Diabetic bone fragility through advanced glycation end product-collagen axis: Mechanisms and therapy of sodium glucose cotransporter 2 inhibitors

Abstract

Type 2 diabetes markedly elevates fracture risk despite normal or high bone mineral density, a paradox reflecting qualitative skeletal deficits rather than loss of mass. Chronic hyperglycemia fosters the accumulation of advanced glycation end products in bone; their nonenzymatic crosslinks stiffen type I collagen, impair mineralization, and erode mechanical strength. By engaging the receptor for advanced glycation end products, these adducts activate nuclear factorκB and mitogen-activated protein kinase cascades, amplifying oxidative stress, inflammation, osteoblast dysfunction, and osteoclastogenesis. This review synthesizes epidemiological data from type 1 and type 2 diabetes, highlights the limits of densitybased skeletal assessment, and details the molecular pathology of the glycation-collagen axis. It also appraises antiglycation therapies, including formation inhibitors, crosslink breakers and receptor antagonists, with a particular focus on sodium-glucose cotransporter 2 inhibitors that couple glycemic control with modulation of the glycation pathway. By integrating recent basic and clinical advances, we propose a mechanistic framework for diabetic bone disease and outline strategies to mitigate glycationdriven skeletal fragility.

Key Words: Advanced glycation end products; Bone mineralization and microstructural heterogeneity; Bone mineral density; Diabetic bone fragility; High-resolution peripheral quantitative computed tomography; Nonenzymatic collagen cross-linking; Oxidative stress; Sodium-glucose cotransporter 2 inhibitors; Type I collagen

Core Tip: This review highlights how advanced glycation end products (AGEs) exacerbate bone fragility in type 2 diabetes by cross-linking type I collagen and disrupting bone mineralization. These findings elucidate the AGE-type I collagen axis as a key pathogenic mechanism underlying the diabetic bone paradox. Emerging evidence suggests that sodium-glucose cotransporter 2 inhibitors may mitigate AGE-related bone damage beyond glycemic control. Targeting AGE formation, cross-linking, and receptor-mediated signaling offers novel therapeutic strategies to improve bone quality in diabetic patients.

INTRODUCTION

Type 2 diabetes mellitus (T2DM) has become a major global public health challenge, currently affecting over 500 million individuals worldwide, with its prevalence and associated healthcare burden expected to continue rising. Patients with T2DM exhibit a significantly increased risk of fractures, particularly at specific skeletal sites, such as the hip, spine, humerus, and foot[1,2]. This elevated fracture risk is associated with prolonged disease duration, the presence of microvascular and macrovascular complications, and the use of certain antidiabetic medications. Notably, fracture risk appears to be more pronounced in female patients with T2DM, underscoring the importance of accurate risk assessment and preventive strategies in this population[3].

Interestingly, many individuals with T2DM present with normal or even increased bone mineral density (BMD), yet their fracture risk remains markedly elevated, a paradoxical phenomenon termed the “diabetic bone paradox”[4]. This observation suggests that conventional BMD-based assessment tools, such as dual-energy X-ray absorptiometry (DXA) and fracture risk calculators (e.g., fracture risk assessment), may underestimate fracture risk in diabetic populations[5,6]. Recent advances in the field have shifted our understanding from a focus on bone quantity to bone quality, encompassing alterations in collagen structure, mineralization processes, microarchitectural integrity, and mechanical properties[7,8].

Under chronic hyperglycemia, advanced glycation end products (AGEs) accumulate in bone tissue, particularly through nonenzymatic cross-linking with type I collagen (COL1), leading to increased matrix stiffness, impaired mineralization, and diminished bone toughness and strength. In addition to direct matrix damage, AGEs interact with their cellular receptor (RAGE), activating downstream oxidative and inflammatory cascades that impair osteoblast function, reduce bone formation, and increase bone resorption. AGEs also disrupt the ultrastructural organization of bone collagen, reducing its extensibility and energy dissipation capacity, thereby increasing fracture susceptibility[9]. Together, these mechanisms constitute the AGE-COL1 axis, a unifying pathogenic framework that may help explain the increased fracture risk observed in patients with T2DM[10].

Despite these insights, current therapeutic strategies targeting diabetic bone disease are inadequate. Traditional antiosteoporotic agents, such as bisphosphonates and parathyroid hormone (PTH) analogs, have not been systematically evaluated for efficacy and safety in T2DM patients[11,12]. Consequently, interest in identifying novel targets and developing metabolic drugs with bone-protective potential is increasing. In recent years, the AGE-COL1 axis has gained attention as a promising mechanistic focus[13-15]. On the one hand, it explains the link between collagen mechanical degradation and mineralization defects; on the other hand, it provides multiple intervention points via its downstream signaling pathways [e.g., RAGE, mitogen-activated protein kinase (MAPK), and nuclear factor-κB (NF-κB)].

Sodium-glucose cotransporter 2 (SGLT-2) inhibitors represent a new class of antidiabetic agents that not only effectively lower blood glucose but also exert cardiovascular and renoprotective effects[16]. However, their effects on skeletal health remain controversial[17]. Some studies suggest that SGLT-2 inhibitors may improve bone quality by reducing AGE formation and exerting antiglycation effects[18]. Conversely, other studies have reported an increased risk of fractures associated with certain SGLT-2 inhibitors, such as canagliflozin, particularly in older adults[19]. As such, the mechanisms underlying the skeletal effects of SGLT-2 inhibitors require further investigation to clarify their potential role in addressing T2DM-associated bone fragility[9,20]. Additional evidence indicates that SGLT-2 inhibitors may attenuate glucotoxicity and AGE accumulation, thereby modulating the AGE-COL1 pathway and improving bone matrix microstructure and mechanical performance[21]. Although the underlying mechanisms remain to be fully elucidated, these potential antiglycation properties offer a novel perspective for repositioning SGLT-2 inhibitors in the management of diabetic bone disease.

This review aims to systematically evaluate the epidemiological features of fracture risk in patients with T2DM, explore the limitations of current fracture risk assessment tools in the context of the diabetic bone paradox, elucidate the molecular mechanisms of the AGE-COL1 axis in diabetes-related bone fragility, and assess the therapeutic potential of SGLT-2 inhibitors in improving skeletal health. By integrating the latest findings from both basic and clinical research, this review seeks to provide new perspectives and strategies for the assessment and management of bone health in individuals with T2DM.

CLINICAL PHENOTYPES OF DIABETIC BONE DISEASE

Epidemiological differences in fragility fractures between patients with type 1 diabetes mellitus and those with T2DM

In early bone metabolism research, fractures were once considered a “distinctive” complication of type 1 diabetes mellitus (T1DM), largely due to early-life exposure to insulin deficiency, impaired peak bone mass acquisition, and chronic metabolic dysregulation[22,23]. However, over the past decade, a growing body of high-quality evidence has demonstrated that T2DM also poses a significant threat to skeletal integrity, although the manifestations differ in terms of relative risk (RR) and absolute burden[24-26] (Table 1).

Table 1 Fracture risk estimates from the key cohort/meta-analyses of type 1 diabetes mellitus and type 2 diabetes mellitus patients.

Ref.	Design and population	Sample size	Fracture endpoint	Effect estimate (95%CI)	Diabetes type
Emanuelsson et al[29]	United Kingdom Biobank + Copenhagen, Mendelian randomization + observational study	507428	Fragility fracture	T1DM: HR = 1.50 (95%CI: 1.19-1.88); T2DM: HR = 1.22 (95%CI: 1.13-1.32)	T1DM; T2DM
Zoulakis et al[1]	Swedish elderly female cohort study	3008	Any fracture	HR = 1.26 (95%CI: 1.04-1.54)	T2DM
Champakanath et al[6]	University of Colorado CACTI cohort	1416	Osteoporotic fracture	HR = 1.08 (95%CI: 1.02-1.04)	T1DM

CACTI: Coronary artery calcification in type 1 diabetes; CI: Confidence interval; HR: Hazard ratio; T1DM: Type 1 diabetes mellitus; T2DM: Type 2 diabetes mellitus.

High-intensity risk in T1DM patients: Studies have shown that T1DM patients exhibit a markedly increased risk of fractures, especially hip fractures, with risk levels up to sixfold greater than those in nondiabetic individuals and that fracture onset occurs approximately 10-20 years earlier[27,28]. A meta-analysis further revealed that the overall RR for any type of fracture in T1DM patients was 1.88 [95% confidence interval (CI): 1.52-2.32], indicating a highly significant increase in fragility fracture risk even among young adults with T1DM (P < 0.001)[27]. A recent large-scale analysis of national registry data from Denmark and the United Kingdom revealed that, even after adjusting for competing mortality risk, T1DM was associated with a significantly elevated hazard ratio (HR) for hip fractures (HR = 2.46, 95%CI: 1.82-3.31), suggesting that the fracture burden in T1DM patients remains inadequately mitigated despite contemporary advances in glycemic control and integrated management strategies[29]. The contributing factors may include early-onset disease resulting in lower peak bone mass, sustained hyperglycemia accelerating AGEs accumulation, and microvascular complications impairing bone perfusion and innervation[30,31].

Widespread burden of T2DM: In contrast, although T2DM confers a slightly lower relative fracture risk than does T1DM, its high prevalence translates into a much greater absolute burden of fractures in clinical practice[32]. A 10-year follow-up of 299104 newly diagnosed T2DM patients from the German Disease Analyzer database revealed a 2.9% greater incidence of any fracture than nondiabetic controls did. After adjusting for comorbidities and baseline characteristics, the HRs remained significant at 1.36 (95%CI: 1.32-1.40) and 1.56 (95%CI: 1.45-1.67), respectively[33]. An analysis of over 580000 T2DM patients in the Swedish National Diabetes Register matched by age and sex to controls reported an overall HR of 1.06 (95%CI: 1.04-1.08) for hip fractures. However, fracture risk significantly increased in subgroups with diabetes duration ≥ 15 years, body mass index < 25 kg/m², or insulin use, indicating dose-response relationships and phenotypic stratification[34]. These findings highlight that fracture risk in T2DM patients is not uniformly elevated across the population but is concentrated in specific high-risk subgroups.

Evidence from elderly women: In 2024, Zoulakis et al[1] reported a prospective cohort study involving 3008 women aged 75-80 years, in which T2DM patients had higher baseline areal BMD (aBMD) but still exhibited an HR of 1.26 (95%CI: 1.04-1.54) for any fracture during 7.3 years of follow-up. Subgroup analyses revealed the highest fracture risk among those with a disease duration > 10 years or who were receiving insulin therapy, whereas no significant increase in fracture risk was detected in patients treated with oral antidiabetic drugs alone. This study underscores that modest microstructural improvements cannot counterbalance physical decline and increased fall risk, providing population-level evidence for the paradox of “high BMD-high fracture risk”.

Comparison between T1DM patients and T2DM patients: According to recent epidemiological data and meta-analyses, both T1DM and T2DM are associated with elevated fragility fracture risk, although they differ in RR magnitude and population-level impact. Emanuelsson et al[29], in a nationwide Danish cohort study combined with Mendelian randomization analysis, reported fragility fracture HRs of 1.50 (95%CI: 1.19-1.88) in T1DM patients and 1.22 (95%CI: 1.13-1.32) in T2DM patients. Notably, the HR for hip fracture was significantly greater in T1DM patients (HR = 2.46, 95%CI: 1.82-3.31) than in T2DM patients (HR = 1.50, 95%CI: 1.33-1.79), with nonoverlapping 95%CIs, indicating greater individual-level risk in T1DM patients. However, from a public health perspective, the total number of hip fractures attributable to T2DM was nearly eightfold greater than that attributable to T1DM (385 cases vs 52 cases), reflecting the overwhelming prevalence of T2DM. This trend aligns with findings from Shah et al’s systematic meta-analysis[35], suggesting that although T1DM patients face a higher individual risk, the overall fracture burden from T2DM is far more substantial on a population scale. Therefore, fracture prevention strategies must address both high-risk individual management in T1DM patients and broad-based risk reduction in T2DM patients to achieve the dual goals of risk intensity and population burden mitigation.

Recent evidence in young adults suggests that diabetes impairs the attainment of peak bone mass. In a Canadian high-resolution peripheral quantitative computed tomography (HR-pQCT) study involving 88 individuals with childhood-onset T1DM (mean age 21 years; mean disease duration approximately 14 years), reductions in trabecular thickness and estimated failure load were observed despite near-normal DXA results, indicating compromised skeletal accrual[36]. Cross-sectional data from 180 United States adolescents with youth-onset T2DM revealed a decline in aBMD Z scores from +1.3 at age 10 to +0.6 by early adulthood, whereas obese controls maintained age-related gains. Notably, visceral adiposity was an independent predictor of lower BMD[37]. Collectively, these findings suggest that both insulin deficiency and severe insulin resistance disrupt bone accrual during adolescence, highlighting the importance of early bone health monitoring and lifestyle interventions in pediatric diabetes patients.

Treatment modality and fracture risk: Insulin therapy is considered a risk enhancer for fractures in T2DM patients. A propensity score-matched cohort study involving 2979 T2DM patients revealed a 38% increase in fracture HR associated with insulin use (adjusted subdistribution HR = 1.38, 95%CI: 1.06-1.80), which remained significant even after adjusting for fall history and chronic complications[38]. Potential mechanisms include hypoglycemia-induced falls, insulin-like growth factor-1 negative feedback due to elevated insulin levels, and disrupted insulin signaling in bone cells.

Discordance between BMD, HR-pQCT microstructure, and fracture risk

For decades, aBMD measured by DXA has served as a cornerstone for fracture risk prediction. However, in diabetic populations, a paradoxical dissociation exists: The majority of individuals with T2DM exhibit normal or even elevated aBMD, yet their risk of hip fracture increases by approximately 40%-50%[39]. This paradox underscores the critical importance of bone quality, beyond bone quantity, as a determinant of skeletal integrity. HR-pQCT has enabled researchers to visualize this subclinical skeletal fragility.

T1DM, diminished volumetric BMD and multidimensional microarchitectural damage: In individuals with T1DM, volumetric BMD is reduced alongside widespread microstructural deterioration. Findings from a long-term HR-pQCT substudy nested within the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications cohort demonstrated that T1DM participants had 3%-11% lower volumetric BMD than healthy controls did, with significantly reduced trabecular thickness and cortical thickness. Multivariate analysis revealed strong associations between poor glycemic control [glycated hemoglobin (HbA1c)], AGE accumulation, macroalbuminuria, and microstructural deficits, suggesting that excess AGE burden and microangiopathy may synergistically accelerate skeletal deterioration in T1DM patients[40].

T2DM, cortical porosity and mechanical decline: In a cross-sectional analysis by de Waard et al[41], T2DM participants with HbA1c > 7% presented approximately 20% greater cortical porosity at the radius, with cortical bone density and thickness reduced by 5% and 16%, respectively; tibial trabecular thickness was also 13% lower. Multivariable-adjusted models revealed linear correlations between HbA1c and these bone microstructure parameters, highlighting the close link between poor glycemic control, cortical deterioration, and trabecular disruption. Similarly, in an HR-pQCT cohort study by Samelson et al[39] involving 1069 participants (mean age 64 years), T2DM patients had significantly lower tibial cortical density (794.8 mg/cm³ vs 812.4 mg/cm³, P < 0.01) and approximately 7% greater cortical porosity (10.7% vs 10.0%, P = 0.02). In contrast, trabecular indices, such as the trabecular number, thickness, and separation, showed no major deterioration in either the tibia or radius and, in some cases, were slightly better than those of the controls. These findings indicate that diabetic bone fragility is characterized primarily by increased cortical porosity and reduced bone strength, whereas the trabecular architecture is relatively preserved. This finding supports the hypothesis that diabetes-related skeletal weakness arises mainly from cortical “perforation” rather than generalized bone mass loss.

Biochemical markers of low bone turnover in diabetes

Bone turnover markers (BTMs) provide real-time insight into the rate of bone remodeling. A systematic review and meta-analysis published in 2024, which included 14 studies and 3650 participants, comprehensively assessed alterations in bone metabolic markers in T2DM patients[42]. The results consistently indicated a lower degree of bone turnover in T2DM patients than in nondiabetic controls. Specifically, the levels of osteogenic markers, procollagen type 1 N-terminal propeptide (P1NP) and osteocalcin (OCN), were significantly reduced, with standardized mean differences of -0.66 (95%CI: -0.76 to -0.55, P < 0.00001) and -0.92 (95%CI: -1.27 to -0.56, P < 0.00001), respectively. Moreover, the bone resorption marker β-CrossLaps of COL1 C-terminal telopeptide (β-CTX) also showed a decreasing trend (standardized mean difference = -0.46, 95%CI: -0.68 to -0.24, P < 0.0001). These findings collectively suggest that both bone formation and resorption are suppressed in T2DM, reflecting a generalized low-turnover skeletal phenotype.

