic suppression is a common feature of antiresorptive treatment.²⁶

Bisphosphonate use in cancer patients

Frequent intravenous infusions (up to every 3 weeks) of the potent bisphosphonate zolendronate are an established practice in patients with bone metastases. Had the atypical fractures been caused in a direct, dose-dependent fashion following bisphosphonate treatment, an “epidemic” of atypical fractures in cancer patients would be expected, which is not the case. In a retrospective review of imaging studies of 327 patients with metastatic bone disease who received a minimum of 24 infusions of pamidronate and/or zoledronic acid between 2004 and 2007, fourteen patients sustained a femoral fracture.²⁷ Four of these fractures were diagnosed as atypical, using less stringent criteria than those proposed by the ASBMR Task Force.²⁸ The duration of treatment in these 4 patients was 68 to 103 months and they had received between 48 and 73 doses of bisphosphonates. Neither the number of doses of bisphosphonates (P = 0.063) nor the duration of treatment (P = 0.148) was significantly different compared to those patients who did not develop a fracture. The authors concluded that the prevalence of atypical subtrochanteric femoral fractures in patients with skeletal malignant involvement who are managed with high doses of intravenous bisphosphonates is very low.

Mechanistic issues

The annual amount of a bisphosphonate taken up by the skeleton from the 70 mg weekly tablet,²⁹ such as alendronate, is estimated at 10.9 mg: 70 mg × 52 weeks × 0.6/100 bioavailability = 21.84 mg. Approximately half of the absorbed dose is excreted through the urine, which would reduce this figure by half to 10.9 mg. The skeletal mineral is estimated in the region of 2 kg. It is therefore highly unlikely that the entire skeleton, or even a fraction of it, is covered by the bisphosphonate absorbed in miniscule amounts over longer periods. For example, the skeletal mineral to which alendronate could bind over 10 years is only 75 mg, or 37.5 parts per million, after taking into account turnover rates and skeletal retention and release.³⁰ Thus, a bisphosphonate cannot possibly coat all resorption surfaces, and the supply of osteoclasts is unlimited, as these cells are constantly being generated from their bone marrow precursors. Diab et al. have shown that alendronate treatment results in similar levels of bone turnover in vertebra (cancellous) and femur (cortical) sites,³¹ suggesting that a “floor” to the reduction in bone turnover by bisphosphonates may indeed exist. In other words, bone turnover is not suppressible to levels below a certain minimum, no matter at what point it initiates.

The therapeutic reduction of bone remodelling activity leads to highly mineralized and homogeneous bone thus resulting in a higher BMD and bone strength. The evidence generally supports a lack of correlation between bone strength and micro-damage accumulation. Notably, despite accumulation of micro-damage in lumbar vertebra of beagle dogs in response to therapy with alendronate and risedronate, bone strength parameters actually improved.³² Similarly, suppressed bone remodelling but increased bone strength was noted for dog ribs when treated with incadronate.³³ Mechanistically, studies on beagle dogs treated with clinically relevant doses of bisphosphonates consistently show that overall bone strength increases with bisphosphonates, despite the accumulation of micro-cracks because of induced increases in bone volume and mineralization.³⁴ Furthermore, any accumulation of micro-cracks occurs early upon therapy and does not increase with time.³⁵ Furthermore, in a 3-year study in beagle dogs, alendronate had no effect on femoral diaphyseal strength or toughness.³⁶

The lack of an association between bone turnover and micro-crack accumulation has also been confirmed in post-menopausal women on bisphosphonates for a mean duration of 6.5 years. Mean crack frequency in these bisphosphonate-treated women (0.13 micro-cracks/mm²) was low and not significantly different from controls (0.05 micro-cracks/mm²), despite the marked reduction in bone turnover.³⁷ There is further evidence that strength parameters specifically in femur do not deteriorate with alendronate treatment, providing direct evidence in humans that alendronate therapy does not make bone more fragile.³⁸

Currently, therefore, there is no evidence that clinically significant micro-crack accumulation is taking place in patients treated with bisphosphonates,³⁹ despite the therapeutic suppression of bone turnover that is the intended outcome of bisphosphonate treatment. Furthermore, only a very small fraction of bisphosphonate-treated patients sustain atypical fractures, whereas bisphosphonate naïve patients suffer similar atypical femoral fractures.⁴⁰

