Diagnosis and management of bone fragility in diabetes: an emerging challenge

Worldwide, one in 11 adults globally is estimated to have diabetes. The global prevalence of type 1 and type 2 diabetes in adults is currently estimated to be close to 425 million, with an expected increase to 629 million by 2045 [1]. In addition, there are an estimated 318 million adults with impaired glucose tolerance.

A meta-analysis including nearly 140,000 subjects with fractures reported a pooled relative risk (RR) of any fracture of 3.16 (95% CI 1.51–6.63; p = 0.002), hip fractures of 3.78 (95% CI 2.05–6.98; p < 0.001), and spine fractures of 2.88 (95% CI 1.71–4.82; p < 0.001) in type 1 diabetes [2]. The RR of a hip fracture in women with type 1 diabetes was 5.19 (95% CI 2.22–12.11, p < 0.001) compared to women without diabetes [2]. Weber et al. showed that increased risk of fractures extended across the life span, with hip fracture incidence occurring 10 to 15 years earlier in patients with type 1 diabetes than in those without [3].

An increased fracture risk has also been reported in some studies of type 2 diabetes [4, 5], but not others [6, 7]. In a middle-aged population of 33,000, diabetes was the strongest predictor of low-energy fracture in both men and women, with RR of 2.38 and 1.87, respectively [8]. In a meta-analysis of patients with type 2 diabetes, the summary RR of fractures at the hip in men was 2.8 [1.1, 6.6] and in women 2.1 [1.6. 2.7] [9].

When contrasting the risk of fractures in type 1 diabetes from type 2 diabetes, Vestergaard reported an odd ratio (OR) for hip fracture of 1.38 [1.2–1.6] in type 2 diabetes compared to 1.70 [1.3–2.2] in type 1 diabetes [4], although the latter was likely to be underestimated. Indeed meta-analyses published by Janhorbani [9] and Vestergaard [10] showed a stronger association and effect size for type 1 diabetes (RR 6.3 and 6.94 respectively) compared to type 2 diabetes (RR 1.7 and 1.38 respectively), in both men and women.

With an OR for osteoporotic fracture in patients with type 2 diabetes of about 1.5, only about 4% of the global osteoporotic fracture burden is statistically attributable to diabetes. However, considering the increasing prevalence of diabetes and the fact it may also be associated with greater risk for (injurious) falls [11], fragility fractures increasingly appear as a serious, yet neglected complication of this disease. Nevertheless, the link between diabetes and skeletal health receives only cursory attention in osteoporosis guidelines and even less in clinical diabetes guidelines [12].

Certain individuals with diabetes seem to be at greater risk of fracture than others. Hence, in type 2 diabetes, age and duration of diabetes are clearly important [4, 13‐15]. In the cohort from Manitoba, Canada, consisting of men and women aged 40 years and older with or without diabetes (n = 6455/55′958), diabetes was a significant independent risk factor for major osteoporotic fractures (MOF) (hazard ratio (HR) 1.32; 95% CI 1.20–1.46). However, age significantly modified the effect of diabetes on hip fracture risk, with younger subjects having a higher relative risk, as the background risk rises in the overall population with aging (adjusted (a) HR age < 60, 4.67 [95% CI 2.76–7.89], age 60–69, 2.68 [1.77–4.04], age 70–79, 1.57 [1.20–2.04], age > 80, 1.42 [1. 10–1.99]; p interaction < 0.001) [15].

A study has suggested that fracture risk is not increased within the first 5 years of diabetes [16], while Ivers et al. observed high risk of any fracture only in patients with diabetes for at least 10 years [17]. A biphasic pattern has been proposed where fracture risk is in fact decreased in newly diagnosed type 2 diabetes, −which could be related to some protective effects of increased fat mass in these subjects–, and only increases significantly after 5 years [18]. In FRAX-adjusted analyses, only duration longer than 10 years was associated with a higher risk for MOF (HR 1.47, [1.30–1.66]) [19].