A combined cross-sectional and longitudinal study conducted in 2019 proposed a “glycemic threshold effect”: When HbA1c reaches 7.4%, the levels of both P1NP and β-CTX exhibit a sharp inflection point decline, implying that glucotoxicity may simultaneously suppress osteoblastic and osteoclastic activity[43]. At the level of the bone-kidney axis, a declining estimated glomerular filtration rate (eGFR) may further compromise bone turnover. A cross-sectional study in older adults with T2DM revealed that patients with diabetic kidney disease (DKD) had lower β-CTX and P1NP levels than their non-DKD counterparts did, although the differences did not reach statistical significance, suggesting a modest influence of renal function decline on BTMs[44].

Disruption of iron metabolism and chronic inflammation also emerge as key modulatory factors. A multicenter analysis in 2024 revealed that serum ferritin was positively associated with β-CTX (β = 0.074) and negatively correlated with 25-hydroxyvitamin D, suggesting a tendency for iron overload and oxidative stress to promote bone resorption signaling[45]. Sustained low turnover may not only impair the repair of trabecular microcracks and accelerate AGE-COL1 cross-link accumulation but also partly explain the clinically observed “blunted” efficacy of bisphosphonates in diabetic patients[46,47]. Therefore, the combined use of anabolic agents such as teriparatide and interventions targeting the AGE-RAGE axis may help restore dynamic bone remodeling and improve skeletal quality[48,49].

BIOCHEMICAL BASIS OF AGES IN BONE TISSUE

Pathways of AGE formation

Multiple formation routes, Maillard reaction, oxidative stress, and dicarbonyl intermediates: The accumulation of AGEs in the bone matrix is not the result of a single chemical pathway but rather a convergence of three interconnected processes: Nonenzymatic glycation of proteins by sugars, oxidative modification mediated by reactive oxygen/nitrogen species, and the accumulation of highly reactive dicarbonyl compounds. Understanding the molecular checkpoints of these three routes is key to elucidating how the AGE-COL1 axis drives bone fragility. Table 2 is for representative chemical reactions, and Table 3 is for major AGE species.

Table 2 Three metabolic pathways of bone matrix advanced glycation end-product formation and their rate-limiting determinants.

Pathway	Key steps	Products/intermediates	Rate-limiting determinants
Classical Maillard reaction	Glucose reacts with epsilon-amino group of lysine, Schiff base, Amadori rearrangement, early Amadori products	Fructosamine, 1-deoxy-1-ketofructose	Blood glucose concentration, temperature, pH
Glyco-oxidative stress	ROS/RNS oxidize sugars or Amadori products, carboxylated side chains	CML, CEL and other “oxidative AGEs”	ROS levels, antioxidant defenses
Carbonyl stress/dicarbonyl pathway	Glucose autoxidation, degradation or lipid peroxidation, MGO, GO, 3-DG, conjugation with lysine/arginine	MG-H1, glucosepane, pentosidine	Glyoxalase-1 activity, glutathione pool

ROS: Reactive oxygen species; RNS: Reactive nitrogen species; MGO: Methylglyoxal; GO: Glyoxal; 3-DG: 3deoxyglucosone; CML: Nepsilon-(carboxymethyl)lysine; CEL: Nepsilon-(carboxyethyl)lysine; AGEs: Advanced glycation end-products; MG-H1: Methylglyoxal-derived hydroimidazolone-1; pH: Potential hydrogen (acidity/alkalinity).

Table 3 Classification and representative structures of advanced glycation end-products.

AGE type	Chemical name	Characteristics	Biological activity
CML	CML	One of the most common AGEs; formed via glycoxidation or lipid peroxidation	RAGE agonist; promotes inflammation and fibrosis
CEL	CEL	Structurally similar to CML; derived from pyruvate and other metabolic intermediates	Associated with insulin resistance
Pentosidine	Crosslink product of reducing sugars with lysine or arginine	Signature structure of AGE crosslinks	Strongly associated with bone fragility and arterial stiffness
MG-H1	MG-H1	Key biomarker of diabetes-related AGE	Promotes apoptosis and mitochondrial dysfunction

CML: Nepsilon-(carboxymethyl)lysine; CEL: Nepsilon-(carboxyethyl)lysine; MG-H1: Methylglyoxal-derived hydroimidazolone 1; AGE: Advanced glycation end product; RAGE: Receptor for advanced glycation end products.

Maillard reaction, the starting line of AGE formation: Under physiological temperature and neutral pH, reducing sugars interact slowly and reversibly with protein side chains to form Schiff bases, which have a short half-life of only a few hours but shift toward accumulation as blood glucose increases[50]. These Schiff bases undergo Amadori rearrangement to yield more stable early glycation products, so-called “pre-AGEs” (e.g., fructosamines). Over a period of weeks to months, Amadori products undergo dehydration, fragmentation, or free radical oxidation to ultimately form advanced structures such as Nepsilon-(carboxymethyl)lysine (CML) and pentosidine[51].

Oxidative stress, the glyco-oxidative coupling accelerator: Reactive oxygen species (ROS) act as critical amplifiers in AGE formation by accelerating glyco-oxidative reactions through two main mechanisms. First, ROS can oxidize glucose and Amadori products to generate highly reactive carbonyl aldehydes, such as glyoxal and methylglyoxal (MGO), thereby introducing new carbonyl moieties and increasing the pool of AGE precursors. Second, ROS can directly oxidize lysine and arginine residues on protein side chains, promoting the formation of oxidized AGEs such as CML and Nepsilon-(carboxyethyl)lysine. This “glycooxidative coupling” mechanism is particularly prominent under diabetic conditions where hyperglycemia and oxidative stress coexist, thereby accelerating AGE accumulation and exacerbating tissue damage[52].

Dicarbonyl stress, a high-flux pathway for cross-linking AGEs: Dicarbonyl stress refers to the accumulation of highly reactive intermediates such as MGO, glyoxal, and 3-deoxyglucosone, which are far more nucleophilic than glucose. These intermediates are major precursors of cross-linking AGEs, including MGO-derived hydroimidazolone-1 (MG-H1) and glucosepane, which rapidly stiffen COL1 and reduce bone toughness[53,54]. The detoxification of MGOs relies on the glyoxalase 1-glutathione system. In diabetes, serum glyoxalase 1 activity is reduced and negatively correlated with MG-H1 Levels, indicating a sustained “carbonyl explosion” that drives ongoing AGE accumulation[55,56]. As such, dicarbonyl stress represents a fast-track and high-yield route for the generation of cross-linked AGEs and constitutes a central driving force behind the AGE-COL1 axis imbalance. Therefore, merely lowering the glucose burden is insufficient to block AGE formation, future interventions must also increase carbonyl clearance or cleavage to protect the bone matrix.

These three metabolic pathways, the Maillard reaction, oxidative stress, and dicarbonyl stress, are intricately nested and synergistically amplify AGE formation, ultimately contributing to bone fragility. The Maillard reaction initiates nonenzymatic condensation between sugars and proteins; oxidative stress enhances carbonyl generation and promotes fragmentation of Amadori products; and dicarbonyl stress rapidly produces highly crosslinked AGEs (e.g., glucosepane) via short-circuit mechanisms at rates far exceeding those of glucose-driven reactions. Together, this tripartite metabolic axis forms a progressive and cooperative cascade, representing the biochemical foundation of diabetic skeletal fragility (Figure 1).

Figure 1 Hyperglycemia-induced biochemical routes to advanced glycation end product-type I collagen cross-linking and their skeletal consequences, together with representative defense strategies. Persistent hyperglycemia initiates three chemically distinct yet convergent pathways that accelerate the accumulation of advanced glycation end products (AGEs) on type I collagen within bone: (1) Maillard reaction (orange panel): The non-enzymatic condensation of reducing sugars with protein carbonyl groups generates reversible Schiff bases, which rearrange into Amadori products and subsequently oxidize them into stable AGEs such as Nepsilon-(carboxymethyl)lysine and pentosidine; (2) Reactive oxygen species-mediated oxidative stress (red panel): Hyperglycemia-driven bursts of reactive oxygen species further oxidize Amadori products or directly modify lysine and arginine side-chains, yielding Nepsilon-(carboxymethyl)lysine and Nepsilon (carboxyethyl)lysine; and (3) Dicarbonyl stress (purple panel): Excess reactive carbonyl species (methylglyoxal, glyoxal) and their downstream 3-deoxyglucosone intermediates form intracellular clusters that can invade the nucleus and react with proteins to produce methylglyoxal-derived hydroimidazolone-1 and glucosepane cross-links. These three routes collectively intensify AGE-type I collagen crosslinking (lower left inset), which stiffens the collagen network, diminishes tissue toughness, and culminates in brittle bone (center). Endogenous and pharmacological counter-measures include the glyoxalase 1-glutathione detoxification system, chemical AGE breakers, antioxidants, and sodium-glucose cotransporter-2 inhibitors (lower right inset). CML: Nepsilon-(carboxymethyl)lysine; AGE: Advanced glycation end-product; COL1: Type I collagen; ROS: Reactive oxygen species; CEL: Nepsilon-(carboxyethyl)lysine; MGO: Methylglyoxal; MG: Glyoxal; 3-DG: 3-deoxyglucosone; MG-H1: Methylglyoxal-derived hydroimidazolone-1; GLO1: Glyoxalase 1; GSH: Glutathione; SGLT-2: Sodium-glucose cotransporter-2; Arg: Arginine; Lys: Lysine.

Enzymatic and nonenzymatic cross-linking of collagen

COL1 is the primary organic component of the bone matrix. The cross-linking between collagen molecules not only affects the tensile strength of the fibrils but also influences the mineralization patterns and overall tissue toughness. These cross-links are broadly categorized into enzymatic and nonenzymatic types, and their reciprocal alterations under diabetic conditions lie at the core of the AGE-COL1 axis pathology (Figure 2).

Figure 2 Dual-pathway schematic of collagen type I crosslinking shift from “tough-ductile” to “stiff-brittle”. A: Enzymatic crosslinking pathway: The lysyl oxidase family catalyzes the oxidation of lysine/hydroxylysine residues on type I collagen, generating aldehydes that form initial divalent crosslinks, hydroxylysine and dihydroxylysinonorleucine, which gradually mature into trivalent pyridinoline crosslinks, including hydroxylysylpyridinoline and lysylpyridinoline, endowing the bone matrix with ductility and postyield toughness; B: Nonenzymatic crosslinking pathway: Hyperglycemia and oxidative stress promote the formation of reactive dicarbonyls such as methylglyoxal, which in turn induce advanced glycation end products (AGEs) to crosslink, such as pentosidine and glucosepane. These AGEs stiffen collagen fibrils and predispose the bone matrix to brittle failure; C: AGEs and their metabolic derivatives (e.g., Nepsilon-(carboxymethyl)lysine, Nepsilon-(carboxyethyl)lysine, pentosidine, methylglyoxal, and glucosepane) further suppress lysyl oxidase gene expression, thereby inhibiting enzymatic crosslink formation and contributing to a maladaptive “gain-loss” imbalance between the two pathways; D: In healthy bone, the enzymatic crosslinking-dominated “tough-ductile” model enables resilience under high load and strain. In contrast, diabetic bone, owing to AGE accumulation and impaired enzymatic crosslinking, presents a “stiff-brittle” mechanical profile characterized by early strength decline and insufficient energy absorption. LOX: Lysyl oxidase; COL1: Type I collagen; HLNL: Hydroxylysine; DHLNL: Dihydroxylysinonorleucine; HP: Hydroxylysylpyridinoline; LP: Lysylpyridinoline; LH2: Lysyl hydroxylase 2; MGO: Methylglyoxal; AGEs: Advanced glycation end products; CEL: Nepsilon-(carboxyethyl)lysine; CML: Nepsilon-(carboxymethyl)lysine.

Enzymatic crosslinking, the lysyl oxidase-lysyl hydroxylase 2b axis establishes toughness: Enzymatic collagen cross-linking, regulated by the coordinated actions of lysyl oxidase (LOX) family enzymes and lysyl hydroxylase 2b (LH2b), plays a critical role in conferring postyield toughness and mechanical integrity to bone. The LOX family, comprising LOX and its homologs LOX-like 1-4, catalyzes the oxidative deamination of lysine and hydroxylysine residues at specific telopeptide regions of COL1, producing highly reactive aldehydes, allysine and hydroxyallysine[57,58]. These aldehydes spontaneously react with adjacent amino groups to form initial divalent cross-links, such as hydroxylysinonorleucine and dihydroxylysinonorleucine, which further mature over time into stable trivalent pyridinoline cross-links, namely, hydroxylysylpyridinoline (HP) and lysylpyridinoline (LP)[57]. The abundance and integrity of these mature cross-links are key determinants of the biomechanical properties of bone. Experimental inhibition of LOX activity via β-aminopropionitrile has been shown to reduce HP and LP contents, even without altering mineral density[59]. Moreover, LH2b-mediated intracellular hydroxylation of telopeptide lysine is essential for promoting HP formation[57]. Thus, the LOX-LH2b axis serves as a vital regulator of collagen maturation and mechanical performance, and its dysfunction may underlie skeletal fragility in aging and diabetes.

Nonenzymatic crosslinking, AGEs induce stiffness and brittleness: In the hyperglycemic and pro-oxidative environment characteristic of diabetes, reducing sugars and Maillard intermediates (e.g., MGO) react nonenzymatically with lysine and arginine residues on COL1 to form AGEs. Among these, pentosidine and glucosepane are the major bifunctional cross-links found in bone tissue[10,60]. These AGE-derived cross-links often localize to specific sites on the collagen triple helix, restricting molecular slippage and altering the distribution of fibrillar stress[60]. Recent nonstretching experiments and molecular dynamics simulations revealed that AGE cross-links induce strain hardening and brittle failure: Collagen molecules rupture at lower strain levels, energy dissipation is reduced, and fibrils lose their toughness. In contrast, enzymatically cross-linked collagen maintains ductility and toughness even under high strain[10]. Therefore, AGE cross-linking not only modifies the physical connections between collagen molecules but also mechanically shifts the bone matrix from a ductile to a brittle state, representing a key nanomechanical basis for diabetic bone fragility[10,60].

Cross-link switch, the chemical root of diabetic skeletal fragility: Importantly, AGE-mediated cross-linking in diabetes involves not only an “add-on” layer to the collagen scaffold but also the active disruption of the physiological enzymatic cross-linking system. AGE accumulation increases nonenzymatic cross-links within collagen fibrils, increasing stiffness, while simultaneously impairing LOX-catalyzed enzymatic activity by competitively modifying lysine residues in collagen telopeptides, thereby suppressing HP and LP formation. Additionally, AGEs and their metabolites downregulate LOX gene expression, further dampening enzymatic cross-linking. This pathological shift from enzymatic to nonenzymatic cross-linking compromises the biomechanical elasticity of collagen fibrils, disrupts the collagen-mineral interface, and severely impairs the postyield toughness and energy dissipation of bone tissue[61]. Both clinical and animal studies have demonstrated a significant inverse relationship between AGE accumulation and the content of enzymatic cross-links in cortical bone under diabetic conditions, which is correlated with reduced energy absorption at fracture[9,62]. In vitro experiments further confirmed that hyperglycemia and AGEs suppress LOX expression and activity in osteoblasts, impairing classical enzymatic cross-linking within collagen fibrils and altering the microstructure and mineralization pattern of the bone matrix[9].

Determinants of physiological-pathological cross-linking balance: During rapid skeletal growth in adolescence, LOX activity is elevated, and glyco-oxidative homeostasis is maintained, resulting in a predominance of enzymatic cross-links (HPs and LPs). In contrast, adulthood, especially under pathological conditions such as diabetes or chronic kidney disease (CKD), is characterized by sustained carbonyl overload and LOX suppression, leading to a marked increase in AGE-derived cross-links. This creates a “glass-like” bone phenotype with increased rigidity and diminished toughness. Physical exercise has been shown to upregulate LOX expression and enhance pyridinoline cross-link content, representing one of the few nonpharmacological strategies capable of partially reversing the collagen cross-link imbalance in diabetic bone[59].