Roschger and co-workers find no negative effects on bone matrix mineralization in patients treated on alendronate for up to 10 years in the FLEX study.⁴¹ With that said, there are several studies that suggest alterations to bone heterogeneity; however, it must be stressed that it has never been demonstrated that less heterogeneous bone is more prone to fracture. In fact, as patients with low bone mass and osteoporosis are known to have increased heterogeneity, the opposite is more likely the case. Therefore, reduced heterogeneity may be an epiphenomenon arising from the pharmacologic action of the drug, with not much, if any, effect on bone strength. Finally, a one-off study from Tang et al. showing changes in matrix glycosylation⁴² has not been reproduced in humans.

Tjhia et al. have recently shown that tissue level properties, as measured by nanomechanical properties and tissue mineral density of iliac bone biopsy samples, are different in patients who have atypical fractures and are on long-term bisphosphonates compared to treatment-naive, age-matched controls.⁴³ Similarly, Bala et al., again using bone biopsies from iliac crest of women treated with alendronate,⁴⁴ have documented increased mineral maturity, as well as changes in crystallinity and micro-hardness that may impact on bone quality. However, it must be stressed that the translation of such findings in iliac crest biopsies to speculations regarding atypical fractures in the femur is premature.

Weighing the facts

Evidence that exists for an association between long-term bisphosphonate use and atypical femur fractures comes from observational studies, namely case-control studies, retrospective cohort studies, and case series. Inherent limitations of such studies are known to include difficulties in estimating exposure precisely, and issues of controlling confounding bias. Two connotations arise from any epidemiologic study that shows more individuals on bisphosphonate therapy having fractures with atypical radiologic features compared with those not on bisphosphonates. First, is that these studies establish unequivocally that patients not on bisphosphonates also sustain atypical femoral fractures. Likewise, it has been shown that atypical fracture risk is increased in patients treated with raloxifene,^{16, 24} a weak anti-resorptive and in bisphosphonate-naive patients.⁴⁰ Secondly, a major reason per se for patients being on bisphosphonates is that they are, at baseline, at a high risk of fracture and therefore are likely to sustain more atypical fractures as well. This means that high fracture risk is itself an inherent confounder (by indication) in all of these studies. Without a quantification of this high fracture risk in every patient, and considering that patients without bisphosphonate therapy also sustain atypical fractures, it seems premature to imply that bisphosphonate use is associated with, let alone being the cause of, atypical femoral fractures.

Even if after each factor contributing to a high fracture risk in a bisphosphonate-treated patient is controlled for, and a residual effect remains, one must question whether a plausible mechanism can tie bisphosphonate use to atypical femur fractures. Several issues underscore a lack of plausibility of such an association. First, the annual cumulative exposure of bisphosphonates being so low (e.g., alendronate ~10 mg/year)²⁹ compared to the vastness of skeleton makes it implausible, if not impossible, for the drugs to coat the skeleton. Additionally, the skeletal uptake of bisphosphonates is not homogenous, with significantly higher levels found in cancellous bones with higher metabolic activity, such as vertebrae, compared with primarily cortical sites, such as the femur shaft.⁴⁵ Furthermore, even when comparing two cortical bone-rich sites, uptake in the subtrochanteric region is considerably lower compared with the femoral neck.^{46, 47} This means that a higher rate of bisphosphonate accumulation occurs at sites that do not develop atypical fractures. Second, there is no evidence of micro-crack accumulation in long bones with therapeutic suppression of bone turnover with bisphosphonates. Even with micro-crack accumulation in the ribs of beagle dogs receiving supra-therapeutic doses of bisphosphonates, bone strength is increased, not decreased. Third, there is no convincing evidence that bisphosphonates change material properties of human bone, and any therapeutic reductions in heterogeneity have not been associated with a high fracture risk; in fact, the reverse is true. Finally, as noted above, in metastatic bone disease where intensive intravenous therapy with zoledronic acid is mandated for up to 15 times a year, instead of once-a-year as in osteoporosis patients, no noticeable increase in the rate of atypical fractures occurs for up to 8 years of follow-up.