A meta-analysis in 2007 [10] did not reveal a clear association between fracture risk and glycemic control. However, recent observational and association studies reported increased fracture risk with worsening control as defined by glycated hemoglobin A1c (HbA1c) levels ≥ 7% [20, 21]. A clinical trial of glycemic control reported that maintaining a median A1c of 6.4% did not reduce fracture risk compared with a median A1c of 7.5% [22], but this trial could not assess effects of poor (> 8%) control on fracture risk. Diabetes has also been shown to be predictive of increased post-fracture mortality risk among patients with hip fractures [23, 24]. Consistent with the notion that longer duration and/or poor glycemic control could further increase fracture risk in diabetes, recent studies have shown that bone microstructural alterations are more prominent among diabetics with microvascular complications (see below) [25].

The relationship between diabetes and bone fragility and therefore the identification of those individuals at increased risk of fracture is further complicated by the variable effects of diabetes medication on the skeleton (Table 1). Although there is no prospective trial on the effects of diabetes medication on bone fragility, results from observational and epidemiological studies and from adverse events in diabetes clinical trials have brought important insights into the potentially beneficial or deleterious effects of these medications on fracture risk.

Most studies have shown that people with type 1 diabetes have lower bone mineral density (BMD) compared with healthy subjects [42]. It might be expected that obesity, which is a strong risk factor for type 2 diabetes, would protect against osteoporosis because of the known positive correlation between body mass index (BMI) and BMD. Indeed, type 2 diabetes is usually associated with a 5 to 10% higher areal BMD than healthy subjects [5, 10, 13, 43], though there is significant heterogeneity between studies [43]. The increase in BMD was more pronounced in younger men, in the presence of higher BMI and—perhaps surprisingly—higher HbA1c levels [43]. The higher BMD was predominantly a feature of the weight-bearing skeleton but not of nonweight-bearing sites such as the forearm [5]. However, the higher BMD noted in type 2 diabetes may also be independent of the increased skeletal loading as higher BMD persists even after adjustment for BMI in numerous cohort studies [43]. An Asian study also reported subjects with type 2 diabetes and hip fracture who are underweight, with a higher BMD compared to non-diabetic counterparts, suggesting other mechanisms for the higher BMD, such as persistent hyperglycemia related to insulin resistance [44].

This relatively higher BMD in those with type 2 diabetes implies that an even lower proportion of subjects with fracture will have a BMD T-score in the osteoporotic range (i.e., T-score ≤ −2.5) than among the non-diabetic population. Schwartz et al. showed that for a given T-score and age, the fracture risk was higher in type 2 diabetes patients compared to patients without type 2 diabetes [45]. Moreover, a T-score in a woman with diabetes is associated with hip fracture risk equivalent to a woman without diabetes with a T-score of approximately 0.5 units lower [45]. Nevertheless, data have clearly confirmed that while BMD systematically underestimates fracture risk, it still stratifies fracture risk in elderly patients with diabetes [46].

Some studies suggest that type 2 diabetes may also be associated with more rapid bone loss, which could also partially explain the increased rate of fractures. Schwartz et al. [47] found that older women with diabetes lose bone more rapidly than those without type 2 diabetes at many skeletal sites, but not the radius. Leslie et al. recently published that in a large registry-based study for Manitoba, women with diabetes had marginally greater BMD loss at the femoral neck but not at other sites compared to a control population without diabetes [48].

Contrarily to BMD, spine trabecular bone score (TBS) tends to be lower among diabetes patients than controls [49, 50]. Moreover, within the type 2 diabetes group, TBS was better in those with good glycemic control compared to those with poor glycemic control. Hence TBS was found to be a BMD-independent predictor of fracture and predicted fractures equally well in those with (aHR 1.27, 95% CI 1.10–1.46) and without diabetes (HR 1.31, 95% CI 1.24–1.38). To be noted, however, that the gradient of risk per 1 SD decrease in TBS remains less than for BMD in diabetic patients, whereas diabetes itself remains an independent risk factor for fractures even after adjustment for BMD and TBS [50]. Recent analyses indicate that TBS as evaluated on Hologic dual-energy x-ray absorptiometry (DXA) devices is inversely related to BMI and abdominal fat [51]. Whether TBS represents alterations of bone structure in diabetes therefore remains unknown.

There exist conflicting results with studies conducted using calcaneal ultrasound. In one study using quantitative ultrasound, speed of sound (SOS) measurements at the radius were significantly decreased in type 2 diabetes compared to controls [52], while another study reported that calcaneal SOS was not different between type 2 diabetes patients with prevalent vertebral fractures (VFs) compared to those without VFs [53].

