Introduction

Diabetes is associated with an increased fracture risk [1, 2], increased mortality, as well as longer time and increased complications during fracture healing [3, 4]. Yet, diabetes is also associated with higher bone mineral density (BMD) [1]. In addition, fracture risk prediction using the FRAX tool underestimates the risk for individuals with diabetes [5]. To date, impaired fasting glucose (IFG) has not been extensively studied in terms of fracture risk.

The increased skeletal fragility associated with diabetes may be due to poorer bone quality including altered material properties [6], or structure, which can be assessed using several different methods, such as magnetic resonance imaging (MRI) or computed tomography (CT); however these are expensive, sensitive to movement artefacts and not readily available [7]. Trabecular bone score (TBS) is a grey-level textural metric that is obtained from lumbar spine dual energy X-ray absorptiometry (DXA) images [8, 9]. Higher TBS indicates stronger bone with a higher fracture resistance, and TBS has been shown to correlate with three-dimensional parameters of bone microarchitecture independently of BMD and clinical risk factors [10,11,12,13]. One strength of TBS is that it can be useful in assessment of fracture risk for individuals who do not have osteoporosis on bone density criteria [14]. This is valuable because most of the population burden of fractures occurs in individuals with moderate deficiency in BMD (osteopenia) [15, 16]. TBS is also useful for fracture risk assessments for individuals with secondary osteoporosis (e.g. diabetes or glucocorticoid-induced osteoporosis), where BMD does not effectively capture the increased fracture risk [8, 9].

The FRAX tool is a fracture risk assessment algorithm developed by the University of Sheffield that utilises clinical risk factors (age, body mass index (BMI), prior fracture, parental hip fracture, smoking, glucocorticoid use, rheumatoid arthritis, secondary osteoporosis and alcohol intake) and femoral neck BMD to assess 10-year fracture probability in men and women aged 40–90 years [14, 17, 18]. TBS has also been shown to improve fracture risk assessments in some populations [8,9,10, 14, 18] and has recently been incorporated into FRAX, allowing a TBS-adjusted score to be calculated [14, 18]. The European Society for Clinical and Economic Aspects of Osteoporosis and Osteoarthritis and the International Society of Clinical Densitometry have endorsed the use of TBS into clinical practice and incorporated it into their recommendations [7, 19].

To date, there have been several studies that have investigated the utility of TBS in distinguishing between individuals with and without diabetes [20,21,22,23,24,25], as well as the ability of TBS adjustment to improve fracture risk predictions using the FRAX tool [26]. These studies, however, have been limited by the number of incident fractures, focussed mainly on postmenopausal women and not all have reported a lower TBS in those with diabetes. Additionally, to our knowledge, no studies have included IFG.

Therefore, this study aimed to describe the relationship between TBS, TBS-adjusted FRAX score and glycaemia status (normoglycaemia, impaired fasting glucose and diabetes) in Australian men and women.

Methods

Participants

This study used data collected as part of the Geelong Osteoporosis Study (GOS), which is a population-based cohort study including residents of the Barwon Statistical Division, located in south-eastern Australia. The region is ideal for conducting epidemiological studies as it has a large population (n ~ 280,000) and includes individuals from a range of social, cultural and geographical settings, which are representative of the wider Australian population. Full details of the cohort have been previously published [27]. GOS participants were randomly selected from Commonwealth electoral rolls, which effectively capture all adults within the region, as voting is compulsory in Australia from age 18 years. The data from this study for men are drawn from baseline (2001–2006, n = 399) or the 5-year follow-up (2006–2011, n = 156) dependent on when the first lumbar spine scan was completed for retrospective assessment of TBS, together with coincident data for determining diabetes status. For women, TBS data were drawn from the 15-year follow-up (2011–2014) as this was the first follow-up where DXA scans were performed on a Lunar Prodigy scanner. However, diabetes status was determined at 10-year follow-up, as there were insufficient data to determine diabetes status at the 15-year follow-up. Prior to this, DXA scans were performed using a DPX-L machine and the software for retrospective determination of TBS is not available for these older DXA devices. Phantom scans were conducted at least three times per week and quality control scans were conducted every day that a participant was scanned. The long-term reproducibility, coefficient of variation, assessed over 180 days using a phantom was 0.23% [28]. All participants provided written, informed consent and Barwon Health Human Research Ethics Committee approved the study.

