Advanced Glycation Endproducts and Bone Material Strength in Type 2 Diabetes

Abstract

Context:

Skeletal deterioration, leading to an increased risk of fracture, is a known complication of type 2 diabetes mellitus (T2D). Yet plausible mechanisms to account for skeletal fragility in T2D have not been clearly established.

Objective:

The objective of the study was to determine whether bone material properties, as measured by reference point indentation, and advanced glycation endproducts (AGEs), as determined by skin autofluorescence (SAF), are related in patients with T2D.

Design:

This was a cross-sectional study.

Setting:

The study was conducted at a tertiary medical center.

Patients:

Sixteen postmenopausal women with T2D and 19 matched controls participated in the study.

Main Outcome Measures:

Bone material strength index (BMSi) by in vivo reference point indentation, AGE accumulation by SAF, and circulating bone turnover markers were measured.

Results:

BMSi was reduced by 9.2% in T2D (P = .02) and was inversely associated with the duration of T2D (r = −0.68, P = .004). Increased SAF was associated with reduced BMSi (r = −0.65, P = .006) and lower bone formation marker procollagen type 1 amino-terminal propeptide (r = −0.63, P = .01) in T2D, whereas no associations were seen in controls. SAF accounted for 26% of the age-adjusted variance in BMSi in T2D (P = .03).

Conclusions:

Bone material properties are impaired in postmenopausal women with T2D as determined by reference point indentation. The results suggest a role for the accumulation of AGEs to account for inferior BMSi in T2D.

“Bone material strength is impaired in type 2 diabetes and is lowest in diabetic patients with increased advanced glycation endproducts as determined by skin autofluorescence.”

Strong evidence, accumulated during the past 2 decades, shows that skeletal fragility is a major complication associated with type 2 diabetes (T2D) (1–5). The risk of hip and other nonvertebral fractures is increased by as much as 3-fold in both men and women with T2D, irrespective of ethnicity (2, 3, 5, 6). Although bone health is frequently eclipsed by more immediate life-threatening complications of T2D, fractures in T2D can have a devastating effect on quality and length of life (7, 8). Yet we do not understand how diabetes affects the skeleton (9). Areal bone density is normal, and in contrast to the classic paradigm of increased bone turnover in postmenopausal women, bone turnover in T2D is instead reduced (2, 10, 11).

Inferior bone matrix properties might be a mechanism for increased bone fragility in T2D. Reference point indentation (RPI) has provided evidence for compromised bone matrix properties in T2D postmenopausal women (12). RPI measures the distance that a probe descends into the bone for a given applied force; the farther the probe indents into the tibia, the worse the bone material strength (bone material strength index [BMSi]) (13–15). This potential mechanism is of interest but has not been widely appreciated. Moreover, until now, studies have not progressed yet to suggest a specific mechanism to account for inferior BMSi in T2D.

One potential mechanism is through an accumulation of advanced glycation endproducts (AGEs) that is known to occur in some tissues of patients with T2D. AGEs have been previously shown to accumulate in preclinical models of T2D (16, 17). However, there are no data showing deleterious effects of AGEs on the skeleton of T2D patients. Furthermore, interpretation of plasma AGE levels in patients is limited by poorly standardized immunoassays (18). Type 1 collagen, a target for AGE accumulation, is present both in skin and bone. Studies with human cadavers show that skin and bone AGE (pentosidine) levels per milligram of collagen are strongly correlated (19). Thus, assessing AGE levels using a noninvasive measure of skin autofluorescence (SAF) is a practical surrogate marker of AGEs in bone collagen. A highly reproducible method to measure skin content of AGEs is based on their fluorescent properties. This novel method effectively measures tissue AGE accumulation and has been validated against direct AGE measurements, including pentosidine, in skin (20). The hypothesis of this study is that bone material properties, measured as BMSi, are reduced in patients with T2D due to accumulation of AGEs.