Since reduced BMD alone does not fully explain bone fragility, particularly not in type 2 diabetes, alteration in “bone quality” is being investigated using various techniques. With magnetic resonance imaging (MRI), Pritchard et al. measured larger holes in trabecular network of type 2 diabetes compared to controls at baseline [54]. Using HR-pQCT (Xtreme CT) at the distal radius and/or tibia, studies in postmenopausal women with or without diabetes suggest that there is a trend towards greater cortical porosity in type 2 diabetes compared to controls [55‐57]. In 99 elderly women with type 2 diabetes and 954 age-matched controls from the Gothenburg Study, Nilsson et al. reported higher cortical porosity at the distal radius but not at the distal tibia in subjects with type 2 diabetes (+ 16%, p < 0.001) [58]. However, they did not find any other alteration in the trabecular or cortical microarchitecture nor decreased estimated bone strength among diabetics in this cohort [58]. Trabecular bone volume is more heterogeneous and is preserved or (apparently) increased [55], though the latter may arise from the trabecularization of the cortex [59]. Furthermore, the increased cortical porosity and larger trabecular heterogeneity is more evident in type 2 diabetes with fractures compared to type 2 diabetes without prevalent fractures [56]. In African-American women with diabetes, cortical porosity was reported to be 26% greater while cortical volumetric BMD (vBMD) was lower compared to controls [60]. A recent study of 52 subjects with type 2 diabetes, of whom 25 had microvascular disease demonstrated, such cortical deficits noted on HR-pQCT were only present in patients with the microvascular complications [25]. Higher cortical porosity in mid-cortical and periosteal layers in type 2 diabetes patients with prior fracture compared to type 2 diabetes without fractures suggests that these cortical sub-compartments might be sensitive to type 2 diabetes-induced toxicity and may reflect microvascular disease [61].

Bone strength estimated by microfinite element analysis (micro-FEA) was shown to be lower in type 2 diabetes compared to controls in association with increased cortical porosity at the distal radius [55, 58]. Furthermore, in type 2 diabetes with fractures, stiffness, failure load, and cortical load fraction were significantly decreased at the ultradistal and distal tibia compared to type 2 diabetes without fractures and this deficit is related to the higher cortical porosity [56]. However, it is unlikely that HR-pQCT will become sufficiently widely available for routine clinical purposes. DXA-derived surrogates for cortical bone volume and strength may provide additional information regarding cortical alterations in diabetes, as well as having potentially widespread accessibility.

Finally, few studies using microindentation of the tibia outer cortex have suggested that the estimated bone material strength index (BMSI) is decreased in type 2 diabetes compared to controls [58, 62], which could reflect alterations in collagen crosslinks by advanced glycation end products (AGEs) and in mineralization (also see below) [63]. These findings are consistent with the concept of “diabetoporosis” as previously suggested to characterize the bone fragility in this particular population [64].

The gold standard for the study of bone turnover is quantitative bone histomorphometry. One of the best estimates of bone turnover rate is the bone formation rate divided by the surface referent (BFR/BS) and this has been shown to be decreased in diabetes at the cancellous, endocortical and intracortical surfaces by 70–80%. In two small studies, reductions in the mineralizing surface and the osteoblast surface (5 patients) and low bone formation (6 patients) have been reported [65, 66].

Most biochemical studies show that bone formation markers, procollagen type I N-terminal propeptide (PINP) and osteocalcin (OC) and the bone resorption markers c-telopeptide (CTX) and tartrate-resistant acid phosphatase 5b (TRAcP5b) activity are usually reduced in type 2 diabetes [38, 67, 68], whereas bone-specific alkaline phosphatase (BSAP) [69] and N-terminal telopeptide (NTX)/creatinine (Cr) [70, 71] are usually normal or slightly elevated.

It is noteworthy that in the context of a low bone turnover, the mechanisms for an apparent increase in the cortical porosity remain unexplained.

No randomized clinical trials have directly evaluated the anti-fracture efficacy of osteoporosis treatment in diabetic patients; management is therefore largely empirical and derives from the good clinical practice and experience of the physician.

The clinical evidence regarding the efficacy of anti-osteoporosis treatments in diabetic patients is therefore provided by post hoc analyses in subgroups from randomized clinical trials that primarily enrolled osteoporosis patients and from a few observational studies (Table 2).

