Introduction

Both type 1 diabetes (T1D) and type 2 diabetes (T2D) increase the risk of experiencing a bone fracture [1–5]. In general, bone fractures are a widespread and costly problem [6, 7]. The expectation is that these costs will increase as the aging population grows [8] and the prevalence of diabetes increases [9], especially since diabetes impairs bone regeneration [10] and is associated with costly complications (e.g., non-unions). In addition to medical costs, fractures deleteriously impact the quality of life. For example, a third of Veterans treated for a hip fracture did not live beyond 1 year: an adjusted mortality rate of 29.8% [11, 12]. The cause for the disproportionate increase in fracture risk among diabetics is not well understood, but herein is a summary of what is currently known about the origins of fracture resistance and how diabetes potentially lowers fracture resistance.

Measured by the primary assessment tool of osteoporosis, dual energy X-ray absorptiometry (DXA) at the hip, areal bone mineral density (aBMD) was found to be lower in subjects with T1D than in age-matched control subjects [13], and in statistical models accounting for gender, race, menopausal status, and disease duration, low aBMD was associated with poor glycemic control [14]. Since aBMD is directly proportional to bone strength as determined from whole bone tests of cadaveric tissue (hip [15], radius [16], and spine [17]), this low aBMD could explain the higher risk of fracture among individuals with diabetes. On the other hand, T2D patients do not necessarily have low aBMD [18, 19], and in some reports, aBMD was actually higher in diabetics [20–22]. One possible cause then for the elevated fracture risk is the greater propensity for diabetics to fall compared to non-diabetics [23, 24]. Yet, in the clinical studies that accounted for history of falls and related risk factors (e.g., impaired vision), the association between T2D and increased fracture risk persisted [25–27]. Moreover, a meta-analysis of clinical studies concluded that people with T2D are at a higher-than-expected risk of a hip fracture than those who did not have the disease [4].

With regard to non-diabetics, the probability of suffering a fracture increases with age, and this increase is independent of the age-related decrease in aBMD [28, 29]. Why this occurs is presently unclear, though the limitations of DXA (e.g., projection method and insensitive to collagen) certainly play a role. To improve the ability of aBMD to predict fracture risk, an online risk assessment tool called FRAX was developed by the World Heath Organization. Based on certain risk factors and aBMD, it calculates the 10-year probability of suffering a major fracture and the 10-year probability of a hip fracture. Despite some improvement in predicting osteoporotic fractures with FRAX [30], T2D patients still have a higher fracture risk than individuals without diabetes for a given FRAX score [31]. In other words, diabetes likely increases fracture risk independent of the risk factors such as aBMD and age, which contribute to the FRAX scores [32]. Taken together, these observations point to the strong likelihood that diabetes affects bone in ways beyond changes in bone mass or density.

Hierarchical Organization of Bone

Bone of course has many interesting features that facilitate its function of sustaining loads. The primary constituents—type 1 collagen, mineral, and water—are organized in a hierarchical fashion with the resistance to fracture arising from each of the length scales (Fig. 1 and described by others [33–35]). At the level of the whole bone (macro-structure), resistance to bending by the femoral neck is dictated by the distribution of tissue about the centroid of the neck region. The quantification of this distribution is called the moment of inertia (aka second moment of area) in which periosteal perimeter has a greater contribution than endosteal perimeter to stiffness and strength [36]. Moving to a smaller length scale (microstructure), intra-cortical porosity or apparent bone density primarily determines the material strength of bone [37]. Material strength is independent of bone size or structure and is typically measured by destructive tests on machined, uniform specimens of bone (40 mm × 4 mm × 2 mm). In the case of the apparent strength of trabecular bone (typically 8 mm diameter cores), the volume of bone within the given total volume under loading (bone volume fraction or BV/TV) is the primary determinant with contributions from the architecture and morphology of the trabeculae [38, 39]. In particular, more plate-like trabeculae aligned with the axis of loading confer greater strength than more rod-like trabeculae aligned orthogonal to loading [40, 41].