Detection of bone AGEs: A multiscale technical spectrum from destructive quantification to nanoscale imaging

Techniques for detecting AGEs in bone span a progressive spectrum from destructive quantification to in situ and superresolution imaging. The gold standard for absolute quantification remains destructive methods such as high-performance liquid chromatography with fluorescence detection or liquid chromatography-mass spectrometry, which analyze demineralized and delipidated bone powder following acid hydrolysis. These methods allow precise measurement of individual AGE species, including CML, Nepsilon-(carboxyethyl)lysine, MG-H1, and pentosidine, but require at least 2 mg of bone material and are time-consuming[63]. For rapid screening, semiquantitative assessment of total fluorescent AGEs can be performed on 5-μm tissue sections via autofluorescence (excitation/emission at 335/385 nm), producing results within minutes. This approach is suitable for biopsy specimens or high-throughput pathology workflows. However, the fluorescence signal is easily confounded by mineral and lipid contents and cannot distinguish among AGE subtypes[64].

At the mesoscale level, confocal Raman microspectroscopy, using a focused laser spot size of approximately 1 μm, enables simultaneous recording of amide I/III and phosphate vibrational peaks. This technique allows for semiquantitative estimation of AGE accumulation on the basis of spectral shape and peak ratio analysis while also providing insight into matrix-mineral coupling. It does not require sample demineralization but is sensitive to the hydration state, and interpretation is experience-dependent[65].

Pushing the spatial resolution further to the nanoscale (20-50 nm), atomic force microscopy-infrared spectroscopy; and nano-Fourier transform infrared spectroscopy couple atomic force microscopy with near-field infrared spectroscopy to visualize amide I subbands and collagen cross-linking density at the level of individual collagen fibrils. These methods reveal the heterogeneous distribution of AGEs across trabecular surfaces and cores. However, their application is limited by high equipment costs, small scanning areas, and, primarily, relative quantification capabilities[66]. In summary, the detection of bone AGEs can be approached via complementary strategies tailored to research objectives and sample availability. Destructive quantification (e.g., liquid chromatography-mass spectrometry) provides absolute values, whereas multiscale spectroscopic imaging techniques (e.g., autofluorescence, Raman, and atomic force microscopy-infrared) enable spatially resolved assessments. Integrating these methods allows for a comprehensive analysis of how the AGE-COL1 axis contributes to skeletal fragility (Table 4).

Table 4 Comparison of advanced glycation end product detection techniques in bone tissue and their application limitations.

Method	Minimum sample size	Spatial resolution	Absolute quantification	Major advantages	Limitations	Ref.
HPLC-FLD/LC-MS	2 mg defatted bone powder		Yes	High sensitivity; can differentiate CML/CEL/MG-H1/Pen	Destructive; requires acid hydrolysis; labor-intensive	[62]
Autofluorescence (Ex 335/Em 385)	5 μm tissue section	Micron level	No (relative)	Rapid, high-throughput; suitable for biopsy screening	Interference from mineral/Lipid autofluorescence; cannot distinguish AGE types	[63]
Raman spectroscopy	Applicable to both in vivo and tissue sections	Approximately 1 μm	No (semiquantitative)	In situ detection; simultaneously captures mineral-matrix information	Sensitive to water; spectrum interpretation requires expertise	[64]
Nano-FTIR/AFM-IR	10 μm tissue section	20-50 nm	No (semiquantitative)	Highest spatial resolution; enables single-fiber localization	Expensive equipment; limited scanning area	[65]

HPLC: High-performance liquid chromatography; FLD: Fluorescence detector; LC-MS: Liquid chromatography-mass spectrometry; Ex: Excitation; Em: Emission; Nano-FTIR: Nanoscale Fourier transform infrared spectroscopy; AFM-IR: Atomic force microscopy-infrared spectroscopy; CML: Nepsilon-(carboxymethyl)lysine; CEL: Nepsilon-(carboxyethyl)lysine; MG-H1: Methylglyoxal-derived hydroimidazolone 1; Pen: Pentosidine; AGE: Advanced glycation end-product.

AGE-COL1 AXIS AND THE MOLECULAR PATHOGENESIS OF DIABETIC BONE FRAGILITY

Effects of nonenzymatic collagen I cross-linking on fibrillar mechanics and mineralization template function

Under hyperglycemic conditions such as diabetes, excessive accumulation of AGEs leads to nonenzymatic covalent cross-linking with lysine residues in COL1 within the bone matrix[67]. These nonenzymatic cross-links markedly increase the stiffness of collagen fibrils while diminishing their viscoelastic properties, rendering the tissue mechanically stiffer but more brittle[68]. Under physiological conditions, moderate cross-linking between collagen molecules enables a balance between molecular slippage and tensile resistance, allowing for energy dissipation under mechanical loading and preventing microdamage accumulation[67]. However, excessive AGE-mediated cross-linking overly restricts molecular mobility, resulting in reduced fibrillar extensibility and increased susceptibility to microcrack formation and propagation, thereby compromising bone toughness[10,67]. Additionally, AGE cross-linking reduces collagen solubility and enhances resistance to proteolytic degradation. This AGE-induced “collagen fixation” impairs the timely turnover of aged or damaged collagen during bone remodeling, further weakening the structural integrity of the bone matrix.

In addition to mechanical alterations, nonenzymatic AGE cross-links also directly disrupt the function of collagen as a mineralization template[68]. Proper collagen assembly and enzymatic cross-linking are essential for regulated hydroxyapatite (HA) deposition within the bone matrix[67]. Aberrant cross-linking patterns, characterized by insufficient enzymatic and excessive nonenzymatic AGE cross-links, can impair the normal physiological mineralization process[69]. Consequently, AGE accumulation in diabetic bone is considered a key contributor to compromised bone quality and increased fragility through its dual effects on the mechanical properties of collagen and its capacity to serve as a scaffold for mineral nucleation and growth.

Mechanisms by which AGEs disrupt intra- and extrafibrillar collagen mineralization

During normal bone formation, minerals are deposited in a tightly regulated manner within (intrafibrillar) and between (extrafibrillar) collagen fibrils. Initial nucleation occurs in the periodic gap zones between collagen molecules, resulting in intrafibrillar mineralization, followed by the deposition of excess minerals on the fibril surface to complete extrafibrillar mineralization[70]. This spatially organized distribution of minerals is critical for maintaining the mechanical competence and toughness of the collagen-mineral composite. However, in diabetic bone with high AGE accumulation, this mineralization pattern becomes markedly disrupted[71].

Studies utilizing high-resolution techniques such as nanocomputed tomography (nano-CT) and transmission electron microscopy (TEM) have revealed irregular mineral distributions in the bones of diabetic mice. A substantial amount of minerals accumulates on the surface of collagen fibrils, whereas intrafibrillar mineralization is notably deficient. These findings suggest that AGE accumulation inhibits the normal process of mineral infiltration and nucleation within collagen fibrils. Mechanistically, AGE-induced cross-linking increases the density and hydrophobicity of collagen fibrils, thereby impeding the diffusion of mineral precursors [e.g., calcium-phosphate complexes or osteogenic noncollagenous proteins (NCPs)] into the fibrillar matrix. As a result, HA nucleation and growth within the fibril interior are suppressed, forcing the surplus minerals to deposit externally, leading to disorganized extrafibrillar mineral networks.

This mislocalization of mineral content results in regions of mineral overloading adjacent to areas of undermineralization, disrupting the microstructural uniformity of trabecular and cortical bone. Three-dimensional reconstructions via nano-CT revealed a highly heterogeneous distribution of mineral density in diabetic bone, in stark contrast to the more homogeneous mineral architecture observed in healthy controls. TEM further confirmed these findings by revealing sparse crystal occupancy within the collagen fibrils and the presence of large, disoriented crystal aggregates on their surfaces. Such patterns of intrafibrillar mineral depletion and excessive surface mineralization compromise the integrity of the collagen–mineral interface, ultimately impairing the biomechanical performance of bone tissue[9]. In summary, AGE accumulation disrupts the collagen mineralization process and leads to nanoscale architectural disorganization of the bone matrix, a phenomenon that is now well supported by advanced nano-CT and TEM evidence.

AGEs reduce HA crystal size and disrupt crystallographic orientation

The adverse effects of AGEs on bone matrix mineralization are not limited to the spatial distribution of mineral deposition but also involve alterations in crystal morphology and orientation. In healthy bone, HA crystals are nanoscale, plate-like structures whose long axis (c-axis) aligns approximately parallel to the longitudinal axis of collagen fibrils. This highly ordered orientation forms a composite structure that confers excellent tensile strength and stiffness to bone while preserving a degree of toughness. However, AGE accumulation disrupts this intricate mineral organization. Studies using techniques such as X-ray diffraction and Raman spectroscopy have demonstrated that HA crystals in diabetic bone exhibit a significantly reduced average size and decreased crystallinity compared with those of controls[72]. For example, in T2DM mice induced by a high-fat diet and streptozotocin (STZ), the average c-axis length of HA crystals is notably shorter than that in healthy controls, suggesting that hyperglycemic, AGE-rich environments hinder crystal growth[72,73]. Impaired intrafibrillar mineralization likely restricts crystal development within collagen gaps, resulting in only small initial crystals; moreover, extrafibrillar crystals, which lack spatial confinement of collagen, tend to be morphologically irregular and fail to achieve sufficient growth or orientation. Furthermore, AGE-mediated dysfunction in osteoblast activity, such as reduced secretion of OCN and chondroitin sulfate proteoglycans, may also contribute to abnormal crystal morphology and alignment.

More importantly, AGE-induced modifications to the collagen matrix contribute to disorganized crystal orientation. Under physiological conditions, collagen fibrils act as templates to guide the aligned deposition of mineral crystals along the fibril axis. In bone enriched with AGEs, however, disordered collagen arrangement and aberrant mineral deposition sites lead to randomly oriented crystals[74]. High-resolution small-angle X-ray scattering and electron microscopy analyses revealed significantly increased angular dispersion of the HA crystal c-axis orientation in diabetic bone, indicating a departure from the uniform longitudinal alignment observed in healthy bone[73,75]. This misalignment reduces the ability of bone to resist directional mechanical loads.

Together, the reduced crystal size and disrupted orientation compromise the mechanical contribution of the mineral phase: Smaller crystals are less effective at load-bearing, and misaligned crystals impede stress transmission. As a result, diabetic bone may exhibit normal or even elevated BMD while possessing inferior material quality in the mineral phase[9]. This mineralogical degradation provides a mechanistic explanation for the so-called “diabetic bone paradox”, normal BMD but heightened fracture risk. In summary, AGE accumulation exacerbates bone fragility by reducing HA crystal size and disrupting HA orientation, thereby impairing the material properties of the bone matrix.

NF-κB/MAPK signaling activation via RAGE and Toll-like receptors: Impacts on bone cells

AGEs not only directly alter the physicochemical properties of the bone matrix but also impair bone cell function through receptor-mediated signaling pathways. The RAGE is the primary pattern recognition receptor for AGEs and is broadly expressed on osteoblasts, osteocyte-like cells, and osteoclasts. The binding of AGEs to RAGE activates intracellular signaling cascades, notably the NF-κB and MAPK pathways[76]. In osteoblasts, AGE-RAGE interactions trigger the nuclear translocation of NF-κB, increasing proapoptotic gene expression and downregulating osteogenic gene transcription[76]. Studies have shown that in high glucose-treated osteoblasts from diabetic rats, AGE accumulation can induce apoptosis via RAGE-mediated caspase-3 activation. Concurrently, sustained activation of the NF-κB and p38/c-Jun N-terminal kinase MAPK pathways suppresses osteoblast differentiation and matrix synthesis[77]. These effects culminate in reduced osteoblast numbers, diminished functional capacity, and lower bone formation rates. In osteocytes, the most abundant terminally differentiated bone cells, AGE-RAGE signaling exerts similar detrimental effects. Elevated levels of AGEs and other RAGE ligands [e.g., high mobility group box 1 (HMGB1) and S100 proteins] can trigger osteocyte apoptosis and promote the release of proinflammatory cytokines[78]. Moreover, AGEs increase the osteocytic expression of sclerostin, an inhibitor of the Wnt pathway, thereby further suppressing osteoblast activity and bone formation. Collectively, AGE-RAGE-NF-κB/MAPK-mediated apoptosis and differentiation blockade in osteoblasts and osteocytes represent critical mechanisms contributing to low bone turnover in patients with diabetic osteoporosis[78].

In addition, AGE-RAGE signaling also modulates osteoclastogenesis and resorptive activity. Various RAGE ligands, including AGEs, S100, HMGB1, and amyloid-β, have been shown to promote the differentiation of osteoclast precursors and enhance the activity of mature osteoclasts[76]. Mechanistically, RAGE activated NF-κB and extracellular signal-regulated kinase signaling crosstalk with the receptor activator of NF-κB ligand (RANKL)-receptor activator of NF-κB axis to amplify the expression of nuclear factor of activated T cells 1, a master transcription factor for osteoclastogenesis. Concurrently, AGEs can stimulate osteoblasts and osteocytes to secrete more pro-osteoclastogenic cytokines, such as interleukin (IL)-6 and tumor necrosis factor α (TNF-α), while NF-κB activation helps maintain osteoclast survival[76]. As a result, AGE accumulation leads to an imbalance between enhanced bone resorption and impaired bone formation, ultimately causing bone loss and structural deterioration.

In addition to RAGE, Toll-like receptors (TLRs), particularly TLR4, also play a significant role in AGE-related bone metabolic disturbances. TLR4 is an innate immune receptor activated by bacterial lipopolysaccharide and various endogenous danger signals. Notably, certain AGE-modified molecules (e.g., glycated albumin or AGE-induced HMGB1) may act as ligands to stimulate the TLR4 pathway[79]. TLR4 primarily activates NF-κB via a myeloid differentiation primary response 88-dependent mechanism, leading to proinflammatory gene expression. Within the skeletal system, the TLR4-myeloid differentiation primary response 88-NF-κB axis is crucial for osteoclastogenesis: Deletion of TLR4 significantly reduces the differentiation of osteoclast precursors and suppresses RANKL-driven osteoclast formation. Under diabetic conditions, chronic low-grade inflammation and elevated levels of endotoxin translocation, adipokines, or AGE-related damage-associated molecular patterns may lead to TLR4 overactivation, further amplifying NF-Κb-mediated bone resorption[80]. Indeed, in STZ-induced type 1 diabetic mice, TLR4-deficient animals exhibit markedly reduced bone loss and ameliorated osteoporotic phenotypes, suggesting that TLR4 hyperactivation contributes to diabetic bone deterioration.

Of particular interest is the crosstalk between TLR4 and RAGE signaling. For example, under hyperglycemic conditions, costimulation of preosteoblastic MC3T3-E1 (mouse calvaria preosteoblast cell line) cells with AGEs and low-dose endotoxin activates the TLR4-phospholipase C gamma 1-c-Jun N-terminal kinase pathway, leading to increased NF-κB activation and excessive production of inflammatory mediators[81]. Therefore, in diabetic bone disease, the RAGE and TLR4 pathways may synergistically amplify inflammatory signaling and suppress osteogenesis. Taken together, AGEs activate the NF-κB and MAPK cascades via RAGE and TLR4 in bone cells, disrupting osteoblast/osteocyte function while enhancing osteoclast activity. This imbalance in bone remodeling ultimately contributes to compromised bone strength and increased fragility.

Interplay between AGEs, oxidative stress, ferroptosis, and iron homeostasis in diabetic bone fragility

Chronic hyperglycemia is closely associated with the accumulation of AGEs, accompanied by elevated oxidative stress, which is recognized as a major driver of diabetic complications. AGEs can directly generate large amounts of ROS through glycoxidation reactions, while the interaction between AGEs and their receptor RAGE further activates ROS-generating enzymes such as nicotinamide adenine dinucleotide phosphate hydrogen oxidase[82,83]. In turn, oxidative stress accelerates AGE formation, forming a vicious cycle known as the “glycoxidation amplification loop”, since ROS promote the oxidation of sugars and lipids, producing more reactive carbonyl intermediates that drive nonenzymatic protein glycation[83].

Within the bone microenvironment, AGEs and oxidative stress jointly impair both cells and the extracellular matrix. ROS can directly damage osteoblast DNA and mitochondrial function, compromising their viability and differentiation; at the same time, ROS promote collagen degradation and mineral loss, weakening the structural integrity of the bone matrix. Together, these effects suppress osteogenic activity while enhancing osteoclastogenesis, leading to a metabolic imbalance in bone characterized by reduced formation and accelerated resorption[84]. Both clinical and animal studies have shown that diabetic osteoporosis is commonly associated with weakened antioxidant defense mechanisms [e.g., decreased activity of glutathione peroxidase and superoxide dismutase (SOD)] and increased lipid peroxidation, which are closely linked to AGE/ROS-mediated cellular damage[85,86].