Clinical Measurements and Questionnaire Data

BMD (g/cm2) was measured using DXA (Lunar Prodigy, GE, Madison, WI, USA). TBS (unitless) was determined retrospectively using TBS iNsight software (Version 2.1, Med-Imaps, Merignac, France). Age referred to the age at TBS assessment. Weight and height were measured to the nearest 0.1 kg and 0.1 cm, respectively. Waist circumference was measured to the nearest 0.5 cm at the minimum circumference between the lowest ribs and the iliac crest. Previous fractures, medication use, rheumatoid arthritis and current smoking were determined by self-report. Ex-smokers were considered as non-smokers. Alcohol consumption was determined from a food frequency questionnaire developed by the Victorian Cancer Council [29] and dichotomised into ≥ 3 or < 3 standard drinks of alcohol per day. A standard glass of beer (285 mL), a single measure of spirits (30 mL) or a medium-sized glass of wine (120 mL) were considered equal to one standard alcohol drink (FRAX Aus guidelines). Physical activity (high/low) was determined by self-report; participants were considered to have “high” physical activity if they had an active lifestyle including light exercise or more undertaken several times per week. Secondary osteoporosis was also identified by self-report, which included insulin-treated diabetes, osteogenesis imperfecta in adults, untreated long-standing hyperthyroidism, malabsorption or chronic liver disease. In this study we were unable to determine chronic malnutrition; instead we applied a previously published method of considering those with BMI less than 18.5 kg/m2 (underweight category) to be malnourished [30].

Socio-economic status (SES) was ascertained using Socio-Economic Indexes for Areas (SEIFA) index scores, based on the 2006 Australian Bureau of Statistics Census data; SEIFA scores were used to determine the level of SES via the Index of Relative Socio-Economic Advantage and Disadvantage (IRSAD). The IRSAD accounts for parameters measured at the area-level, including high and low income, and type of occupation. A low score using the IRSAD identifies the most disadvantaged (quintile 1), while a high score identifies the most advantaged (quintile 5).

Venous blood samples were collected after an overnight fast. Fasting glucose was measured using an adaptation of the hexokinase–glucose-6-phosphate dehydrogenase method [31]. Diabetes was classified if fasting plasma glucose (FPG) ≥ 7.0 mmol/L (126 mg/dL), self-reporting diabetes or use of antihyperglycaemic agents (antihyperglycaemic medication use referred to medications taken regularly and currently at baseline). IFG was considered present if FPG level was between 5.5 and 6.9 mmol/L (100–125 mg/dL); according to the 2003 ADA diagnostic criteria [32]. Medical records and self-report were also used to determine whether any women had developed diabetes since 10-year follow-up. Any women who were diagnosed with diabetes between the 10- and 15-year follow-up were considered to have diabetes.

FRAX Calculations

The Australian version of FRAX (FRAX Aus®) was used in this study and includes the risk factors of age (ranging from 40 to 90 years), sex, weight, height, previous fracture, parental hip fracture, current smoking, glucocorticoid use, rheumatoid arthritis, secondary osteoporosis, alcohol consumption of ≥ 3 units/day and femoral neck BMD. These data were entered into the FRAX (Aus) online tool [33] for each participant and four 10-year probability scores were generated: (i) major osteoporotic fracture (MOF) with BMD, (ii) MOF with BMD, adjusted for TBS, (iii) Hip fracture with BMD and (iv) Hip fracture with BMD, adjusted for TBS. Since FRAX scores are affected by age, and there is a significant age difference between the dysglycaemia groups, ratios were calculated as TBS-adjusted versus unadjusted scores.

Statistical Analyses

The ANOVA or Kruskal–Wallis test was used to determine differences for continuous variables according to glycaemia status. A Chi-Square test or Fisher’s exact test was used for categorical variables.

Using multivariable regression, the relationship between glycaemia status and TBS was assessed for men and women separately, using three different models. The models were first adjusted for age only (age-adjusted model), then for age, weight and height (adjusted model 1). Since TBS is affected by abdominal thickness [34], another model was also developed including age, height and waist circumference (adjusted model 2). Interaction terms were tested in the models to check for effect modification.

Differences between FRAX score ratios with and without TBS adjustment were investigated using a Kruskal–Wallis test for non-parametric data, as FRAX scores were skewed. No clinical measure or lifestyle factor data were missing; however, seven women (6 normoglycaemia and 1 IFG) did not have a femoral neck BMD assessment due to bilateral hip replacement and thus were not included in the FRAX score analysis.