Materials and Methods

Subjects

This cross-sectional study was approved by the Columbia University Medical Center Institutional Review Board, and informed written consent was obtained from all participants. We recruited postmenopausal women through advertisement flyers between December 2014 and December 2015. Menopause was defined as no menses for at least 3 years. T2D was defined as glycated hemoglobin (HbA1c) of 6.5% or greater (21). Subjects were excluded if they had a history of disorders associated with altered skeletal structure or function such as chronic kidney disease, chronic liver disease, active malignancy, acromegaly, Cushing's syndrome, thyroid disease, hyper- or hypoparathyroidism, or organ transplant. Additionally, subjects were excluded if they were currently using teriparatide, estrogen replacement therapy, loop diuretics, anticonvulsive therapies, corticosteroids (>3 wk over the past 3 y), thiazolidinediones, sodium-glucose cotransporter-2 inhibitors, or anticoagulants. Bisphosphonate and/or denosumab use within the past 12 months was also an exclusion criterion. We excluded those of African American ancestry because skin pigmentation is known to influence the measurement of SAF (22). One hundred seventy-one women with or without T2D expressed interest in the study. Fifty were excluded via telephone screen; 45 were not interested once the study was explained to them in detail; 16 were lost to follow-up after the initial phone call. Sixty subjects participated in screening visits including dual-energy x-ray absorptiometry (DXA) and biochemistries. Twenty-five were excluded from analysis because of the following: osteoporosis by DXA in controls (n = 9); loss to follow-up after the screening visit (n = 6); hypercalcemia (n = 2); unmeasurable SAF due to dark skin complexion (n = 2); excessive alcohol use (n = 2); excessive sc fat, which impaired the RPI measurement (body mass index [BMI] 38.4 kg/m², n = 1); glomerular filtration rate less than 60 mL/min (n = 1); severe vitamin D deficiency in a control (25-hydroxyvitamin D 11 ng/mL, n = 1); or withdrawal of consent (n = 1).

Study protocol

All subjects were interviewed for medical history including fracture and diabetic complication history; these data were corroborated by review of medical records when available. The presence of secondary complications was confirmed by medical record review (nephropathy by albuminuria; retinopathy by eye examination; neuropathy by sensory testing; cardiac disease by abnormal stress test, or prior myocardial infarction or coronary artery bypass grafting) in the nine patients for whom such records were available. Anthropometric data were collected on all subjects including height and weight and waist and hip circumference wearing light-weight clothing and no shoes.

Biochemical analyses

Fasting morning blood was drawn and analyzed by standard methods for biochemistries as well as 25-hydroxyvitamin D, PTH, TSH, and HbA1c level. Additional serum was stored at −80°C and was assessed for procollagen type 1 amino-terminal propeptide (P1NP) by RIA (intraassay coefficient of variation [CV] of 6.5% and interassay CV < 8.3%; Immunodiagnostic Systems) and cross-linked C-telopeptide of type I collagen (s-CTx) by chemiluminescence immunoassay (intraassay CV 3% and interassay CV 10.90%; Immunodiagnostic Systems). Urinary pentosidine, a measure of AGEs, was measured in a second morning void by ultraperformance liquid chromatography tandem mass spectrometry (pentosidine sensitivity 2 nmol/L in urine; interassay CV 10%; pentosidine standard was purchased from Toronto Research Chemicals) and was corrected for urinary creatinine.

Reference point indentation

Microindentation was performed with the OsteoProbe (Active Life Scientific). This handheld device is placed over the tibia at 90º and measures the indentation distance increase from the impact of the test probe through the bone, which is then automatically calculated to yield indentation properties (BMSi) by the software (13, 14, 23). Between 10 and 15 indentations were performed per subject to minimize heterogeneity of measurements (12–14, 23). Measurements were performed by one of two operators on a random alternating schedule. All measurements were performed at the midshaft of the tibia, calculated as the midpoint between the lateral malleolus and the head of the fibula. Local anesthesia (1% lidocaine) was administered followed by insertion of the probe through the soft tissue and periosteum until it settled on the bone surface. The device was kept perpendicular to the bone surface and the measurement was taken by compressing the outer housing that allowed the internal primary spring to compress and triggered the impact mechanism. This impact creates a force to drive the probe into the bone, whereas the displacement transducer measures indentation distance increase (IDI) from impact (12–14). The IDI from the impact is converted by a computer to BMSi, defined as 100 times the ratio of the harmonic mean IDI from five separate impacts into a polymethylacrylate plastic calibration phantom relative to the IDI from the impact into the bone (12–14). For each subject, the BMSi was calculated as the average of 10–15 measurements at different midshaft tibia sites separated by approximately 2 mm. The within-subject CV for BMSi was 8.7%. We prespecified the exclusion of all measurements that were outside 2 SD of the mean to remove outliers, as per the manufacturer's recommendation. Measures were also omitted if the operator reported that the probe slipped or if the device did not actuate.