In recent years, increasing evidence has suggested that ferroptosis, an iron dependent, nonapoptotic form of regulated cell death, may participate in the pathogenesis of diabetic bone fragility[87]. This process is characterized by intracellular iron overload and iron-catalyzed lipid peroxidation, leading to membrane rupture and cell death. AGEs may promote ferroptosis via RAGE signaling by upregulating transferrin receptor 1 and divalent metal transporter 1, activating nuclear receptor coactivator 4-mediated ferritinophagy, and downregulating glutathione peroxidase 4 and solute carrier family 7 member 11 expression, thereby triggering free iron accumulation and lethal lipid ROS. Iron chelators (e.g., deferoxamine) or ferroptosis inhibitors (e.g., ferrostatin-1) can reverse these phenotypes and restore impaired osteogenic function in diabetic animal models[88-90].

Extracellularly, AGEs can also carbonylate lysine and arginine residues in COL1 (e.g., forming CML, glucosepane), generating high-affinity binding sites for Fe²⁺/Fe³⁺. These bound iron ions can catalyze Fenton-like reactions, causing collagen fibril fragmentation, new carbonyl group formation, enhanced nonenzymatic cross-linking, and the inhibition of LOX-mediated pyridinoline cross-linking. The resulting iron-sensitive, carbonyl-rich matrix not only accelerates glycoxidation and increases dicarbonyl flux but also activates RAGE-NF-κB signaling to increase cellular iron uptake, forming a pathological “AGE-iron-ferroptosis” positive feedback loop. This leads to a state in which low-turnover osteoblast depletion coexists with a mechanically fragile collagen scaffold[82].

In vitro studies further support this mechanism: In the human osteoblast cell line hFOB1.19, AGE-enriched medium induces hallmarks of ferroptosis, including reduced cell viability, elevated lipid peroxidation, intracellular iron accumulation, and partial activation of apoptotic signals[85]. Serum from patients with diabetic osteoporosis exacerbates these injuries, suggesting that AGE-rich serum enhances ferroptotic phenotypes. Deferoxamine significantly attenuates AGE- and diabetic serum-induced osteoblast damage, reduces iron and lipid peroxidation levels, and delays cell death[91].

Similarly, animal studies revealed ferroptotic alterations in the bones of T2DM rats, such as reduced glutathione peroxidase 4 expression, elevated lipid peroxidation, and disruption of the osteocyte ultrastructure[86]. Ferroptosis not only reduces the number of osteoblasts but also may damage the osteocyte network, impairing its ability to sense and respond to mechanical loading, thereby interfering with bone remodeling. Moreover, iron overload promotes osteoclastogenesis and inhibits osteoblast differentiation, accelerating bone loss. AGEs that react with iron ions generate hydroxyl radicals, further exacerbating tissue damage[85].

In summary, diabetic bone fragility can be conceptualized as a pathological “triangle” involving AGE accumulation, redox imbalance, and disrupted iron homeostasis, with each component reinforcing the others. AGEs induce oxidative stress and ferroptotic tendencies; iron overload catalyzes Fenton chemistry to increase ROS production, which in turn promotes increased AGE formation, forming a vicious cycle. Intervening at any point in this loop (e.g., with antioxidants, iron chelators, or antiglycation agents) may represent an effective strategy to delay diabetic osteoporosis. Ultimately, the AGE-COL1 axis synergizes with oxidative and ferroptotic stress to mediate multilayered damage, substantially compromising bone quality and mechanical performance.

Noncollagenous matrix proteins and diabetic bone mineralization

NCPs constitute approximately 10% of the bone matrix and, despite their minor contribution to stiffness, play critical roles in regulating mineral deposition and bone toughness[92]. Key NCPs include OCN, osteopontin (OPN), and bone sialoprotein (BSP). OCN, the most abundant NCP in bone, is a 49-amino-acid osteoblast product that strongly binds HA via its γ-carboxylated glutamate residues. Together with OPN, OCN helps bond minerals to minerals in the extrafibrillar matrix, forming complexes that allow “sacrificial” slipping between mineral clusters, a nanoscale toughening mechanism dissipating energy during stress. Notably, genetic ablation of either OCN or OPN in mice impairs this sacrificial bonding, leading to significantly reduced fracture toughness[92]. BSP, in turn, is highly expressed during bone formation and contains domains for collagen binding and HA nucleation[93]. Both BSP and OPN localize to the mineralization front, where they initiate and regulate the growth of new mineral crystals[94]. BSP is considered a potent nucleator of HA, promoting crystal formation, whereas OPN binds nascent crystals to modulate their size and prevent premature or excessive growth[94,95]. In contrast, OCN is detected mainly in the later stages of mineralization (the fully mineralized matrix) and can limit overmineralization, as OCN deficiency leads to abnormally high mineral content but poor bone quality[96]. Through their coordinated actions, BSP initiates mineral nucleation, OPN restrains crystal growth, and OCN ensures optimal mineral maturation and mechanical bonding, these NCPs create a normally mineralized yet tough collagenous scaffold.

Diabetic conditions (chronic hyperglycemia with elevated oxidative stress) disrupt the expression and function of these NCPs, thereby disrupting the mineralization process. High glucose exposure alters osteoblast gene programs, often downregulating OCN production and other osteogenic markers[97]. In a hyperglycemic fracture model, the reactive dicarbonyl MGO (a byproduct of high glucose) caused a significant reduction in OCN and BSP expression in the bone callus, resulting in lower mineral density and a brittle callus structure[97]. This suppression is linked to impaired activation of the osteoblast transcription factor Osterix under oxidative/carbonyl stress[97]. Consistently, clinical studies have reported reduced circulating OCN in patients with T2DM[97], reflecting diminished bone formation. In contrast, OPN is often upregulated in diabetes. OPN levels are elevated approximately twofold in individuals with T2DM compared with normoglycemic controls, and they increase further as glucose tolerance worsens[95]. Elevated OPN is also observed in T1DM[95]. Chronic hyperglycemia and inflammation likely drive excess OPN expression as a maladaptive response. Because OPN potently inhibits HA crystal growth by binding to mineral surfaces, its overabundance in diabetic bone can aberrantly inhibit or delay normal mineral deposition[95]. Moreover, BSP dynamics are dysregulated: Preclinical models have shown that hyperglycemia leads to altered (often delayed or prolonged) BSP expression during bone healing[98]. The normal sequential upregulation of BSP may be blunted, and in some contexts, high glucose or AGEs instead decrease BSP availability[97]. Taken together, diabetes skews the timing and levels of NCPs; one study noted that OCN, OPN and BSP have altered expression profiles under diabetic conditions, with in vivo bone showing delayed and/or sustained presence of these proteins[98]. Such dysregulation of extracellular matrix components perturbs the orderly sequence of events in bone mineralization, including collagen fibril formation, mineral nucleation and crystal growth[98]. For example, insufficient BSP in early bone formation can impair initial crystal nucleation, whereas excessive OPN can prevent crystals from growing and maturing properly; the result may be a matrix with patches of unmineralized osteoid alongside regions of overmineralization, as remodeling is impaired. Indeed, diabetic bone often exhibits heterogeneous or abnormal mineralization, sometimes higher mineral density but lower toughness, which is consistent with these NCP imbalances[97,98].

In addition to expression changes, the biochemical integrity of NCPs is compromised by the diabetic milieu. AGEs form not only on collagen but also on NCPs, given the oxidative/carbonyl stress associated with hyperglycemia[92]. OCN has a short half-life, but recent evidence shows that it can undergo glycation: Molecular simulations demonstrate that in T2DM, OCNs may form abnormal covalent cross-links (e.g., between an arginine and the N-terminus) that disrupt its α-helical structure[92]. This glycation-induced unfolding markedly reduces OCN’s affinity for HA and abolishes the normal mechanical role of the protein in absorbing energy during bone deformation[92]. In essence, glycation “deadens” OCNs, preventing them from bridging mineral interfaces to dissipate fracture energy. Although less studied, OPN and BSP likely undergo similar modifications: AGEs can cross-link lysine/arginine residues or induce oxidative damage, potentially impairing OPN’s integrin-binding RGD sequence or BSP’s collagen-binding motifs. These posttranslational modifications further diminish NCP function in organizing the matrix. For example, the glycoxidation of BSP could hinder its ability to nucleate crystals, and heavily glycated OPN may form aggregates that cannot be properly anchored to HA or cells. Along with direct collagen glycation, such changes contribute to what might be called a “matrix quality deficit” in diabetes: Collagen fibers are stiffer and more brittle due to AGE cross-links, and the surrounding NCP network is depleted or functionally impaired and unable to compensate.

Overall, alterations in NCPs under diabetic conditions synergistically amplify the AGE-COL1 axis of bone fragility. The collagen backbone of diabetic bone is already compromised by nonenzymatic cross-links, and the loss or malfunction of NCP regulators removes key control points in mineralization and mechanical buffering. The net effect is a bone matrix that can become hypermineralized yet is structurally unsound, a densely packed mineral without the normal microstructural design that confers toughness. Empirically, diabetic bone can paradoxically have normal or high BMD but still presents inferior material properties and greater fracture risk (the “diabetic bone paradox”)[97,98]. This paradox is explained by matrix glycation coupled with NCP dysregulation: Collagen glycation stiffens the framework, whereas the dysregulated OPN/OCN/BSP axis leads to mislocalized or aberrant mineral deposits and a loss of toughening mechanisms. For example, a healthy bone’s resilience comes from a balance of cross-linked collagen and slip interfaces mediated by OCN/OPN; in diabetes, collagen is overcrosslinked by AGEs, and the slip interfaces are lost due to OCN/OPN damage, yielding a brittle, fracture-prone matrix[92]. Likewise, proper mineral crystal size and distribution (ensured by OPN and BSP) are distorted, so the collagen–mineral composite lacks the optimal architecture for absorbing impacts[98]. In summary, hyperglycemia and AGE accumulation not only target collagen (the focus of prior sections) but also disrupt the NCP scaffold that orchestrates mineralization. This dual hit, on collagen cross-linking and on mineral-regulating proteins, produces a bone matrix that is quantitatively mineralized yet qualitatively fragile. Restoring normal NCP function (for example, by mitigating oxidative stress or AGE formation) is therefore a critical aspect of improving bone quality in diabetes, complementing strategies that target the collagen-AGE axis. By integrating collagen-centric and NCP-centric mechanisms, we gain a more complete understanding of diabetic osteopathy: A condition in which the AGE-COL1-NCP interplay skews the fine-tuned collagen-mineral balance, ultimately leading to bone that is brittle despite its seemingly intact mineral content.

PRECLINICAL ADVANCES IN TARGETING THE AGE-COL1 AXIS

Mechanisms and therapeutic potential of AGE inhibitors

Inhibitors of AGE formation aim to interrupt nonenzymatic reactions between sugars and proteins, thereby reducing AGE accumulation in tissues and protecting the bone matrix. Representative agents include aminoguanidine, pyridoxamine, and metformin. Aminoguanidine is a classical AGE inhibitor that scavenges reactive carbonyl intermediates such as MGO, thus preventing the formation of AGE precursors[99]. In diabetic models, aminoguanidine significantly reduces AGE deposition in bone and improves the cross-linking pattern of COL1[9]. A recent study using db/db mice, a well-established model of T2DM, demonstrated that aminoguanidine administration markedly improved trabecular microarchitecture, biomechanical strength, and fracture healing[9]. These findings suggest that inhibiting AGE formation may partially reverse the paradox of diabetic bone fragility, where BMD is preserved or even elevated despite reduced bone strength[9].

Metformin, a first-line antidiabetic agent, also has antiglycation properties. On the one hand, it indirectly decreases AGE production by improving hyperglycemia; on the other hand, it directly scavenges dicarbonyl compounds such as MGO through irreversible adduct formation, thereby limiting AGE generation[100-102]. Experimental studies have shown that metformin promotes osteogenesis in diabetic rats by activating the AMP-activated protein kinase (AMPK) pathway and reducing AGE accumulation in bone[9]. In the aforementioned db/db mouse study, metformin produced bone improvements comparable to those of aminoguanidine, with similarly increased bone mass and strength[9].

Pyridoxamine, an active form of vitamin B6, has demonstrated strong antiglycation effects as an AGE formation inhibitor. Its mechanisms include scavenging reactive carbonyl intermediates and exerting antioxidant activity[99]. These findings indicate that pyridoxamine effectively reduces AGE levels in diabetic bone tissue and restores osteoblast differentiation. For example, in STZ-induced diabetic mice, pyridoxamine administered via the drinking water (1 g/L) significantly accelerated the healing of femoral defects. Micro-computed tomography and histological analyses confirmed that new bone formation in the pyridoxamine-treated group was markedly greater than that in the untreated diabetic control group[103]. Additionally, pyridoxamine counteracts MGO-induced suppression of osteogenic differentiation in vitro, restoring alkaline phosphatase (ALP) activity and other osteoblastic markers[103]. These results collectively support the therapeutic potential of AGE formation inhibitors in preventing or ameliorating diabetes-associated bone damage. Clinically, the strategy of AGE inhibition is beginning to show promise. A randomized controlled trial (RCT) involving 55 elderly women with T2DM demonstrated that oral pyridoxamine (200 mg, twice daily for one year) increased the bone formation marker P1NP, significantly improved BMD, and reduced HbA1c levels[104]. Although no significant changes in BTMs or skin autofluorescent AGE levels were detected, these findings suggest a potential role for pyridoxamine in improving bone quality via AGE suppression[104].

Biological effects of AGE cross-link breakers and their impact on bone quality

AGE cross-link breakers aim to cleave the covalent bonds already formed between AGEs and structural proteins, primarily collagen, thereby restoring the mechanical integrity of the extracellular matrix. ALT-711 (alagebrium) is the most representative AGE cross-link breaker. As a thiazolium-based compound, it disrupts protein-protein cross-links mediated by AGEs. Animal studies have shown that ALT-711 reduces the AGE burden in various tissues and improves tissue elasticity. For example, in a rat model of CKD-mineral and bone disorder, 10 weeks of ALT-711 treatment significantly reduced the total AGE content in bone tissue[105]. Specifically, ALT-711 at a dose of 3 mg/kg lowered the level of advanced glycation cross-links in bone and partially reduced cortical bone porosity. However, it did not significantly improve the biomechanical strength of the bone in this model[105]. These findings suggest that although cross-linking breakers can partially remove accumulated AGEs from the bone matrix, their monotherapy may be insufficient to reverse skeletal fragility in patients with pathological conditions such as diabetes. Additional strategies to increase bone metabolism may be needed for optimal effects[106].

Notably, AGE cross-linking breakers have also demonstrated anti-inflammatory and tissue-protective effects in models of other complications. For example, ALT-711 was reported to reduce collagen cross-linking and inflammation in the vascular walls of diabetic mice, which may indirectly benefit bone by improving vascular supply and nutrient delivery[107]. Overall, the use of AGE cross-link breakers provides a mechanistically distinct approach from that of AGE formation inhibitors in reducing the AGE burden. Although current evidence shows limited efficacy on bone strength alone, their use in combination therapies, such as calcium supplementation or antioxidant interventions, may hold promise for enhancing bone quality.

Strategies for inhibiting RAGE signaling and their mechanisms of action

RAGE mediates the deleterious effects of AGEs on cellular signaling. In the diabetic state, hyperactivation of the AGE-RAGE axis in bone tissue contributes to chronic inflammation and impaired osteogenesis. Thus, targeting RAGE signaling has emerged as a promising therapeutic strategy for mitigating diabetic bone fragility. Small-molecule RAGE antagonists, such as FPS-ZM1, have shown encouraging results in vitro and in animal models[108]. FPS-ZM1 binds to RAGE with high affinity, effectively blocking its interaction with AGEs. Under high-glucose conditions, treatment with FPS-ZM1 significantly reduces the expression of inflammatory cytokines such as IL-1β and TNF-α in bone marrow-derived mesenchymal stem cells and suppresses excessive activation of the thioredoxin-interacting protein/NLRP3 (NACHT, LRR and PYD domains-containing protein 3) inflammasome[109]. Moreover, FPS-ZM1 restored the impaired osteogenic differentiation induced by hyperglycemia, suggesting that RAGE-mediated inflammatory pathways play a central role in diabetic osteogenic dysfunction[108].