Fig. 1
figure 1

The hierarchical arrangement of bone. Each length scale from mineralized collagen fibrils to the structure of the femoral neck confers resistance to fracture

Full size image

The origins of strength also exist at the tissue level (nano-structure to ultra-structure). With respect to the mineral phase, increases in the degree of mineralization through secondary mineralization are thought to be one-way bisphosphonate therapy, which lowers remodeling activity, increases fracture resistance [42]. Direct correlations between crystallinity and material strength of human bone have been reported [43], but mineral is not necessarily the sole determinant of whether bone can sustain high stress (>125 MPa). Collagen also can influence strength of cortical bone. For example, collagen fibrils oriented in the longitudinal direction tend to confer greater axial strength than those oriented in the transverse direction [44]. Also, disruption of enzymatic collagen crosslinking can lower bone strength independent of any effects on mineral density [45].

An Engineering Perspective of Bone’s Resistance to Fracture

With multiple length scales contributing to fracture resistance, diabetes conceivably increases fracture risk in numerous ways from possibly disrupting cortical microstructure to altering the organic matrix. Complicating matters is that a decrease in bone strength is likely not the only effect of diabetes on fracture resistance. From an engineering perspective, material failure during service often involves fatigue-generated microcracks and the propagation of such cracks or flaws to a critical size upon which fracture occurs. As such, in addition to monotonic load-to-failure tests that measure modulus (linear slope of stress vs. strain curve), strength (peak stress), and toughness (area under the stress vs. strain curve) [36], there are other mechanical tests to characterize the ability of a material or bone to resist fracture such as fatigue (e.g., number of loading cycles to failure) and fracture toughness testing (e.g., resistance to crack propagation). Daily loads on the skeleton generate microdamage in bone with a stress fracture being the clinical manifestation of a fatigue-induced fracture [46, 47]. These fatigue fractures are one possible cause for the bone destruction that occurs with Charcot arthropathy, a common problem among diabetics [48, 49].

Unlike engineering materials, bone in vivo does have a repair mechanism for fatigue microdamage since bone remodeling can be activated to remove microcracks [50] and resorption sites are preferentially associated with microcracks [51]. Nonetheless, microdamage in bone tends to increase with age [52–54] and is inversely related to bone toughness [55]. Thus, how well the bone matrix can resist crack propagation is important to whether bone breaks, and moreover, this resistance is a function of osteonal density, porosity, compositional heterogeneity, collagen integrity, and other factors existing at each hierarchical level of bone’s organization [56–59]. At present, there is dearth of information on how diabetes affects many of these factors that influence microdamage initiation and resistance and ultimately the fracture resistance of bone.

Traditionally, osteoporosis is viewed as a problem of low bone mass causing reduced bone strength, and to some extent, elevated fracture risk with diabetes has been interpreted as an osteoporosis problem [60, 61]. Nonetheless, studies involving the mechanical testing of cadaveric bone have consistently found that aging affects the ability of bone to dissipate energy during failure (bone toughness) to a greater extent than it does the material strength of bone (Fig. 2) [62–65]. Basically, an older bone does not stretch (strain measured in change of length per original length) as far as a younger bone (Fig. 3), and this is related to an inability of the tissue to handle damage that forms during fracture. There are several engineering methods for quantifying the ability of bone to resist crack propagation (damage); and whether measuring the critical stress intensity factor [66] or the strain energy release rate [56], bone’s resistance to crack propagation (i.e., its fracture toughness) decreases with an increase in age. Moreover, the ability of bone tissue to demand greater energy to propagate a crack as the crack grows in length (i.e., R-curve behavior) is lost or reduced with aging [67, 68].