Skin autofluorescence

SAF as measured by the AGE Reader (DiagnOptics Technologies BV; Supplemental Figure 1) was calculated based on the wavelength of light emitted from the forearm skin when it is illuminated with an UV light source (20). Subjects were instructed to not apply sunscreen or self-tanning lotions on the forearm area for 2 days prior to the measurement. The forearm was positioned over the measurement window in a darkened room as recommended by the manufacturer. Calculation of UV light intensity from a reference white reflecting sheet within the device provided the standard to calculate the reflectance of skin. Once the subject placed her forearm on the measurement window, the measurements were made with a white and violet LED light source. A skin surface of 4 cm² guarded against surrounding light and was illuminated with an excitation wavelength of 420–600 nm (fluorescence). The excitation light that was reflected by the skin (300–420 nm) was multiplied by 100 and expressed in arbitrary units known as SAF (20). The measurement was repeated two times for three total measurements at slightly different sites on the forearm for increased accuracy; the mean value was used. SAF has been validated in patients with skin reflection greater than 6% (pale white to moderate brown skin, scale for human skin pigmentation class 1–4 [24]). In patients with darker skin color (class 5–6, dark brown or black), a correction was automatically made to the SAF value if the UV reflectance was 6%–10%. If the UV reflectance was less than 6%, the AGE Reader gave a warning that the signal was too low, in which case the patient's SAF read was excluded from the analysis (22). SAF was calculated by an automated analysis using AGE Reader software version 2.4 and was observer independent. Reproducibility in those with T2D has been reported with an overall Altman error percentage of 5.1% for measurements taken over a single day (20).

DXA and high-resolution peripheral quantitative computed tomography (HRpQCT)

DXA (Hologic QDR 4500) with whole-body composition was performed using our standard methods (25). Vertebral Fracture Assessment was obtained. Body composition was assessed by measurement of trunk fat, total body fat (percentage), and subtotal body fat (total minus head). HRpQCT scans (Scanco Medical) were acquired for all subjects using the standard manufacturer in vivo imaging protocol (26, 27). Each subject was scanned on their nondominant forearm and ankle. The wrist or ankle was placed in a carbon-fiber cast to minimize limb motion during scan acquisition. If the subject had a history of fracture in the nondominant limb, then the dominant limb was scanned. A standard anteroposterior scout view was taken (fixed settings of the machine), and the operator placed the reference line at the endplate of the distal radius or distal tibia (28). Cortical and trabecular volumetric bone mineral density (BMD) and microarchitecture was assessed from HRpQCT images, including cortical porosity; estimated bone stiffness was determined from finite element analysis. At the initiation of the study, subjects were scanned using the first-generation HRpQCT with a nominal isotropic resolution of 82 μm. Midway through the study, we transitioned to the second-generation HRpQCT (XCT2), from the same manufacturer, with a higher nominal isotropic resolution of 61 μm. In a separate study, we compared XCT1 and XCT2 measurements in 51 adults and found excellent agreement between the scanners (R² ≥ 0.91 for almost all measures), which allowed us to calibrate the two machines. Ultimately, 19 subjects were scanned on the XCT1 (13 controls, six T2D subjects) and 16 subjects (six controls and 10 T2D subjects) on the XCT2. For XCT1, a 9.02-mm section (110 slices) was scanned beginning at 9.5 mm proximal to the radius reference line and 22.5 mm proximal to the tibia reference line. For XCT2, a 10.24-mm section (168 slices) was scanned beginning at 9.0 mm proximal to the radius reference line and 22.0 mm proximal to the tibia reference line, thus centered on the XCT1 region.