In addition to the use of small-molecule inhibitors, increasing the level of soluble RAGE (sRAGE) represents another important approach. sRAGE, the truncated soluble isoform of RAGE, acts as a decoy receptor by binding circulating AGEs, thereby preventing their interaction with membrane-bound RAGE. Reduced serum sRAGE levels have been associated with increased osteoporosis risk in patients with T2DM, implying that increased sRAGE levels may confer skeletal protection[110].

In animal models, exogenous administration of sRAGE has been shown to attenuate AGE-RAGE-mediated inflammatory responses. For example, in a diabetic mouse model of periodontitis, blockade of RAGE signaling via sRAGE- or RAGE-neutralizing antibodies markedly reduced alveolar bone loss and local inflammatory infiltration[111]. Moreover, RAGE monoclonal antibodies have been demonstrated to suppress NF-κB activation in macrophages and osteoblasts in preclinical inflammatory models, although their application in bone research remains limited. Overall, multiple RAGE-targeting strategies, including small-molecule inhibitors, sRAGE supplementation, and monoclonal antibodies, can mitigate AGE-induced oxidative stress and inflammation to varying extents[108,111]. These findings support the feasibility of the use of RAGE as a therapeutic target in diabetic bone disease. However, given the broad physiological roles of RAGE, long-term and systemic inhibition requires careful evaluation of its safety and potential off-target effects.

Advances in anti-glycation compounds with bone-protective properties

To date, a variety of antiglycation compounds have demonstrated bone-protective effects in diabetic animal models and in vitro studies. AGE formation inhibitors such as aminoguanidine and pyridoxamine have been shown to reduce AGE accumulation within the bone matrix, promote defect healing, and restore osteoblastic function. Metformin, in addition to its hypoglycemic effect, has been found to improve trabecular microarchitecture and enhance bone strength in diabetic rodents[9]. ALT-711, an AGE cross-link breaker, effectively reduces the total AGE content and cortical porosity in bone tissue; however, its impact on overall mechanical strength remains limited[105]. FPS-ZM1, a potent RAGE antagonist, inhibits inflammasome activation and upregulates osteogenic gene expression in high-glucose environments[108]. Natural compounds such as silibinin and resveratrol have also demonstrated therapeutic benefits: Silibinin alleviates bone loss by modulating the RAGE signaling pathway and inhibiting osteoblast apoptosis, whereas resveratrol enhances matrix mineralization and mitigates bone quality deterioration under diabetic conditions[112,113]. Collectively, these findings highlight the therapeutic potential of targeting the AGE-COL1 axis in the prevention and treatment of diabetic bone fragility (Table 5).

Table 5 Research progress on anti-glycation compounds.

Intervention	Model and type	Dosage and duration	Evaluation indicators	Summary of main findings	Ref.
Aminoguanidine (AGE formation inhibitor)	db/db genetic T2DM mice (in vivo)	100 mg/kg/day, intraperitoneal injection, 8 weeks (estimated)	Femoral BMD, microarchitecture, biomechanical strength	Reduced AGE accumulation in bone matrix; increased BMD and trabecular number; significantly improved 3-point bending load	[8]
Pyridoxamine (AGE formation inhibitor)	STZ-induced T1DM bone defect model (in vivo); MC3T3-E1 osteoblasts (in vitro)	1 g/L in drinking water for 4 weeks; 50-500 μM for cells	Bone defect CT imaging, histology; ALP activity	Accelerated bone defect healing; increased bone density in defect site within 7-14 days; rescued MGO-induced ALP suppression in vitro	[91]
Metformin (AGE inhibition/antihyperglycemic)	db/db T2DM mice (in vivo)	200 mg/kg/day, oral gavage, 12 weeks (estimated)	Bone volume fraction (BV/TV), biomechanical strength	Increased trabecular bone volume and cortical thickness; improved bending strength; inhibited AGE accumulation in bone	[8]
ALT-711 (AGE crosslink breaker)	Cy/+ chronic kidney disease rats (diabetic osteoporosis-like, in vivo)	3 mg/kg/day, intraperitoneal injection, 10 weeks	Bone AGE content, porosity, mechanical strength	Decreased total bone AGE levels and cortical porosity; no significant improvement in biomechanical strength	[104]
FPS-ZM1 (RAGE small-molecule antagonist)	High-glucose-treated bone marrow mesenchymal stem cells (in vitro)	5 μM for 24 hours	Inflammatory markers (e.g., IL-6), osteogenic markers	Inhibited RAGE and TXNIP/NLRP3 inflammasome; reduced IL-1β and IL-6; upregulated ALP and osteogenic gene expression	[107]
Silybin (natural flavonolignan)	STZ-induced diabetic rats (in vivo); MC3T3-E1 cells (in vitro)	50 mg/kg/day intraperitoneal injection, 6 weeks; 100 μM in cells	BMD, bone strength; osteoblast apoptosis rate	Attenuated diabetic bone loss, increased BMD; inhibited AGE-induced apoptosis by downregulating RAGE and mitochondrial pathway	[111]
Resveratrol (natural polyphenol)	STZ-induced diabetic bone defect model (in vivo)	10 mg/kg/day oral gavage, 8 weeks	Bone regeneration (μCT), serum AGE levels	Promoted mineralized bone formation in defect site; reduced AGE deposition in bone; improved bone matrix quality	[112]

AGE: Advanced glycation end-product; T2DM: Type 2 diabetes mellitus; STZ: Streptozotocin; T1DM: Type 1 diabetes mellitus; MC3T3-E1: Mouse calvaria preosteoblast cell line; Cy/+: Heterozygous cyclophilin mutant (chronic kidney disease model); BMD: Bone mineral density; CT: Computed tomography; ALP: Alkaline phosphatase; BV/TV: Bone volume/total volume; IL-6: Interleukin-6; μCT: Micro-computed tomography; MGO: Methylglyoxal; RAGE: Receptor for advanced glycation end-products; TXNIP: Thioredoxin-interacting protein; NLRP3: NACHT, LRR and PYD domains-containing protein 3; IL-1β: Interleukin-1β.

Translational challenges and clinical evidence of AGE-targeting therapies

However, evidence supporting AGE-targeted therapies in humans remains limited[114]. Several candidate compounds have encountered translational barriers. For example, the AGE formation inhibitor aminoguanidine progressed to phase 3 trials for diabetic nephropathy but was discontinued due to toxic side effects[114]. Another agent, pyridoxamine, has demonstrated modest efficacy in DKD and has recently been evaluated for skeletal benefits in diabetes[114]. In a 12-month randomized trial involving older women with T2DM (n = 55), pyridoxamine treatment modestly improved both bone turnover and bone density[104]. Compared with those in the placebo group, the levels of the bone formation marker P1NP were increased by approximately 23% in the pyridoxamine group, and the femoral neck BMD was significantly increased (+2.6% vs -0.9%; P = 0.007). Glycemic control also improved (HbA1c -0.38% vs +0.05%; P = 0.04). Although the levels of bone resorption markers and skin autofluorescence (a surrogate for AGE burden) did not appreciably change, the increase in P1NP was inversely correlated with the reduction in HbA1c. These findings suggest that inhibiting AGE formation may enhance bone formation and improve metabolic control in T2D patients, although larger trials are needed to assess its impact on fracture risk[104].

In contrast, attempts to reverse existing AGE cross-links have been less successful. ALT711, a compound designed to cleave AGE crosslinks, advanced to clinical trials for cardiovascular disease but failed to demonstrate benefit in a phase 3 trial targeting arterial stiffness[114]. Its development was subsequently discontinued, and no clinical studies to date have evaluated its potential for preventing osteoporotic fractures. Even in preclinical settings, ALT711 reduced AGE accumulation in bone but did not improve mechanical strength[114]. Collectively, these limitations highlight the translational hurdles faced by AGE-targeted strategies. Despite the strong mechanistic rationale, demonstrating clear skeletal efficacy and safety in humans remains a significant challenge for future research[104,114].

SGLT-2 INHIBITORS: PHARMACOLOGICAL MECHANISMS BEYOND GLYCEMIC CONTROL IN DIABETES MANAGEMENT

Pharmacodynamic characteristics and structural diversity

SGLT-2 inhibitors reduce renal glucose reabsorption by inhibiting SGLT-2 in the proximal tubules, thereby inducing glycosuria and lowering blood glucose levels[115,116]. This glucose-lowering mechanism is insulin independent, making these agents applicable across different stages of diabetes with a low risk of hypoglycemia[115,117]. Commonly used SGLT-2 inhibitors include dapagliflozin, empagliflozin, and canagliflozin, all of which are C-glucoside derivatives but differ in their pharmacokinetic profiles. Most SGLT-2 inhibitors have an elimination half-life of approximately 10-13 hours, allowing for once-daily administration[118]. Specifically, dapagliflozin and ipragliflozin exhibit a half-life of approximately 12 hours following oral intake, whereas tofogliflozin has a shorter half-life of approximately 5 hours[118].

These agents are rapidly absorbed after oral administration, with high bioavailability (e.g., > 90% for empagliflozin) and strong plasma protein binding (91% for dapagliflozin, 86% for empagliflozin, and 99% for canagliflozin)[119]. The high protein-binding affinity reflects their lipophilicity, with canagliflozin being the most lipophilic and empagliflozin the least lipophilic, which influences tissue distribution and metabolic pathways[119].

SGLT-2 inhibitors are metabolized primarily in the liver and kidneys via uridine diphosphate-glucuronosyltransferase (UGT) enzymes into inactive glucuronide conjugates, which are subsequently excreted in feces and urine[119]. Specific metabolic pathways vary: Dapagliflozin is metabolized mainly by UGT1A9, empagliflozin is metabolized by UGT2B7 and others, and canagliflozin is metabolized by UGT1A9 and UGT2B4[119]. Because unchanged parent compounds are renally excreted at very low proportions (< 2%), most are eliminated in metabolized forms[120]. This implies that in patients with impaired renal function, drug clearance slows, and glycosuric efficacy decreases[121]. Nevertheless, most SGLT-2 inhibitors remain safe for use in mild to moderate renal impairment, although dose adjustment or avoidance is recommended on the basis of the eGFR to ensure pharmacokinetic stability[121].

The selectivity for SGLT-2 vs SGLT-1 varies across different agents, depending on the molecular structure. Empagliflozin has the highest selectivity (approximately 2500-fold for SGLT-2 over SGLT-1), followed by dapagliflozin (approximately 1200-fold) and canagliflozin (approximately 250-fold)[122]. Lower selectivity, as observed with canagliflozin, may lead to partial inhibition of intestinal SGLT-1 at higher doses, delaying postprandial glucose absorption[118]. Proline-canagliflozin is a structural derivative of canagliflozin with an L-proline moiety incorporated. This proline modification enhances molecular stability and hydrophilicity, leading to improved oral absorption and consistent drug exposure. It has been approved in China as a fourth-generation SGLT-2 inhibitor for the treatment of T2DM[123]. In summary, SGLT-2 inhibitors display diverse pharmacokinetic characteristics. Differences in half-life and lipophilicity influence the duration of action and tissue distribution, whereas selectivity variations may confer additional effects or adverse reactions. This structural diversity supports individualized clinical decision-making for diabetes management[118,122].

Effects on insulin-independent glucose disposal, ketone bodies, and mineral metabolism

Insulin-independent glycemic control and ketogenesis: SGLT-2 inhibitors lower blood glucose by promoting urinary glucose excretion, removing tens of grams of glucose daily, without the need for insulin-mediated uptake[115]. This approach is especially beneficial in advanced-stage diabetes patients with β-cell failure or severe insulin resistance, as it reduces hyperglycemia without increasing the risk of hypoglycemia[115,117]. Long-term use also alleviates glucotoxicity, thereby reducing β-cell workload and improving insulin sensitivity in peripheral tissues[116].

The accompanying caloric loss and mild osmotic diuresis trigger a shift in systemic energy metabolism, promoting lipolysis and fatty acid oxidation, which in turn moderately increases the number of circulating ketone bodies[116]. Both clinical and preclinical studies have confirmed that SGLT-2 inhibitor therapy leads to a mild increase in plasma β-hydroxybutyrate (BHB) and other ketones, an effect that is generally well tolerated in most patients with T2DM[116]. As an “alternative fuel” to glucose, ketone bodies can be efficiently utilized by organs such as the heart and kidneys, providing increased energy efficiency and potential cytoprotective effects. Some researchers have proposed the “ketone hypothesis”, suggesting that improvements in myocardial and renal bioenergetics from ketone metabolism contribute to the cardiorenal benefits of SGLT-2 inhibitors.

Moreover, ketones exert antioxidant and anti-inflammatory effects, and their moderate elevation may help suppress chronic complications such as atherosclerosis. However, in rare cases, especially in patients with T1DM or severely compromised insulin levels, SGLT-2 inhibitors may induce euglycemic diabetic ketoacidosis (DKA). This is thought to involve increased glucagon secretion and impaired renal ketone clearance. Therefore, ketone monitoring and avoidance of prolonged fasting are recommended in susceptible populations[124]. Overall, SGLT-2 inhibitors improve systemic energy homeostasis by reducing glucose levels through insulin-independent pathways and inducing mild ketogenesis, which may be advantageous for patients with T2DM and comorbid heart failure[116].

Natriuresis and hemodynamic effects: SGLT-2 inhibitors also influence electrolyte balance and hemodynamic stability. As the SGLT-2 cotransporter reabsorbs one molecule of sodium for every molecule of glucose, its inhibition enhances renal sodium excretion (natriuresis)[103]. Clinical studies have reported that SGLT-2 inhibitors induce mild to moderate diuresis and volume contraction, leading to reductions in both systolic and diastolic blood pressure. This contributes to the alleviation of hypertension and reduced cardiac preload in patients with diabetes[125]. Interestingly, unlike conventional diuretics, SGLT-2 inhibitors are associated with fewer electrolyte disturbances: They exert minimal effects on serum potassium (and may even reduce the risk of hyperkalemia), whereas hyponatremia and hypomagnesemia are also uncommon[126,127]. These differences may be attributed to their diuretic action being located primarily in the proximal tubule rather than directly affecting distal electrolyte transporters[122]. In addition, SGLT-2 inhibitors reduce glomerular hyperfiltration and intraglomerular pressure, thereby mitigating the progression of diabetic nephropathy[122]. Overall, their natriuretic effect produces modest blood pressure and volume reductions with favorable electrolyte safety, forming a key component of their cardiorenal protective mechanisms.

Mineral metabolism and bone turnover: A growing body of evidence suggests that SGLT-2 inhibitors can influence mineral homeostasis, particularly calcium, phosphate, and magnesium, as well as bone metabolic hormones. Short-term studies have shown that treatment with SGLT-2 inhibitors results in mild increases in serum phosphate levels, along with elevations in PTH and fibroblast growth factor-23 (FGF-23), accompanied by a reduction in active vitamin D (1,25-dihydroxyvitamin D) levels[128]. For example, in patients with T2DM, empagliflozin treatment transiently increases serum phosphate, PTH, and FGF-23 while decreasing 1,25-dihydroxyvitamin D levels[129].

Mechanistically, these changes are believed to be associated with altered sodium delivery to renal tubules, which affects phosphate reabsorption. Specifically, the inhibition of SGLT-2 reduces proximal tubule sodium-glucose cotransport. To maintain luminal sodium gradients, sodium-phosphate cotransporters (NaPi-IIa/IIc) in the proximal tubule may be upregulated, thereby increasing phosphate reabsorption and increasing serum phosphate levels[21]. Elevated phosphate then stimulates osteocytes to secrete FGF-23, which acts on the kidneys to suppress phosphate reabsorption and inhibit 1α-hydroxylase activity, reducing the synthesis of active vitamin D. The resulting decline in 1,25-dihydroxyvitamin D impairs intestinal calcium absorption, potentially leading to mild hypocalcemia, which in turn triggers increased PTH secretion. Thus, SGLT-2 inhibitor-induced phosphate dysregulation manifests as an activated high phosphate-FGF-23-PTH axis.