Fig. 2
figure 2

Mechanical properties as function of donor age. Whether measured using tensile tests (a) [61] or bending tests, the toughness of cortical bone decreases with aging to a greater extent the material strength (mean ± SD)

Full size image
Fig. 3
figure 3

Aging differences in bone deformation. Aging reduces by half the degree of failure strain (left) and permanent strain (right) that bone can endure (mean ± SD)

Full size image

Fatigue testing involves subjecting the specimen to repeated or cyclic loads that eventually cause failure, despite the service loads being much lower than the yield force of the material. Such loading of human tissue samples has found that aging affects the fatigue life of cortical bone in bending (compression and tension modes) [69, 70] as well as the fatigue strength in the shear mode [71]. The early studies on the fatigue properties of bone examined variables such as frequency of the applied cyclical load, specimen geometry, loading mode, and species [72], but certain determinants were identified. For example, Haversian porosity was inversely related to fatigue life [73]. The expectation is that diabetes like aging would affect the fatigue life of human cortical bone, but this is currently an untested hypothesis.

Possible Effect of Diabetes on Mechanical Properties of Bone

Since access to bone of sufficient size from humans with diabetes is extremely limited, little is still known about the full effect of diabetes on many of the mechanical properties of human bone. In two studies investigating bone from humans with diabetes, mechanical properties of bone were not different between middle-aged diabetics and elderly non-diabetics [74, 75]. Specifically, the modulus and strength of the metatarsals—acquired from foot amputations or an allograft bank—were not different between non-diabetic (72.3 ± 10 years old) and diabetic (51.3 ± 8 years old) donors as assessed by three-point bending [75]. In the subsequent study, Fleischli and co-workers tested machined, beam specimens from the tibia with the diabetic donors being between 46 and 61 years old and the non-diabetic donors being between 67 and 85 years old. There were no differences in material strength and fracture toughness suggesting that the effect of diabetes on the fracture resistance of human is akin to accelerated aging, but without age-match controls, this is only a supposition. In one other study investigating the mechanical behavior of cadaveric cortical bone from donors of varying age including one male, 74 years old donor with diabetes, the crack growth toughness was 14% less and crack initiation toughness was 40% less for the diabetic bone than for healthy young bone samples [68]. These percent differences were slightly greater (i.e., worse) than the percent differences between young and aged healthy bone. Given that an age-related decline in collagen integrity is associated with decreasing fracture resistance, diabetes could be a problem of bone brittleness, not just a loss of bone strength.

Clinical Observations of Diabetic Effects on Bone Structure

There is evidence that individuals lacking the ability to properly generate insulin may develop smaller bones. Using peripheral quantitative computed tomography (pQCT), Saha et al. [76] found that the radius and tibia of T1D adolescents, especially boys, had a smaller cross-sectional area (CSA) than did these bones of appropriately matched non-diabetic adolescents. A smaller CSA translates to weaker bones assuming no difference in inherent tissue quality. Another pQCT study by Bechtold et al. [77] indicated that as T1D adolescents reached 14 and 15 years of age, their cortical CSA normalized, becoming equivalent to the cortical CSA of non-diabetics at the same age.

Structural deficits in bone are not necessarily apparent in individuals with T2D. In one study, pQCT-derived cortical thickness (Ct.Th) and CSA were similar between T2D subjects and non-diabetic controls [78], but in another study with a larger cohort, cortical CSA was smaller in the T2D subjects, albeit this was offset by a higher volumetric BMD, than for the controls [79]. Assessing trabecular bone of the spine, a QCT study reported that differences in volumetric BMD between T2D and non-diabetic women and men were not independent of differences in body mass index [80]. As for postmenopausal women, those with T2D did not develop bones with a markedly different structure compared to non-diabetics other than a smaller cortical area at the distal tibia as measured by high-resolution (HR) pQCT [81]. For another relatively small cohort of elderly women in which bones were assessed by HR-pQCT, Burghardt et al. [82] observed greater cortical porosity in the distal radius of T2D subjects compared to bones of otherwise healthy women. In all likelihood, poorly controlled diabetes can increase fracture risk of T1D patients through structural changes and affect T2D patients through reductions in the ratio of structural strength to body weight, which is typically elevated with diabetes.