Statistical analysis

Data are expressed as mean ± SEM unless otherwise specified. All variables were tested for skewness and kurtosis; plots and regression models were used to check the data for normality, linearity, outliers, and potential influential observations. Because all variables satisfied the requirements for parametric statistics, transformations were not performed. Demographic and clinical characteristics as well as anthropometric, biochemical, and bone parameters were compared between the T2D and control groups using two-sample t tests and χ² tests as appropriate. Further comparisons of bone parameters were made using an analysis of covariance model adjusted for age or the presence of complications of diabetes (nephropathy, retinopathy, neuropathy, cardiac disease). Associations of SAF, duration of T2D, HbA1c level, and bone turnover markers with BMSi were examined using Spearman correlations, which are appropriate for smaller sample sizes and more robust to potential outliers and a regression analysis. Testing was performed at a significance level of P < .05 (two tailed).

Results

Patient characteristics

The T2D (n = 16) and control (n = 19) groups were not different in age, years of postmenopause, height, weight, BMI, or waist circumference (Table 1). Other than fasting glucose and HbA1c levels, there were no differences in serum biochemistries between T2D and controls (Table 2). Four T2D subjects could not be fasting prior to the morning laboratory blood testing because of their diabetes medication regimen, so their glucose levels were excluded; the other 31 subjects (89%) were fasting. Average HbA1c levels were significantly higher in subjects with T2D compared with controls (8.3% ± 0.35% vs 5.8% ± 0.08%, P < .0001, Table 2) as per entry criteria. Urinary pentosidine levels did not differ between groups. Body composition by DXA confirmed no difference between groups in trunk fat, percentage total body fat, and subtotal fat (Table 3).

Table 1.

Demographic Characteristics of T2D and Control Subjects (±SD)

	Control (n = 19)	T2D (n = 16)	P Value
Age, y	65.6 ± 1.2	65.4 ± 2.4	.94
Duration of T2D, y	N/A	14.3 ± 2.0	N/A
Postmenopause, y	16.3 ± 1.5	17.4 ± 2.5	.69
History of smoking cigarettes, n, %	3 (15.8)	4 (25)	.37
Previous fragility fractures, n, %a	2 (11)	3 (19)	.64
Prevalent vertebral fractures by vertebral fracture assessment, n	0	0	N/A
PPI use, n, %	3 (15.8)	4 (25)	.36
SSRI use, n, %	4 (21.1)	5 (31)	.51
Height, cm	158 ± 2	156 ± 2	.30
Weight, kg	75.8 ± 3	77.5 ± 4.0	.74
BMI, kg/m2	30.5 ± 1.3	31.5 ± 1.6	.59
Waist circumference, cm	94.2 ± 3	103 ± 3	.06
Nephropathy, n, %	N/A	5 (31.2)	N/A
Retinopathy, n, %	N/A	2 (12.5)	N/A
Neuropathy, n, %	N/A	4 (25)	N/A
Diabetes therapy with insulin, n, %	N/A	8 (50)	N/A
Diabetes therapy with metformin, n, %	N/A	16 (100)	N/A
Diabetes therapy with other oral agents, n, %b	N/A	5 (31.2)	N/A

Abbreviations: N/A, not available; PPI, proton pump inhibitor; SSRI, selective serotonin reuptake inhibitor.

^aFracture locations: controls: one ankle, one wrist; T2D: two ankles, one wrist.

^bOral agents include sulfonylureas (n = 4) and dipeptidyl peptidase IV inhibitors (n = 2).

Table 2.