Although these changes are typically modest and transient, mostly occurring during early treatment, they may affect bone turnover[21,130]. A meta-analysis of RCTs involving 22828 patients revealed that SGLT-2 inhibitors significantly increased serum PTH and the C-terminal telopeptide of COL1 (CTX) while slightly decreasing bone-specific ALP, suggesting enhanced bone resorption and reduced bone formation[130]. However, their effects on BMD and fracture risk appear to be neutral: The same meta-analysis revealed no significant impact of SGLT-2 inhibitors on BMD and no increased risk of fractures[130]. Animal studies have yielded mixed findings. In nondiabetic mice, long-term genetic deletion of SGLT-2 did not significantly alter the serum levels of PTH, FGF-23, or 1,25-dihydroxyvitamin D, although it did lead to a slight increase in bone fragility[131]. In contrast, in diabetic models, SGLT-2 inhibitors may exert dual effects: On the one hand, they improve hyperglycemia and mitigate AGE-induced damage to the bone matrix, conferring potential bone-protective effects (discussed below); on the other hand, they may transiently induce a high phosphate-PTH state that is detrimental to bone anabolism. These opposing effects may counterbalance each other, possibly explaining the overall neutral fracture risk observed clinically[130]. Further long-term studies in specific populations are needed to clarify the net impact of SGLT-2 inhibitors on skeletal health[132]. At present, it is recommended that patients with T2DM who are at high risk of fracture undergo regular monitoring of bone metabolic markers and ensure adequate intake of calcium and vitamin D during SGLT-2 inhibitor therapy[21,130].

Mechanisms by which SGLT-2 inhibitors modulate AGE burden

AGE accumulation under hyperglycemic conditions is considered a key driver of diabetic complications. SGLT-2 inhibitors reduce the systemic AGE burden through multiple mechanisms, including the suppression of AGE precursor formation, enhancement of AGE clearance, and downregulation of RAGE-mediated pathogenic signaling.

Reduction in AGE formation: SGLT-2 inhibitors substantially improve chronic hyperglycemia and glycemic variability, thereby reducing the generation of AGE precursors at the source. In particular, reactive carbonyl compounds such as MGO, a byproduct of glucose metabolism outside the tricarboxylic acid cycle, are major precursors of AGEs. MGO reacts with lysine and arginine residues on proteins to form AGE adducts such as MG-H1. Studies have shown that SGLT-2 inhibitors such as empagliflozin enhance glucose utilization and fatty acid oxidation, thereby decreasing the metabolic flux toward MGO and other reactive dicarbonyls[116,133]. A prospective clinical trial comparing SGLT-2 inhibitors and dipeptidyl peptidase-4 (DPP-4) inhibitors demonstrated that only the SGLT-2 inhibitor group exhibited a significant reduction in plasma MG-H1 Levels after 3 months of treatment (along with a reduction in HbA1c), whereas the DPP-4 inhibitor group, despite improved glycemic control, showed no significant change in MG-H1[121]. These findings suggest that the AGE-lowering effect of SGLT-2 inhibitors extends beyond glycemic control and may involve the suppression of carbonyl stress or enhanced detoxification of MGOs[134]. Similar reductions in MG-H1 and other AGE precursors have been observed in animal models, where SGLT-2 inhibitor treatment decreased the AGE content in diabetic rat serum and tissues[135]. These findings indicate that SGLT-2 inhibitors mitigate the upstream accumulation of AGE precursors by correcting hyperglycemia and metabolic imbalance.

Promotion of AGE clearance and detoxification: In addition to their ability to suppress formation, SGLT-2 inhibitors may facilitate the clearance and detoxification of AGEs. One major route of AGE elimination is renal filtration followed by catabolism in proximal tubular cells[136]. In diabetes, however, excessive AGE production combined with progressive renal dysfunction impairs this clearance pathway, promoting systemic AGE accumulation. By reducing glomerular hyperfiltration and preserving renal function, SGLT-2 inhibitors indirectly support renal AGE clearance capacity[122,137]. Studies have shown that AGE filtration flux increases several-fold in individuals with diabetes but decreases with the onset of kidney injury[138]. Thus, by slowing the progression of diabetic nephropathy, SGLT-2 inhibitors may help prolong the systemic clearance of AGEs over time.

In addition, SGLT-2 inhibitors enhance endogenous antioxidant defense systems, including the upregulation of glucuronosyltransferases and glutathione-related enzymes, which promote the detoxification and excretion of water-soluble metabolites[139]. Although direct clinical evidence of increased renal AGE excretion following SGLT-2 inhibitor use remains limited, the renoprotective effects of SGLT-2 inhibitors support the hypothesis of improved AGE elimination. However, further studies are needed to clarify whether SGLT-2 inhibitors enhance endocytic degradation of AGE-RAGE complexes or activate lysosomal degradation pathways.

Downregulation of RAGE expression and attenuation of oxidative stress: RAGE signaling is the principal mediator of AGE-induced inflammation and cellular injury. In diabetic animal models and in vitro studies, SGLT-2 inhibitors have been shown to downregulate RAGE expression across multiple tissues, thereby weakening the pathological AGE-RAGE axis. For example, in STZ-induced diabetic rats, empagliflozin treatment significantly reduced renal expression of both AGEs and RAGE, accompanied by a decrease in the oxidative stress marker 8-hydroxy-2’-deoxyguanosine[139]. Inflammatory and fibrotic gene expression was also suppressed, suggesting attenuation of AGE-RAGE-induced inflammatory responses.

Similarly, in high glucose-induced human renal proximal tubule cells, SGLT-2 inhibitors reduce intracellular AGE accumulation and RAGE protein levels. In a separate study on podocytes, exogenous AGE stimulation upregulated both SGLT-2 and RAGE expression, leading to cellular injury. Pretreatment with dapagliflozin counteracted these effects and restored the expression of the podocyte marker synaptopodin[140]. These findings suggest that SGLT-2 inhibitors may disrupt the vicious feedback loop between AGEs and RAGE, thereby protecting cells from AGE toxicity.

In addition to their effects on the kidney, SGLT-2 inhibitors have also demonstrated anti-AGE-RAGE effects in atherosclerosis models, where they decrease RAGE expression and oxidative stress in macrophages, mitigating vascular inflammation[141,142]. Suppression of oxidative stress is a core component of the “beyond-glycemic” effects of SGLT-2 inhibitors. These drugs not only reduce ROS generation via AGE-RAGE blockade but also activate intracellular antioxidant pathways such as the AMPK/sirtuin 1 pathway, leading to increased expression of antioxidant enzymes[139]. In diabetic animals, SGLT-2 inhibitors increase renal levels of superoxide dismutase and peroxidases while reducing the deposition of the nitrative stress marker 3-nitrotyrosine[139]. These changes help restore redox homeostasis and protect tissues from hyperglycemia-induced oxidative damage.

In summary, SGLT-2 inhibitors diminish the systemic AGE burden by restraining AGE formation, accelerating their renalhepatic clearance, downregulating RAGE expression, and limiting oxidative stress, actions that collectively underpin their cardiovascular, renal, and skeletal benefits (Figure 3)[134,139]. Coupled with their hallmark glycosuric and natriuretic effects, the attendant shift toward ketone and lipid utilization, and direct modulation of the AGE-COL1 axis, these pharmacodynamic attributes provide the mechanistic framework for interpreting the bone-specific outcomes evaluated in the next section.

Figure 3 Schematic illustration of the mechanisms by which sodium-glucose cotransporter 2 inhibitors alleviate diabetic bone fragility via multitarget modulation of the advanced glycation end product-type I collagen-receptor for advanced glycation end products axis. Persistent hyperglycemia accelerates the accumulation of advanced glycation end product (AGEs) in both the vasculature and bone matrix, promoting nonenzymatic type I collagen (COL1). This process increases collagen fibril stiffness and brittleness, thereby compromising the mechanical integrity of the bone matrix. Excess AGEs bind to their receptor, triggering downstream inflammatory and apoptotic pathways, including nuclear factor-κB and mitogen-activated protein kinase signaling, and amplifying reactive oxygen species-mediated oxidative stress. These cascades impair osteoblast function, promote osteoclast differentiation, and lead to reduced bone formation, enhanced bone resorption, and degradation of the COL1 structure, ultimately resulting in disrupted bone remodeling and osteoporosis. Sodium-glucose cotransporter 2 inhibitors not only reduce blood glucose levels but also suppress AGE formation, inhibit AGE-COL1 crosslinking, and downregulate receptor for advanced glycation end products signaling. Collectively, these effects help preserve bone structure and function in diabetic conditions. AGEs: Advanced glycation end products; SGLT-2: Sodium-glucose cotransporter 2; COL1: Type I collagen; RAGE: Receptor for advanced glycation end products; NF-κB: Nuclear factor-κB; MAPK: Mitogen-activated protein kinase; IL-6: Interleukin 6; TNF-α: Tumor necrosis factor α; ROS: Reactive oxygen species.

PROTECTIVE EFFECTS OF SGLT-2 INHIBITORS ON THE SKELETAL SYSTEM IN T2DM: PRECLINICAL EVIDENCE

Building on the pharmacological framework established in section 6, this section synthesizes translational evidence linking SGLT-2 inhibitors to bone turnover, microarchitectural integrity, and fracture risk. We first examine preclinical data that connect glomerular hemodynamic changes, altered phosphate and calcium handling, and reduced AGE formation with improvements in osteoblast function and collagen quality. We then critically appraise major clinical trials, highlighting apparent discrepancies such as CANVAS (Canagliflozin Cardiovascular Assessment) vs CREDENCE (Canagliflozin and Renal Events in Diabetes with Established Nephropathy Clinical Evaluation), and discuss how early osmotic diuresis, long-term AGE attenuation, and other context-specific factors may together shape net skeletal outcomes.

Animal studies: Improvement of bone microarchitecture and mechanical strength in diabetic models

A growing body of animal research indicates that SGLT-2 inhibitors can partially reverse diabetes-induced skeletal fragility in T2DM rat and mouse models[143,144]. Untreated diabetic animals often exhibit compromised trabecular and cortical bone architecture, including reduced BMD and bone volume (BV) fraction, cortical thinning with increased porosity, and diminished biomechanical strength[143]. For example, in TallyHo diabetic mice, hyperglycemia significantly reduces the femoral cortical cross-sectional area and thickness, increases cortical porosity, and leads to a lower trabecular BV fraction [BV/total volume (BV/TV)] and trabecular thickness in the distal femur and vertebrae. These structural impairments translate into decreased bone mechanical properties, such as yield force and ultimate load, in three-point bending tests[143,145].

Treatment with SGLT-2 inhibitors can attenuate these bone deficits. In a study by Thrailkill et al[143], 12 weeks of canagliflozin treatment in diabetic mice significantly improved femoral cortical bone structure, by increasing cortical area and thickness, and partially prevented trabecular BV/TV loss. However, the authors noted that even after normalization of glycemia and body weight, the parameters of bone strength did not fully recover, suggesting that the improvement in diabetic bone fragility was incomplete. These findings indicate that although SGLT-2 inhibitors can ameliorate bone microarchitecture, they may not fully restore the mechanical competence of diabetic bone.

Similarly, Londzin et al[144] compared the effects of dapagliflozin and canagliflozin in a high-fat diet plus STZ-induced T2DM rat model. Both agents effectively reduce blood glucose and partially mitigate reductions in bone mass and density, although they correct only a subset of diabetes-induced skeletal impairments. After 4 weeks of treatment, osteoporotic features such as femoral and tibial BMD loss and trabecular rarefaction were alleviated. Notably, the skeletal effects varied by anatomical site: Canagliflozin resulted in greater improvement in the cortical bone mechanical strength of the femoral shaft, whereas dapagliflozin had a relatively greater effect on the trabecular microarchitecture (e.g., trabecular thickness)[146]. Nevertheless, neither drug fully reversed the mechanical weakening caused by diabetes, highlighting the need for longer treatment durations or combination therapeutic strategies.

Interestingly, some studies suggest that the skeletal effects of SGLT-2 inhibitors may depend on the diabetic state. In nondiabetic rats, the administration of dapagliflozin or canagliflozin paradoxically led to deterioration in trabecular parameters, such as decreased BV/TV and trabecular thinning[21], implying that SGLT-2 inhibitors may exert different metabolic effects on bone under normoglycemic conditions. These findings warrant cautious interpretation when animal data are translated to clinical settings[144]. Overall, in diabetic rodent models, appropriate dosing and duration of SGLT-2 inhibitor treatment generally improve trabecular architecture (e.g., increased BV/TV and trabecular number) and cortical thickness, thereby partially enhancing bone biomechanical properties, including yield strength in bending tests[21,143]. These preclinical results provide experimental support for the use of SGLT-2 inhibitors in the prevention and treatment of diabetes-associated osteoporosis.

Cellular and molecular mechanisms: Effects on osteoblasts/osteoclasts and coupling regulation

At the cellular level, SGLT-2 inhibitors exert beneficial effects by promoting osteoblastogenesis and inhibiting osteoclastogenesis, primarily through the modulation of energy metabolism pathways and alleviation of glucotoxic damage. Under high-glucose culture conditions, the addition of canagliflozin significantly promoted the differentiation of osteoblast precursors, such as the murine preosteoblast cell line MC3T3-E1[147]. Mechanistically, this effect is associated with increased phosphorylation of AMPK in osteoblasts, which activates the key transcription factor Runt-related transcription factor 2 (RUNX2), which is involved in osteoblast differentiation[147]. Upregulation of RUNX2 subsequently enhances the expression of ALP and the formation of mineralized nodules, indicating enhanced osteogenic maturation[147]. Notably, the osteogenic effects of canagliflozin are partially abrogated by the AMPK inhibitor compound C, while the AMPK agonist 5-aminoimidazole-4-carboxamide ribonucleoside mimics and potentiates this response, reinforcing the central role of the AMPK-RUNX2 axis[147]. Consistent with in vitro findings, similar changes have been observed in bone tissue from diabetic mice: Compared with untreated animals, those treated with canagliflozin presented higher levels of phosphorylated AMPK and RUNX2 in bone tissue[147]. These results suggest that activation of the AMPK-RUNX2 pathway is a key mechanism through which SGLT-2 inhibitors enhance osteogenesis and improve bone formation.

Conversely, AGE accumulation under chronic hyperglycemia and insulin resistance plays a pivotal role in diabetic bone fragility. Excessive deposition of AGEs within the bone matrix induces abnormal collagen cross-linking, which stiffens the matrix and impairs its mechanical integrity, ultimately rendering the bone more brittle and fracture prone[148]. Moreover, AGEs bind to their receptor RAGE, triggering downstream oxidative stress and inflammatory signaling pathways, thereby suppressing osteoblast differentiation (e.g., reduced ALP and COL1 synthesis) and potentially enhancing osteoclast activity[148]. By improving glycemic control, SGLT-2 inhibitors reduce the overproduction of AGEs and downregulate RAGE expression in target tissues[18]. Some reviews have shown that antidiabetic agents, including SGLT-2 inhibitors, can attenuate AGE formation, suppress RAGE expression on cell surfaces, and increase soluble RAGE levels, thereby mitigating AGE-RAGE-mediated skeletal damage[18]. These findings suggest that SGLT-2 inhibitors may protect bone cells from glucotoxicity and reduce AGE-induced matrix brittleness and cellular dysfunction.

In addition to promoting osteoblast activity, SGLT-2 inhibitors also exert inhibitory effects on osteoclastogenesis. Diabetes is often associated with uncoupled bone remodeling, characterized by decreased osteoblast activity and/or increased osteoclast activation. Osteoclast precursor cells (e.g., RANKL-induced RAW264.7 cells) exposed to high-glucose conditions exhibit increased expression of osteoclast markers, which can be attenuated by the addition of canagliflozin to the culture medium. This treatment downregulates genes associated with osteoclast differentiation, including tartrate-resistant acid phosphatase (TRAP) and cathepsin K[147]. In the femoral bone tissue of T2DM rats, the number of TRAP-positive osteoclasts was significantly lower in the canagliflozin-treated group than in the untreated control group. These findings suggest that SGLT-2 inhibitors suppress osteoclast formation and activity in hyperglycemic environments, potentially by reducing AGE-RAGE signaling in osteoclast precursors and improving the bone marrow microenvironment. Furthermore, improved glycemic control and enhanced osteogenesis may increase the ratio of osteoprotegerin/RANKL secreted by stromal cells, thereby inhibiting osteoclastogenesis and reestablishing balanced bone remodeling.

Taken together, SGLT-2 inhibitors protect the skeleton through multiple mechanisms: Activating energy metabolism pathways in osteoblasts to promote bone formation (AMPK-RUNX2), alleviating glucotoxicity and oxidative stress to preserve bone matrix integrity (via suppression of AGE-RAGE signaling), and inhibiting excessive osteoclast activation under hyperglycemic conditions. Converging evidence from animal and cell-based studies suggests that SGLT-2 inhibitors restore the coupling balance of bone remodeling, upregulate the expression of bone formation markers (e.g., OCN), and reduce the expression of bone resorption markers, thereby mitigating diabetes-induced bone loss and mechanical weakening.