Potential Effects of Diabetes on Bone Tissue

At the nano-length scale, non-enzymatic collagen crosslinks or advanced glycation end products (AGEs) have been implicated in the diabetes-related decrease in fracture resistance. As animals age, AGEs accumulate in a variety of connective tissues [83, 84]—including bone [85–87]—even though bone undergoes turnover throughout life. The formation of these crosslinks involves the Maillard reaction with sugars resulting in less soluble collagen [88]. As such, hyperglycemia promotes glycation-mediated crosslinking in the tissues of individuals with poorly controlled diabetes. Such excessive crosslinking of the organic matrix could potentially increase fracture risk [89]. In a prospective study of 432 elderly Japanese women, urinary pentosidine (a biomarker for AGEs) was a significant predictor of vertebral fracture in addition to traditional risk factors such as aging and areal BMD [90]. In a French cohort of 396 postmenopausal women, fracture risk was higher for women with high urinary pentosidine, but this biomarker was not independent of the other osteoporosis risk factors [91]. Moreover, two different research groups report that a high level of pentosidine in urine or serum was associated with greater fracture risk in patients with T2D [92, 93]. Lastly, bone taken from hip fracture patients was found to have higher concentration of pentosidine in the tissue than bone of age-matched, postmortem controls [94].

Given these clinical observations, the question arises of whether AGEs mechanistically reduce bone’s resistance to fracture. Mechanical tests of cadaveric tissue consistently find that the age-related decrease in both bone toughness and bone fracture toughness correlate with an increase in pentosidine [65, 85, 95]. Moreover, inducing non-enzymatic crosslinking by incubating cortical or trabecular bone in high concentrations of sugar in vitro reduced fracture properties related to post-yield energy dissipation mechanisms [87, 96], but not strength [97]. This glycation process increases the stiffness of the collagen phase (i.e., demineralized bone) [98] and affects the ability of the collagen to dissipate energy [87]. In vivo, the non-enzymatic crosslinking could occur while mineralization is still an active process, and unlike enzymatic crosslinks that facilitate fibrillation and mineralization [65], these AGEs could potentially disrupt proper collagen–mineral interactions. The nature of such disruptions is presently unknown.

Hypothetically, AGEs cause the collagen to behave in a brittle fashion (i.e., fail at low strain, though resistant to stress) reducing energy dissipation by collagen fibril stretching and sliding within the mineral phase. With the loss of energy dissipation at the local tissue level, microcracks more readily propagate through bone tissue with higher concentration of AGEs. Several studies provide supporting evidence of this hypothesis: collagen fibrils undergo more deformation than mineral crystals at the same apparent tissue strain before and after the yield point of bone (onset of damage) [99, 100], this collagen fibril strain is lower in aged bone with higher AGEs than in young cortical bone [101], and the local tissue strain immediately surrounding a propagating crack is lower in aged bone than young cortical bone such that the extent of crack deflection at lamellar interfaces is reduced with aging [102]. Since crosslinking is likely affecting the contribution of collagen to energy dissipation mechanisms, diabetes likely affects fracture resistance through time-dependent changes to the organic matrix.