Biochemical Parameters of T2D and Control Subjects (±SE)

	Control (n = 19)	T2D (n = 16)	Normal Values	P Value
Fasting glucose, mg/dL	87.0 ± 1.6	150 ± 20a	65–99	.0003
HbA1c, %	5.8 ± 0.1	8.3 ± 0.4	<5.7	<.0001
Cr Cl, mL/min	90.9 ± 5.6	95.3 ± 8.6	88–128	.66
TSH, mIU/L	1.53 ± 0.2	2.60 ± 0.9	0.40–4.50	.24
Alkaline phosphatase, U/L	74.5 ± 5.7	84.2 ± 8.6	33–130	.34
AST, U/L	20.7 ± 1.2	25.6 ± 4.2	10–35	.24
ALT, U/L	20.3 ± 2.0	24.5 ± 2.5	6–29	.19
PTH, pg/mL	39.6 ± 5.6	41.5 ± 5.3	14–64	.83
Calcium, mg/dL	9.5 ± 0.1	9.5 ± 0.1	8.6–10.4	.86
Phosphorus, mg/dL	3.6 ± 0.1	3.7 ± 0.1	2.5–4.5	.59
25-Vitamin D, mg/mL	33.7 ± 3.2	26.2 ± 2.5	30–100	.09
P1NP, μg/L	51.2 ± 3.7	38.4 ± 2.4	19–83	.009
s-CTX, ng/mL	0.478 ± 0.034	0.322 ± 0.028	0.112–0.738	.002
Urinary pentosidine, pmol/mg Cr	72.0 ± 18	52.9 ± 14		.43

Abbreviations: ALT, aminotransferase; AST, aspartate aminotransferase Cr Cl, creatine clearance. Bold values denote statistical significance.

^aFour T2D patients were not fasting; their glucose levels are excluded.

Table 3.

BMSi, Regional BMD, and Body Composition by DXA and HRpQCT-Derived Bone Parameters at the Distal Radius and Tibia in Patients With T2D and Age-Matched Nondiabetics Controls

Bone Parameter	Control (n = 19)	T2D (n = 16)	P Value
In vivo microindentation testing
BMSi	70.12 ± 1.9	63.69 ± 1.9	.02
Distal radius parameters (HRpQCT)
Cortical porosity, %	2.4 ± 0.4	2.0 ± 0.3	.35
Cortical volumetric BMD, mg HA/cm3	804.3 ± 19.9	784.6 ± 30.7	.58
Cortical thickness, mm	0.87 ± 0.04	0.92 ± 0.06	.51
Cortical area, mm2	49.9 ± 1.8	54.9 ± 2.7	.13
Trabecular bone volume fraction	0.102 ± 0.005	0.118 ± 0.005	.04
Trabecular number, 1/mm	1.85 ± 0.07	1.93 ± 0.07	.41
Trabecular thickness, mm	0.059 ± 0.002	0.065 ± 0.016	.02
Trabecular separation, mm	0.53 ± 0.03	0.54 ± 0.85	.85
Trabecular volumetric BMD, mg HA/cm3	122.3 ± 6.7	136.4 ± 7.6	.17
Trabecular area, mm2	195.1 ± 11.2	187.1 ± 10.4	.61
Total area, mm2	244.1 ± 10.8	240.3 ± 9.7	.80
Stiffness kN/m	31 208 ± 1551	38 039 ± 1763	.01
Distal tibia parameters (HRpQCT)
Cortical porosity, %	6.1 ± 0.5	6.6 ± 0.7	.57
Cortical volumetric BMD, mg HA/cm3	783.4 ± 15.6	744.3 ± 29.6	.23
Cortical thickness, mm	1.24 ± 0.05	1.22 ± 0.08	.89
Cortical area, mm2	105.7 ± 3.8	106.4 ± 6.6	.92
Trabecular bone volume fraction	0.13 ± 0.004	0.13 ± 0.007	.25
Trabecular number, 1/mm	1.93 ± 0.07	1.81 ± 0.09	.30
Trabecular thickness, mm	0.06 ± 0.002	0.07 ± 0.002	.05
Trabecular separation, mm	0.47 ± 0.02	0.51 ± 0.04	.32
Trabecular volumetric BMD, mg HA/cm3	149.0 ± 5.4	154.8 ± 8.4	.56
Trabecular area, mm2	542.2 ± 25.9	568.4 ± 30.1	.51
Total area, mm2	647.2 ± 24.4	672.8 ± 27.5	.49
Stiffness kN/m	103 462 ± 3901	116 339 ± 6148	.08
Regional BMD (DXA)
Lumbar spine (L1-L4), g/cm2	0.978 ± 0.02	0.95 ± 0.03	.48
T-score	−0.86 ± 0.29	−0.9 ± 0.3	.63
Femoral neck, g/cm2	0.69 ± 0.01	0.77 ± 0.03	.03
T-score	−1.3 ± 0.1	−0.72 ± 0.3	.06
Total hip, g/cm2	0.87 ± 0.02	0.97 ± 0.04	.02
T-score	−0.5 ± 0.1	0.23 ± 0.3	.03
One-third radius, g/cm2	0.65 ± 0.01	0.65 ± 0.02	.89
T-score	−0.7 ± 0.2	−0.63 ± 0.33	.89
Body composition (DXA)
Subtotal fat, kg	33.3 ± 2.2	32.2 ± 2.1	.58
Trunk fat, kg	15.3 ± 1.1	18.0 ± 2.0	.49
Total body fat, %	45.1 ± 1.45	43.2 ± 1.1	.37