Comparative bone effects of different SGLT-2 inhibitors

Different types of SGLT-2 inhibitors may exert varying degrees of skeletal protection. Recent head-to-head animal studies have directly compared the effects of dapagliflozin, empagliflozin, and canagliflozin on bone parameters in diabetic models, revealing differences in their potency and specific targets of action[146,147]. Overall, in certain studies, canagliflozin has shown more pronounced improvements in bone microarchitecture, including increases in trabecular volume and cortical thickness[144]. Dapagliflozin also appears to enhance the biomechanical properties of bone, particularly in terms of improving bone material strength[21]. In contrast, the skeletal effects of empagliflozin are generally neutral or modest, with some models showing no significant improvements in bone parameters[144]. Importantly, these differences may be influenced by variations in experimental models, dosing regimens, and treatment durations and thus cannot be directly extrapolated to clinical outcomes. Nevertheless, the findings highlight subtle distinctions and potential advantages among individual SGLT-2 inhibitors in the regulation of bone metabolism. A summary of the differential effects of various SGLT-2 inhibitors on bone metabolism and mechanical properties in diabetic bone fragility models is provided in Table 5.

CLINICAL EVIDENCE OF SGLT-2 INHIBITORS IN BONE METABOLISM

Patients with T2DM frequently experience comorbid osteoporosis or an increased risk of fractures, making the impact of antidiabetic agents on bone metabolism a major clinical concern[149]. Given their unique mechanism of lowering blood glucose via renal glucose excretion, SGLT-2 inhibitors have raised safety concerns regarding bone health. The United States Food and Drug Administration issued a warning linking canagliflozin to an elevated fracture risk, which was echoed by the CANVAS trial reporting an increased incidence of fractures in the canagliflozin group[150,151]. However, subsequent large-scale trials and observational studies have produced inconsistent findings.

RCTs: Effects on BTMs and BMD

Multiple RCTs and meta-analyses have shown that SGLT-2 inhibitors exert minimal or neutral effects on BTMs and BMD. A systematic review and meta-analysis involving 20 RCTs and 12764 participants revealed no significant alterations in major BTMs associated with SGLT-2 inhibitor use, including the bone resorption marker CTX and the bone formation marker procollagen P1NP[152]. Specifically, the weighted mean difference for CTX was +0.04 (95%CI: -0.02 to -0.09), and for P1NP, it was +1.06 (95%CI: -0.44 to -2.57), with no statistically significant differences compared with controls. Additionally, there were no significant differences in the serum levels of PTH, calcium, phosphorus, or other bone metabolism-related indices between the treatment and control groups.

With respect to BMD, the same meta-analysis revealed no substantial effect of SGLT-2 inhibitors on the lumbar spine, femoral neck, total hip, or distal radius BMD. These findings suggest that, in the short to medium term, SGLT-2 inhibitors do not cause detectable bone loss in RCTs. Notably, while most RCTs have not reported adverse skeletal events, the CANVAS trial revealed that canagliflozin was associated with a 60% increase in fracture risk (13.0 vs 8.3 per 1000 patient-years)[150]. However, this risk was not consistently replicated in later studies. For example, in the CREDENCE trial involving patients with DKD, fracture incidence was not significantly different between the canagliflozin and placebo groups (HR = 0.98, 95%CI: 0.70 to -1.37)[153]. Overall, RCT evidence since 2020 suggests that SGLT-2 inhibitors exert a neutral effect on bone turnover and bone mass without confirming a clear risk of bone loss or abnormal bone remodeling[152].

Large-scale observational studies and real-world evidence on fracture risk

Real-world studies provide complementary evidence regarding the fracture risk associated with SGLT-2 inhibitors. Most large-scale cohort studies have shown that SGLT-2 inhibitors do not significantly increase the overall incidence of fractures. In a new-user cohort study of the United States Medicare population aged ≥ 65 years, fracture risk was compared between new initiators of SGLT-2 inhibitors and those of DPP-4 inhibitors (DPP-4is) or glucagon-like peptide-1 receptor agonists (GLP-1RAs). No increased risk of fractures was observed in the SGLT-2 inhibitor group. Specifically, the HR for fractures was 0.90 (95%CI: 0.73-1.11) for DPP-4is and 1.00 (95%CI: 0.80-1.25) for GLP-1RAs, with no statistically significant differences. These findings were consistent across subgroups stratified by sex, frailty status, and age[17].

Similarly, a propensity score-matched cohort study based on Taiwan’s National Health Insurance Research Database (with a maximum follow-up of approximately 2 years, n = 21155) reported no increase in overall fracture risk associated with SGLT-2 inhibitors compared with that associated with DPP-4is. The HR for major osteoporotic fractures (hip, vertebrae, radius, and shoulder) was 0.89 (95%CI: 0.80-1.00), and for nonmajor fractures, the HR was also 0.89 (95%CI: 0.81-0.98). These findings remained consistent even after a 180-day lag for fracture events was implemented to exclude the potential early impact of osmotic diuresis, further supporting the skeletal safety of SGLT-2 inhibitors[154].

Moreover, integrated analyses of RCTs suggest that the transient increase in fracture risk was largely limited to the CANVAS cardiovascular outcome trial. In that study, the incidence of fractures in the canagliflozin group began to rise within the first few weeks of treatment, whereas fracture rates were similar between the treatment and control groups in subsequent non-CANVAS trials. At the longest exposure time point, the fracture HR for canagliflozin approached 1.0. Researchers speculate that this early risk peak may be attributed to treatment-induced osmotic diuresis, orthostatic hypotension, and fall-related injuries. Over time, the fracture risk appears to converge with that of the control group[155].

Heterogeneity in patient subgroups has also been reported in observational studies. A nationwide cohort study in Korea focusing on elderly women (mean age: 72 years) revealed that users of SGLT-2 inhibitors had a greater incidence of vertebral fractures than nonusers did (19.2 vs 13.8 per 1000 person-years), with an HR of 1.40 (95%CI: 1.00-1.96, P = 0.049), suggesting a potential increase in vertebral fracture risk[156]. However, no significant differences were observed in hip or other nonvertebral fractures in the same study[156]. In summary, most real-world studies have not shown an increased risk of fractures, including hip and vertebral fractures, with SGLT-2 inhibitors. Some studies even suggest that the risk may be lower than that associated with certain comparator drugs. Nevertheless, signals of increased vertebral fracture risk in specific subpopulations, such as elderly women, highlight the need for enhanced monitoring in these groups.

Network meta-analyses: Comparative fracture risk across antidiabetic agents

Network meta-analyses comparing SGLT-2 inhibitors with other major antidiabetic drugs provide valuable insight into the risk ratio (RR) of fractures across drug classes. Overall, SGLT-2 inhibitors, GLP-1RAs, and DPP-4is confer comparable fracture risk, whereas thiazolidinediones (TZDs) are associated with a significantly increased risk[149]. For example, a network meta-analysis by Chai et al[157], integrating 177 RCTs involving over 160000 patients, reported no significant differences in fracture risk among newer drug classes compared with insulin. The odds ratios for fractures with SGLT-2 inhibitors vs DPP-4is, GLP-1RAs, or insulin were all approximately 1, supporting an overall neutral bone safety profile for these agents in RCT settings[157].

Another network meta-analysis focusing on real-world cohort data yielded interesting findings. Mostafa and Alrasheed[158] included 13 population-based cohort studies encompassing more than 1.06 million patients and compared the incidence of fractures associated with SGLT-2 inhibitors, GLP-1RAs, and DPP-4is. The use of SGLT-2 inhibitors was significantly associated with a reduced fracture risk, especially when SGLT-2 inhibitors were used in combination with other antidiabetic medications, with an estimated 87% reduction in risk. As monotherapy, SGLT-2 inhibitors were associated with a 67% risk reduction, GLP-1RAs with approximately 60%, and DPP-4is with approximately 55%[158]. The authors hypothesized that these newer agents may confer a lower overall fracture rate than traditional drugs such as sulfonylureas or insulin. However, they also cautioned that the observed benefits might reflect prescription bias or patient selection effects, necessitating cautious interpretation.

When findings across studies are integrated, TZDs consistently have the most detrimental impact on bone health, whereas SGLT-2 inhibitors do not seem inferior to GLP-1RAs or DPP-4is. In a Bayesian network meta-analysis, GLP-1RAs were ranked as having the lowest fracture risk, whereas TZDs were associated with the highest fracture risk. Compared with TZDs, GLP-1RAs reduce fracture risk by approximately 50% (RR = 0.50, 95%CI: 0.31–0.79), and sulfonylureas reduce fracture risk by approximately 44% (RR = 0.56, 95%CI: 0.41-0.77)[159]. Notably, this analysis also indicated that SGLT-2 inhibitors might confer a slightly greater fracture risk than GLP-1RAs (RR = 1.50, 95%CI: 1.05-2.16), although the fracture risk relative to that of neutral comparators such as DPP-4is or placebo did not reach statistical significance. These results suggest a spectrum of fracture risk across antidiabetic agents: TZDs confer the greatest harm, GLP-1RAs may offer modest skeletal protection, whereas SGLT-2 inhibitors and DPP-4is generally exhibit a neutral safety profile[149]. Table 6 summarizes the RR estimates for fractures across different drug classes and the corresponding levels of evidence.

Table 6 Relative effects of common antidiabetic medications on fracture risk.

Drug class	Effect estimates and 95%CI	Study type	Ref.
SGLT-2 inhibitors vs placebo	HR = 0.98 (95%CI: 0.70-1.37)	Randomized controlled trial	Perkovic et al[153]
SGLT-2 inhibitors vs DPP-4 inhibitors	HR = 0.90 (95%CI: 0.73-1.11)	Cohort study	Zhuo et al[17]
SGLT-2 inhibitors vs GLP-1 RAs	HR = 1.00 (95%CI: 0.80-1.25)	Cohort study	Zhuo et al[17]
GLP-1 receptor agonists vs placebo	OR = 1.27 (95%CI: 0.88-1.83)	Systematic review/network meta-analysis	Chai et al[157]
DPP-4 inhibitors vs placebo	RR = 1.44 (95%CI: 1.04-1.98)	Network meta-analysis	Tsai et al[159]
TZDs (e.g., pioglitazone) vs placebo	RR = 1.21 (95%CI: 1.01-1.45)	Systematic review/meta-analysis	Azhari and Dawson[162]

SGLT-2: Sodium-glucose cotransporter 2; DPP-4: Dipeptidyl peptidase-4; GLP-1 RA: Glucagon-like peptide-1 receptor agonist; TZD: Thiazolidinedione; CI: Confidence interval; HR: Hazard ratio; OR: Odds ratio; RR: Risk ratio.

Bone safety in special populations

Postmenopausal women: As postmenopausal women are inherently at increased risk for osteoporosis and fractures, they represent a key population for evaluating skeletal safety. A nationwide Korean study specifically assessed fracture risk among postmenopausal women with diabetes and compared SGLT-2 inhibitors with other antidiabetic agents. In women with a mean age of 60 years, SGLT-2 inhibitors were not associated with increased fracture risk; in fact, they were linked to a lower risk than DPP-4is were (HR = 0.78, 95%CI: 0.72-0.84) and showed comparable risk to GLP-1 receptor agonists[160]. This large real-world study offers valuable reassurance regarding the use of SGLT-2 inhibitors in postmenopausal patients. However, another study focusing on older women (aged ≥ 65 years) warrants caution. After patients on GLP-1RAs and TZDs were excluded, SGLT-2 inhibitor users presented a slightly elevated risk of vertebral fractures (HR approximately is 1.40, P approximately is 0.049), although no significant difference was observed for hip or other nonspinal fractures[156]. Such discrepancies may stem from differences in study populations and comparator drugs. Overall, the current evidence does not indicate a substantial increase in overall fracture risk among postmenopausal women using SGLT-2 inhibitors, but vigilance is advised in elderly women with heightened skeletal fragility, especially concerning vertebral fractures.

Patients with CKD: Individuals with CKD commonly experience mineral and bone disorders, which may alter their skeletal response to antidiabetic medications. CKD itself increases fracture risk, and impaired renal function could exacerbate phosphate and calcium metabolism disturbances potentially linked to SGLT-2 inhibitors. Cowan et al[161] analyzed health data from Ontario, Canada, focusing on elderly patients with T2DM and comorbid CKD. Among nearly 38000 propensity score-matched new users of SGLT-2 inhibitors and DPP-4is, no significant difference in 1-year cumulative fracture incidence was observed (weighted HR = 0.88, 95%CI: 0.88-1.00), with comparable risk across different strata of renal function (no significant interaction). These findings suggest that even among CKD patients, including those with eGFRs as low as 30 mL/minute/1.73 m², SGLT-2 inhibitors do not confer increased fracture risk relative to DPP-4is. Similarly, the CREDENCE RCT by Perkovic et al[153] did not observe increased fracture events during periods of impaired renal function. Collectively, the current evidence supports the skeletal safety of SGLT-2 inhibitors in patients with moderate CKD. However, owing to limited usage in advanced CKD patients (eGFR < 30), data remain scarce, and further research is warranted.

High-fracture-risk phenotypes: Patients with a long duration of diabetes, established osteoporosis, or a prior history of fractures theoretically warrant closer scrutiny regarding skeletal safety. Such high-risk individuals are often excluded or underrepresented in clinical trials. Indirect evidence suggests that even among elderly patients with baseline frailty or comorbid insulin use, SGLT-2 inhibitors do not appear to significantly increase fracture risk. A JAMA cohort study demonstrated no significant fracture risk differences between SGLT-2 inhibitor users and controls across subgroups defined by frailty status or concurrent insulin use[17]. This implies that even in high-risk populations with prolonged disease duration and multiple comorbidities, SGLT-2 inhibitors do not present disproportionate skeletal risk compared with other glucose-lowering drugs. In contrast, studies on pioglitazone have highlighted the need for extra caution in patients with preexisting fractures or bone disease[162], but no such adverse signals have been reported for SGLT-2 inhibitors to date. While direct evidence remains limited, it is reasonable to conclude that in patients with elevated fracture risk (e.g., prior fractures or severe osteoporosis), SGLT-2 inhibitors should not be categorically avoided. Rather, a comprehensive risk-benefit assessment should be performed, including close monitoring of BMD and biochemical markers. Notably, the cardiovascular and renal benefits of SGLT-2 inhibitors may outweigh potential skeletal concerns in the context of holistic diabetes management.

Mechanistic insights and risk signals related to bone safety

Volume depletion and falls: SGLT-2 inhibitors lower blood glucose through osmotic diuresis, which can result in mild fluid volume loss and reductions in blood pressure[163]. Theoretically, this may increase the risk of dizziness and falls, thereby indirectly increasing the risk of fractures, particularly among elderly patients. As previously mentioned, a transient increase in fracture incidence during early treatment has been attributed to falls induced by initial volume depletion[155,160]. However, long-term follow-up data have not revealed a persistent peak in fall-related fractures. Patients should be advised to maintain adequate hydration and to avoid postural hypotension to mitigate fall risk[152,164].

DKA: SGLT2 inhibitors confer a relative increase in DKA risk, yet the absolute incidence in T2DM is very low, approximately 1-6 events/1000 patientyears[165,166]. A metaanalysis including 2.96 million individuals reported a 33% higher risk among users (HR = 1.33, 95%CI: 1.14-1.55)[166]. Similarly, a 2024 network metaanalysis of 20 RCTs (7183 participants) identified 135 DKA events (1.88%), with dapagliflozin 5 mg, empagliflozin 10 mg and both sotagliflozin doses exhibiting the strongest signals[167]. During DKA, systemic acidosis (pH < 7.30) mobilizes bone salts to buffer excess hydrogen ions. Classical and contemporary studies have demonstrated that metabolic acidosis dissolves carbonaterich apatite, releases calcium and phosphate, and activates osteoclastic resorption via protonsensing receptors[168,169]. Small clinical series in adults hospitalized for DKA support these mechanisms: Markers of resorption (βCTX, TRAP5b) peak during acidosis and normalize after treatment, whereas the formation marker P1NP rebounds later, indicating a transiently uncoupled highturnover state[170,171].The chronic skeletal consequences of SGLT-2 inhibitor therapy remain under investigation. Canagliflozin has been linked to increased bone turnover, hip BMD loss, and micro-architectural deterioration in some preclinical and clinical studies, effects likely mediated indirectly by changes in calcium-phosphate balance, weight loss, and secondary hyperparathyroidism[172]. Despite these findings, most large meta-analyses have not shown a class-wide increase in fracture risk[172]. Regarding DKA specifically, available evidence suggests that episodes are rare, transient, and not followed by sustained skeletal deterioration. A recent meta-analysis reported that DKA patients who were SGLT-2 inhibitor users had lower glucose and lactate levels and no differences in hospital stay, intensive care unit admission, or mortality than non-users did, implying a favorable prognosis once acidosis was corrected[173]. The current data indicate that SGLT-2 inhibitor-associated DKA is an uncommon, acute complication with predominantly transient effects on bone. Its chronic skeletal sequelae remain unclear. Prospective, long-term studies integrating bone density, micro-architecture, and turnover markers are warranted to clarify whether any residual risk persists beyond the acute phase.