In a model adjusted only for age, we detected a glycaemic status and age interaction term (p = 0.009), and thus both the multivariable regression and FRAX score analyses were also assessed in two age categories; < 65 and ≥ 65 years. All statistical analyses were performed using Minitab (version 16, Minitab, State College, PA, USA).

Results

There were 555 men and 514 women aged 40–90 years with complete TBS, glycaemia status and FRAX clinical risk factor data included in this study. There were 318 men (57.3%) and 381 (73.7%) women with normoglycaemia, 172 (31.0%) men and 85 (16.5%) women with IFG and 65 men (11.7%) and 48 women (9.3%) with diabetes.

For both sexes, those with IFG and diabetes were older than those with normoglycaemia (Table 1), and had higher mean weight and BMI. Mean femoral neck BMD or lumbar spine BMD was not different between the glycaemia groups; however, there was a trend for a higher mean lumbar spine BMD in women with diabetes (p = 0.055). Few men and women were using bisphosphonates, and many of those with diabetes were treated using medication (43.1% of men and 62.5% of women).

Table 1 Descriptive characteristics of men and women included in the study, stratified by glycaemic status (normoglycaemia, impaired fasting glucose (IFG) and diabetes)
Full size table

MEN: TBS and Glycaemia Status, All Ages

The mean age-adjusted TBS was higher in men with normoglycaemia, compared to men with IFG and diabetes (Table 2).

Table 2 Trabecular bone score values for men and women with normoglycaemia, impaired fasting glucose (IFG) and diabetes
Full size table

In the model adjusted for age, weight and height, the relationship between TBS and glycaemia status was attenuated for IFG. However, men with diabetes had lower mean TBS compared to the normoglycaemia group (Table 2).

When adjusted for age, height and waist circumference, results were similar to the model adjusted for age, weight and height (Table 2).

We also detected an association between femoral neck BMD and TBS in men (unadjusted: r = 0.25, p < 0.001), which persisted after adjustment for age, weight and height (r = 0.41, p < 0.001).

WOMEN: TBS and Glycaemia Status, All Ages

In the age-adjusted model for women, diabetes was associated with a lower mean TBS compared to the normoglycaemia group (Table 2); however, there was no difference observed between IFG and normoglycaemia.

In the model adjusted for age, weight and height, women with diabetes had a lower mean TBS compared to normoglycaemia. There was again no difference between TBS for women with IFG compared to normoglycaemia (Table 2).

The model adjusted for age, height and waist circumference showed the same results; compared to normoglycaemia, those with diabetes had a lower TBS, and those with IFG were not different.

There was also an association between femoral neck BMD and TBS in women (unadjusted: r = 0.32, p < 0.001), which also persisted after adjustment for age, weight and height (r = 0.49, p < 0.001). The pattern of the association was similar to that in men; higher femoral neck BMD values corresponded to higher TBS values.

MEN: TBS and Glycaemia Status, Stratified at Age 65 years

When the analysis was stratified by age, a similar pattern was observed. In the age-adjusted model, younger (< 65 years) men with diabetes had lower mean TBS than the normoglycaemia group (Table 2); however, there was no difference between normoglycaemia and IFG. The associations remained the same after adjustment for age, weight and height. In the model adjusted for age, height and waist circumference, the association for lower TBS in those with diabetes was attenuated (p = 0.051).

For older (≥ 65 years) men, mean TBS was not different for IFG or diabetes compared to normoglycaemia in either the age-adjusted or two adjusted models (Table 2).

An association was also observed between femoral neck BMD and TBS in younger men (unadjusted: r = 0.16, p = 0.023), which remained after adjustment for age, weight and height (r = 0.44, p = 0.021). A similar association was also detected for older men (unadjusted: r = 0.21, p < 0.001; adjusted: r = 0.29, p < 0.001).

WOMEN: TBS and Glycaemia Status, Stratified at Age 65 years

In younger (< 65 years) women, age-adjusted mean TBS was lower in those with diabetes compared to normoglycaemia, but not for those with IFG (Table 2). This association was also observed after adjustment for age, height and weight/waist circumference.

For older women (≥ 65 years), TBS values for those with IFG or diabetes were not different to the normoglycaemia group in any of the models.

There was an association detected between femoral neck BMD and TBS in younger (< 65 years) women (unadjusted: r = 0.27, p < 0.001), as well as older (≥ 65 years) women (unadjusted: r = 0.18, p = 0.007), where higher BMD corresponded to a higher TBS. These associations were sustained after adjustment for age, weight and height in younger (r = 0.46, p < 0.001) and older (r = 0.34, p = 0.006) women.