Bold values denote statistical significance.

Reference point indentation

Compared with controls, T2D had 9.2% lower BMSi (63.69 ± 1.9 vs 70.12 ± 1.9; P = .02, Table 3 and Figure 1A). BMSi was lowest in those with the longest duration of diabetes (r = −0.68, P = .004; Figure 1B). The difference in BMSi persisted after adjusting for the presence of diabetes-related complications (nephropathy, retinopathy, neuropathy, cardiac disease). The addition of 25-hydroxyvitamin D as a covariate did not change the estimate of the between-group difference in BMSi. HbA1c levels correlated inversely with BMSi in T2D (r = −0.59, P = .02) but not in controls (r = −0.13, P = .61). There was no relationship in either group between BMSi and total body fat (controls: r = −0.32, P = .17; T2D: r = −0.4, P = .88), trunk fat mass (controls: r = −0.17, P = .48; T2D: r = −0.30, P = .26), or subtotal body fat (controls: r = −0.25, P = .30; T2D: r = −0.12, P = .26).

Figure 1.

Bone material strength in type 2 diabetes. BMSi was lower in postmenopausal T2D as compared with matched nondiabetic controls (A) and was worse with longer duration of T2D (B) (±SE). *, P = .02.

Skin autofluorescence

SAF was higher in T2D as compared with controls (T2D: 2.8 ± 0.1 vs controls: 2.2 ± 0.1; P < .001) and correlated with HbA1c in the entire group (n = 35; r = 0.53, P = .001), although not in T2D (r = 0.07, P = .81) or controls (r = 0.22, P = .36) alone. In T2D, BMSi correlated inversely with SAF (r = −0.65, P = .006, Figure 2). Among controls, there was no relationship between BMSi and SAF (r= −0.02, P = .92, Figure 2). SAF also predicted worse BMSi in T2D after adjusting for age. In T2D, age and SAF accounted for 43% of the variance in BMSi, with age accounting for 17.5% and SAF accounting for 26.3% of the total variance. For each unit increase in SAF in T2D, BMSi decreased by 9.9 ± 4 (P = .04).

Figure 2.

Bone material strength, SAF levels, and P1NP. BMSi correlated inversely with SAF in postmenopausal T2D (A) but not in matched nondiabetic controls (B). SAF correlated inversely with P1NP levels in postmenopausal T2D (C) but not in matched nondiabetic controls (D).

Bone structure

Areal BMD was higher in T2D than controls at the femoral neck and total hip but did not differ at the lumbar spine and distal one-third radius site (Table 3). By HRpQCT, trabecular bone volume fraction and estimated bone stiffness were greater at the radius but not at the tibia, whereas trabecular thickness was greater at both sites in T2D patients (Table 3). No other differences between groups were detected in volumetric BMD or microarchitecture. Cortical porosity at both the radius and tibia did not differ between T2D and controls (Figure 3); there was also no difference in those scanned on the XCT2 (Scanco Medical) alone (six controls and 10 T2D subjects). There was no correlation between any measures of bone structure by DXA or HRpQCT and BMSi, SAF, or HbA1c in either T2D or controls.