In the Fracture Intervention Trial (FIT), postmenopausal women including diabetic participants with a femoral neck T-score < −1.6 were randomly treated with alendronate or placebo for 3 years. In a post hoc analysis, Keegan et al. [89] reported that diabetes did not alter the effect of alendronate on BMD gain vs placebo. Similarly, two relatively small observational studies showed than alendronate improved lumbar spine BMD but not hip BMD similarly in postmenopausal osteoporotic patients with and without diabetes [96, 97]. Data extracted from the Danish national prescription registry reported that diabetes, with or without complications, did not influence fracture risk in patients who adhered to alendronate [90]. Another Danish cohort study found no difference in the anti-fracture efficacy of alendronate or etidronate at the hip, lumbar spine, and forearm [91]. Furthermore, this study concluded that risk of hip fracture with these treatments was similar in type 1 diabetes, type 2 diabetes, and non-diabetic patients [91]. In osteoporotic Japanese women with diabetes in 3 phase III trials, risedronate treatment showed similar responses on lumbar spine BMD and bone markers between diabetic and non-diabetic patients [92]. There are no data regarding IV bisphosphonates (ibandronate, zoledronic acid) in diabetic patients; renal impairment may limit the utility of these therapies in diabetics. Data are not currently available regarding the anti-fracture efficacy of denosumab or the effects of discontinuation in those with diabetes. Considering that anti-resorptive treatments decrease bone turnover and increase the degree of mineralization, their effects on whole bone strength and fracture risk without low BMD remain to be ascertained.

In the MORE trial, univariate analysis showed a higher efficacy of raloxifene in reducing vertebral fracture risk in diabetic women compared to those without diabetes (p = 0.04) [93]. Anti-fracture efficacy of raloxifene was similar between patients with and without diabetes in the RUTH (Raloxifene Use for The Heart) trial and in a Danish cohort [91, 94].

Post hoc analyses of the DANCE study (Direct Analysis of Non-vertebral Fractures in the Community Experience) assessed the effects of teriparatide (20 μg/d SQ up to 24 months) on skeletal outcomes in patients with and without type 2 diabetes. Teriparatide treatment had a similar effect in diabetic vs non-diabetic persons on vertebral and total hip BMD. Interestingly, the effect on femoral neck was greater in the diabetic treated patients compared to those without diabetes. Incidence of non-vertebral fracture at 6 months was similar in both groups [95]. Nevertheless, because complicated diabetes could be associated with cortical porosity and teriparatide has been reported to increase cortical porosity [98], the effects of teriparatide on bone strength and fracture risk in severe diabetics remain to be specifically evaluated.

Criteria to establish a diagnosis of osteoporosis are based on the presence of fragility fracture and/or a low BMD. These strict diagnostic criteria have to be differentiated from treatment thresholds. Since prior fracture predicts risk for future fracture as strongly in diabetic as in non-diabetic patients [15], treatment should be initiated when a patient with diabetes meets the intervention guidelines for the general population (Fig. 1). Otherwise, treatment should be considered at more favorable FRAX and BMD values in diabetic than in non-diabetic patients, as both BMD and FRAX may underestimate the risk of fracture in this population. Alternatively, FRAX estimates should be adjusted upwards in diabetics (see below).

Patients with diabetes are at increased risk of fragility fractures. While the pathophysiology of bone fragility in these patients is not entirely clear, it is likely multifactorial. Longitudinal studies have established that FRAX and BMD T-score predict fracture risk in those with type 2 diabetes but both require adjustment for diabetes to avoid underestimation of risk. The optimal approach to management of patients with diabetes has not yet been established based on prospective clinical studies. Hence, our currently proposed algorithm should be considered as a consensus among some experts which may change over time as more evidence will be gathered. Data would suggest that if a patient has indication for therapy based on criteria developed for non-diabetes patients, these patients should be treated with osteoporosis drugs. In absence of established osteoporosis though, these medications may be used with caution though, as the effects of these drugs in situations where bone fragility is mainly due to alterations in bone quality remain to be thoroughly evaluated. Future studies should continue to evaluate the structural determinants (microstructure, material properties, …) of bone fragility and refine the fracture prediction algorithms by including disease-specific determinants of fracture (Table 3). New trials will have to prospectively investigate the efficacy and safety of osteoporosis treatment in diabetics with and without low aBMD.