Electrolyte disturbances and hormonal changes: SGLT-2 inhibitors can induce alterations in mineral metabolism through renal effects. One of the most notable effects is reduced phosphate excretion. Studies have shown that SGLT-2 inhibitors enhance phosphate reabsorption in the proximal tubules, resulting in a mild elevation in serum phosphate levels[174]. Elevated phosphate subsequently stimulates PTH and FGF-23, leading to secondary hyperparathyroidism and decreased 1,25-dihydroxyvitamin D levels, which could promote bone resorption[21]. Short-term clinical trials have revealed that several weeks of dapagliflozin treatment significantly increased the serum phosphate, PTH, and FGF-23 Levels, whereas the serum calcium and vitamin D levels decreased. These findings suggest that SGLT-2 inhibitors may promote bone turnover through disruptions in the mineral-hormonal axis. Interestingly, biochemical outcomes vary: Some studies reported elevated resorption markers such as CTX alongside reduced bone-specific ALP[130], whereas others reported no significant changes[152]. Overall, phosphate retention, hypocalcemia, and compensatory increases in PTH induced by SGLT-2 inhibitors are considered potential bone safety signals[161]. However, the magnitude and duration of these hormonal responses appear limited and insufficient to cause overt bone loss, which is consistent with the lack of significant changes in BMD or fracture risk observed in clinical trials and real-world studies. Nevertheless, clinicians should monitor the calcium-phosphate balance in long-term users and consider the periodic assessment of BTMs for the early detection of abnormalities.

Other metabolic effects: The weight loss associated with SGLT-2 inhibitor therapy presents a double-edged sword. On the one hand, reduced body weight may diminish mechanical loading on the skeleton, potentially impairing bone mass maintenance. On the other hand, reduced adiposity may alleviate adipose tissue-derived inflammatory cytokine interference with bone metabolism[130,152,175]. Preclinical studies suggest that SGLT-2 inhibitors may reduce the accumulation of AGEs under hyperglycemic conditions, thereby improving diabetic bone microarchitecture[147]. Thus, within the context of diabetic bone disease, the net skeletal effect of SGLT-2 inhibitors may hinge upon the balance of multiple mechanisms: Short-term factors (e.g., diuresis, mineral disturbances) may slightly promote bone resorption, whereas long-term benefits (e.g., glycemic control, reduced AGE accumulation) may support bone quality. Current human evidence suggests an overall neutral skeletal profile, indicating that these opposing mechanisms may counteract or exert only modest effects[152].

MULTIPLE MECHANISMS BY WHICH SGLT-2 INHIBITORS REGULATE THE AGE-COL1 AXIS TO AMELIORATE DIABETIC BONE FRAGILITY

Reduction in hyperglycemia and carbonyl stress

Persistent hyperglycemia accelerates the accumulation of AGEs within the bone matrix, particularly in bone collagen, where it leads to the formation of cross-links such as pentosidine and CML. These excessive AGEs directly increase collagen fiber rigidity, rendering bone tissue more brittle and concurrently triggering oxidative stress imbalances in the bone microenvironment[160]. By lowering blood glucose levels, SGLT-2 inhibitors reduce the overproduction of AGE precursors such as MGO, thereby suppressing the formation of AGEs and their interaction with the RAGE at the source[176]. This action effectively mitigates the “carbonyl stress” driven by the accumulation of reactive carbonyl intermediates. As the AGE burden decreases, the extent of nonenzymatic glycation of bone collagen is alleviated, leading to deactivation of the aberrantly activated AGE-COL1 axis and a consequent reduction in biomechanical impairment of the bone matrix.

Attenuation of oxidative stress and lipid peroxidation

The interaction between AGEs and RAGE excessively increases the intracellular production of ROS and proinflammatory mediators, thereby increasing oxidative stress levels within bone tissue[160]. This oxidative cascade, initiated by the AGE-RAGE axis, contributes to further damage to both the bone matrix and resident bone cells. SGLT-2 inhibitors alleviate oxidative stress at their source by reducing AGE accumulation and interfering with RAGE-mediated signaling pathways under hyperglycemic conditions. For example, in diabetic animal models, treatment with empagliflozin significantly downregulated the expression of AGEs and RAGE in bone tissue and reduced the levels of oxidative damage biomarkers such as 8-hydroxy-2’-deoxyguanosine[139]. Concurrently, SGLT-2 inhibitors have been shown to restore the function of endogenous antioxidant systems by increasing the activity of key antioxidant enzymes, including SOD, glutathione peroxidase, and catalase[139]. As ROS overproduction is suppressed and antioxidant defenses are reinforced, the accumulation of lipid peroxidation end products such as malondialdehyde is also markedly reduced[139]. The overall reduction in oxidative stress and lipid peroxidation helps to protect bone cells and the extracellular matrix from further oxidative damage in the diabetic milieu.

Regulation of ketone levels and mineral metabolism

SGLT-2 inhibitors can induce a state of mild ketosis characterized by a moderate increase in circulating BHB levels, particularly during fasting. Notably, this mild increase in ketone bodies is considered potentially beneficial for bone metabolism. Studies have shown that BHB can reduce bone loss and improve osteoporosis by promoting osteoblast differentiation and inhibiting osteoclast formation[177]. The underlying mechanisms may involve the modulation of key cellular signaling pathways, for example, the suppression of the transcription factor nuclear factor of activated T cells 1, which is essential for osteoclast differentiation, and the attenuation of inflammatory responses during bone resorption via the inhibition of the activation of the NLRP3 (NACHT, LRR and PYD domains-containing protein 3) inflammasome[177]. In addition, as an alternative energy substrate, BHB reduces reliance on glucose metabolism and the associated formation of harmful byproducts, thereby potentially decreasing AGE accumulation and the risk of lipid peroxidation. Evidence from cardiovascular disease models has confirmed that SGLT-2 inhibitor-induced BHB elevation is accompanied by reduced ROS production and amelioration of inflammatory responses[176]. Thus, moderate ketonemia may contribute to bone homeostasis through multiple protective pathways.

Harmful effects of hyperglycemic conditions on the skeleton

On the other hand, the impact of SGLT-2 inhibitors on mineral metabolism warrants close attention. These agents act on the renal proximal tubules to increase phosphate reabsorption, resulting in a mild elevation in serum phosphate levels and stimulating osteoblasts to secrete FGF-23[21]. FGF-23, in turn, inhibits the activity of renal 1α-hydroxylase, leading to a decrease in the production of active vitamin D (1,25-dihydroxyvitamin D), which reduces intestinal calcium absorption[20]. Deficiencies in vitamin D and hyperphosphatemia may induce compensatory changes in PTH secretion. In the long term, these alterations may impair bone mineralization and increase fracture risk. Therefore, during SGLT-2 inhibitor therapy, monitoring the calcium-phosphate balance and the FGF-23/PTH axis is essential to prevent potential mineralization disorders.

In contrast, several studies and clinical trials have reported a mild increase in serum magnesium levels among patients with T2DM treated with SGLT-2 inhibitors. Hypomagnesemia is a common complication of diabetes, primarily due to renal magnesium wasting, and can impair osteoblast activity, increase osteoclastogenesis, and promote chronic inflammation through elevated proinflammatory cytokines such as TNF-α and IL-6. Since approximately two-thirds of total body magnesium is stored in bone, restoring magnesium levels may directly support bone formation and improve mechanical strength. Consequently, by alleviating magnesium deficiency, SGLT-2 inhibitors may help improve the inflammatory and mineralization milieu of the diabetic bone microenvironment, potentially counterbalancing the adverse effects of phosphate retention.

Anti-inflammatory effects via downregulation of the RAGE pathway

The interaction between AGEs and RAGE not only drives oxidative stress but also amplifies chronic inflammation in bone tissue by activating intracellular inflammatory cascades. The binding of AGEs to RAGE and certain pattern recognition receptors, such as TLR4, triggers key signaling pathways, including NF-κB and MAPK, leading to the upregulation and secretion of proinflammatory cytokines such as IL-1, IL-6, and TNF-α[178]. This state of chronic low-grade inflammation is considered a critical contributor to diabetic bone fragility. SGLT-2 inhibitors, in addition to lowering the AGE burden, have been shown to downregulate the expression of RAGE and TLR4 in target tissues, thereby significantly suppressing abnormal NF-κB activation associated with the AGE/RAGE axis[178]. Notably, this anti-inflammatory effect may occur partially independently of glucose reduction. SGLT-2 inhibitors can directly modulate inflammatory signaling pathways to reduce cytokine production[179]. For example, canagliflozin has been found to lower TLR4 expression in macrophages and block NF-κB activation, thereby inhibiting the secretion of proinflammatory mediators. By attenuating the RAGE-NF-κB/MAPK signaling axis, SGLT-2 inhibitors effectively alleviate the chronic inflammatory microenvironment in diabetic bone tissue, helping preserve the physiological functions of osteoblasts, osteocytes, and osteoclasts. This contributes to the reestablishment of balanced bone remodeling under diabetic conditions[18,142,180].

TRANSLATIONAL PROSPECTS OF SGLT-2 INHIBITORS IN DIABETIC OSTEOPATHY

Diabetic osteopathy is characterized by bone fragility resulting from multiscale damage ranging from molecular alterations to tissue-level structural deterioration. However, clinical indices capable of comprehensively capturing this “multiscale” pathological process are still lacking. The accumulation of AGEs in the bone matrix is a central mechanism, and pentosidine, a typical AGE cross-linking marker, has been associated with fracture risk. Nonetheless, it has not yet been incorporated into routine risk stratification or therapeutic efficacy assessments[110,181]. Conventional BTMs (e.g., procollagen type I N-terminal propeptide, OCN, TRAP5b, and CTX) are generally suppressed in diabetic patients, limiting their predictive value for fractures and rendering them insufficient to reflect AGE-COL1-mediated deterioration of bone quality[110].

Advanced imaging techniques such as HR-pQCT have revealed microarchitectural damage in T2DM, predominantly characterized by cortical thinning and increased porosity, whereas trabecular bone appears relatively preserved[182]. This aligns with the paradox of “normal BMD but impaired bone quality” frequently observed in diabetic individuals. However, the broader clinical utility of HR-pQCT is limited by accessibility and standardization issues, whereas higher-resolution modalities such as nano-CT and micro magnetic resonance imaging remain largely confined to research settings and lack large-scale clinical validation[181].

With respect to direct evidence, human data remain scarce due to the invasive nature of bone biopsy. Limited studies of iliac crest biopsies from patients with T1DM have confirmed that bone tissue accumulation of pentosidine correlates with fracture history[110], but the sample sizes are small, and even fewer data exist for patients with T2DM. Future efforts should include obtaining bone samples in surgical settings such as diabetic hip fractures, enabling systematic quantification of AGE-COL1 cross-links and microcracks. Coupled with molecular and cellular indicators, this may facilitate the construction of a robust validation framework. In summary, research on diabetic bone disease urgently requires the following: (1) The establishment of an integrated biomarker panel spanning molecular, cellular, and tissue levels; (2) The standardization and clinical implementation of high-resolution imaging technologies; and (3) Human tissue-based validation of the true impact of the AGE-COL1 axis on bone toughness. These breakthroughs are essential for enabling precision risk stratification and targeted therapeutic strategies.

FUTURE DIRECTIONS

Current interventions targeting AGE-mediated stiffening of COL1 in diabetic bone are increasingly directed toward a dual objective: Reversing established cross-links and limiting their further formation. Chemical AGE breakers such as ALT-711 have been shown to significantly reduce the total AGE burden in animal bone tissues; however, their ability to improve bone mechanical properties remains limited, highlighting the urgent need for more selective and efficacious targeting strategies[105]. Enzyme engineering is considered a promising frontier, potentially enabling the development of high-affinity, low-toxicity hydrolases capable of cleaving glucosepane and related AGE cross-links directly within the bone matrix. Given the limitations of single-agent interventions, a “dual-pathway” strategy has been proposed: Upstream glycemic control via SGLT-2 inhibitors to reduce the flux of AGE precursors, coupled with the downstream use of AGE inhibitors or breakers to eliminate or block established cross-links[172,183]. Achieving true synergism will require rational optimization of treatment sequencing and dosing, along with careful monitoring of renal and hemodynamic effects.

Moreover, mechanistic research has revealed novel cellular targets for intervention. For example, AGE-induced ferroptosis in osteoblasts may be mitigated through iron chelators; the suppression of the Wnt/β-catenin signaling pathway by AGEs could be counteracted by dickkopf 1 or sclerostin antagonists or through direct pathway activation. Additionally, senolytic therapies such as dasatinib plus quercetin have shown potential in rejuvenating the diabetic bone marrow microenvironment and enhancing osteogenesis[85,184,185].

One unresolved issue concerns the skeletal impact of chronic low-grade ketosis induced by SGLT-2 inhibitors. Some studies suggest that increased acid load and bone marrow adiposity may impair osteogenesis, especially during developmental stages. Conversely, other studies have reported that BHB suppresses osteoclast differentiation and inflammatory resorption, indicating potential bone-protective effects[132,177,186]. Future investigations should systematically assess the bidirectional role of BHB in bone formation and resorption, clarify its dose-response dynamics and epigenetic influence, and ultimately establish a safety threshold for SGLT-2 inhibitor use in skeletal contexts. Overall, emerging interventions, ranging from AGE cross-link reversal and pharmacologic synergy to osteocellular targeting and energy metabolism modulation, are converging to construct a multidimensional and systems-based therapeutic framework for diabetic skeletal fragility. Long-term animal studies and clinical trials are essential to validate the feasibility and efficacy of these novel strategies.

CONCLUSION

A central “paradox” of diabetic osteoporosis lies in the observation that patients often exhibit normal or even elevated BMD but remain at increased risk of fractures. Growing evidence supports the AGE-COL1 axis as a core mechanism underlying this paradox. Chronic hyperglycemia promotes the nonenzymatic glycation of COL1, resulting in the excessive accumulation of AGEs and the formation of abnormal cross-links. These changes increase the brittleness of the bone matrix, compromising bone quality even when bone mass is preserved. This mechanism plays a key role throughout the progression of diabetic bone disease: AGE cross-linking not only directly impairs collagen elasticity but also activates downstream oxidative stress and cellular dysfunction through interaction with RAGE, thus explaining why elevated fracture risk in diabetes patients often eludes detection via conventional BMD measurements.

SGLT-2 inhibitors, an emerging class of antidiabetic agents, have shown potential benefits for skeletal health. Preclinical and early clinical studies suggest that SGLT-2 inhibitors do not significantly exacerbate bone loss despite promoting glycemic control and weight reduction. In contrast, improvements in bone mass and mechanical strength have been observed in diabetic osteoporosis models. These findings suggest that SGLT-2 inhibitors may confer bone benefits indirectly through comprehensive metabolic modulation. However, current evidence remains largely limited to short-term observations and post hoc analyses. The relative contributions of specific mechanisms, such as reduced AGE accumulation, improved adipose distribution or altered renal calcium-phosphate handling, remain to be fully delineated.

Before expanding the use of SGLT-2 inhibitors into the domain of osteoporosis treatment, mechanism-driven and multiparameter clinical validation is necessary. First, prospective studies in representative patient populations are needed to comprehensively assess changes in bone turnover, biomechanics, microarchitecture, and mass to confirm the durability and clinical relevance of any improvements in bone quality. Second, a combination of imaging modalities and biochemical markers should be used to elucidate the drug’s regulatory effects on key bone metabolic pathways, including the AGE-RAGE axis, Wnt signaling, inflammatory mediators, and marrow adiposity, thereby providing mechanistic proof of benefit. Finally, careful evaluation of safety and therapeutic boundaries is essential. In older adults or individuals with preexisting high fracture risk, routine monitoring of BMD and fracture incidence is warranted during SGLT-2 inhibitor therapy. In summary, the identification of the AGE-COL1 axis offers a novel perspective on diabetic bone fragility, and SGLT-2 inhibitors represent a promising interventional approach. Only through robust mechanistic and clinical validation can the extension of these agents into the osteoporosis treatment landscape be grounded in a solid evidence base.