Figure 3.

HRpQCT images of the radius and tibia. Cortical porosity did not differ at the radius (upper panels) or tibia (lower panels) between T2D (representative images on left) and controls (representative images on right).

Bone turnover markers

Both markers of bone formation (P1NP) and resorption (s-CTx) were lower in T2D (Table 2). Within the T2D cohort, increased SAF correlated with reduced P1NP but not in controls (Figure 2); SAF did not correlate with s-CTX.

Discussion

Increased risk of fracture is a known complication of T2D (1–5), but underlying mechanisms are not well understood. This report implicates, for the first time, a role for AGEs as a potential mechanism to account for reduced bone material strength in T2D. This study confirms worse BMSi in T2D and demonstrates the lowest bone material strength in patients with diabetes with the greatest AGE accumulation, as measured by SAF. Furthermore, this association was specific to the patients with diabetes, with an inverse correlation with BMSi in T2D but not in nondiabetics. Taken together, these data help to define a biochemical mechanism that could explain the BMD-independent increase in bone fragility seen in T2D patients.

Understanding how BMSi is inferior in T2D is an important issue in accounting for increased bone fragility in T2D. We and others have shown that increased bone AGEs in diabetic mice alter the biomechanical properties of the bone matrix (16, 17, 29, 30). Although AGEs have been presumed to accumulate in bone in human subjects with T2D, data showing a deleterious effect of AGEs in bone in T2D patients are almost nonexistent (31). Only one study has reported AGE accumulation in human diabetic bone specimens (32). Immunoassay measurement of circulating AGEs, such as pentosidine, is limited by poor assay validity and standardization (18). Consistent with this known limitation, we did not detect a difference in urinary levels of the AGE pentosidine. Moreover, circulating AGEs might not accurately reflect skeletal AGE content because intracellular glycation is a major local source of AGEs (33, 34). This point makes accurate noninvasive quantitation of skeletal AGEs (short of direct measurement in bone biopsies) a major challenge.

We have used a noninvasive measure of SAF, a highly reproducible method based on the fluorescent properties of AGEs. This new technique effectively measures tissue AGE accumulation and has been validated against direct AGE measurements, including pentosidine, in skin (20). Moreover, studies with human cadavers have demonstrated that skin and bone AGE (pentosidine) levels per milligram of collagen are strongly correlated with each other (19). SAF is also inversely correlated with quantitative ultrasound of calcaneal bone in healthy men (35). In contrast to HbA1c, which reflects recent (≤3 mo) glycemic control, SAF reflects a long-term hyperglycemic metabolic memory (36).

We found that higher HbA1c levels reflected worse BMSi to a similar extent as did SAF. Farr et al (12) similarly found that the 10-year cumulative HbA1c measures correlated inversely with BMSi. It is thus possible that higher HbA1c is a clinical marker for increased bone fragility risk. Our data also indicate that greater AGE accumulation is related to reduced bone formation; this is consistent with preclinical data relating AGEs to decreased osteoblast activity (37). We also considered that factors other than T2D might be explaining the difference we observed in BMSi. However, even after adjusting for vitamin D levels (that were nonsignificantly lower in T2D) and the presence of diabetes-related covariates, the lower BMSi persisted. Obesity has independent adverse effects on bone and in the work of Farr et al T2D had greater BMI than controls (12, 38). Our subjects, in contrast, did not differ in BMI, and our body composition data showed no differences in measures of body fat or relationships between body mass measures and BMSi. It is also possible that diabetes medications, rather than T2D per se, contributed to worse BMSi. Although we cannot definitively rule out this possibility, it seems unlikely because all of our T2D patients were on metformin, which has beneficial skeletal effects, whereas none were on known skeletal harmful agents (eg, thiazolidinediones or sodium dependent glucose transporter-2 inhibitors) (39).