Further adjustment for other variables such as smoking status, alcohol consumption and physical activity did not affect the relationship between glycaemia status and TBS in men or women.

FRAX (Aus) Scores for Men and Women with Impaired Fasting Glucose and Diabetes

All Ages

For both men and women, the assessment of the ratios for TBS-adjusted versus unadjusted MOF and hip FRAX scores resulted in higher values for those with diabetes, compared to normoglycaemia (Table 3). This indicates that TBS adjustment has increased FRAX scores in the individuals with diabetes to a greater extent than in those with normoglycaemia or IFG, which more accurately reflects fracture risk across the groups.

Table 3 FRAX score ratios (TBS-adjusted vs. unadjusted) for major osteoporotic fracture (MOF) or hip fracture for men and women with dysglycaemia
Full size table
Stratified by Age 65 years

In both younger (≥ 65 years) and older (65 + years) men, the ratios for TBS-adjusted versus unadjusted MOF and hip FRAX scores were not different across the three groups. However, in women, there was a difference between the glycaemia groups in the MOF and hip FRAX score ratios for younger (≥ 65 years) individuals (Table 3), but not those in the older age group (65 + years). This could indicate that TBS-adjusted FRAX scores may be more effective at differentiating between the glycaemia groups in younger (< 65 years) women.

Discussion

This study reports that in both men and women, age-adjusted TBS was lower in those with diabetes, particularly in younger individuals. However, there were no differences observed between IFG and normoglycaemia. The ratios of TBS-adjusted versus unadjusted FRAX scores were different between glycaemia groups in both men and women. For women, ratios were higher for those with diabetes in the younger age group (≤ 65 years), but not the older age group (65 + years).

Several other studies have examined TBS in individuals with diabetes. One such study by Dhaliwal et al. [23] reported that TBS was lower in women with type 2 diabetes (age-adjusted mean ± SD: 1.228 ± 0.140) compared to controls (1.298 ± 0.132, p = 0.013). In a study of 141 overweight and obese men, Romagnoli et al. [34] also report significant associations between measures of HbA1c and TBS, as well as FPG and TBS in participants with BMI > 35 kg/m2. Similar to this study, other research has shown that the effect of diabetes on hip fracture risk and bone fragility is stronger at younger ages [35]. A study by Kim et al. [25] showed this result in women, reporting that TBS was lower in women with diabetes compared to controls for the younger age group (aged < 65 years), but there was no difference observed in TBS for the older age group (≥ 65 years).

Bonaccorsi [21] have reported that TBS-adjusted FRAX scores did not differ compared to TBS alone (0.74 vs. 0.71, p = 0.65) in predicting fractures in women with diabetes. This indicates that TBS may be a useful for fracture prediction in women with diabetes, compared to unadjusted FRAX. Another study by Leslie et al. [22] also investigated the ability of TBS to aid in fracture risk prediction for women aged 50 years or older and reported that TBS predicted fractures in those with diabetes, independent of BMD.

This study has some major strengths. Our study population is randomly selected from the Australian population. We also were able to include both men and women in this study, where many other studies examined only women. We had sufficient information to calculate FRAX scores, including self-report of parental hip fracture, which was not available in some other studies. This study also has some limitations. First, diabetes status for women was determined at the 10-year follow-up, whereas the other data were obtained at the 15-year follow-up. However, this is unlikely to have impacted the number of women that we classified as having diabetes. We have shown across a 10-year time period, that women with diabetes did not revert to IFG or normoglycaemia [36]. However, we may have misclassified some women with IFG. Additionally, we were not able to determine which individuals had type 1 or type 2 diabetes, but it is likely that most had type 2 diabetes. As we did not have information about the duration of diabetes, it is possible that diabetes-related bone changes had not had enough time to manifest in those with newly diagnosed diabetes. The statistical power for age-stratified analyses is lower due to smaller sample size, and may have impacted the reported associations. Finally, it has been shown that diabetes does not affect fracture risk the same across different ethnicities [37], and thus our study, which included mainly Caucasian individuals [27] may not be generalisable to other populations.

To conclude, there was no difference detected in BMD for men and women with diabetes compared to normoglycaemia; however, TBS was lower. We did not detect a difference in TBS between normoglycaemia and IFG. Fracture risk assessment using the FRAX (Aus) tool was better able to distinguish between glycaemia groups after TBS adjustment, particularly in younger women (< 65 years). Overall, these data underscore the importance of halting or reversing the transition from IFG to diabetes to prevent the increase in fracture risk due to diabetes.