Diabetic Effects on Rodent Bone

As with many preclinical studies of disease, rodents have been widely used to study the effects of diabetes on bone (recent findings summarized in Table 1). Starting in 1952, diabetes (induced by alloxan) was found to affect bone structure in rats [103]. Consistently since then, the loss of insulin production (more commonly induced by streptozotocin or STZ) causes a reduction in the structural strength of rodent long bones (mice [104] and rats [105–109]). That is, bones from T1D rodents are narrower with a thinner cortex (Fig. 4), and therefore, break at a lower force or torque than bones from control rodents. There is also a loss of trabecular bone with diabetes [104, 110]. Inconsistently, STZ-induced diabetes decreased the degree of mineralization of cortical bone in male mice [104] and female Wistar rats [111] but did not affect the mineralization density of the cortical tissue (Ct.TMD) in male Fischer 344 and Sprague–Dawley rats [109]. In addition, T1D has not always affected material strength or toughness across studies (Table 1), and the loss in fatigue life was related to the diabetic effect on bone structure, not tissue quality [109]. Variations in the duration of T1D among studies may explain these observed differences in the diabetic effects on the biomechanical properties of bone. We recently found that with increasing duration of T1D, the strength vs. structure relationship (i.e., peak force vs. moment of inertia) changes from that of normal mice, and bones exhibit less post-yield work-to-failure, a measure of brittleness (Fig. 5) [104].

Table 1 Summary of the recent literature showing percent differences in cortical bone properties, as determined by primarily three-point bending tests and DXA or CT analysis of the long bones, between diabetic and non-diabetic rodents
Full size table
Fig. 4
figure 4

Micro-CT-derived renderings of the femur mid-shaft. The bones of the diabetic rodent are smaller with thinner cortices than those of the control rodents

Full size image
Fig. 5
figure 5

Diabetic effects on bone brittleness. With increasing duration of diabetes, bone from male mice became brittle relative to control, as measured by post-yield work-to-fracture (mean ± SEM)

Full size image

There are a variety of rat models of T2D including the Zucker Diabetic Fatty (ZDF) [112, 113], the Zucker Diabetic Sprague–Dawley (ZDSD) [113], Goto-Kakizaki [114], and the WBN/Kob [115], and each exhibits a difference in various bone parameters compared to non-diabetic controls (Table 1). The long bones of the ZDF male rats, which become hyperglycemic by 13 weeks of age, have similarities to those of STZ-treated rats in that they are shorter and narrower with a lower structural strength than controls [109, 112, 113]. This is also the case for male ZDSD rats, though the impairment in growth was reported to be less [113]. The cortical and trabecular bone tissue also have lower mineralization density (mgHA/cm3, as measured by peripheral or pQCT) in these diabetic rats relative to controls. Interestingly, when accounting for structural differences using the moment of inertia (also measured by pQCT), the estimated material strength of long bones was not different between ZDF or ZDSD rats and respective controls [112, 113]. The cause of this dissociation between BMD and material strength is not known, but likely involves heretofore unmeasured changes in the collagen phase of bone interacting with changes in the tissue mineral density. A study involving obese and non-obese T2D rats by Reinwald et al. [113] reported that the post-yield energy to failure, as determined by a three-point bending test of the femur, was markedly lower for the ZDF rats at 33 weeks of age than for the age-matched controls. Increases in AGEs within the bone tissue have been documented for STZ-treated rats [109] and the WBN/KoB rats [115], and it is likely that a similar increase occurs in the ZDF and other T2D rats and mice.

Clinical Directions

While diabetes increases the risk of suffering a fracture, the therapeutic options for preventing a fracture are limited. There is no definitive evidence that bisphosphonates are an effective treatment for individuals with diabetes. Nonetheless, a retrospective review of fracture cases in Denmark investigated the use of bisphosphonates in the context of diabetes and found that T1D or T2D did not appear to affect the ability of bisphosphonates to lower fracture risk [116]. With regard to other diabetic complications involving bone, anti-resorptive therapy has had some success in alleviating the symptoms of Charcot neuroarthropathy [117, 118], although recent studies found no beneficial effects of bisphosphonates on time to resolution or total immobilization time [119, 120]. Since the higher-than-expected fracture incidence among those with T2D is not associated with low BMD, there is no clear rationale to treat T2D patients with a bisphosphonate if the patient’s BMD values are not low. There is however a clear rationale for investigating how diabetes lowers bone’s resistance to fracture. Doing so could lead to effective therapeutic options for lowering the fracture risk of diabetics.