Increased cortical porosity has been implicated in bone fragility in T2D (40–42). Our measures of cortical porosity fell into the range of previously reported values in T2D, with 2.4% at the radius (prior reports: 0.8–4.3% [12, 40–44]) and 6.1% at the tibia (prior reports: 3.0–10.1% [12, 40–43]). Nevertheless, we did not detect a difference from controls, possibly due to our small sample size and/or because, unlike others, we did not recruit postmenopausal diabetic women with fragility fractures, in whom abnormal findings would be more likely. Furthermore, absolute differences in cortical porosity between individuals with and without diabetes are small (<2.5%) and may not themselves negatively affect whole-bone strength (40, 41). However, the role of changes in small cortical pore sizes less than 100 μm has never been examined. Had we measured a greater number of patients on the second-generation XCT2 (Scanco Medical), we might have detected small pores near the periosteal surface in T2D, which may in fact be markers for reduced BMSi (40–42). Cortical porosity has indeed been related to worse BMSi in nondiabetics (45); future studies would be necessary to examine this issue. It is also possible that if we had scanned at more proximal sites at the radius and tibia, the greater cortical content at those areas would have allowed for the detection of cortical porosity differences.

We found that trabecular thickness and bone volume fraction were greater in the diabetic patients. Other studies using HRpQCT have similarly found that trabecular microarchitecture is preserved (12, 25, 42) or increased (40) in T2D. Most recently, T2D patients were found to have greater bone tissue stiffness at the tibia as compared with type 1 diabetes patients (46). However, studies using other tools suggest that in T2D trabecular microarchitecture may be impaired. Using magnetic resonance imaging, postmenopausal women with T2D were found to have larger holes within the trabecular network at the distal radius than nondiabetic women (47). Moreover, trabecular bone score, a technique that uses spine DXA images to perform a gray-level textural analysis of trabecular bone microarchitecture, was abnormal in T2D in a number of reports (48–50). It thus remains to be determined whether mechanical competence is truly enhanced in the trabecular compartment in T2D.

This study has several limitations that are associated with the technologies we used. Although we demonstrated abnormal BMSi in T2D, it is not clear exactly how BMSi relates to traditional bone mechanical properties (such as toughness and elastic modulus) (51). Furthermore, the area studied using RPI is limited to the outer 100–200 mm of tibial bone and may not represent properties at fracture sites such as the femoral neck and spine. However, evidence that BMSi is a good measure of bone material properties comes from testing three calibration blocks, namely polymethylacrylate, brass, and Noryl with the OsteoProbe (Active Life Scientific) and a Rockwell Hardness tester, which showed that BMSi increased with increasing hardness (13). Furthermore, in vivo evidence makes it clear that BMSi does provide a good, if not perfect, measure of bone material properties (23, 52). The consistency of our results with those of Farr et al (12), in which BMSi was 9.2% lower in T2D than in controls after adjustment for covariates, further supports the value of RPI in T2D. An additional concern about RPI is user variability and how the operator determines whether data should be removed (51). We prespecified exclusion criteria to minimize this issue (excluding measurements outside 2 SD of the mean measurement or if the probe slipped or the device did not actuate). Importantly, despite the fact that a large difference in BMSi between T2D and nondiabetics needed to be seen to surpass the threshold of the within-subject CV, we were able to detect that large difference. A limitation of SAF technology mandated that we exclude patients with dark skin because skin pigmentation is known to influence the measurement of SAF (22). Finally, despite our crossover study, it remains possible that using two generations of HRpQCT technology injected some variability in our results and prevented us from seeing cortical pores. A limitation unassociated with technology arises from the fact that T2D is a heterogeneous disease with variable underlying mechanisms for insulin resistance; further studies are necessary to determine whether BMSi differs based on the underlying etiology of T2D.

In summary, reference point indentation confirms that bone material strength is impaired in T2D. RPI offers information beyond that available from areal BMD testing by DXA, making it a potentially valuable tool for the robust clinical assessment of bone quality in T2D patients. The data further demonstrate that BMSi is most impaired in T2D with increased AGEs, as determined by SAF. We conclude that the data support the hypothesis that AGEs play a potential role in the development of skeletal abnormalities in human T2D. Future studies are necessary to investigate longitudinal changes in BMSi in T2D and to assess their value at predicting fracture risk.